The Morphology of Proto-Indo-European

Jesse Lundquist & Anthony D. Yates

University of California, Los Angeles
jlundquist@[Link] & adyates@[Link]

§1 Introduction
§2 PIE nominal morphology
§3 PIE morphophonology
§4 PIE verbal morphology
§5 Conclusions

§1 Introduction

This chapter aims to provide an updated overview of Proto-Indo-European (PIE) morphology, broadly establishing
the typological properties of the reconstructible system, and offering some new perspectives on certain contro-
versial aspects of this reconstruction. In this respect, we hope to make this chapter both relevant and accessible
to several audiences: to students of IE languages looking to understand which categories are reconstructed for the
proto-language and what their formal exponents looked like, so that they may see the daughter languages in the
light of their diachronic developments; to specialists in IE linguistics, who may be interested in a “state-of-the-art”
assessment of long-standing issues in PIE morphology and, to a lesser extent, the proposals we advance here; and
to general linguists pursuing typological, historical, or theoretical questions who wish to see what kinds of mor-
phological categories are reconstructed for the IE languages, on what basis they are reconstructed, and what types
of analyses have been proposed.
Considerations of length prohibit a comprehensive survey of PIE morphology, a subject which, even more
than phonology and much more than syntax, has received tremendous attention in the long history of the field. In
a treatment of this size we simply cannot do justice to the wealth of reconstructed PIE morphology; consider that
as of 2017 the projected coverage of morphology in the series “Indogermanische Grammatik” ([Link]. Lindner; see
[Link]) encompasses six volumes! Similarly, we cannot provide full discussion of the breadth of
vigorous and informed controversy that envelops certain areas of PIE morphology; only salient features will be
examined, with in-depth treatment reserved for areas of particularly great controversy. Readers looking for an
introduction to the state of the field may consult Fortson (2010) and Clackson (2007), and for a more extensive
overview Meier-Brügger (2010), the last with rich bibliography. The most extensive handbook of PIE morphol-
ogy to date remains Brugmann and Delbrück (1906), although it is necessarily antiquated (especially in lacking
evidence from Anatolian and Tocharian) and is currently in the process of being replaced by the volumes of the
aforementioned “Indogermanische Grammatik;” note finally that a massive collection of bibliography on IE mor-
phology has been assembled by Heidermanns (2005).

§1.1 Methodological preliminaries In this chapter, we aim to describe the morphology of the last stage of the
proto-language that is the ancestor of all the IE languages (including the Tocharian and Anatolian branches of the
family), and that is thus directly reconstructible by application of the Comparative Method (e.g. Meillet 1925; Weiss
2014). We reserve the label PIE for this directly reconstructible stage, thereby distinguishing it from the common
ancestor of the non-Anatolian IE languages (including Tocharian), an entity referred to here as Proto-Nuclear-
Indo-European (PNIE), whose inner articulation remains difficult to define (our PNIE is equivalent to what other
scholars call “core PIE,” Germ. Restindogermanisch). We distinguish the label PIE from the still earlier stage of the
language reached via internal reconstruction on PIE data, which we refer to as pre-PIE. We will repeatedly have
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 2

occasion to consider the evidence for a given reconstruction on which almost all IE languages converge, with the
persistent exception of the Anatolian branch and, in some cases, the Tocharian languages as well. Some major
examples discussed below include, in the nominal system, the reconstruction of grammatical number and gender
distinctions, and in the verbal system, the status of many fundamental PNIE verbal categories, including the *s-
aorist, the optative, and (perhaps above all) the perfect. These divergences may indicate that Anatolian was the
earliest branch to “hive off” (in Watkins’ (1998:31) memorable phrase) from the ancestor of the other IE languages,
whose period of common unity after the departure of Anatolian allows for the possibility of shared innovations
that can thus be reconstructed for PNIE. This position amounts to a version of the “Indo-Hittite” hypothesis, first
proposed by Sturtevant (1929, 1933) and later championed by Cowgill (e.g. 2006:25). It is usually held that Tocharian
was, in turn, first to depart from PNIE; this assumption will be broadly followed here, although the view is less
universally held (see the discussions by Ringe and Jasanoff, both in this volume).

§1.2 Conventions We employ in this chapter the linguistic conventions and abbreviations standardly used in
Indo-European scholarship, which can be found in any of the handbooks noted in §1 above; we note here only
a few terms and symbols whose usage is not uniform across the field or which we use in a way that may depart
from standard practice. The most significant difference relates to the use of slant brackets (/…/). In some IE lan-
guages — in particular, those which are attested in non-alphabetic scripts (e.g. Mycenean Greek, Hittite) — it is
customary to cite linguistic forms in transliteration together with a “phonological transcription” enclosed in slant
brackets. We do not follow this practice here, instead reserving slant brackets for indicating underlying phonolog-
ical representations (in the generative sense; cf. Byrd, this volume); when forms cited in ordinary transliteration
require further clarification (as often with these non-alphabetic orthographies), approximate IPA transcriptions
enclosed in square brackets ([…]) are also provided. The symbols “→” and “←” indicate phonological mappings
between underlying and surface forms; the distinction between these two levels of representation becomes im-
portant, especially, in the discussion of PIE morphophonology in §3 below. Synchronic word-formation processes
are indicated by the symbols “⇒” and “⇐”; “>>” and “<<” denote that a combination of phonological and analogical
changes have occurred between two historical stages. The symbol “x ” marks a form or meaning that never existed
at any historical stage, but might be expected under a different phonological or morphological analysis. Finally,
we use a preceding asterisk “*” to mark reconstructed word forms, a preceding doubled asterisk “**” for word forms
in internally reconstructed state(s) of the proto-language (i.e. pre-PIE), and an asterisk following a word (i.e. “x*”)
to indicate that the particular form is not attested, but its existence is securely inferred from other attested forms.

§2 PIE nominal morphology

This section outlines the reconstructed morphology of the PIE noun as well as related nominal categories, such
as adjectives, pronouns, and adverbs. It is broadly organized as follows. We begin with the noun itself, discussing
inflectional morphology in §2.1. Pronouns are addressed in §2.2, and numbers, adverbs, and other related nominal
topics in §2.3. We discuss the derivational morphology of the PIE noun in §2.4, then turn in §2.5 to the closely-
related topic of PIE adjectives. The section concludes in §2.6 by examining nominal compounds.

§2.1 PIE nominal inflection The PIE nominal system was characterized by rich inflectional morphology. All
nouns and adjectives were grammatically specified for number, gender, and case. Number and case were expressed
by fusional inflectional case endings in which no separate markers for number or case can be distinguished. These
case endings (or “desinences”) were sensitive to the grammatical gender of the nominal stem to which they were
suffixed; for instance, grammatically neuter nouns selected different endings than animate nouns in a subset of
case forms. Adjectives modifying nouns generally exhibit concord with respect to all three nominal inflectional
features, and in the same vein nominal subjects trigger number agreement on verbs. As will become clear below,
this type of agreement plays a significant role in assessing certain difficult questions about the grammatical status
of number and gender distinctions in PIE.
PIE inflectional endings are also an important locus of morphophonological alternations: in certain nominal
classes, inflectional suffixes may induce changes in word-final segments (e.g. assimilation), in stem-internal vowels
(ablaut), and in the position of the (single) surface word accent. Throughout this section we offer brief remarks
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 3

on these issues when they arise; a fuller discussion — especially of word accent and its relationship to ablaut (or
“apophony’’; see also Byrd, this volume) — is presented in §3 below.

§2.1.1 Case PIE nouns inflected for case, the case ending of a noun denoting its grammatical role in the sen-
tence — e.g. subjects are marked with nominative case. Eight cases are standardly reconstructed for PIE, primar-
ily on the basis of Vedic Sanskrit where they remain formally distinct. These cases (and their basic functions) are:
nominative (subject); accusative (direct object); instrumental (means, accompaniment, agent); dative (recipient,
beneficiary, goal); ablative (source, separation); genitive (possessor); locative (in-/adessive, illative); and vocative
(direct address). All case endings are fusional exponents of case and number, e.g. athematic genitive singular *–
(o/e)s vs. plural *–oh1/3 om, which show no morpheme segmentable as an exponent of case [genitive] or of number
[singular, plural]. The nominative, accusative, and vocative are generally referred to together as the “strong” (or
“direct”) cases, in contrast to the rest, which are termed “weak” (or “oblique”); this terminology is useful for de-
scribing noun classes in which the strong cases share a single “strong stem” allomorph that differs from the “weak
stem” allomorph in accent or ablaut, and is thus standardly employed in analyses of PIE morphophonology (see
§3 below).
PIE noun inflection is characterized by a formal opposition between “thematic” stems (i.e. ending in the theme
vowel *–e/o–) and “athematic” stems (not ending in the theme vowel *–e/o–). Examples include thematic nomina-
tive singular *h2 ŕ̥tk̑-o-s ‘bear’ (Hitt. ḫart(a)gg-a-š, Ved. ŕ̥kṣ-a-s, Gk. árkt-o-s) vs. athematic *gw ów-s ‘cow’ (Ved. gaú-
s, Gk. boû-s).Thematic case endings are often transparently derived from athematic, such as nominative singular
*–os from *–o– + *–s, but there are some instances in which they diverge. For instance, the neuter nominative-
accusative singular in thematic nouns was realized by an ending *–om (e.g. Gk. zug-ón, Ved. yug-ám, Hitt. yuk-an
‘yoke’), while in athematic nouns, this case was zero-marked (Gk. dóru, Ved. dā ́ru, Hitt. tāru ‘wood’). However,
it is worth emphasizing that this distinction was purely formal: both athematic singular *–ø and thematic *–om
have the same function despite their phonological differences. Conversely, although ablative and genitive singular
are marked identically in athematic stems (*–o/es), these cases remain functionally distinct. This point is clearly
shown by the separate formal mergers of the ablative in the daughter languages, e.g. with the genitive in Greek,
but more unexpectedly, with the instrumental and locative in Latin; in the latter, a continuant of the thematic ab-
lative ending (likely *–oh1 ad; see discussion below) marks all three functions in thematic stems (athematic dative
*–ey functions similarly in athematic stems), while the genitive has its own marker. Nouns containing the PNIE
“feminine” suffixes *–ih2 /yeh2 – or *–(e)h2 exhibited unique inflectional patterns, similar but not identical to other
athematic nominal classes; their inflection is discussed separately in §2.1.3.
The table in (1) provides reconstructed PIE athematic and thematic inflectional paradigms for those case end-
ings whose status is relatively uncontroversial. Thematic inflection is presented with case endings joined to the
thematic vowel. As a cursory glance shows, there is significant agreement on the reconstruction of singular case
endings and the plural endings of the structural cases (nominative, accusative, genitive), although each has prob-
lematic or controversial aspects which are discussed below. The remaining plural endings are more uncertain;
these too are discussed below.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 4

Athematic Thematic
Animate Neuter Animate Neuter
Nominative *–s *–os
*–ø *–om
Accusative *–m *–om
Instrumental *–h1 , *–eh1 *-oh1

Singular
Dative *–ei *–ōi
(1) Ablative *–s, *–es, *–os; *–ti? *–oh1 ad
Genitive *–s, *–es, *–os *–os, *–osyo, *–oso
Locative *–i, *–ø *–oi
Vocative *–ø *–e
Nominative *–es *–ōs
Plural

*–h2 *–eh2
Accusative *–ms *–oms?
Genitive *–oh1/3 om, *–om? *–oh1/3 om
The PIE athematic animate nominative singular ending must be reconstructed as *–s, which is also the source
of thematic singular *–o-s (e.g. Gk. hípp-os, Ved. áśv-as ‘horse’; Hitt. išḫ-aš, Lat. er-us ‘master’; Lith. vil̃k-as, Goth.
wulf-s ‘wolf’), and appears to be continued in certain obstruent-final stems in several archaic IE languages, e.g.
Lat. rēx ‘king’ (< PIE *h3 rḗg̑-s); Gk. klṓp-s ‘thief’ (< *klṓp-s); Hitt. šīwaz ‘day’, CLuw. D tiwaz ‘Sun-god’ (< *díw-
ot-s). However, the ending has a phonologically restricted distribution — in particular, it is generally absent in
sonorant-final (or *s-final) stems, where the only case marker is a lengthened suffixal vowel: Gk. patḗr, Ved. pitā ́,
OLat. patēr < *ph2 tḗr ‘father’; Gk. kuṓn ‘dog’, Ved. ś(u)vā < *k̑(u)wṓn ‘dog’; Ved. úṣās, Aeol. Gk. aúōs < *h2 (é)us-ōs.
This lengthened vowel is due to Szemerényi’s Law, which is traditionally understood as a pre-PIE phonological rule
deleting *s in word-final vowel + sonorant (or *s) +*s sequences with compensatory lengthening of the preceding
vowel, i.e. pre-PIE **–V{R, s}s# > PIE *–V̄ {R, s} (cf. Szemerényi 1996:113–9 with references).
The PIE athematic nominative plural ending is straightforwardly reconstructible as *–es (never zero-grade x –
s, despite the consistent lack of surface accent; cf. §3 below). It is found after obstruents, e.g. Gk. pód-es ‘feet’,
Ved. pā ́d-as* (< PIE *pṓd-es; cf. §3.1 below) and sonorants, e.g. Gk. patér-es, Ved. pitár-as ‘fathers’ (< *ph2 tér-es);
Hitt. arki-eš ‘testicles’ (<< *h1 org̑ h -ey-es). The thematic plural ending is *–ōs, the long vowel likely arising from a
prehistoric contraction of **–o-es; it is reflected in e.g. Ved. vīr-ā ́s ‘men’; Goth. hund-os ‘dogs’; Osc. Núvlan-ús ‘men
of Nola’; Pal. mārḫ-aš ‘guests’. This ending has been replaced in many IE languages by the pronominal nominative
plural ending *–oi, e.g. Gk. hípp-oi, TB yakw-i ‘horses’; Lat. vir-ī ‘men’; OCS grad-i ‘cities’. This replacement is just
one example of the interplay between nominal and pronominal inflection that is characteristic of the development
of the IE languages.
The PIE athematic animate accusative singular ending is also securely reconstructed as *–m, which is evi-
dent in nouns with stem-final glides *i or *u, e.g. Ved. matí-m ‘mind’ (< PIE *mn̥ -tí-m ‘thinking’), Ved. gántu-m,
Lat. (ad)ventum (< PIE *gw (e)m-tu-m ‘going’), although the original state is partly obscured by the regular sound
change of *m > [n] in word-final position observed in several IE branches (Greek, Anatolian, Germanic). After
an obstruent, it surfaced as vocalic *–m̥ , e.g. Gk. póda, Lat. ped-em ‘foot’ (< PIE *pó/éd-m̥ ). The corresponding
thematic marker was *–om, e.g. Ved. áśv-am, OLat. equ-om, Gk. hípp-on ‘horse’ (< PIE*h1 ék̑w-om); Hitt. išḫ-ān
‘master’.
The thematic animate accusative plural ending is typically reconstructed as *–oms (or *–ons, although the
Anatolian accusative plural forms appear to require unassimilated *m), which would thus be a surface exception
to Szemerényi’s Law (and is therefore used as evidence for situating this rule in pre-PIE). Yet the (non-)assimilation
issue is just one question that complicates the reconstruction of this ending’s phonetic realization in PIE. A more
significant problem is that, while *–oms appears to be directly continued in dialectal (Gortynian Cretan) Gk. –ons
and Goth. –ans with retained *s, as well as in Dor. Gk. –ōs/Att. Gk. –ous and Lat. –ōs with further simplification
of this phonotactically problematic cluster, the long vowel in Ved. –ān points to a pre-form *–ōm, which looks
like an effect of Szemerényi’s Law (and does not easily submit to analogical explanation). The difficulty in recon-
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 5

ciling these disparate outcomes with a single PIE surface form may in fact stem from the relatively transparent
morphophonemic analysis of the ending as a composite of the thematic vowel + accusative marker *–m + plural-
izing *–s (in other words, internally reconstructed **–o-m-s). If this ending were still synchronically analyzable as
such within the history of the daughter languages (in generative terms, underlyingly */–o-m-s/), it is possible that
the attested endings that appear to continue *–oms are post-PIE innovations, while the lengthened vowel in the
pre-form *–ōm reflected by Ved. –ān is due to the synchronic operation of Szemerényi’s Law in PIE (cf. Sandell
and Byrd 2014), and the Vedic ending’s sandhi variant –āṃs derived by further recharacterization with pluralizing
*–s. The surface form of the ending in any case depends on what phonological processes can be reconstructed for
PIE, which itself calls for further research.
The PIE athematic animate accusative plural ending was *–ms, in all likelihood a composite of accusative *–m
+ pluralizing *–s, and by further additifon of the thematic vowel *–o–, the source of thematic *–oms discussed
above. It is usually assumed that this ending is directly continued (modulo assimilation) in glide-final stems in
Germanic, e.g. Goth. gasti-ns ‘guests’; Goth. sunu-ns ‘sons’. Yet like its thematic counterpart, accusative plural
*–ms is more frequently eliminated as a surface sequence, e.g. Hitt. ḫašš-uš* ([Link]Š-uš) ‘kings’; Lat. hostīs
‘strangers, enemies’; Ved. agnīń ‘fires’; Lat. manūs ‘hands’; Ved. sūnū ́ n ‘sons’. After a consonant, the ending was
realized as *–m̥ s, the diverse reflexes of which include Ved. pad-ás, Lat. ped-ēs (via *–ens) ‘feet’; Gk. kún-as, Lith.
šun-į̀s ‘dogs’; and Goth. broþr-uns ‘brothers’.
Nominative and accusative were not formally distinguished in PIE neuter nouns; both are zero-marked in
the singular of athematic noun classes, e.g. Ved. jā ́nu, Gk. gónu, Lat. genū, Hitt. gēnu ‘knee’ (< PIE *ǵó/én-u-ø),
and marked by *–om in thematic nouns, e.g. Lat. iug-um, Ved. yugám ‘yoke’ (< PIE *yug-óm). Nominative and
accusative similarly shared a (“collective”/set; cf. §2.1.2) plural ending *–h2 . This ending yielded a long vowel in
glide-final stems, e.g. OHitt. āššū ‘goods’, Ved. madhū ‘sweet’ (adj.) (< *–uh2 ); Ved. śúcī ‘shining’ (adj.) (< PIE
*–ih2 ). In sonorant- and s-final stems, the original situation is probably reflected in Hitt. widār ‘waters’ (< *wed-ṓr;
cf. synchronically sg. Gk. húdōr ‘water’) and OAv. manā ̊ ‘thoughts’ (< *men-os-h2 ); the long vowel in the final
syllable of these words can then be ascribed to Szemerényi’s Law just as in the animate nominative singular, the
environment for the change unified by the assumption that *h2 was — like *s — a fricative (per Kümmel 2007:227–
36, probably [χ]). In PNIE, at least, the neuter plural of obstruent stems was subject to “laryngeal vocalization” (i.e.
vowel epenthesis; see Byrd, this volume) which yielded –i in Indo-Iranian, and –a in Greek, e.g. [Link]. Ved.
–ant-i, Gk. –ent-a (< PIE *–ent-h2 ). However, reconstructing a single PIE (surface) form is in this case difficult,
since Anatolian probably deleted final *h2 after an obstruent (see Byrd 2015:96); as always, this split between PNIE
and Anatolian raises questions about the PIE-level reconstruction. No such problems arise in the reconstruction
of the thematic nominative-accusative neuter plural ending, which was *–eh2 . This ending is directly continued
in Ved. yugā ́, OCS iga ‘yokes’ (< PIE *yug-eh2 ); Hitt. kunn-a ‘right-hand’ (adj.). In Greek and Latin the case ending
unexpectedly surfaces with a short vowel –ă, which is usually assumed to reflect its replacement by the athematic
ending (see further Weiss 2011:211).
Outside the nominative and accusative, case endings were the same for animate and neuter nouns. Most of the
evidence for the reconstruction of the instrumental comes from Indo-Iranian, where the instrumental is produc-
tively continued as a distinct paradigmatic case form. An athematic singular ending *–h1 is generally reconstructed
to account for the long vowel observed in Indo-Iranian glide-final stems, e.g. Ved. matī ́ ‘with thought’, OAv. rāitī
‘with liberality’ (< PIE *–ih1 ); OAv. xratū ‘with wisdom’ (< PIE *–uh1 ). However, the much more common athe-
matic ending was *–eh1 , e.g. (anim.) Ved. pad-ā ́ ‘with the foot’, OAv. zərədā(-cā) ‘with heart’ (< PIE *–eh1 ); (neut.)
Ved. mánas-ā, OAv. manaŋh-ā ‘with thought’ (< PIE *mén-es-eh1 ). This ending, moreover, is diachronically oust-
ing the *–h1 instrumental ending — compare Ved. krátv-ā (< *–eh1 ) with OAv. xratū (< *–h1 ). The corresponding
thematic ending was *–oh1 , which is reflected directly by archaic Indo-Iranian forms like Ved. yajñ-ā ́, OAv. yasn-ā
‘with sacrifice’. (< PIE *h1 yag̑n-óh1 ); Lith. výr-u ‘with (a) man’; OSax. word-u, OHG wort-u ‘with (a) word’. Instru-
mental singular *–eh1 is also likely to be continued elsewhere in certain adverbs, e.g. Lat. vald-ē ‘very’, dialectal
(Elean) Greek taut-ē ‘here’, and is according to Jasanoff (1978, 2003b) the pre-PIE source of the formally identical
PIE stative suffix *–eh1 (-ye/o)– (cf. §4.3.1).
The PIE athematic dative and locative singular endings are uncontroversially reconstructed as *–ei and *–i
respectively, e.g. dative VOLat. rec-ei ‘for the king’ (= Lat. rēg-ī), Ved. mātr-é ‘for the mother’, OCS synov-i ‘for the
son’ (< PIE *–ey); and locative Ved. pad-í ‘at the foot’ (< *–i). These two cases often underwent formal syncretism,
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 6

as in Greek, where all first millennium dialects use –i (< *–i) to express dative (e.g. Di(w)í ‘for Zeus’) and locative
(pod-í ‘for/at the foot’) functions, although traces of the old dative ending are preserved in Mycenaean (cf. di-we
[diwéi] ‘for Zeus’). The same exact syncretism has occurred in Hittite, where –i (< PIE *–i; never x –e in Old Hittite)
continues both dative (e.g. Hitt. šiun-i ‘for the deity’) and locative functions (Hitt. nēpiš-i ‘in heaven’); however,
this development must also be recent, since CLuw. –ī < *–ei (e.g. tappaš-ī ‘id.’) points to an independent merger
of dative and locative with the case ending of the former as exponent. Italic also continues locative *–i, but in yet
another functional role: it has become the marker of the ablative singular in consonant stems, e.g. Lat. –e, Umbr.
–e (see Vine, this volume). The PIE athematic locative singular could also be realized with a zero-ending, the so-
called “endingless locative”. This type is best represented in Anatolian, e.g. Hitt. šīwat (< PIE *–ot-ø) ‘on the day’,
and in Indo-Iranian, e.g. Ved. nā ́m-an ‘in name’ (< *–men-ø); OAv. dąm ‘in the house’ (< *dṓ/ém-ø). In some cases,
endingless locatives show a lengthened vowel in their stem-final syllable, as in the Avestan form cited above.
The PIE thematic dative and locative singular endings were *–ōi and *–oi respectively. Dative *–ōi – with
long diphthong via prehistoric contraction from **o-ei, as in the thematic nominative singular (see above) — is
continued in e.g. Gk. hípp-ōi ‘for the horse’; VOLat. dven-oi ‘for a good (man)’ (= Lat. bon-ō); OAv. ahur-āi ‘for
the lord’; Lith. výr-ui ‘for a man’, while locative *–oi is reflected in Ved. yajñ-é, OAv. yesn-ē ‘in the sacrifice’; OCS
grad-ě ‘in the city’; (adv.) Gk. oík-oi ‘at home’. The fact that the thematic locative ending was transparently derived
by adding the athematic ending *–i to the thematic vowel –o– is crucial to its subsequent development in Greek
and Balto-Slavic, both of which require that the ending was disyllabic in the shallow prehistory of these languages
(Jasanoff 2009).
Several allomorphs are reconstructible for the PIE athematic genitive singular : *–s, *–es, and *–os. The last —
arguably an intrusion from thematic nouns — is observed in Greek, e.g. patr-ós ‘of the father’, while certain Balto-
Slavic forms require *–es, e.g. OLith. szird-es ‘of the heart’; OPr. kermen-es ‘of the body’. Both of these endings
are attested in archaic Latin inscriptions, e.g. apolon-es ‘of Apollo’ (< –es); nomin-vs ‘of name’ (< *–os), although
only *–es is continued (as –is) in the classical language (cf. Apollōn-is, nōmin-is) (see Vine, this volume). Due to
the Indo-Iranian merger of *e and *o, Ved. –as and OAv. –as(-cā) could reflect either *–es or *–os. The distribution
of the zero-grade *–s ending in the daughter languages is more limited, and generally confined to sonorant-final
stems. It is best established in *i-stems (e.g. Ved. agné-s ‘of the fire’, OPers. cišpai-š ‘of Teispes’; OCS kostei-s ‘of
the bone’; Osc. aet-eis ‘of part’ < *–ey-s) and *u-stems (Ved. sūnó-s, Goth. sunau-s, OCS synou ‘of the son’ < *–
ew-s). Zero-grade *s is also found in r-stems in Indo-Iranian and Germanic (e.g. YAv. žaotar-š, Ved. hótur ‘of the
offerer’; Ved. pitúr, Anglian OE fadur, ON fǫður ‘of the father’ < *–r̥ s); some Avestan n-stems (e.g. OAv. rāzə̄ṇg ‘of
the ruler’ < *–on-s; cf. Ved. rā ́jñ-as) and Hittite verbal nouns in –waš (< *–wen-s); and to an *m-final root noun in
the likely PNIE collocation *dém-s páti– ‘master of the house’ (OAv. də̄ṇg paiti–, Ved. dán páti–, Gk. despótēs). For
the possibility that these zero-grade forms reflect a PIE syncope rule that applied in the final syllable of sonorant
stems, see Kümmel (2014). The single example of *s in an obstruent-final stem is Hitt. neku-z [nekw t-s] (cf. Lat.
noct-is, Gk. nukt-ós ‘of the night’), which occurs only in the fixed phrase nekuz mēḫur ‘time of evening; twilight’
(Schindler 1967); the fact that it is not attested in any other Anatolian obstruent-final stems suggests that it is a
pre-PIE archaism, and argues against reconstructing it as a PIE allomorph in obstruent-final stems.
The PIE thematic genitive singular also has multiple reconstructible allomorphs: *–os, *–osyo, and *–oso. Plain
*–os — homophonous with thematic nominative singular *–os and clearly formed by the addition of athematic
genitive singular *–s to the thematic vowel — is found only in Anatolian, e.g. Hitt. išḫ-āš ‘of the master’. The
ending *–osyo is continued in HLuw. [–asi] (e.g. DEUS-na-si-i [massan-asi] ‘of the deity’) as well as probably Car.
–ś) (pleq-ś ‘of Peldēkos (PN)’; on both points, see Melchert 2012a with references); *–osyo is also well-represented in
the NIE languages: Gk. –oio (e.g. Myc. i-qo-jo, Hom. hípp-oio ‘of a horse’); PIIr. *–asya (Ved. áśv-asya, OP aspahyā
‘id.’; OAv. ahur-ahiiā ‘of the Asura’), PItal. *–osyo (VOLat. valesiosio ‘of Valerius (PN)’; Fal. kaisi-osio ‘of K– (PN)’),
Arm. –oy (mard-oy ‘of a man’). As for the ending *–oso, its reflexes can be seen in Germanic (OSax. dag-as, OE
dag-æs ‘of the day’), perhaps Balto-Slavic (OPr. deiw-as ‘of god’ (for an alternative view see Olander 2015:134–6),
as well as Anatolian, cf. Lyc. –Vhe, e.g. Xerig-ahe ‘of Xeriga (PN)’; Car. –s, ntro-s ‘of “Apollo”’ (see Melchert 2002,
2012a). Since Germanic nominal forms cited in support of an ablaut variant *–eso (Goth. dag-is, OHG tag-es ‘of
the day’) are more likely analogical (Ringe 2006:201–2), there is no positive evidence for its reconstruction in the
nominal system. An innovative thematic genitive ending *–ī is found in Italic (e.g. Lat. equ-ī ‘of the horse’) and
Celtic (e.g. Ogam Ir. maqq-i ‘of a son’); this ending may be historically related to the derivational suffix *–ih2 – (the
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 7

́
so-called vr̥kī-suffix), on which see Nussbaum (1975) and the discussions in §2.1.3 and §2.4 below.
The athematic ablative singular and genitive singular were formally syncretic in PNIE. However, the Anatolian
languages show instead a formal merger of instrumental and ablative, one exponent of which is the ending (undif-
ferentiated for number) *–ti (> Hitt. abl.-instr. –z; Hitt. –az, CLuw. –ati, HLuw. –adi/–ari, Lyc. –edi < *–o-ti with
inner-Anatolian thematization). Melchert and Oettinger (2009) argue that this ending was the marker of ablative
singular and plural in PIE, and that this situation was inherited into Anatolian, while the syncretism of ablative
and genitive singular was a PNIE development (see also Oettinger, this volume); however, it is just as plausible that
the PNIE situation is archaic and Anatolian innovative, with a new formal marker of the ablative(-instrumental)
developing independently in Anatolian just as it likely did in Armenian (e.g. i get-oy ‘from the river’) and perhaps
also in Tocharian (e.g. TA āsān-äṣ ‘from the throne’).
The PNIE ablative singular of thematic nouns, in contrast, had a distinctive ending, which has traditionally
been reconstructed as *–ōd (OLat. –ōd, Ved. –āt). Yet this reconstruction is problematized by the Lithuanian
genitive singular –o, which requires Proto-Baltic *–ād (thus likely also OCS –a < PBS –*ād; see [Olander 2015: 134–
6]). In order to reconcile these outcomes, it is generally assumed that the ending was disyllabic, with the pre-PIE
agglutination of an element reconstructed as either *h2 ed or *ad that is also the source of various prepositions,
adverbs, and local particles in the daughter languages (e.g Lat. ad ‘to’; Goth. at ‘at’; see Dunkel 2014:II.8–18).
Of these possibilities, the Hittite (singular/plural) instr. ending –(i)d is phonologically straightforward only from
the latter: Melchert and Oettinger (2009:55) derive the ending via resegmentation of PIE *–oh1 -ad — i.e. the
thematic instrumental ending plus postpositional *ad — whence Pre-Hitt. *–ad (PIE *–o(h1 )-h2 ed would have
yielded x –aḫ(ḫ)ad); *–ad was then reanalyzed as Hitt. /–a-d/, a combination of thematic vowel + –d (alternatively,
*d may come directly from pronominal inflection; cf. §2.2.1). The development of PIE thematic ablative *–oh1 ad —
perhaps indifferent to number as in Hittite — would thus follow a cross-linguistically well-established trajectory
whereby new case endings emerge via accretion of adverbial elements (see generally Kulikov 2009, and on the
Tocharian “secondary” cases, Kim 2013a and Pinault, this volume).
The PNIE athematic instrumental plural ending is typically reconstructed as *–bh is, for which Indo-Iranian
(Ved. –bhis, OAv. –bīš) provides both formal and functional support; this ending is also directly continued in
Celtic (Gaul. –BI, OIr. dat. pl. –ib). Possible further reflexes of the ending include Arm. instr. pl. –bkc /–ovkc /–
owc (beside instr. sg. –b/–v/–w; see Olsen, this volume) and Myc. Gk. instr. –pi [–ph i(s)], although these may
rather be traced back directly to the adverb-forming suffix *–bh i (Hom. Gk. –ph i, e.g. (w)ĩ-ph i ‘by force’), which is
historically contained in the ending *–bh is and which must be reconstructed for PIE (Hitt. kuwa-pi ‘when; where’).
Germanic *–mis (> Goth. dat. pl. –m, ON –m(r)) and (with unexpected long vowel) Balto-Slavic *–mīs (> Lith.
instr. pl. –mì, OCS –mi; secondarily Lith. instr. sg. –mì, OCS –mĭ) are also generally derived from *–bh is via the
so-called “Northern IE” substitution of *bh by *m (itself likely an adverbial suffix, e.g. Lat. ill-im ‘from there’; HLuw.
abl.-instr. pron. zin ‘from/with this’). The absence of Anatolian evidence for any *bh -initial case endings strongly
suggests that *–bh is is a PNIE innovation, and Jasanoff (2008) has argued that the PIE instr. pl. was rather *–is.
He identifies this suffix in a set of adverbs (e.g. Gk. móg-is ‘with toil; hardly’; Ved. āv-ís, YAv. āuu-iš ‘manifest’)
and, more significantly, in the PNIE pronominal instr. pl. ending *–ōis (see below), and derives PNIE –bh is by its
addition to adverbial *–bh i. This scenario has a plausible parallel in the development of the PNIE dative-ablative
plural ending *–bh (y)as (see below), but is complicated by the lack of external support from Anatolian for the *–is
suffix itself.
The PNIE thematic instrumental plural ending is straightforwardly reconstructible as *–ōis, e.g. Indo-Iranian
(Ved. hást-ais, OAv. zast-āiš ‘with the hands’); Gk. dat. pl. th e-oĩs ‘for/by the gods’ (unless from [Link]. *–oisu;
see below); Italic (VOLat. dat./abl. pl. SOKI-OIS ‘for the friends’ (> Cl. Lat. soci-īs); Osc. Núvlan-úís ‘for the
men of Nola’); PBS *–ōis (Lith. výr-ais ‘with men’; OCS grad-u ‘with cities’). This ending appears to contain a
post-thematic i-element original to the pronominal declension (cf. §2.2.1) just like the PNIE thematic locative
plural and (arguably) dative plural (on both, see below); to this base in *–oi– was added, according to Jasanoff
(2008), a suffix *–is, which may have been the PIE athematic instrumental ending (see above). In Anatolian, the
instrumental plural was syncretic with the ablative, both formally marked by a historical exponent of the ablative;
it thus presents no evidence for or against the PIE status of *–ōis.
The PIE dative plural ending is likely reconstructible as *–os, which is directly reflected in Anatolian (e.g. Hitt.
–aš, CLuw. –aš, Lyc. –e), but whether this ending was original to thematic or athematic nouns is uncertain. It is also
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 8

highly probable that the PNIE athematic dative-ablative ending generally reconstructed as *–bh (y)as is derived by
addition of this *–os to the adverb-forming suffix *–bh i. The phonologically expected outcome of their fusion is
*–bh yas, which is continued in Indo-Iranian (e.g. Ved. viḍ-bhyás, OAv. vīži-biiō ‘to/for/from the clans’) and usually
held to be the basic PNIE form of this syncretic ending. The functionally equivalent yod-less ending *–bh os is found
in Italic (e.g. Lat. dat. pl. rēg-ibus ‘for the kings’ (with intervening i analogically spread from *i-stem paradigms))
and Celtic (Gaul. Matrebo ‘for the (divine) mothers’). Corresponding thematic endings were formed by adding
the athematic ending either to the thematic vowel (e.g. Ven. louderobos ‘for the children’) or — less likely at the
PIE stage — to stem-final *–oi– under the influence of the pronouns (cf. §2.2.1 below), as in Indo-Iranian (Ved.
ukth-ébhyas, OAv. uxð-ōibiiō ‘to/for chants’). PBS shows athematic *–mos and thematic *–omos with *m instead
of *bh just as in the athematic instrumental plural (see above), e.g. (athematic) OLith. dat. pl. sunú-mus ‘to/for
sons’, OCS kostĭ-mŭ ‘to/for bones’; (thematic) Lith. výr-ams ‘to/for men’, OCS grad-omŭ ‘to/for cities’. The same
*–(o)mos may occur in PGmc. *–(a)m (Goth. –am, ON –mr), although it may instead reflect PGmc. instr. pl. –miz
(see above).
The PNIE ablative plural was, as noted above, syncretic with the dative plural in both thematic and athematic
nouns. For the possibility that in PIE ablative case was marked by number-indifferent endings — athematic *–ti
and thematic *–oh1 ad — see the discussion of the ablative singular above.
The PIE genitive plural ending in athematic nouns is much disputed, either *–om or *–oh1/3 om (as in thematic
nouns; see below). The Anatolian languages (Hitt. –an, Lyc. –ẽ) are uninformative, as they could reflect either
ending. Within PNIE, there is incontrovertible evidence for *–oh1/3 om in athematic nouns in Indo-Iranian (e.g.
Ved. padā ́m ‘of the feet’, frequently with disyllabic scansion), as well in Greek and Baltic (e.g. Gk. pod-ō ñ ‘id.’;
Lith. akmen-ų̃ ‘of stones’, both with circumflex accent). Nevertheless, on structural grounds the disyllabic ending
*–oh1/3 om is aberrant in athematic nominal inflection, and is thus reasonably assumed to originate historically in
thematic paradigms; the question, then, is whether there was wholesale replacement of the “short” ending *–om
(or possibly *–h1/3 om) by the “long” ending *–oh1/3 om already in P(N)IE, in which case there should be no definitive
trace of *–om in the daughter languages, or if instead athematic *–oh1/3 om is an innovation in the shallow prehis-
tory of those branches in which it is attested. The answer to this question depends largely on the interpretation
of the Slavic evidence. Jasanoff (1983) has contended that PS *–ŭ (e.g. OCS dŭšter-ŭ ‘of daughters’) can be derived
from *–oh1/3 om (via *–ōm); however, Olander (2015:255–9) maintains the older view of Meillet (1922) that PS *–ŭ
must continue “short” *–om. The matter remains unresolved at present.
The PIE thematic genitive plural ending was *–oh1/3 om. Besides its possible Anatolian outcomes mentioned
above (which likely rule out *h2 for the medial laryngeal by their lack of a consonantal reflex), it is productively
continued in this nominal class in Greek (e.g. hípp-ōn ‘of the horses’), Sabellic (SPic. raeli-om ‘of the Raelii (PN)’;
Umb. pihakl-u ‘of the purification rites’), and Baltic (Lith. lang-ų̃ ‘of windows’; Latv. tę̃vu ‘of fathers’). In several
languages, the inherited ending has been analogically remodeled, e.g. Latin de-ōrum ‘of the gods’ on the basis of
the feminine genitive plural (Pre-Lat. *–āsōm ̆ ) (see Weiss 2011:208, 224; cf. §2.2.1 below) and PIIr. *–ānaam (Ved.
yajñ-ā ́nām, OAv. yasnanąm ‘of the sacrifices’) on the basis of *n-stems (cf. Kümmel, this volume); yet PIE *–oh1/3 om
survives in both Latin and Vedic in relic forms: Lat. de-um; Ved. devā ́ñ ( jánma) (RV VI.11.3b) ‘(race) of the gods’.
The PNIE athematic locative plural is generally reconstructed as *–su, which is continued in Indo-Iranian (e.g.
Ved. vik-ṣú ‘among the clans’; OAv. naf-šu ‘among the descendants’) and Balto-Slavic (dial. Lith. aki-sù ‘in (the)
eyes’; OCS gostĭ-xŭ ‘among guests’). Greek dat. –si (e.g. nau-sí(n) ‘to/for/on the ships’) likely reflects the same
ending with analogical *–i from the locative singular. Adding *–su to the pronominally influenced base *–oi–
(cf. §PIE-N-pron-infl below) yielded the thematic ending *–oisu, e.g. Ved. márt(i)y-ešu, OAv. maṣǐ i-aēšu ‘among
mortals’; OCS grad-ěxǔ ‘in cities’; and with the same analogical development, Gk. dat. th e-oĩsi ‘to/for/among the
gods’. This ending –oisi may also be the source of the shorter Greek thematic dat. ending –ois (via apocope), unless
it instead reflects thematic instr. pl. *–ōis (see above). The only alleged trace of *–su in Anatolian is as an adverb
in the Luwic languages (CLuw. 3-šu, HLuw. ta-ra/i-su ‘thrice’; perhaps also Milyan trisu); thus if the PA syncretic
dat.-loc. ending *–os was originally a dative marker, it would be be possible to reconstruct a distinct PIE locative
plural ending *–su.
The PIE athematic vocative singular was zero-marked (*–ø), e.g. Hitt. d Kumarbi ‘(O) Kumarbi’; Gk. páter, Ved.
pitar, Lat. (iup)-piter ‘(O) (sky-)father’; Ved. sūno, Goth. sunau, Lith. sūnaũ, OCS synu ‘(O) son’ (< *–ew-ø, with full-
grade of the derivational suffix in glide-final stems). The vocative of PIE thematic nouns was marked by *–e, e.g.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 9

Gk. lúk-e, Lat. lup-e, Lith. vil̃k-e ‘(O) wolf’; Ved. dev-a, OPr. deiw-e ‘(O) god’; OCS bož-e ‘id.’. The use of nominative
singular for vocative singular — very likely by analogy to the plural (see below) — is also found to various degrees
in many languages (especially in athematic nouns).
The PIE vocative plural was identical to the nominative plural in both athematic and thematic nouns. This
situation is continued into all of the daughter languages except Old Irish, where the distinctive vocative (e.g. (á)
ḟir-u ‘(o) men’) reflects the inherited PIE nominative plural ending *–ōs, which has been replaced qua nominative
by innovative *–ī from the pronouns (see above).
Some scholars argue for the reconstruction of a ninth PIE case, the allative (or “directive”), which signified
movement in a direction or toward a goal. The status of this case is disputed, as is its formal marker (possibilities
include *–e/oh2 , *–h2 e, and *–o). Evidence for the reconstruction of the allative comes principally from Old Hittite,
where it is a productive case ending, while in the PNIE languages there are certain adverbs that could be relic case
forms, e.g. Gk. kh amaí ‘to the ground’. Most regard this evidence as insufficient to justify the reconstruction of an
additional PIE case, which implies its synchronic status in Hittite reflects an innovation (for arguments in support
of this scenario, see Melchert fthcm. c).
Similar attempts have been made to reconstruct another case form attested exclusively in the Anatolian lan-
guages, the ergative. In Anatolian, when a neuter noun is the subject of a transitive verb, it receives ergative case.
Singular and plural ergative endings are securely reconstructible for Proto-Anatolian in view of agreement between
Hittite and the Luwic languages: (sg.) Hitt. –anza [–ant͡s], CLuw. –antiš, HLuw. –antis, Lyc. pre-nasalizing –ti;
(pl.) Hitt. –anteš, Luw. –antinzi [–antint͡si]. However, at least since Garrett (1990) it has been generally agreed that
these endings — and the Anatolian syntactic feature, split-ergativity, that they would imply — are post-PIE inno-
vations (cf. Melchert 2011b), and according to most recent hypotheses, the ergative endings have grammaticalized
from an animacy-increasing (or “individuating”) derivational suffix, perhaps PIE *–e/ont– (see Goedegebuure 2013
and Oettinger, this volume).
In addition to the case endings associated with singular and plural number, a limited reconstruction of dual
markers is possible. For the athematic animate dual, a syncretic nominative-accusative ending *–h1 e is plausibly
reconstructed on the basis of Greek (e.g. pód-e ‘two feet’) and Lithuanian (OLith. žmûn-e ‘two men’) evidence,
although this reconstruction is somewhat complicated by the fact that glide-final stems appear to continue just
*–h1 , e.g. Ved. kav-ī ́ ‘two poets’, OCS gost-i ‘two guests’; Ved. sūnū ́ , OCS syn-y, Lith. sū ́ n-u ‘two sons’. More formally
secure is an athematic neuter dual ending *–ih1 , for which there is at least one lexical match between these two
branches (Gk. ósse = Lith. akì ‘two eyes’), as well as agreement within Indo-Iranian (e.g. Ved. vácasī, OAv. vacahi(-
cā) ‘two words’). Thematic forms were formed by addition of the athematic endings to thematic vowel *–o–, thus
likely neuter *–o-ih1 (e.g. Ved. yug-é ‘two yokes’; OCS měst-ě ‘ two places’), and animate *–oh1 e which, according
to Jasanoff’s (1988a:73–4) proposal, would have yielded accent-conditioned variants *´–oh1 (e.g. Gk. hípp-ō, Ved.
áśv-ā ‘two horses’) and *–óh1 u (e.g. Ved. dev-aú. On the possible Anatolian reflexes of these endings, see §2.1.2
below.
Very little can be said with certainty about the reconstruction of the oblique case forms of the dual, yet two
points are fairly clear. First, the oblique cases of the dual were also almost certainly syncretic: in Vedic, instrumen-
tal, dative, and ablative functions are marked by –bhyām), while genitive and locative share the marker –os; but
even within Indo-Iranian there are differences, since Avestan has distinct genitive (OAv. –ā ̊) and locative (OAv. –
ō) endings. Similarly, PBS probably had one case ending for dative and instrumental, and another for genitive and
locative (see Olander 2015:205–20 for discussion). It is also clear that the oblique dual endings were built out of
adverbial elements just like the non-structural plural case markers (in some cases, the same elements, e.g. *–bh i),
but the details of their reconstruction are even more uncertain. On the more general question of the status of the
dual as a PIE category, see again §2.1.2 below.

§2.1.2 Number A three-way nominal contrast for number (singular, dual, plural) is securely reconstructible for
PNIE. This number system — a cross-linguistically common type (Corbett 2000:20) — is synchronically operative
in the oldest stages of Indic, Iranian, Greek, Baltic, Slavic, Tocharian, Celtic, and to a lesser extent in Germanic
(mainly Gothic); the other PNIE languages have lost the dual as a living category, retaining traces in the numeral
system (e.g. *dwoh1 > Lat. duo ‘2’) or elsewhere. The reconstruction of number in PIE is problematized, on the one
hand, by the absence of the dual as a living number category in Anatolian, and on the other, by the vexed question
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 10

of the neuter plural. Many scholars would trace the neuter plural back to an original singular “collective” — in part
because of the formal affinity between its marker *–(e)h2 and the suffix *–eh2 that primarily marks feminine nouns
in the NIE languages (cf. §2.1.3 below), and in part because of the singular verbal agreement patterns observed with
neuter plural subjects in several ancient IE languages. We take up these issues in turn below.
As discussed in §2.1.1, formal markers for the (nominative-accusative) dual are securely reconstructible for
PNIE. Nouns marked with dual number refer to exactly two distinct real world entities (and by implication, the
plural for three or more such entities). In the IE branches in which the dual is preserved (Indo-Iranian, Greek,
Celtic, Balto-Slavic, Gothic, Tocharian) it is most frequently used with naturally occurring pairs — one widespread
example is Gk. ósse, Ved. akṣī,́ YAv. aši, Lith. akì, OCS oči, TB eśane ‘two eyes’ — as well as with items at the highest
end of the animacy hierarchy (see Corbett 2000:55ff.), thus especially when a noun’s referents are human, e.g. Gk.
anth rópō ‘two men’, Lith. žmûne ‘id.’. In addition, it appears that the IE dual had certain idiosyncratic uses — for
instance, as an associative marker in the “elliptical dual”, e.g. Ved. Mitrā ́ ‘Mitra and his companion Varuna’; Hom.
Gk. Aíante ‘Ajax and his companion Teucer’ (Wackernagel 1877 [= 1953b:538–45]).
The dual was lost in many IE languages, in some cases within the historical period (e.g. post-classical Greek).
This extensive loss may be easier to explain if it is assumed that in P(N)IE the use of the dual for two referents was
optional — or more standardly “facultative” — as already observed in Homeric Greek, which regularly allows the
plural in these contexts. However, since facultative use of the dual is found in many languages in which the category
remains productive (Corbett 2000:42–53), this need not in itself be viewed as an indication of the incipient loss of
the grammatical category.
Projecting the dual back from PNIE to PIE itself is complicated by the limited evidence for dual number in
the Anatolian nominal system. The dual exists as a synchronic grammatical category in none of the Anatolian
languages. Possible support for PIE inheritance is restricted to lexicalized relics, forms denoting natural pairs that
may have escaped the loss of the dual by reanalysis as set (or “collective”; see further below) plurals due to their
(synchronic) formal identity with members of this productive category (Rieken 1994:52–3). Potential traces of the
dual in animate nouns include CLuw. tāwa ‘eyes’, iš(ša)ra ‘hands’, and pāta* (GÌ[Link]Š-ta) ‘feet’ (< *–oh1 (e)), while
the neuter dual ending in *–ih1 may be continued in Hitt. GIŠ ēlzi ‘scales’, mēni– ‘face’, and a few other lexical items;
see Melchert (fthcm. a) with references. It is generally thought that additional evidence for the dual in Anatolian
comes from the verbal system — in particular, the PA 1st plural ending *–wen(i) — but see further discussion in
§4.2.2 below.
A separate, much discussed question concerns the PIE status of number in neuter nouns: did plural number
exist as a grammatical category in neuter nouns, or did neuter nouns instead form only a grammatically singu-
lar “collective”? Advocates of the latter position typically point to the formal affinity between the marker of the
neuter plural *–(e)h2 and that of the PNIE feminine-forming suffixes *–ih2 /*–yeh2 –, *–ih2 , and in particular *–eh2
(see §2.1.3 below), whose derivatives have a (remarkably *s-less) nominative singular, e.g. PNIE *h2 widh éw-eh2
> Ved. vidhávā, Lat. vidua ‘widow’ (cf. LIV 2 , 294). This ending is phonologically identical to the ending which
characterizes neuter plurals in the daughter languages (e.g. PIE *yug-éh2 > Ved. yugā ́, Lat. iuga ‘yokes’).
There is a general consensus, then, that this formal agreement reflects a prehistoric connection between neuter
plural and feminine, but the exact nature of this relationship is much disputed — in particular, whether the neuter
played a role in the genesis of the feminine (see §2.1.3 below) — and has given rise to an enormous literature (for
a range of recent opinions, see the papers collected in Neri and Schuhmann 2014). However, the question for the
directly reconstructible stage of PIE amounts to a simpler one: Is there any compelling evidence that neuter nouns
marked with *–(e)h2 were grammatically singular in the IE languages?
That the PNIE descendants of neuter *–(e)h2 nouns are synchronically plural is undisputed: they regularly refer
to multiple individuated entities, and except for the nominative-accusative case, have the same plural inflectional
endings as animate nouns. The analysis of the Anatolian evidence is more often called into question — for in-
stance, it has repeatedly been claimed (e.g. Harðarson 1987, 2015; Matasović 2004:156) that neuter *–(e)h2 nouns
show singular agreement with predicate adjectives and pronouns. However, this claim is false for Old Hittite, and
the New Hittite examples cited in support are demonstrably innovations (van den Hout 2001). Moreover, even
Anatolian pluralia tantum of this type in which an original singular value might be detected — e.g. Hitt. warpa
‘enclosure’, Lyc. arawazija ‘memorial’ — are grammatically plural, as shown by their resumption in discourse with
unambiguously plural case forms (dative-locative plural Hitt. warpaš, Lyc. arawazije; see Melchert 2011: 396).
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 11

The remaining alleged evidence for the erstwhile singular status of IE neuter plurals comes from verbal agree-
ment patterns in Anatolian, Greek, and, on a more limited basis, Indo-Iranian: in contrast to animate plurals,
neuter plural subjects in these languages take singular verbal agreement morphology. This phenomenon — now
generally (although anachronistically) referred to as the “tà zō î -a trékh -ei rule” — was recognized for Greek already
by the ancient grammarians; for its parallel operation in Hittite see Hoffner and Melchert (2008:240). The singular
verb marking in this type is held to reflect a stage at which these neuter nouns were grammatically singular, and
thus singular verb agreement was appropriate. Yet there is no need for recourse to such a prehistoric stage to ex-
plain this agreement pattern, since it is typologically common that low animacy nouns morphologically marked
as plural fail to trigger plural agreement on verbs (cf. Comrie 1989:190–191); outside of IE itself, such patterns are
observed in Georgian (Smith-Stark 1974) and Turkish (Bamyacıet al 2014), as well as Muna (Austronesian) and
Ngalakan (Australian) (see Corbett 2000:71, 188–189 with references). Per Patri (2007:62), Anatolian verbs may
therefore “default” to singular in the absence of an (animate) plural controller; the same analysis could account
for Greek and, in turn, be extended to PIE, in which case there is no need to assume that neuter *–eh2 nouns were
singular in PIE, nor necessarily at any earlier period (see Melchert 2011a for arguments to this effect).
Yet in contrast to the evidence for their singularity, there is strong support for the notion that *–(e)h2 marked
“collectives”, or perhaps more precisely, “set plurals” (cf. Eichner 1985:142, Melchert 2014c:257–8; on the problem-
atically “variegated” usage of the term “collective,” see Gil 1996:66–70). It has long been known that there are a
number of cases in the NIE languages of three-way splits in animate nouns, where a continuant of *–(e)h2 is at-
tested beside ordinary singular and plural forms, e.g. Gk. kúklos ‘wheel’, kúkloi ‘wheels’, kúkla ‘wheel-set’; Gk. mērós
‘thigh’, mēroí ‘thighs’, mēr̂ a ‘(sacrificial) thigh-pieces’ (on the accentual variation, see Probert 2006b:158–163); Lat.
locus ‘place’, locī ‘places’, loca ‘literary passages’ (although the distinction between the latter two is debatable; cf.
Weiss 2011:196; Clackson 2007:101–3). These examples — most clearly, Greek kúkla — are consistent with the
idea that animate *–eh2 nouns denoted multiple distinct entities that were conceptualized as constituting a set.
Eichner (1985:148) identified similar Hittite examples, e.g. alpaš ‘cloud’, alpēš ‘clouds’, alpa ‘cloud-bank’. Supple-
menting the Hittite data collected by Eichner (e.g. Hitt. palšaš ‘path,’ palšeš ‘paths’, palša ‘path comprised of
ritual materials’) and adding Lycian and Luwian comparanda, Melchert (2000:62–7) argues that the relatively ro-
bust Anatolian evidence is indicative of a productive grammatical process; thus in contrast to the NIE languages,
where the marginality and generally specialized meaning of animate set plurals allows them to be plausibly ana-
lyzed as lexicalized relics (cf. Harðarson 1987; Tichy 1993), the Anatolian situation is best explained by assuming
that PIE animate nouns could regularly form either a count plural (marked with *–es/–ōs) or a set plural (marked
with *–(e)h2 ), whereas neuter nouns lacked the grammatical category of count plural.
PIE would thereby distinguish at least two grammatical numbers, singular and plural, and according to most
researchers, a third, the dual; in addition, the plural had two distinct sub-classes (Melchert 2000:62, 67 n. 38), count
plural and set plural (cf. Eichner’s (1985) Komprehensiv, though reconstructed as a fourth category), although only
in animate nouns was the morphological distinction realized. This system is outlined in (2) with the nominative
case endings reconstructible for each category:

Singular Dual Plural

Set Count
(2)
Animate *–s/*–os *–h1 e *–h2 / *–eh2 *–es/*–ōs
Neuter *–ø/*–om *–ih1 *–h2 / *–eh2 –
This reconstruction gives rise to a number of questions. There are, for instance, neuter nouns reconstructible for
PIE for which the notion of count plural was semantically appropriate (e.g. *pédom ‘place’; cf. §2.1.3 below) — how
was this expressed by PIE speakers? Moreover, while it is clear that by PNIE the morphological contrast between set
and count plural had been eliminated and that the resulting undifferentiated category was marked by an exponent
of the original count plural for animate nouns and of the set plural for neuter nouns, the details of the diachronic
pathway that led to this situation remain to be worked out. In addition, the hypothesized number system in (2) —
with its morphological gap for neuter count plural — merits further consideration from a typological perspective.
Still more uncertain are questions about the deeper prehistory of this system — in particular, about the de-
velopment of the PIE set plural suffix *–eh2 , which must ultimately be traced back to a pre-PIE derivational suffix
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 12

that also yields the PNIE feminine suffix of the same shape (see §2.1.3 below). For intriguing discussion of how the
pre-PIE suffix **–(e)h2 may have separately grammaticalized as the marker of both set plural and of feminine gen-
der, see Melchert 2000, 2014c, Luraghi 2009a,b, 2011, and Nussbaum 2014a. For more traditional opposing views,
see the references in §2.1.3 below.

§2.1.3 Gender Just as for number, a three-way grammatical gender split is securely reconstructible for PNIE:
masculine, feminine, and neuter. Yet even before the discovery of Hittite and the other Anatolian languages the
gender system of PNIE displayed numerous features suggesting that this three-way gender division might have
replaced an older bipartite system that distinguished only between animate and neuter (cf. Brugmann 1891). When
it eventually became clear that Hittite attests just animate and neuter genders (see further below), it seemed all
but confirmed that Anatolian reflects the older PIE situation, and that the diachronic development of the feminine
gender was a crucial innovation of PNIE. Such a position is now the majority view; see Ledo-Lemos 2000:41–94
and Matasović 2004:36–41 with references. However, concerning the details of the feminine’s development, there
is very little agreement — in particular, it is disputed how the formal affinities between the feminine suffixes *–
ih2 /*–yeh2 –, *–ihx –, and *–eh2 – and the set plural suffix *–(e)h2 (cf. §2.1.2) should be reconciled. In this section,
we outline the principal evidence for the PNIE innovation of the feminine, and discuss some recent hypotheses
about its origin.
The PNIE three gender system — a cross-linguistically common type, occurring in approximately 23% of the
languages surveyed by Corbett (2013) that have grammatical gender (more common is two) — is observed intact
in the oldest stages of most of its language branches: Albanian, Celtic, Greek, Indo-Iranian, Italic, Germanic, and
Slavic. All nouns are specified for masculine, feminine, or neuter gender, and trigger gender agreement on attribu-
tive and predicative modifiers (adjectives, pronouns). In adjectival agreement PIE gender exhibits inflectional
character (cf. Luraghi 2014:199): agreement is obligatorily realized on adjectives with inflectional endings — for
masculine and neuter adjectives, by the addition of PIE animate and neuter nominal case endings respectively (cf.
§PIE-N-case), and for feminine adjectives, by suffixes that generally combine a marker of the feminine with PIE
animate case endings, e.g. feminine accusative singular *–eh2 -m (Ved. –ām, dialectal Gk. –ān, Lat. –am, Goth. –a,
OCS –ǫ, Lith. –ą).
Grammatical gender assignment in PNIE was sensitive, on the one hand, to the animacy and individuation of
a noun’s referent (cf. Ostrowski 1985:316, Matasović 2004:196–203), and on the other, to its sex (Luraghi 2009a).
Prototypical neuter nouns referred to inanimate and weakly individuated entities, thus especially mass nouns, e.g.
*h1 ésh2 -r̥ ‘blood’ (> Gk. éar, TB yasar); *mélit ‘honey’ (> Gk. méli, Goth. miliþ, Alb. mjaltē); that these nouns often
have exact cognates in Anatolian (CLuw. āšḫar ‘blood’; Hitt. milit ‘honey’) suggests they belonged to the inherited
core of PIE neuter nouns (cf. §2.4.3 below). Still, neuter nouns referring to countable entities are reconstructible
for PNIE, e.g. *w(e)rdh om ‘word’ (> Lat. verbum, Goth. waurd); *h2 erh2 trom ‘plow’ (> Gk. árotron, OIr. arathar,
OCS ralo); and in a number of cases, likely further back to PIE, e.g. *pédom ‘place’ (> Hitt. pēdan, Gk. pédon);
*yugóm ‘yoke’ (> Gk. zugón, Ved. yugám Lat. iugum, Hitt. yukan).
In contrast, highly animate and individuated entities like human beings and large animals were generally as-
signed to either masculine or feminine gender depending on the sex (or “natural gender”) of the referent, e.g.
masculine *ph2 tḗr ‘father’ (see §2.1.1 above); *wĺk̥ w os ‘(he-)wolf’ (> Ved. vŕ̥kas, Goth. wulfs, Lat. lupus) vs. feminine
*méh2 tēr ‘mother’ (> Ved. mātā ́, OIr. mathair, TB mācer); *wl ̥kw íhx s ‘she-wolf’ (> Ved. vr̥kīs,́ ON yglr). Yet the
non-neuter genders also take in less prototypically animate members — e.g. masculine *pód– ‘foot’ ([Link]. in Gk.
pód-a, Ved. pā ́dam, Lat. pedem); feminine *nókw t– ([Link]. in Gk. núkta, Lat. noctem, Goth. naht) — while ex-
cluding others that refer to living beings, but are weakly individuated: for instance, *pék̑u ‘livestock’ (> Ved. páśu,
Goth. faihu, Lat. pecū) is neuter. Examples of this kind suggest that referential (or “natural”) animacy and gram-
matical animacy were partially independent, and that factors like individuation (and relatedly, topic-worthiness;
see Comrie 1989:189–95) played an role in gender assignment (cf. Luraghi 2011). Similarly, the fact that words for
‘child’ in the daughter languages are often neuter (e.g. Gk. téknon, OHG kind, OCS dětę) shows that referential
animacy is not a sufficient condition for grammatical animacy.
Although some feminine nominal formations in the PNIE languages are formally indistinguishable from mas-
culines (e.g. *méh2 tēr ‘mother’ cited above), the majority contain a suffix *–ih2 /*–yeh2 –, *–ihx –, or *–eh2 – (referred
to as Motion suffixes in German scholarship). Words containing these suffixes are overwhelmingly feminine in the
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 13

NIE languages, and in many cases, appear to be derived from masculine nominals — in particular, from masculine
*o-stem nouns, where the feminine suffix is traditionally analyzed as replacing the thematic vowel. Exact word
equations support the reconstruction of this process for PNIE, e.g. *wĺk̥ w os ‘(he-)wolf’ ⇒ *wl ̥kw -íhx s ‘she-wolf’
(cited above); *h1 ék̑wos ‘horse’ (Ved. áśvas, Lat. equus) ⇒ *h1 ék̑w-eh2 (Ved. áśvā, Lat. equa, OLith. ašvà). In view
of its productivity, however, it is possible that some of these words were formed independently in the daughter
languages, especially in a case such as *h1 ék̑weh2 , where an older strategy is likely reconstructible (see below). The
basic strong and weak stem inflection of these suffixes are illustrated in (3) with their outcomes in Vedic Sanskrit;
note that the long vowel of the accusative singular is due to Stang’s Law (see Byrd, this volume):

‘goddess’ ‘she-wolf’ ‘mare’

PNIE Vedic PNIE Vedic PNIE Vedic
(3) [Link]. *deiw-íh2 > devī ́ *wl ̥kw -íhx -s > vr̥kīś *h1 ék̑w-eh2 > áśvā
́
[Link]. *deiw-īm > devīm ́ *wl ̥kw -īm
́ > vr̥kīm ́ *h1 ék̑wām > áśvām
[Link]. *diw-yéh2 -e/os >> devyā ́s *wl ̥kw -íhx -e/os > vr̥kíyas *h1 ék̑w-eh2 -e/os >> áśvāyās

As evident in (3), feminine nouns derived with these suffixes broadly resemble athematic non-neuter nouns,
adding the same animate inflectional endings (e.g. [Link]. *–m) to the suffixed stem (cf. §2.1.1). The only major
point of departure is in the nominative singular of the *–ih2 /*–yeh2 – and *–eh2 – paradigms, which strikingly lacks
the characteristic final *–s of other athematic non-neuter nouns. The accentual patterns shown by the *–ih2 /*–
yeh2 – suffix is further discussed in §3.2 below.
The suffixes *–ih2 /*–yeh2 – and especially *–eh2 – are also associated with the formation of PNIE feminine ad-
jectives. The reconstruction of *–ih2 /*–yeh2 – in adjectival formation is supported by trigeneric (m./f./n.) cognate
sets like ([Link].) Ved. pr̥ thús, pr̥ thivī,́ pr̥ thú; Gk. platús, plateîa, platú ‘broad’ (< PIE *pl ̥th2 ús, *pl ̥th2 wíh2 , *pl ̥th2 ú);
in this set, suffixing *–ih2 /*–yeh2 – to the masculine stem forms the feminine stem, to which are added athematic
animate inflectional endings, as in the noun. Even more well-established are cognate sets in which the feminine
adjectival stem appears to be derived, again as in the noun, by substitution of *–eh2 – for the thematic vowel of a
masculine *o-stem, e.g. Ved. návas, návā, navam; Gk. né(w)os, né(w)ā, né(w)on; Lat. novus, nova, novum ‘new’ (<
PNIE *néwos, *néweh2 , *néwom); this pattern is productively continued in most NIE branches that preserve the
PNIE three gender system intact, including Albanian, Celtic, Greek, Indo-Iranian, Italic, and Slavic.
Yet while adjectival inflection confirms that PNIE had a fully grammaticalized gender system distinguishing
masculine, feminine, and neuter, it also gives one important clue that this three-way division does not reflect the
oldest situation. Evidence for its non-antiquity comes from certain NIE branches that have, in addition to the
three-way adjectival sets cited above, other adjective classes that exhibit only a two-way split, making no formal
distinction between masculine and feminine, while neuter is differentiated from both (in strong case forms) by its
characteristic inflectional endings. This situation is not infrequently observed in athematic noun classes across
the NIE languages — for instance, in compound *s-stem adjectives in Vedic and Greek (m./[Link]. Ved. su-
mánās, Gk. eu-menḗs; n. Ved. su-mánas, Gk. eu-menés ‘good-minded; kindly’) and in most adjectives of the 3rd
declension in Latin (e.g. m./f. immortalis; n. immortale ‘immortal’) — but also, more strikingly, in Greek “two
termination” thematic adjectives, where the endings canonically associated with masculines marks both mascu-
line and feminine gender. Greek has a number of simplex two-termination adjectives, e.g. m./f. ph orós, n. ph orón
‘bearing’; m./f. pátrios, n. pátrion ‘hereditary’, and most o-stem compound adjectives are two-termination, e.g.
m./f. á-dikos, n. á-dikon ‘unjust’; m./f. kh rusó-th ronos, n. kh rusó-th ronon ‘having a golden throne’. Some regularly
two-termination adjectives are also attested with distinctive feminine forms (e.g. Att. Gk. patríā), but these forms
are demonstrably innovative in Greek; this innovation further recommends the possibility that the corresponding
Latin and Sanskrit adjective classes, which regularly have distinct feminine forms, have independently undergone
the same development in their shallow prehistory (see Wackernagel 1926-8 [2009]:460–3). Moreover, the pronom-
inal systems of these languages likely show parallel developments: in Greek, the interrogative pronoun has one
form for masculine and feminine (tís < PIE *kw ís), another for neuter (tí < PIE *kw íd), while Latin and Sanskrit have
developed distinct feminine forms: m./f./n. Lat. quis, quae, quid; Skt. kás, kā ́, kím.
A similar situation is also observed in thematic nouns. Although canonically associated with masculine gender
in PNIE, thematic nouns nevertheless may be grammatically feminine in Greek and, to a lesser extent, Latin (on
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 14

the latter, see Vine, this volume). Numerous feminine thematic nouns are attested in Greek, e.g. ([Link].) hodós
‘road’; kópros ‘excrement’; ph ēgós ‘oak’ (cf. Lat. fāgus ‘beech’, also feminine), as well as thematic nouns that are
grammatically feminine when the sex of their referent is female, e.g. (hḗ) troph ós ‘nurse’; (hē) aoidós ‘female singer’.
Just as in two-termination adjectives, there is a tendency in Greek to create new overtly marked feminine forms (in
–ā/–ē < *–eh2 ) for these female entities; as a result, some dialects use innovative (hē) theā ́ ‘goddess’ (e.g. Hom. Il.1.1)
against the older situation observed in, e.g., thḗleia theós ‘female god’ (Il.8.7; for an analogous usage in Old Latin, cf.
Ennius’ lupus fēmina ‘female wolf’ (Ann. 65, 66 Skutsch)). If the same tendency were occurring in the prehistory
of the other NIE languages, it might explain how the congenitor of Gk. (hē) híppos ‘mare’ (< PNIE *h1 ék̑wos ‘id.’)
was replaced by *–eh2 -characterized *h1 ék̑weh2 in these languages.
In nouns, adjectives, and pronouns, then, there is an observed tendency in the NIE languages for feminine
forms to be secondarily differentiated, often via further characterization of the masculine stem with one of the
PNIE feminine suffixes. This pattern suggests that the largely sex-based division between masculine and feminine
in PNIE was subordinate to a primary split between animate and neuter genders. When it was eventually estab-
lished that the Anatolian languages have a two-gender system of this kind, opposing just animate (traditionally
“common”) and neuter genders, two possible diachronic scenarios presented themselves: either PIE had a skewed
system similar to PNIE and the relatively less entrenched feminine gender was lost as a grammatical category in
Anatolian; or the two-gender animate-neuter opposition attested in Anatolian reflects the original PIE system, and
the emergence of the feminine gender is an innovation of PNIE (possibly excluding Tocharian; see below). The
issue has long been a source of significant debate, although over the last decade, a general consensus has emerged
that the Anatolian situation is archaic (see Melchert fthcm. a and Jasanoff, this volume).
This conclusion stems from a reassessment of evidence previously held to indicate that the feminine gender
was lost in the prehistory of Anatolian. Earlier scholarship had identified apparent traces in Anatolian of the formal
markers associated with the PNIE feminine, which were taken as support for the category’s inheritance (similar to
the relic forms held to show inheritance of dual number; cf. §2.1.2). However, some of these alleged traces were
later shown to be spurious. A case in point is “i-mutation,” a phenomenon observed in Luwian, Lycian, and to a
lesser degree Lydian and Carian, in which some noun and adjective classes have common gender nominative and
accusative forms that, in contrast to other paradigmatic forms, show an –i– inserted between stem and inflectional
endings (Starke 1990:54–85). This feature was argued to be either a reflex of *–ihx – (Starke 1990:85–89) or of *–
ih2 /*yeh2 – (Oettinger 1987; Melchert 1994b); however, it has now been demonstrated by Rieken (2005) that “i-
mutation” likely has nothing to do with either of these suffixes, and instead reflects the analogical influence of
ablauting *i-stem paradigms. Other traces of the PNIE feminine suffixes were correctly identified, but in functions
that give little reason to identify them with an erstwhile feminine gender (cf. Hajnal 1994, Melchert 2014c). The
suffix *–ihx – is likely contained in the Hittite adjective nakkī– ‘heavy; weighty’ (< PIE *h1 nók̑-ihx – ‘burdensome’
⇐ *nók̑-o– ‘burden’; cf. Widmer 2005), but nakkī– is an ordinary adjective with no special synchronic association
with any particular gender or sex, and its derivation can in any case be explained by assuming that *–ihx – was used
in its original function as an appurtenance suffix (e.g. Lohmann 1932:67–70, 81–83, Balles 2004) rather than as a
feminine marker.
The suffix *–eh2 is much better attested in Anatolian, but clearly absent is the PNIE sex-based semantic cor-
relation with female referents. This suffix is found, especially, in Lycian (Melchert 1992b, Hajnal 1994), where
it forms concrete and abstract nouns of animate gender (e.g. χupa– ‘tomb’; arawa– ‘freedom’), and is also con-
tained in the productive complex suffix –(a)za– (< PIE *–tyeh2 ) that marks animate nouns referring to professions
(asaxlaza– ‘governor’, wasaza– ‘(kind of priest)’, zxxaza– ‘fighter’). Although some Lycian *–eh2 nouns do have
female referents (e.g. Lyc. lada– ‘wife’, χñna– ‘grandmother’), still more refer to such (primarily male) professions
or else to naturally inanimate entities (i.e. concrete objects or abstract concepts). The other Anatolian languages
present a similar picture. The same *–tyeh2 suffix may be attested in Luwian, e.g. CLuw. urazza– ‘great’; wašḫazza–
‘sacred’ (the latter potentially a direct cognate of Lyc. wasaza–; see Sasseville 2014/15:108–9, but for a different
view, Yakubovich 2013a:159–61). A few animate concrete and abstract derivatives of *–eh2 are also attested in
Hittite, e.g. ḫišša– ‘hitch-pole’, ḫāšša– ‘hearth’, wārra– ‘help’. Although the derivation of these Hittite nominals
is partly obscured by various morphophonological developments, the *eh2 -origin of wārra– ‘help’ is assured by
CLuw. warraḫit– ‘id.’ (a derived neuter abstract in –it– preserving the final *h2 of its base) and for the other two
cited forms by (near) word equations within Anatolian or with PNIE feminine nouns: Hitt. ḫāšša– ‘hearth’ = Lyc.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 15

(abl-instr.) χaha-di ‘id.’; Lat. āra, Osc. aasa– ‘altar’ (< PIE *h2 ó/éh1/3 s-eh2 ); Hitt. ḫišša– ‘hitch-pole’ = Ved. īṣā ́ ‘id.’ (<
PIE *h2 ih1/3 s-eh2 ).
Two final arguments speak against inheritance of the feminine into Anatolian. First, Hittite and Luwian show
clear evidence of a different, perhaps even more archaic strategy for deriving nouns that refer exclusively to female
entities, viz., use of a derivational suffix based on PIE *sor ‘woman’ (on the development of which, see recently
Harðarson 2014). While PIE *sor is attested only in traces in the NIE languages — with further characterization, as
a word for ‘woman’ (Ved. strī,́ OAv. strī), in terms for females, e.g. PNIE *swésor– ‘sister’ (> Ved. svásar–, Lat. soror–),
and in feminine case-forms of certain numerals (see §2.3 below) — it appears to have developed in Anatolian into
a somewhat productive suffix, which is attested in oppositional male-female pairs such as Hitt. išḫā– ‘lord’ : išḫa-
ššara– ‘lady’ and (derived adjectives) CLuw. nāni(ya)– ‘brotherly : nāna-šr-i(ya)– ‘sisterly’. The other, still more
important point, is that inheritance of morphemes used to derive nouns with female referents does not imply
inheritance of the feminine gender as a grammatical category (cf. Hajnal 1994; Melchert 2014c). Grammatical
gender is defined by syntactic agreement (e.g. Corbett 1991:4–5), and there is no synchronic evidence for uniquely
feminine agreement in the Anatolian languages. Noticeably absent is adjectival agreement of the productive PNIE
type *néw-os, *néw-eh2 , *néw-om (except possibly as a marginal innovation in Lycian; see Melchert 1994b:236–7);
rather, Anatolian *–eh2 -nouns as well as *sor-suffixed female nouns behave just like other animate stem classes
with respect to adjectival and pronominal agreement patterns, which therefore provide no evidence that these
nouns were grammatically feminine.
Both the Anatolian and PNIE-internal facts are therefore best explained by the hypothesis that PIE had a two-
gender animate-neuter opposition, and that the feminine gender was a PNIE innovation, or perhaps even later,
during the period of NIE unity subsequent to the departure of Tocharian (in support of this hypothesis, see Kim
2009 and Hackstein 2011, and against, Fellner 2014; cf. Pinault, this volume). Setting aside the issue of Tocharian,
most recent scholarship has adopted the position that the three-gender system was an innovation. Accordingly,
more attention has been paid to the vexed question of the origin of the feminine gender (see the papers collected
in Neri and Schuhmann 2014). In this respect, opinions fall principally into two camps: (i) the feminine devel-
oped (primarily) via the reanalysis of PIE neuter “collectives” (i.e. set plurals; see §2.1.2); or (ii) the feminine arose
(primarily) from within the animate gender.
The first view is driven, above all, by the formal affinity between the the PIE set plural suffix *–(e)h2 and the
markers of the PNIE feminine, (arguably) all of which contain *–h2 . In particular, the phonological identity of the
*s-less nominative singular of PNIE feminine *–(e)h2 -nouns and PIE neuter “collectives” (noted above) was taken
already by Schmidt (1889) to indicate the historical relatedness of these formations, and subsequent scholars (e.g.
Harðarson 1987, 2015, Tichy 1993, Matasović 2004, Litscher 2014) have argued that the former developed directly
via reanalysis of the latter. Under this view, the core of the feminine gender was constituted by a subset of erst-
while *–h2 -marked neuter “collectives” that became semantically specialized with reference to females, e.g. PNIE
*h2 widh éw-eh2 ‘widow’ (cited above) from an original meaning **‘(set of) dead person’s relatives’ (Tichy 1993:16);
the suffix *–eh2 in these nouns was then reinterpreted as the formal marker of feminine gender.
Yet while this hypothesis has the virtue of explaining the remarkable phonological shape of PNIE feminines, it
suffers from a number of serious issues (see Luraghi 2009b, 2011; Melchert 2014c). First, only a few words with any
claim to antiquity are plausible candidates for the semantic development from “collective” to feminine, and in each
case, the original collective meaning for these nouns is entirely conjectural: the daughter languages provide no
evidence that (e.g.) PNIE *h2 widh éw-eh2 meant anything other than ‘widow’. It is therefore questionable whether
such a development occurred (repeatedly), and if so, whether the number of items affected was sufficiently robust
to constitute the core of a new grammatical category. Even more problematic, however, is that these accounts
generally assume that the reanalysis of these neuter “collectives” as feminine singulars was facilitated by the fact
that they were grammatically singular, and so exhibited singular agreement patterns; however, as discussed in
§2.1.1, these “collectives” were grammatically plural already in PIE.
The alternative account assumes that the PNIE feminine arose primarily out of the animate gender. This hy-
pothesis — strongly advocated already by Meillet (1931) — explains the close affinities between the PNIE mascu-
line and feminine gender discussed above, especially their formal identity in some stem classes, via their common
descent from PIE animate nouns; grammatically feminine nominals belonging to the undifferentiated classes were
archaisms in the grammar of PNIE, and predictably, were subject to re-characterization in the daughter languages,
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 16

a pattern that, as noted above, is observed within the attested period of several NIE languages. Also explained un-
der this hypothesis are the word equations between Anatolian animate singular and PNIE feminine singular nouns
cited above (e.g. Lat. āra ‘altar’ = Hitt. ḫāšša– ‘hearth’), the latter of which developed from original animates when
the suffix *–eh2 — together with *–ih2 /yeh2 – and *–ihx – — became associated with the feminine gender after
the separation of the Anatolian branch. Luraghi (2009a, 2011) has adduced typological support for such a gender-
based split at the high end of the animacy hierarchy, as well as for Meillet’s (1931:19) proposal that a crucial step
in the grammaticalization of the feminine gender was the extension of feminine marking (*–eh2 ) to the animate
demonstrative pronominal stem *so/to– (i.e. the creation of *seh2 /teh2 –; see §2.2.2). Exactly how the PNIE fem-
inine suffixes came to be associated with the feminine gender is uncertain. Luraghi 2011 and Melchert (2014c)
present detailed proposals, both of which posit a core of PIE *–(e)h2 -marked animate nouns with female referents
as the starting point; however, numerous open questions remain — such as which nouns played a pivotal role, or
what mechanisms gave rise to agreement (cf. Luraghi 2014) — that call for further research.
Significantly, this latter account departs from the former by situating the historical connection between the
PIE set plural suffix and the PNIE feminine markers in pre-PIE. Although all the markers involved likely originate
from a unitary (probably derivational) suffix *–h2 , already by PIE this suffix had become an inflectional marker of
neuter (set) plural, and given rise to the (animate) derivational suffixes that eventually developed into the major
exponents of the PNIE feminine gender. On the chronology of these developments, see especially Melchert 2014c,
and generally on the prehistory of *–h2 , Nussbaum 2014a.

§2.2 PIE pronouns

§2.2.1 PIE pronominal inflection “Pronominal inflection” refers to the distinct inflectional properties of the
pronouns (personal and deictic/anaphoric), as well as determiners, wh-words (interrogative and relative) and
(some) quantifiers against nouns and adjectives. A number of formal peculiarities motivate a special treatment of
pronominal inflection: the neuter nom./acc. singular case ending *–d, e.g. deictic/anaphoric *tó-d (e.g. Lat. istud
‘that’, Hitt. apāt ‘that’); the affix *–sm– in masc. and [Link]. forms, e.g. [Link]. *tó-sm-ōi ‘to that one’ (> Ved.tásmai,
Go. þamma), and its feminine counterpart in *–sy–, e.g. [Link]. *to-sy-eh2 -ei (> Ved. tásyai, Go. þizai); [Link].
in *–oi instead of nominal *–es, e.g. deictic/anaphoric *toi (> Ved. té, Go. þai, or Hitt. anim. [Link]. kē ‘these’);
a segment *–s– appears in the [Link]., e.g. [Link]. *teh2 -s-ōm (> Ved. tā ́sām, Hom. Gk. táōn). These inflectional
features are all peculiar to pronominal inflection, although later in the development of the IE languages the in-
teraction of nominal and pronominal inflection led to a diffusion of forms (see, e.g., §2.1.1 above on [Link].). For
some of these idiosyncrasies internal reconstructions have been proposed: the affix *–sm– might be the numeral
‘one’ *sem–, and on that basis (and with more daring) fem. *–sy– might have arisen via deletion of *m in pre-PIE
**–sm-y– (Ringe 2006:55).

§2.2.2 Deictic/anaphoric pronouns A number of deictic/anaphoric stems can be reconstructed for PIE; we
illustrate in (4) some points of pronominal inflection with the *so/to– deictic/anaphoric pronoun using masculine,
feminine, and neuter, singular and plural forms (omitting the dual, whose reconstruction is highly uncertain). Note
that we give the oblique cases with hesitation between *o- and *e-grades; Cowgill (2006:524–7) held that only *o-
grade was found in the paradigm, but this is not certain. For further details see Weiss (2011:335–54) and on the
evidence from Indo-Iranian — a key source for this reconstruction — see Gotō (2013:67–73).

Singular Plural
m. f. n. m. f. n.
nom. *só *séh2 *tód *tói *téh2 es *téh2
acc. *tóm *téh2 m *tód *tóms *téh2 ms *téh2
(4) instr. *tó/éh1 *té(h2 )ih2 eh1 *tó/éh1 *tṓis *téh2 bh i(s) *tṓis
dat. *tó/ésmōi *tó/ésyéh2 *tó/ésmōi *tóibh os *téh2 bh os *tóibh os
abl. *tó/ésmōd *tó/ésyeh2 es *tó/ésmōd *tóibh os *téh2 bh os *tóibh os
gen. *tó/ésyo *tó/ésyeh2 es *tó/ésyo *tóisōm *téh2 sōm *tóisōm
loc. *tó/ésmi *tó/ésyeh2 i *tó/ésmi *tóisu *téh2 su *tóisu
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 17

The absence of the *so/to– pronoun in Anatolian is a puzzle: the pronoun might have originated as an inno-
vation of PNIE (n.b. the paradigm is found in Tocharian, e.g. TB se, sā, te; oblique ce, tā, te, etc.). However, the
persistent idea that the source of the PNIE *so/to– pronoun is to be localized in the clause initial conjunctions
seen in Old Hittite (not elsewhere in Anatolian) šu, ta is untenable (see Jasanoff this volume, Melchert (fthcm.
a)). Within the history of numerous daughter languages deictic/anaphoric pronouns became articles (see esp.
Wackernagel (1926-8 [2009]:555–8) on their development); for PIE we reconstruct a language with no articles.

§2.2.3 Relative pronoun A division in the formal exponence of the relative pronoun splits the IE world: there
are languages that mark their relative clauses with reflexes of *hx yo– (Greek, Indo-Iranian, Phrygian, Celtiberian,
etc.); and languages with reflexes of *kw i–/*kw o– (Italic, Anatolian, etc.). This division reflects a diachronic change
in the latter set: *hx yo– was the formal exponent of the PIE relative pronoun, while *kw i–/*kw o– was an indefinite
and interrogative pronoun that came to mark relative clauses in Italic, Anatolian, and elsewhere. The develop-
ment of relative markers from interrogative pronouns — more typologically plausible than from indefinites —
is especially well-attested in languages of Europe (Probert 2014:146–9). Probert (2015:444–8) reconstructs the fol-
lowing prehistoric system underlying relative clauses in Ancient Greek: free and semi-free relative clauses; relative
correlative sentences; restrictive postnominal relative clauses; “paratactic” relative clauses. It is highly likely that
this range of relative clauses was in place in PIE and was marked by the relative pronoun *hx yo–: the same system,
albeit marked with innovative *kw i–/*kw o–, underlies Old Hittite (Probert 2006c) and Anatolian (Melchert 2016c),
as well as other IE languages. For further discussion of relative clause morphosyntax see the helpful summary by
Clackson (2007:173-6), as well as Hale (this volume) on Indo-Iranian, Huggard (2015) on Hittite. The inflection of
the relative pronoun was the same as that of the *só/tó– pronoun (minus stem suppletion), as witnessed by the
Ved. paradigm yás, yā ́, yád, with pronominal inflection fully intact (e.g. masc. [Link]. yá-sm-ai, loc. yá-sm-in,
[Link]. yé etc.), for which Gotō (2013:74-5) presents a diachronic overview.
A number of languages reflect the formal marker of the relative but in changed roles: for instance, Baltic and
Slavic have a suffixed pronoun built on the stem *hx yo– used in marking definite adjective declension; Insular
Celtic has forms of *hx yo-, continued as the relative endings of the simple verb (cf. Watkins 1994:22–30 [= 1963:24–
32]) (of Celtic languages, only Celtiberian attests an inflected relative pronoun, io– < *hx yo–); and in Germanic (as
well as Baltic and Slavic) are found complementizers and other subordinating conjunctions built to the relative
stem, e.g. Go. jabai ‘if’.

§2.2.4 Interrogative-indefinite pronoun The stem *kw i–/*kw o/e– (just mentioned) had two uses in PIE: as an
interrogative when accented (*kw ís > Gk. tís ‘who?’) and as an indefinite when enclitic (Gk. tis ‘someone’, Lat. sī
quis ‘if someone’). Robust evidence may be quoted for both an o-stem, e.g. Go. hv-as, f. hv-o, n. hv-a < *kw o–,
and an i-stem, e.g. Gk. t-í-s, t-í ‘who?’, Lat. qu-i-s, Hitt. ku-i-š, ku-i-n, n. ku-i-d < *kw i/e–. It is likely that the formal
distinction overlays an older functional one: perhaps the o-stem was originally adnominally used, the i-stem as
a full nominal, an idea rooted in the teaching of Warren Cowgill: see Sihler (1995:395–400) and Ringe (2006:56).
Note that in a number of traditions the interrogative takes over the function of the relative pronoun (for reasons
why, see just above on relatives); such a transferral occurred in Italic, Baltic, Slavic, Iranian, Hittite, and Tocharian.
An indefinite use marked by a doubling of the pronoun is familiar from Lat. quisquis ‘whoever’, Hitt. kuis kuis
(further uses may be found in Weiss 2011:350–3)

§2.2.5 Personal Pronouns Personal pronouns have been well characterized as the “Devonian rocks” of PIE mor-
phology (Watkins 2011:xxii) and they tend to be repositories for linguistic archaisms in the IE languages. The recon-
struction of the personal pronouns poses many unique problems, which cannot be addressed within a treatment
of this scope: pronominal topics are most fully dealt with by Katz (1998), which remains unpublished; overviews
representative of different schools of thought may be found in Sihler (1995:369–82), Meier-Brügger (2010:361–4),
Beekes and de Vaan (2011:232–6), and Dunkel (2014).
The personal pronouns show stem suppletion of nominative vs. oblique cases (cf. Eng. I vs. me) that recalls the
*so/to– pronoun; furthermore, the singulars, duals, and plurals are formed from different elements (Eng. I vs. we).
Case marking is realized idiosyncratically in the personal pronouns — for instance, the [Link]. of the first person
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 18

pronoun is reconstructed as *(h1 )eg̑oh2 (e.g. Gk. egṓ) with no recognizable marker of [nominative], and the [Link].
*méne (> OCS mene) has no clear exponent of [genitive]. Pronouns were not distinctively marked for gender, a
feature already noted by the ancient grammarians (see Wackernagel 1926-8 [2009]:405 with references); a notable
exception is Tocharian, which does distinguish between masculine and feminine in the 1sg., i.e. [Link]./acc. TA
näṣ /TB ñäś/ñiś), [Link]./acc. ñuk (see the explanation of Jasanoff 1989).
As is common cross-linguistically, PIE had tonic and clitic forms of the pronouns outside the nominative sin-
gular. A special development is the inner-Anatolian creation of subject enclitic pronouns for unaccusative verbs
(i.e. intransitive verbs whose argument is not semantically agentive), as proposed by Garrett (1990, 1996) and re-
cently maintained by Goedegebuure (2013). On the development of clitics in Vedic (and cross-linguistically) see
Hale (2007:255–88). PIE probably did not have third person personal pronouns, but rather employed demonstra-
tives. A reflexive pronoun *swe– (and/or *se) is often reconstructed (cf. Lat. [Link]. sē, etc.), and is seen as the
basis for the reflexive adjective *swo– ‘one’s own’. Kiparsky (2011) argues that PIE had no reflexive pronoun, but
*swe– was an adjective meaning ‘own’ (grammaticalized to a possessive reflexive in certain languages), *se– was
a referentially independent demonstrative pronoun (weakened to an anaphoric pronoun and then in certain lan-
guages grammaticalized to a reflexive). The pronominal stems of the first and second person pronouns form the
basis for inflecting the reflexives of the first and second persons in Greek, Germanic, Latin, and Slavic (Petit 1999).
We provide in (5) a representative sample of first and second singular and plural forms to illustrate the sup-
pletion and unique forms characteristic of this area of IE morphology (clitic forms are preceded by “=”):

(Hom.) Gk. Ved. Hitt. Goth.

1 [Link] egṓ(n) ahám ūg ik
1 [Link] emé, =me mā ́m, =mā ammuk, =mu mik
2 [Link] sú, tū ́ nē t(u)vám zik þū
(5) 2 [Link] sé, =se tvā ́m, =tvā tuk þuk
2 [Link] hēmeîs vayám wēš weis
2 [Link] hēméas asmā ́n, =nas anzās, =nas unsis
3 [Link] hūmeîs yūyám šumēš jūs
3 [Link] hūméas yuṣmā ́n, =vas šumāš, =šmaš izwis

§2.3 Other PIE nominal categories

§2.3.1 Numerals In PIE, the cardinal numbers ‘1’ to ‘4’ were declined, higher numbers ‘5’ to ‘10’ were indeclinable.
The IE languages offer evidence for at least two candidates for ‘1’. One root is seen in *(h1 )oi-no– (Lat. ūnus, Go.
ains, OCS inǔ ‘a certain one’, Gk. oínē ‘the ace on dice’), to which root some languages add a different suffix —
for instance, Ved. é-ka– and Very Old Indic ai-ka– in the Kikkuli-tracts reflect *(h1 )oi-ko–, while OAv. aē-uua–, OP
ai-va– and Gk. oîos ‘alone’ continue *(h1 )oi-wo–. Another root is *sem–, whose outcomes include Gk. heîs, hen– ‘1’,
f. mía (*sm-ih2 ), TA sas, TB ṣe, Lat. adv. semel ‘once’ etc. Whatever nuances these different formations bore in PIE
do not seem recoverable (but cf. Dunkel 2014:2.588-9, 673). Recent research indicates that a stem *syo- ‘1’ should
also be reconstructed, since it has been identified in Hittite (Goedegebuure 2006), Tocharian (Pinault 2006b) and
now Indo-Iranian (Kümmel 2016).
The number ‘2’ is unsurprisingly inflected in the dual: *d(u)woh1 e m. (> Gk. dúō, Ved. dváu/ā ́, Lat. duo etc.),f.
*d(u)weh2 -ih1 , n. *dwo-ih1 . The number ‘2’ had a form used in compounds *dwi–, e.g. Gk. dí-pod-, Ved. dvi-pad-,
Lat. bi-ped- all ‘two-footed’. Cowgill (1985b in Cowgill 2006:433–40) raises the possibility that there existed as well
an uninflected form *dwó for at least PNIE. A stem *bh o– ‘both’ can also be reconstructed (cf. Go. bai etc.), and
within a compound of *h2 ent– ‘face’ it occurs as TA masc. āmpi, TB antapi ‘both’, Gk. ámphō ‘both’, Lat. ambō
(Jasanoff 1976).
The number ‘3’ clearly inflected as an *i-stem, cf. Hitt. teri–, Ved. tráy-aḥ, n. Ved. trī,́ Gk. tría, etc. < anim.
*tréy-es, n. *trí-h2 . The *i-stem basis is seen clearly too in the combining form *tri– (Gk. trí-pod– ‘tripod’, etc.).
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 19

Interestingly ‘3’ (and ‘4’) show an archaic feminine derivation in Indo-Iranian and Celtic, where a morpheme *–
sr– appears instead of the common feminine-deriving *–h2 formants: Ved. ti-sr-áḥ (via dissimilation from *tri-
sr-es) and OIr. téoir (cf. Wackernagel 1905:349–51). The suffix *–sr– likely derives from the lexeme *ser– ‘woman’,
identifiable within Hittite (and elsewhere in Anatolian) as a suffix –(š)šara– for deriving feminines from nouns
denoting human (or divine) males (Hoffner and Melchert 2008:59), e.g. ḫaššuš ‘king’ > ḫaššuššaraš ‘queen’. On
the Celtic evidence see Kim (2008).
‘4’ shows a similarly archaic inflection, masculine and neuter *kw etwores, *kw wetworh2 respectively, but the
feminine again suffixes *–sr–: Ved. cáta-sr-aḥ and OIr. cethéoir, both < *kw éte-sr-es.
Subsequent numerals up to ‘10’ were indeclinable (though daughter languages often introduce plural inflec-
tion). The reconstructed items are *pénkw e ‘5’, *swék̑s ‘6’, *septḿ̥ ‘7’, *ok̑tṓ(u) ‘8’, *(h1 )néwn̥ ‘9’, dék̑m̥ ‘10’.
The higher cardinals ‘11’ to ‘17’ were dvandva compounds based on the uninflected numeral plus ‘10’, so Ved.
dvā ́-daśa ‘two-ten, 12’. Diverse methods of forming certain cardinals were employed in the daughter languages, so
e.g. Gk. hek-kaí-deka ‘six-and-ten, 16’, subtraction in Lat. un-dē-vīgintī ‘one-from-twenty, 19’ or multiplication in
Welsh deu-naw ‘two-nine, 18’.
PIE derived “decads” (‘20’, ‘30’ etc.) with the neuter plural of the numeral plus a decad-deriving suffix based
on ‘10’, probably *–dk̑omth2 (cf. Gk. –konta). The cardinal number ‘100’ is a neuter derivative of ‘ten’, *dék̑m̥ ‘10’ ⇒
**dk̑m̥-tó-m > *k̑m̥tóm (with onset cluster reduction) ‘100’ (e.g. Lat. centum, Gk. he-katón with analogical he– from
the stem hen– ‘1’). A numeral ‘1000’ may be reconstructed as *(sm̥ )-g̑ h eslo-, which is reflected in Ion. Gk. kh eílioi,
Ved. sa-hásra–, Lat. mīlle).
Ordinals were inflected adjectives. The adjectives expressing ‘first’ and ‘second’ are not based on the cardinals
‘1, 2’ — for instance, Lith. pìrmas, Go. fruma, Eng. fore-most continue PIE *pr̥h2 -mo– ‘first’. The ordinals ‘third’ and
above are based on the cardinals, e.g. *tri– ‘3’ provides the base for *tri-tiyo– ‘third’, Umb. terti– ‘third’, Av. θritiia–,
Go. þridja, Eng. third, etc. Other similar derivations are also found, e.g. Gk. trítos < *tri-to-.
For further works on numerals, see the collection of papers in Gvozdanović (1992); handbook treatments in-
clude Ringe (2006:52–5), Weiss (2011:364–76), Meier-Brügger (2010:368–73) and Beekes and de Vaan (2011:237–42).
Rau (2009a:9–64) is an extensive treatment of decads.

§2.3.2 Adverbs In the oldest IE languages inflected nouns and adjectives could be used adverbially (cf. Del-
brück 1888:184–8 on the accusative so used). Additionally one could form adverbs with distinct adverbial mor-
phology. Denominal adverbial suffixes include instrumental-locatival *–bh i, allatival *–e/oh2 ? , and ablatival *–m;
the formal and functional differences between adverbs derived with these suffixes and “adverbial” inflected case
forms were probably minimal, as suggested by the subsequent grammaticalization of each of these suffixes as a
fully productive (pro)nominal case ending in one or more of the daughter languages (see §2.1.1 above). Two more
local adverb-forming suffixes plausibly reconstructed for PIE are ablatival *–tos (e.g. Ved. hr̥ t-tás ‘from the heart’,
Lat. caeli-tus ‘from heaven’, Gk. en-tós ‘from within’) and locatival *–en (Ved. jmán ‘on the earth’).
In some cases, inflected nominal case forms “petrify” in these adverbial functions, surviving synchronically in
the individual languages as adverbs even after the loss of their nominal stem (e.g. Ved. mŕ̥ṣā ‘in vain’ < PIE [Link].
*–eh2 ), of the case itself as a distinct inflectional category (Gk. oíkoi ‘at home’ < PIE [Link]. *–oi), or even of both
(OIr. ís ‘underneath’ < PIE [Link]. *pēd-su ‘at the feet’). Erstwhile case endings can also be the source of productive
adverbial morphology: for instance, it is likely that the Latin deadjectival adverbial suffix –ē (e.g. Cl. Lat. rēct-ē
‘correctly’ : rēctus ‘straight’; cf. Umb. rehte ‘id.’) continues the PIE [Link]. suffix *–eh1 , although the instrumental
case itself is no longer synchronically distinct in the Italic languages.
Adverbs expressing degree or quantity in the daughter languages are often identical to — or else closely re-
semble — neuter nom./acc. adjectival forms, e.g. Lat. multum, Ved. máhi, Gk. méga, Hitt. mekki ‘much’; Lat.
paulum, Hitt. tēpu ‘a little’; the usage is inherited. Temporal and spatial adverbs are frequently indistinguishable
from nominal case forms, of which locative and ablative are especially frequent. It is likely, too, that PIE speakers
could use full repetition of such case forms — āmreḍita, in the terminology of the Sanskrit grammarians — to
form quantificational adverbs that signal unlimited iteration of an event or action, e.g. Ved. divé-dive, Cyp. Gk.
[āmati-āmati], Cl. Arm. awur awur, Hitt. šiwat šiwat* (UD-at UD-at) ‘on day after day; every day’. Full reduplica-
tion of this kind is reasonably well attested in Vedic (see Klein 2003), but fairly limited elsewhere, with few lexical
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 20

matches across languages; yet in view of the (near) cross-linguistic universality of the type (e.g. Stolz et al 2011), it
is plausibly assumed for PIE.
The evidence of the daughter languages does not converge in the reconstruction of a suffix used to derive
manner adverbs. PIE speakers probably used the instrumental singular of an abstract noun (e.g. Ved. sáhas-ā
‘with might; mightily’), or else possibly neuter accusative case-forms of adjectives (e.g. Ved. drav-át; Lat. facile
‘easily’). The oldest languages tend to retain these strategies, but also innovate new denominal suffixes specific to
manner adverbs. The development of Cl. Lat. –ē was noted above; similarly, Greek developed deadjectival –ōs,
e.g. soph -ō ŝ ‘wisely (: soph ós ‘wise’), while Hittite speakers created the denominal suffix –ili, e.g. ḫaran-ili ‘eagle
(ḫaran–)-like; swiftly’ or luwili ‘in the Luwian (URU luwiya–) language/way’. It appears to be characteristic of the
ancient IE languages that such new adverbial morphology coexists with inherited adverb-forming processes.

§2.3.3 Adpositions Adpositions occur as pre- and post-positions in the oldest daughter languages and such
usage is reconstructible for PIE. In some cases, the etymology is obvious. One particularly interesting example is
*h2 enti ‘in front of’. It is clearly related to the noun seen in Hitt. ḫant– ‘forehead’ (whose adverb is ḫanta ‘in front’),
but in PNIE forms an adverb derived from the [Link]. *h2 ent-i > Gk. antí ‘over against, facing’ (governing gen. case)
and Lat. ante ‘before’ (a prep. governing acc., as well as an adv.), adv. Ved. ánti ‘before, facing’. This use of *h2 ent-i
may represent a common innovation of PNIE.

§2.3.4 Particles Finally we note a motley collection of items loosely labelled “particles,” such as Gk. ge, Hitt.
kan, etc. The meanings of these items are hard to pin down in the ancient (and indeed modern) languages, their
reconstructible semantics elusive. At least one interjection is securely reconstructible, an expression of pain and
suffering: Lat. vae, Hitt. uwai, Eng. woe; its expressive meaning (and the issue of reconstructing registers) is
discussed by Watkins (2013). For a comprehensive collection of forms with etymological interpretations see now
Dunkel (2014).

§2.4 Nominal derivation: Overview IE nominal derivation is highly affixing. The majority of affixes are deriva-
tional suffixes added between the root/stem and inflectional endings, yielding a canonical shape that is schema-
tized R(oot)-S(uffixes)-E(nding). There is no theoretical limit on how many derivational suffixes may be added and
it is not uncommon to find more than one in the formation of a given nominal. Traditionally a distinction is made
between so-called “primary” derivational suffixes, which are added directly to the root, and “secondary” suffixes,
which are added to an already derived stem. The distinction is widely employed in IE studies and we maintain it
here. An example of a primary derivative is Ved. śráv-as– ‘fame, report’, derived by adding the primary suffix –as–
(< PIE *–o/es–) to the root śrav– ‘hear, listen’ (< PIE *k̑lew–). To this stem could be further added a secondary suffix
like –yá– (< PIE *–yé/ó–), which forms denominative verbs, to produce Ved. śravas-yá-ti ‘is seeking fame.’
The majority of derived nouns in PIE represent lexical nominalizations of verbal roots. In a cross-linguistic
survey Comrie and Thompson (2007) identify two major types of nominalizations: those of event nouns, form-
ing nouns of action or state from active or stative verbal roots or adjectives respectively, and those nominalizing
semantic arguments of the verb, such as agent, instrument or location. We follow this syntactic-semantic distinc-
tion in our survey of nominal suffixes. Besides these nominalizations, PIE had a core stock of concrete referential
nouns, which are analyzable to varying degrees. This situation is typologically common: languages commonly (al-
ways?) have a class of underived nouns at the core of the lexicon which refer to concrete, everyday entities, such
as humans, body parts, flora, fauna, and celestial or man-made objects (cf. Dixon 2004:3-4).
Besides affixes a few other types of derivation may be mentioned. There is (limited) evidence in the IE lan-
guages for nominal reduplication. One widespread example is *kw e-kw l-o– ‘wheel’ (to PIE *kw el(hx )– ‘turn’), which
is seen in Gk. kúklos, Ved. cakrám, TA kukäl ‘chariot’. Much better is the evidence for nominal compounding,
and we devote space below to a discussion of the main types of compounds (§2.6). No infixes are found in nom-
inal derivation (but there is one in the verbal system; see §4.3.1 below). In addition to affixal morphology, PIE
had non-concatenative (or “transformational”) processes of derivation. In particular, new formations could be
derived through changes only in morpheme-internal vowels (i.e. ablaut) or in accent. For instance, certain types
of derivatives were associated with particular vowel grades: deverbal event/result nouns could be formed with an
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 21

*o-grade root and a thematic vowel suffix — for instance, to the root *g̑enh1 – ‘to engender’ was formed a result
noun *g̑ónh1 -o– ‘what is begotten, child’ (> Gk. gónos). Another non-concatenative process may be analyzed as
conversion, where derivation operates with a shift in accent but no overt affixation. Similarly, Vedic attests pairs
like the neuter noun bráhman–, bráhmaṇas ‘sacred formulation’ beside m. brahmán–, brahmáṇas ‘one possessing
the sacred formulation, sacred formulator’ or neuter noun yáśas– ‘glory’ beside adj. yaśás– ‘glorious’. Such pairs
appear broadly comparable to English diatonic pairs like (noun) cónvert : (verb) convért (on conversion see the
collection of papers in Bauer and Valera 2005). This process is known in the literature as “internal derivation’’ (viz.
as opposed to being derived with an “external” affix); although it clearly existed as a derivational process in Vedic
and Greek, its status in the proto-language is controversial and competing assessments have been advanced, e.g.
by Widmer 2004 and Rau (2009a), and in a similar vein Kim 2013b, differently Kiparsky 2010a.
Some scholars reconstruct an additional word-formation process for PIE whereby new nouns and adjectives
were derived directly from inflected nominal case forms (e.g., the proposal by Nikolaev 2009). The process is re-
ferred to in the literature as “decasuative” derivation (from Lat. cāsus ‘case’). For example, Ved. dámya– ‘domestic’
(in RV metrically dámiya–) would derive from a [Link]. *dóm-i ‘located at/belonging to the home’. However, none
of the ancient IE languages show compelling evidence for productive “decasuative” derivation; rather, commonly
adduced examples are drawn from reconstructed stages of these languages, as is the case for Ved. dámya– (no
direct reflex of PIE *dóm-i is attested in the RV). More problematically, “decasuative” derivation in its strong for-
mulation directly challenges the proposed typological universal that inflection does not feed derivation (Green-
berg’s (1963) Universal 28). Some proposed examples of decasuative derivation are more likely derived directly
from adverbs, for example the Vedic adjective purā-ṇá– ‘old’ from the adverb purā ́ ‘formerly’ (itself historically the
petrified [Link]. of a root noun). Deadverbial derivation is found in the history of many IE daughter languages —
for instance, Lat. intrāre, OE innian ‘enter’ from Lat. intrā, OE inne ‘within’ or Gk. heōth i-nó– ‘early, in the morning’,
an adjective from the adverb ēōt̂ h i ‘at dawn’ (Probert 2006b:273). As a synchronic process, deadverbial derivation
is typologically paralleled in languages such as Khalkha Mongolian (see Aikhenvald 2011:240 for examples and
further references). Unless comparable cross-linguistic parallels for synchronic “decasuative” derivation can be
identified, it may be necessary to limit the “decasuative” hypothesis to forms that could plausibly have passed
through an intermediate historical stage in which the inflected case-form had grammaticalized as an adverb (and
whose derivation would thus be deadverbial). Further typological and diachronic research addressing this issue
is required.
To date no comprehensive treatment of PIE nominal derivation exists. Perhaps the fullest treatment of PIE is
still Brugmann and Delbrück (1906); more up-to-date surveys with bibliography include Meier-Brügger (2010:321–
73, 416–36) and Lühr and Balles (2008); Lühr and Matzinger (2008). The reconstructed PIE noun is agglomerated
into a (non-comprehensive) lexicon by Wodtko et al (2008). Of the older IE languages see Wackernagel and De-
brunner (1954) on Old Indic, Chantraine (1933) on Ancient Greek, Weiss (2011:266–324) on Latin, Casaretto (2004)
on Gothic, de Bernardo Stempel (1999) on Old Irish, and Matasović (2014) on Slavic. Incorporating Hittite and
Tocharian into the PIE picture remains an ongoing project. Rieken (1999) treats many aspects of Hittite noun
formation (cf. Hoffner and Melchert 2008:51–63) and on Luwian see Melchert (2003:194–200); a monograph on
Tocharian nominal derivation remains a desideratum. The chapters of Lieber and Štekauer (2014) offer diverse the-
oretical perspectives on derivational morphology, and on nominal derivation note especially the chapter therein
by Alexiadou.

§2.4.1 Action/state nominalization One common way to form deverbal event/result nouns was via the the-
matic vowel added to an o-grade root, schematically R(ó)-o– (action or result noun, of masc. gender). This type is
known in the literature as tómos nouns after the eponymous Gk. tómos ‘a thing cut, a slice’ (to the verbal root in
Gk. tem– ‘cut’ < PIE *temh1 –). It is securely reconstructible for PIE: in Tocharian the type remains productive, e.g.
TB traike ‘confusion’ (to AB root trik–, TB prs. triketär ‘is confused’), TB pautke ‘a share, tribute’ (to the Tocharian
root putk– ‘divide’ < *put-sk̑é/ó–). Within Anatolian examples include Hitt. ḫarga– ‘destruction’ ( < *h3 órg-o–, to
ḫarg– ‘perish’), ḫarpa– ‘heap, pile’ (< *h3 órbh o–, to ḫarp– ‘separate’). Beside these *o-grade forms a number of
archaic-looking neuter nouns with e-grade are found, e.g. *wérg̑-o-m neut. ‘work’ (> Gk. (w)érgon, ON werk, Av.
varǝzǝm), and *péd-o-m neut. ‘ground, place’ (> Gk. pédon, Hitt. pēda– ‘place’, Umb. peřum, etc.).
tomós agent nouns (schematically *R(o)-ó-) are found beside (and may be derived from) the tómos action
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 22

nouns. Out of context it is often unclear whether they are nouns or adjectives: for instance in Gk., where the
type is strongly represented, phorós ‘bearer, bearing’ (to phérō ‘I bear’) or trophós ‘nurse, rearing’ (m. and f. in
early Gk., to tréph ō ‘I rear, nourish’) could be agent nouns or adjectives. Some further likely examples of the type
are the BS cognates OCS drugǔ ‘companion’, Lith. draũgas, Latv. dràugs ‘friend’ (all < *dh rough ó–, cf. verbal root
in Go. driug-an ‘to serve (militarily)’), and in Italic there is Lat. coquus ‘cook’ (*pokw ós with assimilation) and
procus ‘suitor’ (< *prok̑-ó-). The Indo-Iranian data is complicated by the (non)-operation of Brugmann’s Law, by
derivatives being attested with both active and passive meanings (e.g. OP asa-bāra- ‘horse-borne’), and by the lack
of direct accentual data in ancient Iranian languages; compare examples like Av. aēša– ‘powerful’ to the root is–
‘rule’ probably from earlier *aik̑-á- (on the evidence see Tucker 2013). A possibly related type has zero-grade root
(*R(ø)-ó-), e.g. *yug-ó-m neut. ‘yoke’, Hitt. iukan, Ved. yugám, Gk. zugón, Lat. iugum, OCS igo (see Malzahn 2013
for further discussion of the type and its Tocharian reflexes).
Derivationally related to tómos nouns are the tomḗ nouns, named after the Greek eponym tomḗ ‘the thing
cut off, the stump’, schematically R*(o)-éh2 (feminine event/result nouns). The type is widely found and remained
productive in certain branches. Examples include Gk. ph orā ́ ‘tribute, thing borne’, Lat. toga ‘covering; a toga’ (to the
root teg–), mola ‘ground-grain’ (to the root mol–), OCS pa-toka ‘flowing’ (tek– ‘run, flow’). Whether this formation
existed in Anatolian is disputed: the strongest candidate is Hitt. ḫāssa– ‘hearth’ and Lyc. xaha-di– ‘altar’ beside
Osc. aasa– ‘altar’, Lat. āra ‘id.’ (< PIE *h2 ó/éh1 seh2 ), but the form is not the feminine counterpart of any identifiable
masculine noun or adjective. Two related subtypes are built with the suffix *–eh2 : (i) zero-grade forms like Gk.
ph ug-ḗ ‘flight’ (pheúg-ō ‘I flee’), Lat. fuga ‘id.’, or Go. wulw-a ‘robbery’ (wilwan ‘to rob’); (ii) lengthened-grade forms
such as Gk. kṓmē ‘unwalled village’ (the origin of the lengthened grade in these formations is disputed; see Vine
1998b for one analysis).
*–ti– formed deverbal action (or process) nouns of feminine gender, e.g. *men– + –tí– → *mn̥ -tí- ‘thinking’
(> Ved. ma-tí- f. ‘thinking, thought’) or *bh er– + –tí– → *bh r̥-tí- ‘bearing’ (> Ved. bhr̥-tí- f. ‘bearing; gift’). In Vedic
this formation regularly shows zero-grade of the root (matched by Greek, e.g. dó-si-s ‘giving, delivery’ < *dh3 -tí-
s) and, in the earliest Vedic texts, consistent suffixal accent. The suffix *–ti– has been internally reconstructed
as inducing mobile accent and vowel reductions (e.g. Rix 1992:146), but the evidence for this reconstruction has
been questioned in recent years; see the discussion of the Vedic evidence with references in Lundquist (2015) (and
§3 below). In some languages reflexes of *–ti– were incorporated into the verbal paradigm — for instance, the
Balto-Slavic infinitive reflects PIE *–tēi, the locative singular case (Olander 2015:171–2). However, distinguishing
between event nominalization and infinitive is often difficult (cf. §4.4.2). One noteworthy extension of this suffix
was –ti-ōn– in Latin, which gives a highly productive class of process nouns in all periods of Latin, e.g. nā-tiōn-
‘birth, origin’.
A suffix *–i– formed nouns especially to thematic adjectives, e.g. Lat. ravus ‘hoarse’ > ravis ‘hoarseness’. The
productivity is currently a subject of research; [Link] 2013 and Grestenberger (2014b) for one approach. Like-
wise the suffix *–ti– (just discussed) may be a derivative of adjectival *–to–, for which possibility see Schindler
(1980:390), Nussbaum (1999:399-400) and Jasanoff (2003b:148n.36).
*–men– formed deverbal neuter nouns, e.g. Ved. dhā ́-man–, [Link]. dhā ́-man-e ‘establishment’ (< *dh éh1 -
men–), Lat. agmen, agminis ‘course, progression; battle-line’ (< *h2 ég̑-men–). The suffix’s inflectional allomorphs
are conditioned by syllable structure: before endings beginning with consonants and word-finally the suffix is in
the zero-grade (e.g. Ved. [Link]. dhā ́-ma-bhis < PIE *dh éh1 -mn̥ -bh is) and before vowels in the full-grade (e.g. Ved.
[Link]. dhā ́-man-e < *dh éh1 -m(e)n-ei); cf. de Saussure (1879:187–8, 205), recently Kümmel (2014:169–70). The pre-
history of this suffix has been internally reconstructed as showing intraparadigmatic accent alternations (Schindler
1975b:263–4, differently Nussbaum 1986:280–1); however, within the IE languages (and by extension within PIE),
this class regularly shows fixed root accent, just as in neuter *–es-stems (on which type see §3.1 below). Most schol-
ars also treat the widespread word for ‘name’ as a *–men-stem (Gk. ónoma, Lat. nōmen, TB ñem, Hitt. lāman– with
dissimilation, etc.), although no verbal root can be unambiguously identified, and it is difficult to derive all reflexes
from a single PIE paradigm.
*–es-stem neuter event nouns represent a type with numerous cross-familial comparanda, e.g. Gk. génos-,
géne-os ‘race, stock, kin’, Ved. jánas–, jánas-as, Lat. genus–, gener-is (all from PIE *g̑énh1 -os–, g̑énh1 -es-e/os, deverbal
to the root *g̑enh1 – ‘generate, become’). The suffix *–es– nominalizes especially “property-concept” roots (see §2.5
below), so (e.g.) in Vedic the root pr̥ th– ‘be broad, expansive’ has a stative nominalization n. práth-as– ‘breadth,
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 23

extension’ (beside the adjective pr̥ th-ú/áu– ‘broad, expansive’). Two recent works have been devoted entirely to
this suffix (and related phenomena), Meissner (2006) and Stüber (2002). Schindler (1975b) is a celebrated internal
reconstruction of the pre-PIE forebear of the *s-stems (cf. §3.3 below).
*t-stems are abstract nouns in the parent language. A very old example is *nókw -t-, a primary t-stem to the
root *nekw /gw – ‘get dark’ (Hitt. nekuzzi), whose reflexes are seen in e.g. Lat. nox, noctis ‘night’, Hitt. [Link]. nekuz
(mēḫur) ‘(time of) evening’ (< *nékw -t-s) (cf. §2.1.1 above). The (same?) t- may be seen in complex suffixes such
as deadjectival *–tāt– (< *–teh2 -t-?), e.g. Lat. līber-tāt– ‘liberty’, Hom. Gk. andro-tḗt– ‘manhood’ (Pike 2011). Com-
parable is denominal *–tūt– (< *–tuhx -t–?), e.g. Lat. senec-tūt- ‘old age, elderliness’. Noteworthy is the t-suffix in
determinative compounds with a root-noun second member, regular in Indo-Iranian (e.g. Ved. –kr̥-t– ‘doer’) but
seen only in trace quantities elsewhere, e.g. Lat. sacer-dō-t– ‘priest’ < (OLat.) sakro– ‘holy, sacred’ + the root noun
*deh3 -t– ‘giver’ (or *dh eh1 -t– ‘placer’). A recent work devoted to the t-stems of IE is Vijūnas (2009).
*–tu– is another deverbal nominalizer, e.g. *men-tu– ‘advice; advisor’ to the PIE root *men– ‘think’ (> m. Ved.
mán-tu–, OAv. maṇtu–). Vedic attests an infinitive built to this suffix in –tavaí (a [Link]. historically) as well as the
Cl. Skt. inf. –tum ([Link].); moreover, in a number of languages the [Link]. has been specialized as a complement
with verbs of motion (directive use of the accusative), traditionally a “supine.” Comparable are the Italic supine
reflected in Lat. –(t)um (Weiss 2011:444–5) and Balto-Slavic in Lith. –tų.

§2.4.2 Participant Nominalization Besides the various event nominalizing suffixes outlined in §2.4.1, PIE pos-
sessed various means to derive “participant” nouns, traditionally “nomina agentium” but participant more accu-
rately reflects their range of syntactic-semantic roles (Alexiadou 2014). Root nouns are formed by adding an in-
flectional ending directly to the root without addition of an overt derivational suffix. The term “root noun” is thus
a solely formal one. Root nouns derived from active verbal roots display agentive meaning, such as Gk. ph ṓr ‘thief,
one who carries off’ < PIE *bh ór-s to the root **bh er– ‘carry (off)’. This derivation was common when root nouns
served as second members of determinative compounds, e.g. Ved. –(g)han– ‘smasher’ (e.g. vr̥ tra-hán- ‘Vr̥tra-
smasher’) to the root han– ‘smash’ (< PIE *gwh en–). Other root nouns are formally identical but are less clearly
agentive, e.g. Gk. poús ‘foot’ < *pod– ‘foot’ (< **‘the goer’?); still others are not obviously agentive and may belong
in the lexical field of underived “basic stock.” Some roots show instead object/result readings, such as *dom–/dem–
‘building, house’ to *dem– ‘build’ (Arm. [Link]. tun < *dōm, [Link]. in OAv. dǝ̄ṇg < *dém-s). A few formal subtypes
of root nouns may be mentioned here, following the reconstructions proposed by Schindler (1972b) (cf. §3.3). One
type appears to have had *o-grade in the strong cases, *e-grade in the weak. Although no IE language synchron-
ically preserves these intraparadigmatic alternations, the fact that a given lexeme shows up with *e-grade in one
language and *o-grade in another suggests a once unified paradigm with alternating *o/e (*dom–/dem– above is
a case in point). What conditions the ablaut grades of the root is not known, although a direct causal connec-
tion with accent can be excluded since root nouns are reconstructed with accented *ó and accented *é (as well as
*ḗ; see further §3.3 below). Studies devoted to root nouns include Schindler’s (1972a) unpublished but influential
dissertation on Old Indic, Kellens (1974) on Avestan, and Griepentrog (1995) on Germanic.
*–tér– and *–tor– agent nouns are found in a number of IE languages. Two varieties are reconstructed for PIE
and may be distinguished on formal and functional grounds, as accented *–tér– (e.g. Gk. dotḗr, dotēr̂ os ‘giver’),
vs. unaccented *–tor- (e.g. Gk. dṓtōr, dṓtoros ‘id.’). Precisely what functional difference underlies this formal
dichotomy has been much disputed. It has recently been proposed by Kiparsky (2016) — following the ancient
Sanskrit grammarian Pāṇini — that the Indo-Iranian reflexes of *–tor– express present/ habitual agency, while
accented *–tér– expresses unrestricted agency. According to this proposal, present agency *–tor– would be more
verb-like in inheriting transitivity from its verbal root (it assigns structural accusative case to objects) and in being
modified by adverbs vs. unrestricted agency *–tér-, which behaves nominally and takes an NP complement in the
genitive case. However, the semantic and morphological properties of this class in Vedic and PIE have not been
settled (cf. the different account by Tichy (1995)). Derivatives of this suffix may be used as adjectives, as (e.g.)
́
in the Latin phrase exercitus victor ‘victorious army’. The feminine of agent nouns was derived via the devī-suffix
(*–ih2 –/*–yeh2 –), e.g. Ved. –trī, Gk. –teira < *–ter-ih2 , Lat. –trīx (with k-extension).
A number of *–n-suffixes form animate participant nouns (denominal and deadjectival). One such suffix is
*–on–, e.g. *gw reh2 w-on– ‘pressing stone, millstone’ > Ved. grā ́v-an– m., Eng. quern, OIr. brau, broon ([Link].),
Lith. gìrnos (from the adj. *gw r̥h2 -ú/éw– ‘heavy’ > Ved. gurú–, Gk. barús, etc.). For the suffix *–en–, cf. Ved. [Link].
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 24

vŕ̥ṣ-aṇ-am ‘bull’ < *–en-m̥ = Gk. [Link]. (w)árs-en-a ‘male’ and Ved. ukṣā ́, [Link]. ukṣ-án-am ‘ox’ = OE [Link].
yxen ‘oxen’ < *–en-es. Another –n-suffix, *–hx on–, formed denominal possessive adjectives. This suffix is known as
the “Hoffmann suffix,” after its discoverer (see Hoffmann 1955 in 1976:378-83). An example is *h2 óyu n. ‘life-force’
(Ved. ā ́yu) ⇒ *h2 yu-hx on– ‘one full of life, youth’ (Ved. yúvan–), weak cases with zero-grade of the suffix *h2 yu-
hx n-es (Ved. yū ́ nas). Another example is PIIr. *mantra-Han– ‘possessing a religious formulation, a mantra’ (OAv.
mąθrān–). A notable use of this suffix was to derive deadjectival and denominal adjectives from thematic and
athematic base forms, exemplified by Lat. caput, capitis ‘head’ ⇒ capit-ōn– ‘having a big head, Capitō (PN)’ or Gk.
gastḗr ‘paunch, belly’ ⇒ gástrōn ‘pot-bellied’. The suffixed adjectives often referred to individuals and so became
popular in onomastic use (in Latin terms, cognōmina). At least some examples of an *n-stem suffix *–mon– appear
to be animate derivatives to deverbal neuter nouns in *–men-, e.g. Gk. térma ‘boundary’ (< *–men–) beside térmōn,
térmonos ‘boundary’ (for one explanation of these pairs see Nussbaum 2014b). Other examples of *–mon– are less
clear, such as the widespread *h2 ek̑-mon– ‘stone; sky’ > Ved. [Link]. áś-mān-am ‘stone’, OP as-mān-am [Link].m.
‘sky’, Gk. ák-mōn ‘anvil, meteoric stone’, Lith. akmuõ, -eñs ‘stone’ etc. (on the semantic discrepancy see Mayrhofer
1986-2001:I.137–8).
Instrument nouns, traditionally “nomina instrumenti,” were formed via the “tool suffix,” which built primary
neuter nouns. Neuter gender was likely correlated with the inanimate nature of the objects. A number of forms
are reconstructible: *–tro-m, *–dh ro-m, *–tlo-m, *–dh lo-m. These suffixes may be partially reconciled under the
assumption that they represent allomorphs conditioned by assimilation of laryngeal features (i.e. Bartholomae’s
Law; see Byrd, this volume 5.1). This assimilation would not, however, account for *r vs. *l. One widespread exam-
ple is *h2 erh3 -tro-m > Gk. áro-tro-n ‘plough’, OIr. arathar, Arm. (h)arawr, Lat. arā-tru-m (rebuilt to the verb arāre),
Lith ár-kla-s. Feminine derivatives to these suffixes are made with *–eh2 , i.e. *–tr-eh2 etc. (cf. exemplification in
Weiss 2011:281–4).

§2.4.3 Concrete Nouns Besides the event and participant nominalizations treated above the PIE lexicon in-
cluded a core stock of referential nouns (cf. Kölligan, this volume). These nouns are analyzable to varying degrees.
For instance, *–r/n-stems form an archaic inflectional class in PIE whose declension is known as “heteroclitic,’’
meaning they decline with an *–r in the strong cases but are suppleted by a stem in *–n– in the weak cases. A num-
ber of neuter *–r/n-stems constitute the basic stratum of the lexicon: words for body parts, ‘water’, ‘fire’, etc. These
archaic words appear to be built directly to roots often no longer recognizable as such. Examples include Hitt. wā-
tar, witen-as ‘water’ (to a weakly attested root *wed–), paḫḫur, paḫḫwen-as ‘fire’, ēšḫar, išḫ(a)n-ās ‘blood’. Beyond
deriving nouns that are neuter in gender, it is unclear what semantics the derivational suffix adds. Evidence for
secondary *–r/n-stems comes especially from Hittite, where there are a variety of suffixes of the shape *–Cer/n–:
*–mer/n–, *–ter/n–, *–wer/n–. All such stems are neuter gender; no adjectives belong to this class (Hoffner and
Melchert 2008:124–30). At least one *–l/n-stem heteroclite existed beside *–r/n–, namely the word for ‘sun’, as in
OAv. [Link]. huuarə̄, [Link]. xv ə̄ṇg ( < *sh2 wen-s), etc.; the *l — although not the heteroclitic inflection — may
be seen more clearly in languages which have not merged PIE *r/l, e.g. Lat. sōl (on the complicated evidence for
this word see Wodtko et al 2008:606–11).
Kinship terms include *ph2 tḗr, ph2 tr-ó/és ‘father’ (Lat. pater, patr-is, Ved. pitā ́, pitúr with remade [Link].),
*b réh2 tēr, bh réh2 tr-o/es ‘brother’ (Ved. bhrā ́tā, bhrā ́tur), *méh2 tēr, méh2 tr-o/es ‘mother’ (Gk. mḗtēr, mētrós, Ved.
h

mātā ́r, matúr, with secondary final accent after ‘father’), and *swésōr, swésr-o/es (Ved. svásā, [Link]. svásre; Lat.
soror, sorōris, TA ṣar, TB ṣer) and dh ugh2 tḗr (Ved. duhitā ́, Gk. th ugátēr, th ugatrós, TA ckācar, TB tkācer, tkātre).
The word for ‘daughter’ is the only kinship term with clear Anatolian cognates: Lyc. kbatra, CLuw. duttariyata/i–,
HLuw. tuwatra/i–. Efforts to etymologize and morphologically segment kinship nouns have been attempted since
the dawn of IE studies but have not met with notable success (compare the very different analyses in Tremblay 2003
and Pinault 2006a). One common analysis segments the word ‘father’ as an agent noun to the root *peh2 – ‘protect’,
but no such analysis is available for *bh réh2 tēr ‘brother’ or méh2 tēr ‘mother’; another analysis would separate a
kinship suffix *–h2 ter– (Sihler 1995:289) but this leaves the awkward “stem” *p– for ‘father’. In our view the kinship
terms are best treated as underived formations in PIE. Notice that *ph2 tḗr, ph2 tr-és ‘father’ and *bh réh2 tēr, bh réh2 tr-
es ‘brother’ share identical ablaut patterns, although the surface accent can be securely reconstructed respectively
on the suffix (*ph2 tḗr) vs. on the root (*bh réh2 tēr).
A number of terms for fauna follow familiar inflectional types and are again analyzable to different degrees.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 25

For instance, the word for ‘sheep’ *h2 ówis (CLuw. hāwī–, Ved. ávi–, ávyaḥ, Lat. ovis, etc.) could contain a suffix
*–i–, but the function of the suffix is not clear, nor is a root identifiable. Thematic inflection may be represented by
*h2 ŕ̥tk̑-os ‘bear’ in Anatolian (Hitt. ḫart(a)gga–) and PNIE (Gk. árktos, OIr. art, etc.); formally the noun is thematic,
though again it cannot be decomposed further into root and suffix. For some items an analysis has been proposed:
*gw ów-s ‘cow’ is a widely attested root noun (e.g. Ved. gáuḥ, Gk. boûs, OIr. bó, TA [Link]. ko) and may go back to
a root *gw eh3 – ‘feed’ (cf. Gk. bó-skō ‘I feed, tend’). This example and the other terms for fauna were presumably
once segmentable.
PIE lexical items that may be early borrowings from non-IE sources get adapted to PIE inflectional categories
rather than showing anomalous non-IE morphology. For instance, *pelek̑u– ‘axe’ (> Gk. pélekus, Skt. paraśú–) has
a non-canonical root shape and is all but certainly a borrowing, but it behaves like an ordinary PIE *u-stem noun.

§2.5 Adjectives Adjectives prototypically attribute qualities (deverbal adjectives) and/or relations with an en-
tity (denominal adjectives) to a referent. PIE adjectives differ from nouns in their agreement patterns (they agree
with a head noun), in their morphology (they are gradable) and in their complements (adverbial modification), but
the dividing line between noun and adjective is not always clearly drawn — for instance, it was observed in §2.4.2
that the agentive *–tor- suffix, seemingly anchored in nounhood, can nevertheless show clear adjectival usage. We
offer here a brief, annotated inventory of salient adjectival suffixes, first of qualitative then of relational adjectives.
Just as for the PIE noun, no comprehensive treatment of the PIE adjective exists: Meier-Brügger (2010:353–60)
provides an up-to-date survey, Rau (2009a:65–186) touches on many aspects of the PIE adjective. Balles’s (2006)
claim that PIE was underspecified in nominal word-formation and knew no categorical distinction between noun
and adjective has not won acceptance.
An important subset of PIE adjectives are those whose roots denote “property-concepts” in the terminology
of Dixon (2004), such as psychological states (‘be thirsty, drunk, happy’), physical properties (‘be heavy, thick,
strong’), values (‘be good, bad’) and internally conditioned changes of state (‘bloom’). These property-concept
roots formed qualitative adjectives in PIE, and continue to do so in the daughter languages. Cross-linguistically
this is the most basic adjectival class, i.e. if a language has only one class of adjectives, this will be it. Qualitative ad-
jectives are prototypically gradable, and this is also the case in PIE; we treat gradation of adjectives below. Dixon’s
(2004) classification of adjectives has been applied to PIE by Balles 2008 and Rau 2009a, 2013; the derivational
relationships between qualitative adjectives and change of state morphology has been examined in a typological
perspective by Koontz-Garboden (2006). As this class of adjectives occupies an important position in PIE and a
central place in current studies of the PIE adjective under the rubric of “Caland suffixes” or “the Caland system,”
we expand on the history of research in this domain before continuing with our survey of suffixes.
Named after its discoverer Willem Caland (1859-1932), “Caland’s Rule” refers to the suffix substitution observed
by Caland between derivatives like compounds with a first member ending in -i- such as YAv. xruu-i-dru- ‘with a
bloody (xruu-i-) spear (dru-)’ and related adjectival forms lacking the -i- element, viz. YAv. xrū-ra-, xrū-ma- ‘bloody’.
Caland demonstrated that the use of -i- in these compounds of xruu-i- vs. the adjectives without it represents a re-
curring pattern; another example is YAv. dǝrǝz-ra- ‘strong’ beside the exocentric compound dǝrǝz-i- + raθa- ‘whose
chariot is strong’ (Caland 1892:266-8 and Caland 1893:592). Wackernagel (1897 (= 1953a:769-75)) then showed that
parallel formations exist in Greek, and so the rule of derivation by suffix “substitution” should be reconstructed
for PIE. His prime example was a compound with a first member with –i-, Gk. arg-i- ‘shining’, beside an adjective
lacking the -i- in argós ‘shining, glistening’ < *arg-ró-s (via dissimilation), matched in the Vedic compound (a per-
sonal name) r̥jí-śvan- ‘who has swift dogs’ and the adjective r̥j-rá- ‘shining, glistening’. Wackernagel referred to this
pattern of suffix substitution as a “rule” (Germ. Regel), though it was — and often still is — called “Caland’s (or
Caland’s and Wackernagel’s) Law” (Germ. Gesetz).
Lively and informed debate continues in the field concerning the nature of “Caland” morphology, which now
extends far beyond the analyses of Caland and Wackernagel, thanks largely to two important works produced in the
1970s, Risch (1974) and Nussbaum (1976). The latter work is an unpublished but widely disseminated dissertation
in which the author proposes to consider the relationships holding between the Caland suffixes as part of a greater
system of root-based derivational morphology. He defines the “Caland system” as a set of parallel derivatives “...all
equally primary and derived more or less simultaneously (in the most remote synchrony which we can actually
recover) as an immediately possible set, one formation implying the others, whatever the starting point of this
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 26

implication” (Nussbaum 1976:5). He illustrates the parallel derivatives with the root *dh ebh – ‘small’ which forms
an adjectival derivative in *–ro- as reflected by Ved. dabh-rá– ‘small’ but also a derivative in *–u– seen in Hitt. tēp-
u– ‘small’ (and in Ved. ádbhuta– ‘unharmed; uncanny’ < *n̥ -dh bh u-to–). Rau (2009a) follows Nussbaum’s approach
and has sought to augment the Caland system to encompass more verbal material (see esp. Rau 2013); a recent
work with extensive bibliography is Dell’Oro (2015). Readable overviews on the history of research in this domain
(with a number of different points of view) may be found in Meissner (1998) and Meissner (2006:14-26). Although
Nussbaum’s dissertation remains unpublished, one can read his treatments of Caland material in more recent
papers, e.g. Nussbaum (1999).
One widespread athematic suffix used to derive qualitative adjectives is *–u/ew–, universally considered cen-
tral to “Caland” morphology (a detailed study of this suffix is de Lamberterie 1990). Typical examples of *u-stems
include Ved. gur-ú-s, gur-áv– ‘heavy, weighty’ = Gk. bar-ú-s, bar-é(w)– ‘id.’ = Go. kaur-u-s < *gw r̥h2 -éw-, and Ved.
ur-ú-s, –áv– ‘wide, broad’ = Gk. eur-ú-s, –é(w)– ‘id.’ This suffix is well attested in Anatolian: Hitt. park-aw– ‘tall,
high’, dašš-aw– ‘heavy; strong’. The Greek and Vedic evidence align in showing regular zero-grade of the root and
accented suffix (see further §3.2 below). Feminines to these stems are formed with the “devī-suffix” ́ (discussed
above), at least incipiently in PNIE. An equation is (e.g.) Ved. ur-v-ī ́ ‘wide’, Gk. eureîa (< PGk. *eur-éw-ya, with
full-grade of the suffix likely analogical after the adjective’s masculine forms); another example is Ved. pr̥ thi-v-ī ́
‘broad (earth)’, Gk. Plataiaí toponym (< *pl ̥th2 -(e)w-yéh2 –, beside the regularized [Link]. plateîa). A number of
old examples show the same form for the masculine and the feminine (so-called “epicene’’ adjectives), such as
Old Lith. platus ‘broad’ and at least one example in Gothic, þaursus ‘withered’ (ad Luke 6.6), apparently relics
predating the introduction of derived feminines (cf. de Lamberterie 1990:886–8).
The thematic suffix *–ro– is closely allied to *–u/ew–, since it also derives qualitative adjectives to property-
concept roots, e.g. Gk. eruth -ró-s ‘red’, Lat. ruber, TB rätre (all < *h1 rudh -ró-s ‘red’). Both Greek and Vedic provide
strong evidence for an inherently accented suffix */–ró-/ as proposed by Probert (2006b:289–94). It is not un-
common to find one language reflecting *–u/ew–, another language *–ro–, both built to the same root: beside
Ved. svād-áv- ‘sweet’ and Gk. hēd-é(w)- ‘id.’ is found Pre-Toch. *swād-ro– as TA swār, TB swāre (in general *–ro–
appears in place of *–u/ew– in Tocharian). Assigning priority to *–ro– or *–u/ew– in such cases is not always fea-
sible; whatever original distinction(s) might have existed between these two adjectival suffixes remains unclear.
Rau (2009a:161–78, 183) discusses the material at length and suggests (Rau 2009a:173 with n.132) that “originally” (in
pre-PIE) both suffixes were denominative to different classes of nouns and were then reinterpreted as deverbative.
Two related *–ro- formations may be mentioned here. Nussbaum (1976:105-10) argues that there was additionally
a category of *–ro– nouns in the proto-language distinct from the adjectives, an example of which would be Gk.
ksu-ró-n n. ‘razor’. Secondly, Vine (2002) has argued that some of the attested full-grade formations in fact reflect a
suffix *–reh2 which derived collectives, of which Gk. mēr̂ a ‘(heap of) thigh-pieces’ and Gk. ágrā ‘the hunt; quarry’
(< *h2 ég̑-r-eh2 ) would be examples.
The status of (pre-)PIE *i-stem adjectives is less clear. i-stem adjectives are well attested in the individual IE
languages, especially Anatolian and Italic, and known elsewhere in the family (e.g. Indo-Iranian, Greek). Many
of these adjectives are made to roots with primary verbal forms, hence are deverbative *i-stem adjectives. Exam-
ples include Ved. śúc-i– ‘gleaming’, Hitt. ḫark-i-š, [Link]. ḫark-ay-aš ‘white, bright’. According to some scholars
(e.g. Meissner 2006:20–5) these adjectives may be connected to first compound members in –i–, so Hitt. ḫarkiš
‘bright’ to Gk. arg-i– ‘bright, shining’ and Ved. r̥j-i– (see above on “Caland morphology”). Other scholars, however,
categorically exclude deverbal *i-stem adjectives from PIE: Rau (2009a:177n.143) finds that “there is no coherent
deverbative i-stem adjectival type reconstructible for the proto-language’’ (so too Grestenberger 2009:8–10).
A series of thematic, deverbal adjectives are best considered together: *–to–, * –lo–, *–no–, *–mo–. The func-
tion of these suffixes may be reconstructed as building resultative adjectives connected to their verbal bases; in
the individual languages they may become participles (see Lowe 2015 on the status of adjective vs. participle). An
example is *–to–: Ved. śru-tá– ‘heard, famed’ (to the root śrav– ‘hear’), Av. sru-ta-, Gk. klu-tó-s, Lat. in-clu-tu-s,
OIr. ro-cloth (< *(pro)-klu-ta-s), all from *k̑lu-tó-s to the root *k̑lew-). In a number of traditions *–to– is integrated
into the verbal system as the past participle, such as Lat. –tus/–sus (cf. Weiss 2011:437–43). Usually (though not
exclusively) an intransitive or passive reading is available. One also finds adjectives with modal-passive meaning,
which express the possibility or necessity of undergoing a particular event (like Eng. adjectives in –able), e.g. Gk.
tlē-tó-s (< *tl ̥h2 -tós) may be both active ‘enduring’ and potential ‘endurable’. The deverbal suffix *–to– is surpris-
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 27

ingly absent from Anatolian; probably deverbal *–to– flourished after Anatolian’s departure from PIE. In its place
Anatolian uses a suffix –nt– which everywhere else in IE forms active participles: the significance of this distri-
bution and the diachronic developments it entails are not fully understood. Within PIE a denominal suffix *–to–
is also found, known as the barbātus type, from Lat. barbātus ‘bearded’; the adjective is built to the noun barba
‘beard’ (there is no verb x barbāre). This denominal type does occur in Anatolian. One noteworthy extension of
*–to– is deverbal *–eto-, studied by Vine (1998a), who finds that it was used in negative compounds and had modal
meaning, e.g. Gk. á-sp-eto-s ‘unspeakable’ from *n̥ -skw -eto-s. The other suffixes in the set also became participles
in the daughter languages; for instance, in Slavic and Armenian *–lo– becomes a past participle.
Denominal relational (or “referential”) adjectives express that a semantic relation holds between the base
noun of the adjective and its head noun; such adjectives are not usually gradable because they denote relations be-
tween entities, not gradable properties (Booij 2012:209–15, Fábregas 2014:279-86). One subclass is qualitative pos-
sessive adjectives, which describe an entity as possessing the notion of the base noun, as in Ved. mádhu– ‘honey’
⇒ mádhu-mant– ‘having honey, honeyed’ or YAv. raii– ‘wealth’ ⇒ YAv. raē-uuaṇt– ‘wealthy’ (athematic suffix
*–ment–, *–went–). Another subclass is adjectives of material, e.g. Att. Gk. kh rūsoûs ‘golden’ (< PGk. *kh rūs-éyos)
from kh rūsós ‘gold’, Ved. hiraṇy-áya- ‘golden’ from híraṇyam ‘gold’ (inherited thematic suffix *–éye/o-). Another
subclass is relational possessive (or “genitival”) adjectives, which express relations also marked by the genitive, e.g.
Lat. patr-ius ‘paternal’ from pater, likewise Ved. pítr(i)ya- from pitár- (thematic suffix *–yo-/*-iyo–, also *–ihx -o–,
cf. Meier-Brügger (2010:417-20)). A suffix *–i(hx )no- makes denominal genitival adjectives from thematic stems,
for instance Lat. dīvīnus ‘divine’, Osc. deivinais. Within the history of many IE languages relational adjectives
compete with genitive nouns, since relational adjectives themselves may express genitival meaning (illustration
in Wackernagel 1926-8 [2009]:485–93).
The thematic vowel could derive denominal possessive adjectives and in this role is accented. An important
reflection of the possessive use of the thematic vowel is a process known by the Sanskrit name vr̥ddhi “strengthen-
ing’’ (sc. of vowel grades). Descriptively an adjective is derived by “strengthening” the ablaut grade of the base and
adding an accented thematic vowel. An example from within Ved. is the athematic neuter noun bráh-man– ‘sacred
formulation’ ⇒ brāh-maṇ-á- ‘one relating to the sacred formulation, Brahmin’. Vr̥ddhi is certainly a derivational
process in Old Indic and examples from Iranian (de Vaan 2003:86–90) demonstrate Indo-Iranian inheritance, but
fewer examples can be drawn from other IE languages. The best examples include reflexes of *deyw-ó– ‘skyling,
sky-god’, derived from the weak-grade *diw– ‘sky’; cognates include Arch. Lat. deivōs ([Link].) ‘gods’, Ved. devá–,
OAv. daēuua–, Lith. diẽvas (Mayrhofer 1986-2001:s.v. devá–, I.742f.). One possible inner-Balto-Slavic example is
Lith. várna ‘crow’ beside Lith. var̂nas ‘raven’ (< PBS *wornós), though the example is disputed (pro: Pedersen
(1933:55) and e.g. Jasanoff (2011); contra: Kortlandt apud Derksen 2014:s.v. varna). The status of vr̥ddhi as a PIE
process has been disputed, especially by scholars from Leiden (cf. Beekes and de Vaan 2011:181–2), primarily on
the grounds that word equations across IE languages are too few for PIE reconstruction. Vr̥ddhi remains, how-
ever, accepted by most scholars today as a synchronic morphological process in PIE; see e.g. Fortson (2010:130),
Meier-Brügger (2010:420), and for one analysis of vr̥ddhi’s historical antecedents see Ringe (2006:13–4).
Gradable adjectives formed comparative and superlative stems by adding suffixes directly to the adjectival
root. The comparative was made with an s-stem suffix *–yos-/-is–, probably of elative or intensified meaning in
PIE, i.e. “is exceptionally X” (Cowgill 2006:193). This suffix is added to the root (not the stem), e.g. Ved. svād-ú/-áv–
m./n. ‘sweet’ ⇒ svā ́d-īyas- ‘sweeter’, matched by Gk. hēd-ú/-éw– and its comparative hēdíō [Link].m. < PGk. *swā ́d-
(i)yos-m̥ (the usual Cl. Gk. form, hed-íon-a, shows innovative n-stem inflection). In Vedic the accent surfaces on
the root (usually in its full-grade) corresponding to leftmost accent in Greek. Both Vedic and Greek show variation
between forms reflecting *–yos and *–iyos (and Vedic also –īyas– with a long vowel): the variation may be due to
Sievers’ Law (see Byrd, this volume), though the details have proven elusive (full discussion in Barber 2013:145–86).
Both Meier-Brügger (2010:356) and Rau (2014) seek to explain the deeper prehistory of the primary comparative
via a further segmentation of the suffix into a nominal *–i– + denominative *–os–.
The most widespread superlative was formed with the suffix *–isto–, whose witnesses include Ved. svā ́d-iṣṭha–
‘sweetest’ = Gk. hḗd-istos ‘id.’, Goth. reik-ists ‘mightiest’ (< PGmc. *–istaz). Its disyllabic shape strongly suggests
that the suffix is composite, and it could be segmented as an agglutinative comparative *–is– + adjectival *–to–
(the aspirated stop in Ved. –iṣṭha- suggests *–isth2 o–, but there is little other evidence to recommend this recon-
struction).
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 28

The PIE suffix *–(t)eros– was contrastive, e.g. *dek̑si-teros ‘right side (opposed to left)’ (> Gk. deksi-terós, Lat.
dexter) and was used with adverbs and pronominal stems, e.g. Gk. póteros ‘which of two’ (< *kw o-teros). The
suffix’s use as a comparative to adjective stems becomes productive only in certain IE dialects, e.g. Gk. díkaios
‘just’ ⇒ dikaió-tero– ‘more just’ (Dieu 2011:680–4). Other formations of the secondary superlative are also found
in the daughter branches, such as Gk. –tatos (< PGk. *–tm̥ -to–) and Italic and Celtic *–is-m̥ mo–, but are not recon-
structible for the proto-language (Cowgill 2006:193–7).
The primary comparative and superlative are not found in Anatolian, Tocharian, Armenian, or Albanian.
While the situation in the latter two branches may be easily attributed to loss, their absence in Anatolian and
Tocharian is more striking, and raises the possibility that the primary comparative and superlative were not only
post-PIE innovations, but were created within NIE after the departure of Tocharian. Both Anatolian and Tochar-
ian express comparison by employing the positive degree of the adjective plus the “yard-stick” (or “standard”) of
the comparison inflected in an appropriate case-form (cf. Hoffner and Melchert 2008:271–6 for Hittite; Carling,
this volume for Tocharian). This strategy for expressing comparison is common cross-linguistically (Stassen 2013).
Given that gradation with affixes is rarer among the world’s languages (Cuzzolin and Lehmann 2004:1215), the de-
velopment of the primary comparative and superlative affixes would represent a significant innovation of PNIE.
Finally, we note that numerous IE languages attest suppletion of comparative and superlative adjectives, fa-
miliar enough from English examples like good, better, best. Suppletion in the forms of gradation is widespread
in the IE languages but specific lexical matches appear to be lacking; see further the overview of Meier-Brügger
(2010:355–60) with references, to which may now be added the full-scale treatment by Dieu (2011).

§2.6 Compounds PNIE is reconstructed as rich in compounds based in the first place on agreements in pro-
ductive compound types (rarely lexical matches) principally between Ancient Greek and Indo-Iranian. The com-
pounds in the earliest stages of these two branches usually consist of two lexemes, although recursion in com-
pounds (structurally [A [B, C]]) was clearly possible and becomes well evinced in the notoriously enormous com-
pounds of the Classical Sanskrit language. Compounding is much more restricted in other IE languages, such as
Latin and Old Irish. More seriously complicating the picture of a proto-language rich in compounds is Anatolian’s
poverty in this regard (cf. Oettinger, this volume §2.2.2); and while Tocharian texts abound in nominal compounds
(see Adams, this volume §5.2) many are thought to represent loan translations from Sanskrit, and Pinault (this vol-
ume) finds that nominal composition does not seem to have been common as a “genuine Tocharian feature.” The
extent to which nominal composition was a productive process at the PIE level is accordingly hard to ascertain.
The topic of compounding is a highly complex one and we can only survey here some prominent types likely to be
reconstructible for the proto-language. The deeper origins of nominal compounding lie beyond our present focus;
we seek to outline the major types of nominal composition found in the ancient IE languages and to reconstruct
the morphology of compounds in the PIE period. Further treatments include Lindner (2011-2016), a discursive
and (to date) unfinished treatment of the history of research on IE compounds which attempts to provide an en-
cyclopedic point of reference; cf. Meier-Brügger (2010:427–30) and Tribulato (2015:13–130) with special reference
to Ancient Greek. A helpful survey of compounding from a number of different theoretical perspectives (though
lacking much discussion of ancient IE languages) may be found in Lieber and Štekauer (2009).
Before beginning our survey, we note a few limitations. Linguists working with ancient languages are always
at the mercy of written evidence and this fact may hamper what is counted as a compound, since vital prosodic
data has often been lost. In the best of circumstances ancient writing systems can actually illuminate the mor-
phological segmentations made by native speakers (for one such case see Morpurgo Davies 1987:268–9), but this
is the exception that proves the rule. The question of how to classify types of compounds is a topic of consider-
able debate in the linguistics literature. With regard to IE studies the problem of classification is exacerbated by
the bewildering diversity of terminology in the literature. For present purposes we adopt a classification based
on syntactic and semantic criteria (following Scalise and Bisetto 2009) and take the basic compound types to be
subordinated (where one member is subordinate to another) vs. coordinated compounds. These two macrotypes
may be subdivided into the morphological/semantic types based on headedness: endocentric (containing a head)
vs. exocentric (not containing a head). To facilitate use of secondary literature we cross-reference terminology as
much as possible.
Compounds in which one member is syntactically/semantically subordinated to another (“determinatives” of
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 29

various stripes) are attested across the IE languages and may be reconstructed for PIE. Determinative compounds
are mostly right-oriented endocentrics (i.e. the category-determining head is one constituent of the compound)
with a nominal or adjectival head, cf. the classic Vedic example rāja-putrá– ‘king-son, son of a king’ or with an
adjectival head Gk. th eo-eíkelos ‘god-like’. When the first member is a noun modifying the second member in
the role of an oblique case, the type is often known by the Sanskrit name tatpuruṣa, for instance Ved. havir-ád–
‘oblation-eating’, Gk. oinó-pedon ‘lit. wine-land’ (‘land for wine’), vineyard’. Determinative compounds where the
first member attributes a property to the second member, i.e. modifies the head element adjectivally (or with a de-
verbal head adverbially) are often known by the Sanskrit name karmadhāraya (also “attributive’’ or “descriptive’’),
although many theoretical models do not include this as a special type. Examples include [AN] compounds like
Ved. kr̥ ṣṇa-śakuní– ‘black-bird, crow’, Gk. akró-polis ‘high-city, citadel’, Gaul. (Latinized) medio-lānum ‘middle-
plain, Milan’, or [A/Adv.-N] as Ved. āśu-pátvan– ‘swift(ly)-flying’. On determinative compounds in Hittite see
Brosch (2010:266–72).
One important type of determinative compound is “synthetic” (“verbal-nexus” or “verbal governing,” Germ.
“Verbale Rektionskomposita,” Skt. upapada). The head is a deverbal noun, either a root noun or an action noun.
Root noun examples include Ved. havir-ád– ‘oblation-eating’, vr̥ tra-hán- ‘Vr̥tra-smashing’ etc. (cf. Scarlata 1999);
action noun examples include Ved. amitra-dámbhana– ‘foe-belittling’. Examples with a second member formed
by *o-grade root and the thematic vowel include Gk. andro-ph ónos ‘man-slaying’ or psukh o-pompós ‘soul-conductor’;
on the complicated evidence for *o-grade in Indo-Iranian, see Tucker (2013). Structurally comparable are Eng.
truck-driver, church-goer or Germ. Macht-haber ‘power-holder, ruler’. “Synthetic” compounds derive their dever-
bal head in the process of compounding. For instance, the aforementioned havir-ád– ‘oblation-eating’ is based on
a potential but not established lexeme (a root noun) ad– ‘eater’, or go-ghná– ‘cow-slaying’ on a potential lexeme
ghn-á– ‘slayer’ which is not established (i.e. it does not exist as a simplex so far as extant records allow). Two
main analyses have been proposed to understand these compounds synchronically: they could be understood as
related to “noun-incorporation,” i.e. a detransitivizing, syntactic process where the patient argument of the verb
is compounded (“incorporated”) with the verb. On this analysis the compounds would be of the structure [[NV]
-er]N ([truck-drive]er ). However, such an analysis would predict unattested and grammatically questionable [NV]
structure for finite verbs such as Eng. X ‘truck-drives’ or Ved. X havir-átti ‘oblation-eats’. An alternative and perhaps
preferable analysis would treat the compounds as [NN] adjunction where the deverbal noun inherits the transitive
argument structure of the base verb, so [N + (deverbal)Ner ]. One problem with the latter option is that it would
predict possible but not established words such as Eng. goer in church-goer or a root noun in Ved. ad– ‘eater’.
It appears that synthetic compounds show a simultaneous use of compounding and derivation, but the issue is
not resolved in a wider theoretical context (cf. Olsen 2014:41–3). See further the clear and influential accounts of
Vedic by Wackernagel (1905:174–232), of Ancient Greek by Risch (1974:189), and note Uhlich (2002) for discussion
and examples from compound-poor Old Irish/Celtic.
“Bahuvrīhis” (in older literature “relativa”) are exocentric, possessive compounds. Exocentricity is not an in-
herent property of the compound (thus not a prime for analysis) but may arise from use in context. As Whitney
(1889:501) put the matter: “A compound having a noun as its final member very often wins secondarily the value
of an adjective, being inflected in the three genders to agree with the noun which it qualifies, and used in all the
constructions of an adjective.” Eng. knuckle-head could be understood as an endocentric compound (‘a head
that is like a knuckle’) but it is used only as an exocentric ‘who has a knuckle-like head’. Examples of bahuvrīhis
include Ved. [A-N] ugrá-bāhu- ‘strong-armed’, [N-N] bāhú-ojas- ‘whose strength is in his arms’ or [prefix-N] su-
mánās ‘good-minded, kindly’, with a numeral first member Lat. bi-dēns ‘having two-teeth’. In other bahuvrīhis an
adjectival suffix may be added to the nominal stem, e.g. Gk. agrió-ph ōnos ‘rough-voiced’ whose second member
–phōn-os is based on the noun phōnḗ ‘voice’. Because the second member “changes” earlier scholars sometimes
called them “mutata” compounds (e.g. Debrunner 1917:54-6). Although there are examples of [AN] compounds,
Schindler (1986) argued that PIE had a further morphological restriction and did not allow material or denominal
adjectives as first members. Instead, the nominal stem was used: Gk. kh rūsó-thronos ‘gold-throned’, where the stem
kh rūso– is used in place of the adjective of material Hom. Gk. kh rū ́ seos ‘golden’, would be an example of an inher-
ited morphological restriction. Functionally, bahuvrīhis are adjectives, attributing a property to a referent outside
the compound. The term “possessive” is a common one, but inadequate: many bahuvrīhis do not express posses-
sion, or express more fine-grained nuances (see the treatment by Schindler (1986)), such as ‘who has a B that has
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 30

A’ (“double possessives”) or ‘who makes/provides XY’ (in the literature “factitive bahuvrīhis”; more generally see
Scalise et al (2009) on types of exocentricity). In general, the semantics may be thought of as “R[elationship holds
in] (B, A),” whose precise delineation of meaning would be guided by speakers’ interpretations (Booij 2012:210–5).
It is sometimes held that the IE languages show trends towards exocentricity and, projecting this trend back, PIE
exocentrics would be considered older than endocentric compounds. The Greek evidence for this position has
been challenged by Tribulato (2015:80–1), who argues that both types are old: they have the same structure but
may be used as determinative nouns or as exocentric adjectives.
In “dvandva” compounds (or “copulative,’’ “co-ordinating,’’ or “co-compounds’’) neither member is subordinate
and its constituents are linked by a conjoining “and’’ relationship. Dvandvas may refer to the aggregate of two
coordinated elements, as in numerals like Gk. duṓ-deka ‘two-ten, twelve’ or [AA] compound like Gk. glukú-pikros
‘bitter-sweet’ where two adjectival properties are attributed to one entity. Dvandvas may refer to a super-ordinate
term: Ved. mātárā-pitárā– ‘lit. mother (du.) - father (du.); parents’ associates two terms without reference to
either one and means ‘parents’ (not x ‘two fathers and two mothers’). Dvandvas where each member is marked
with dual inflection (like Ved. mātárā-pitárā–) may have arisen from the associative/elliptical dual, i.e. where
Ved. mitrā ́ in the dual number was used for Mitra and his conventional associate Varuṇa (cf. §2.1.2 above), and
this construction was then transferred to the dual dvandvas like Ved. mitráyor-váruṇayoḥ ([Link]) ‘of Mitra and
Varuṇa as a pair’ (not x ‘of two Mitras and two Varuṇas’); see further Corbett (2000:228-31) and Kiparsky (2010b).
Dvandvas may also be endocentric (in which case splitting them from determinatives becomes tricky), e.g. Gk.
iatró-mantis ‘physician-seer (of Apollo)’. On the endocentric type in Ancient Greek and IE see Tribulato (2015:63–
7), and for further discussion of dvandva compounds and their cross-linguistic analysis, cf. Olsen (2001).

§3 Morphophonology of PIE

At the heart of PIE nominal morphophonology is the relationship between “ablaut” — i.e. morpheme-internal
alternations in vowel quantity (*Vː : *V : *∅) and quality (*o : *e) — and “accent,” a term traditionally used to
refer to the single word-level accentual peak, whose primary phonetic correlate in PIE was probably high pitch
as in Vedic Sanskrit and Ancient Greek (cf. Byrd, this volume). The collective evidence of the oldest daughter
languages shows a correlation between these variables — in particular, between *e : ∅ vowel alternations and the
presence or absence of accent; in none of these languages, however, can these qualitative or quantitative vowel
alternations be explained by a purely phonological process conditioned by the position of the accent.
The attempt to understand the opaque relationship between accent and ablaut in the IE languages, and in
turn, what should be reconstructed for the proto-language has exercised scholars since the beginning of IE stud-
ies. In this section, we begin by situating the PIE accentual system in typological perspective and discussing the
morphophonological principles by which word accent in PIE was determined. The core features of this accentual
system are outlined in §3.1, while §3.2 turns to issues that arise in complex derivation, where more open questions
remain. Finally, §3.3 takes up the still more difficult problem of the relationship of accent and ablaut.
Readers should be aware that the analysis of PIE word accent laid out in §§3.1–3.2 diverges from the traditional
“paradigmatic” approaches to this problem that are presented in most standard handbooks of the field (Fortson
2010:119-23, Weiss 2011:257-62, Meier-Brügger 2010:336-53, i.a.). One important way in which our discussion differs
is that it does not take ablaut patterns as evidence for word accent at the PIE stage as reached by the comparative
method; rather, it assumes that accent and ablaut were independent variables already at this stage (cf. Watkins
1998:62). We thus focus instead on the position of word accent and the principles by which it is determined in the
ancient languages and as it can be reconstructed for their immediate ancestor. These issues are discussed more
extensively in §3.3 below.

§3.1 PIE lexical accent: The basic system The principal languages generally held to contribute to the recon-
struction of PIE accent — Ancient Greek, Lithuanian, Russian, Proto-Germanic, Hittite, and above all, Vedic San-
skrit — all have prototypical lexical accent systems (on this term see van der Hulst 2014, and in more detail, Re-
vithiadou 1999 and Alderete 2001b). The definitive feature of word-prosodic systems of this kind, which have also
been identified and studied in such diverse languages as Thompson Salish (Salishan; Revithiadou 1999:250–77),
Tokyo Japanese (Japonic; e.g. Poser 1984, Kubozono 2011), Chamorro (Austronesian; Chung 1983), and Cupeño
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 31

(Uto-Aztecan; Alderete 2001c; Yates 2017), is that a word’s accent is not determined by its purely phonological
properties (such as syllable weight or metrical structure), but is rather dependent on what morphemes it contains
and how they are combined. In these systems, certain lexically specified morphemes may “attract” the accent,
either to itself or to an adjacent syllable, while others may be “neutral,” exerting no effect on the position of the
accent. Three such typologically well-established accentual features are securely reconstructible for PIE: inher-
ently accented morphemes, which prefer to host the word’s single surface accentual peak (per above, high tone in
PIE); preaccenting morphemes, which prefer the accentual peak to fall on the immediately preceding syllable; and
inherently unaccented morphemes, which neither attract nor repel the accentual peak. For the sake of consistency
with previous scholarship, we employ the term “underlying accent” or “inherent accent” for this abstract lexical
feature, and maintain the traditional use of unmarked accent to refer to the single surface accentual peak (more
common in the theoretical literature is “accent” for the lexical feature and “stress” for its surface realization; cf.
van der Hulst 2014:4–6).
An example of an inherently accented morpheme is the PIE adjectival suffix *–nó–, whose derivatives regularly
bear suffixal accent in Vedic and Greek, e.g. Gk. hag-nó-s ‘holy’, Ved. yaj-ñá-s ‘sacrifice’ (< PIE *h1 yag̑-nó-s). The idea
that the accentuation of this and other thematic adjectival classes (e.g. –ro–, –to–) should be attributed to some
accentual property of the suffixes themselves was suggested already by Bopp (1854:163–8); generative frameworks
formalize this insight by treating this property as a lexical feature on the suffix marking it as accent-preferring (i.e.
/–nó–, –ró–, –tó–/). Such PIE adjectives would therefore have been derived as in (6):

(6) PIE */h2 erg̑ - ró - (o)s/ → *h2 r̥g̑-ró-s ‘shining’ ([Link])

PIE */k̑lew - tó - (o)s/ → *k̑lu-tó-s ‘heard (of); famous’ ([Link])
The PIE forms in (6) develop into attested Ved. r̥j-rás, Gk. argós (via dissimilation of *r; cf. §2.5 above), and
Ved. śrutás, Gk. klutós. Note that the derivations in (6) assume that quantitative ablaut (i.e. *e/ø) is to some ex-
tent operative as a synchronic phonological process in PIE: the verbal roots have underlying full-grade (*/h2 erg̑/,
*/k̑lew/), but suffixal accent causes their /e/ vowel to be deleted on the surface (*h2 r̥g̑–, *k̑lu–). The IE languages
provide relatively robust evidence for a rule deleting /e/ in syllables immediately preceding the surface accent as
in (6) (according to Kiparsky’s (2010a) proposal, preceding underlyingly accents). However, even this restricted
formulation of the ablaut rule has exceptions in the very same morphological categories — some likely recon-
structible (e.g. Gk. gnō-tós, Ved. jñā-tá– < PIE *g̑neh3 -tó– ‘known’; cf. Vine 2004:360–6), others uncertain due to
mismatches in the daughter languages (e.g. Dor. Gk. dā-rós vs. Ved. dū-rá– < PIE *dw(e)h2 -ró– ‘long’) — as well
as elsewhere in the system (e.g. [Link]. Ved. dhar-tā ́r-am ‘supporter’ ⇐ /dhar/ ‘hold, support’). Further complicat-
ing the issue is evidence for *e/ø ablaut in non-pretonic environments (see, e.g., the discussion of the 3pl. ending
/–énti/ in §4.2.5). Additional research is required to determine under what phonological and morphological con-
ditions quantitative — as well as qualitative — ablaut applied at the PIE stage (Kiparsky (2010a, fthcm.) provides
one in-depth synchronic treatment). Ablaut problems are discussed more fully in §3.3 below.
As in (6), inherently accented morphemes generally receive the surface accent; however, since morpholog-
ically complex words may contain multiple inherently accented morphemes, or alternatively, no inherently ac-
cented morphemes, lexical accent systems have language-specific (morpho)phonological principles that deter-
mine which underlying accent will receive surface accent or else assign a “default” accent in the absence of under-
lying accents. Such principles are employed in analyses of lexical accent systems to model synchronic accentual
variation within morphological categories and across lexemes. In the IE languages an important locus of such
variation is the class of root nouns, some of which are accented on their inflectional endings in their oblique case
forms (e.g. Ved. pad-ā ́ ‘with the foot’), while others show persistent root accent (cf. Ved. gáv-ā ‘with the cow’).
While the surface accent of the former is straightforwardly analyzed as in (7a) as resulting from attraction to the
inherently accented instrumental case ending (Ved. /–ā́/ < PIE /–éh1 /), the latter can be treated as containing
an inherently accented nominal root /gáv/ (an idea foreshadowed by de Saussure 1879:199, further developed by
Kiparsky 2010a:141–4); the virtual competition between the lexical accents of the root /gáv/ and the weak case in-
flectional suffixes is then decided by a phonological principle of “accent resolution.” Similar principles of accent
resolution are standardly assumed to be operative in Tokyo Japanese and Cupeño, where they account for the con-
trast between (e.g.) Jap. yon-dára ‘if (he) calls’ vs. yón-dara ‘if (he) reads’ and between (e.g.) Cu. max-qáʔ ‘(he)
gives’ and ʔáyu-qa ‘(he) wants’ (see Alderete 2001b:49–51, 99). Analyses of accent resolution in all three languages
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 32

are laid out in parallel in (7b):

(7) a. Ved. /pad – ā́/ → pad-ā ́ ‘with the foot’ ([Link].)

Jap. /yob – tára/ → yon-dára ‘if (he) calls’ (call-cond)
Cu. /max – qá/ → max-qáʔ ‘(he) gives’ ([Link])

b. Ved. /gáv – ā́/ → gáv-ā ‘with the cow’ ([Link].)

Jap. /yóm - tára/ → yón-dara ‘if (he) reads’ (read-cond)
Cu. /ʔáyu – qá/ → ʔáyu-qa ‘(he) wants’ ([Link])

However, not all surface accents correspond to underlyingly accented morphemes. For instance, it is evident
from (7a) that roots like Ved. /pad/ ‘foot’ and Cu. /max/ ‘give’ have no underlying accent, since the inherent accent
of the inflectional ending attracts the surface accent; nevertheless, these roots receive the surface accent in other
paradigmatically related forms, e.g. Ved. [Link]. pā ́d-as* ‘feet’ (cf. attested [Link]. Ved. pā ́d-am), Cu. máx-wənə
‘(they) give’. The accentuation of such forms is generally assumed to be the result of a phonological principle of
“default” accentuation, a grammatical process that operates when a word contains no inherently accented mor-
phemes, assigning an accent to a phonologically unmarked position in order to fulfill the typologically common
requirement that all words bear an accent (the “obligatoriness” parameter; see, e.g., Hyman 2006). In Vedic (and
Cupeño), default accent surfaces on the word’s leftmost syllable as in (8a) (cf. Kiparsky 2010a:144, Yates 2017),
while (8b) shows that this default accentual pattern does not arise in words containing the same suffixes if there
is already an accented morpheme present:

(8) a. Ved. /pad – as/ → pā ́d-as* ‘feet’ ([Link].)

Cu. /max – wənə/ → máx-wənə ‘(they) give’ ([Link])

b. Ved. /marút – as/ → marút-as ‘Maruts’ ([Link].)

Cu. /təwáʂ – wənə/ → təwáʂ-wənə ‘(they) lose’ ([Link].)

One important Vedic accentual phenomenon that emerges from (7–8) is the synchronic distinction between
“mobile” root nouns — i.e. that show surface accent on the root in the strong cases, on inflectional suffixes in the
weak — like pā ̆d– ‘foot’, and those with “fixed” (i.e. consistent) root accent like gā ̆v– ‘cow’ (on the strong/weak
case distinction, cf. §2.1.1 above). Root nouns with mobile accent are the dominant type (e.g. nāv– ‘boat’, pur–
‘stronghold’, yudh– ‘fight’), while the minority fixed accent pattern is instantiated by a handful of other lexical
items in addition to gā ̆v–, including nar– ‘man’ ([Link]. nár-e) and raṇ– ‘pleasure’ ([Link]. ráṇ-e). By applying the
same tools used to model similar accentual alternations in Tokyo Japanese and Cupeño, it is possible to arrive
at an explanatory account of the different accentuation of these classes, which falls out directly from a minimal
contrast in the underlying accentedness of the relevant roots (/gáv/ ‘cow’ vs. /pad/ ‘foot’) and affixes ([Link]. /–ā́/
vs. [Link]. /–as/). If Vedic here largely preserves the PIE situation (as is generally assumed), the PIE derivation of
root nouns with mobile vs. fixed accent can be represented as in (9):

(9)
root nouns in PIE
fixed mobile
[Link]. */gw ów – es/ w
→ *g ów-es ‘cows’ */pod – es/ → *pód-es ‘feet’
[Link]. w w
*/g éw – éh1 / → *g éw-eh1 ‘with the cow’ */ped – éh1 / → *ped-éh1 ‘with the foot’

Under this analysis, accentedness and unaccentedness respectively are properties of the Vedic roots /gáv/ (<
PIE */gw ów/) and /pad/ (< PIE */pod/), not properties of their basic (i.e. root noun) inflectional paradigms. In
contrast to the paradigmatic approaches discussed in §3.3, which reify the status of intraparadigmatic accentual
(im)mobility, this analysis takes the respective fixed and mobile accentual patterns of these nouns to be emer-
gent from the lexical properties of their roots. It thus predicts that the underlying accentual contrast between
these roots will recur in derivation, resulting in differences in the surface accentuation of certain morphologi-
cally related forms. In this case, the prediction is borne out: when /gáv/ and /pad/ are further suffixed by Ved.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 33

–mant– or –vant– (< PIE *–ment–/*–went–) — two possessive adjectival suffixes with similar accentual behavior
that probably descend from a single morpheme at some stage of the proto-language (cf. Wackernagel and Debrun-
ner 1954:781–2) — the resulting complex forms show a minimal contrast in surface accent: root-accented gómant–
vs. suffix-accented padvánt–. Similarly, the peninitial accent of /marút/ is retained in its derivative marútvant–.
One potential analysis of these derivatives is presented in (10) below (for an alternative, see Sandell 2015:184–9):

(10) /gáv – mánt – am/ → gómantam ‘possessing cattle’ ([Link])

/pad – vánt – am/ → padvántam ‘possessing feet’ ([Link])
/marút – vánt – am/ → marútvantam ‘accompanied by the Maruts’ ([Link])

Ved. gáv-ā in (7b), as well as gómantam and marútvantam in (10), show a consistent pattern of accent res-
olution: when multiple inherently accented morphemes compete for the single surface accent in Vedic, accent
falls on the inherently accented morpheme closest to the word’s left edge (also cf. [Link]. Ved. pad-vát-e ← /pad
– vánt – é/). Combining this generalization about accentual resolution with the pattern of leftmost “default” ac-
centuation observed in (8a), Kiparsky and Halle (1977) proposed that Vedic accentuation is governed by the Basic
Accentuation Principle (BAP), which can be stated as in (11) (cf. Kiparsky 2010a):

(11) Basic Accentuation Principle (BAP):

If a word has more than one accented syllable, the leftmost of these receives word stress. If a word has no
accented syllable, the leftmost syllable receives word stress.

Kiparsky and Halle (1977) present evidence from the accentual systems of Balto-Slavic and Ancient Greek in
support of the BAP and, on the basis of their convergence, argue that it should be reconstructed for PNIE. This
hypothesis is now corroborated by evidence from Anatolian, where Yates (2016) contends that the BAP is syn-
chronically operative, accounting (e.g.) for the Hittite contrast in the mi-conjugation between primary verbs that
are accentually mobile (i.e. show accent on the root in the singular and on inflectional endings in the plural) and
those with fixed root accent. Mobile accent is the majority pattern in this category, instantiated by common verbal
roots like šeš– ‘sleep’, while a few roots — such as wek– ‘demand’ — exhibit fixed root accent. Just as in the root
nouns in (9) above, the accentual contrast between these verbs can be derived by assuming: (i) the singular verb
endings are inherently unaccented (e.g. Hitt. [Link]. /–zi/); (ii) the plural endings are inherently accented (3pl.
/–ánzi/); (iii) the roots differ underlyingly in accentedness (/wék/ vs. /šeš/); and (iv) the operation of the BAP. This
derivation is represented in (12):

(12) Primary Verbs in Hittite

fixed mobile
[Link] /wék – zi/ → wēk-zi [wéːkt͡si] ‘demands’ /šeš – zi/ → šēš-zi [séːst͡si] ‘sleeps’
[Link] /wék – ánzi/ → wek-anzi [wé(ː)kant͡si] ‘demand’ /šeš – ánzi/ → šaš-anzi [sasánt͡si] ‘sleep’

Vedic attests an identical contrast in primary verbs between mobile and fixed accentual types. Mobile accent
is observed in most Vedic root presents, including Ved. [Link] sás-ti / 3pl. sas-ánti ‘sleep(s)’, which is directly
cognate with the Hittite forms of šeš– cited in (12). This perfect equation suggests that their PIE congenitors were
derived in exactly the same way as in Hittite — in other words, that the corresponding PIE morphemes had the
same accentual properties (*/ses/ ‘sleep’; [Link]. */–ti/; 3pl. */–énti/) and underwent the same interaction with
the BAP, i.e. (13) (for the accentuation of Ved. sas-ánti*, cf. imp. sas-ántu):

(13) PIE */ses – ti/ → *sés-ti ‘sleeps’ ([Link].) > Hitt. šēš-zi [séːst͡si], Ved. sás-ti
PIE */ses – énti/ → *sə s-énti ‘they sleep’ ([Link].) > Hitt. šaš-anzi [sasánt͡si], Ved. sas-ánti*

The fixed accent type in (12) also has a parallel in Vedic, where it is similarly a minority pattern. An example is the
Vedic root takṣ– ‘fashion’ with fixed accent, as in the 3pl. tákṣ-ati (the accent of the [Link]. tāṣ-ṭi is unattested, but
would be tā ́ṣ-ṭi*). The fixed root accent can be derived by assuming that the root itself is inherently accented (i.e.
/tákṣ/), like Hitt. /wék/ ‘demand’.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 34

The existence of inherently accented (verbal) roots in Vedic and Hittite raises the question of whether they
should also be reconstructed for PIE. In this respect, it is notable that Hitt. wēk-zi and Ved. 3pl. tákṣ-ati are verbal
forms analyzed by LIV 2 as “Narten presents,” a type of PIE root present characterized by lengthened grade of the
root in singular active forms and fixed root accent (see §4.3.1 below). If the special phonological behavior of this
type is due to the fact that they are formed from “Narten roots” (Schindler 1994; Jasanoff 2012b; Villanueva Svens-
son 2012, i.a.), it may be the case that lexical accent was one property of these exceptional roots. An alternative
possibility — consistent with Kümmel’s (1998) and Melchert’s (2014b) arguments that “Narten presents” were a
derived category in PIE — is that all PIE verbal roots were inherently unaccented, and that fixed accent in “Narten
presents” was due to the presence of an additional derivational morpheme (albeit one with no segmental con-
tent), much as in thematic presents (see below), *s-aorists, and other verbal categories with fixed accent. If so, the
emergence of accented roots in the daughter languages might be attributed to the loss of Narten derivation as a
productive morphological process, at which point the fixed accent associated with this category was reanalyzed
as a lexical feature of the verbal root. Further research may shed light on these questions.
In addition to accented and unaccented morphemes, PIE also had preaccenting morphemes, which place a lex-
ical accent on the final syllable of the preceding morpheme. Strong candidates for PIE preaccenting morphemes
include the neuter event noun-forming suffix *–o/es– (cf. §2.4.1 above) and, in the verbal system, the *–e/o– suffix
that forms PIE simple thematic presents (cf. §4.3.1 below). Nouns and verbs derived with these suffixes show fixed
root accent and (generally) full-grade of the root (see further discussion of *–o/es– in §3.3 below). Under the preac-
centing analysis, the accent on the root in these items is the surface realization of a lexical accent sponsored by
the immediately following suffixes, PIE */–ˊo/es–/ and */–ˊe/o–/. This analysis of several securely reconstructible
nominal and verbal examples is given in (14a) and (14b) respectively:

(14) a. PIE */wekw – ˊes – é/ós/ → *wékw -es-e/os ‘of speech’ ([Link].) > Ved. vác-as-as, Gk. (w)ép-e-os
PIE */men – ˊes – é/ós/ → *mén-es-e/os ‘of thought’ ([Link].) > Ved. mán-as-as, Gk. mén-e-os
b. PIE */h2 eg̑ – ˊe – té(-)/ → *h2 ág̑-e-te(-) ‘you drive’ ([Link]) > Ved. áj-a-tha
PIE */bh er – ˊe – té(-)/ → *bh érete(-) ‘you bear’ ([Link]) > Ved. bhár-a-tha

In (14), the lexical pre-accent “wins” over the lexical accent of the athematic genitive ending */–é/ós/ and of the
[Link]. ending */–té(-)/ due to the BAP, which assigns surface accent to the lexical accent that it is closer to the left
edge of the word. Certain other potential analyses of these forms are not tenable. For instance, surface accent on
the root cannot emerge by default, since athematic [Link]. */–é/ós/ and 2pl. */–té(-)/ must be inherently accented:
[Link]. */–é/ós/ — like [Link]. /–éh1 / — attracts the surface accent in mobile root nouns (e.g. Ved. pad-ás ‘of the
foot’ ← /pad – ás/), and similarly, 2pl. */–té(-)/ in mobile root presents (e.g. Ved. ha-thá ‘you smash’ ← /(g)han
– thá/). Nor can surface root accent in (14) arise because the roots are themselves inherently accented, since the
action/process-noun forming suffix *–ti/tey– regularly attracts the surface accent when suffixed to these roots, i.e.
PIE *mn̥ -tí– ‘thinking; thought’, *bh r̥-tí– ‘bearing’ (> early Ved. matí–, bhr̥ tí–; see further discussion of this class in
§3.2 below).
However, just like the lexical accent of accented morphemes, the lexical accent sponsored by a pre-accenting
morpheme does not always receive the surface accent. Vedic shows a clear synchronic contrast between examples
like (14), where the lexical pre-accent “wins,” and those like (15b), where the principles of accentual resolution
prefer a different accented morpheme. The same contrast is observed with preaccenting morphemes in (e.g.)
Cupeño and Japanese; examples that parallel the Vedic data are laid out in (15a) and (15b) respectively (Japanese
data from Kawahara 2015, Cupeño from Hill 2005):

(15) a. Ved. /śrav – ˊas – ás/ → śráv-as-as ‘of fame’

Jap. /yosida – ˊsi/ → yosidá-si ‘Mr. Yoshida’
Cu. /pə – tama – ˊŋa/ → pə-tamá-ŋa ‘in his mouth’

b. Ved. /prá – [ śrav – ˊas ] – ás/ → prá-śrav-as-as ‘of him whose fame is advancing’
Jap. /nisímura – ˊsi/ → nisímura-si ‘Mr. Nishimura’
Cu. /pə – sáʔi – ˊŋa/ → pə-sáʔi-ŋa ‘in his belly’
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 35

As is evident from (15b) the accentuation of bahuvrīhi compounds (cf. §2.6 above) like Ved. práśravasas is con-
sistent with the BAP. The inherent accent of the first member (1M) — in this case, the preverb Ved. /prá/ — is
assigned surface accent because its lexical accent is closer to the word’s left edge than that of 2M /śrávas–/, whose
initial accent is due to the preaccenting neuter event noun suffix /–ˊas–/. First member accent is the inherited rule
in Greek’s exocentric compounds as well. In its cognate class of *s-stem adjectives, Greek has a number of relic for-
mations that reflect first member accent, thus making it plausible to reconstruct PIE compounds like *pró-k̑lewes–
(> Ved. prá-śravas–) with 1M surface accent due to the BAP (cf. with details and references Lundquist 2016).
Productively formed Greek s-stem compounds have suffixal accent (m/[Link]. –ḗs), which reflects a historical
change from denominal to deverbal derivation in this class of adjectives (cf. Meissner 2006:161–215).
More generally, an analysis along these lines can be extended to other types of bahuvrīhi compounds which,
even more clearly than in other categories, require a principle of accent resolution to determine which of the
accents that their members bear as free-standing words will receive the single surface accent of the compound. In
Vedic — and in all likelihood, in PIE — the surface accent of these compounds is that of their 1M (cf. Wackernagel
1905:§113–5), provided that the 1M contains an inherently accented morpheme. This pattern is again predicted by
the BAP; simplified derivations for Vedic bahuvrīhi compounds of several structural types are given in (16) below
(stem-stem compounding is assumed here, but see Kiparsky (2010a:170–6, fthcm.) for more detailed analysis with
extension to other compound types):

(16) a. Noun + Noun:

/bāhú + ójas/ → bāhú-ojas– ‘having strength in one’s arms’ (arm + strength)
/kaví + krátu/ → kaví-kratu– ‘having the will of a poet’ (poet + will)
/sóma + kā́ma/ → sóma-kāma– ‘desirous of soma’ (soma + desire)

b. Adjective + Noun:
/ugrá + bāhú/ → ugrá-bāhu– ‘mighty-armed’ (mighty + arm)
/dabhrá + cétas/ → dabhrá-cetas– ‘small-witted’ (small + perception)
/sahásra + dákṣina/ → sahásra-dakṣina– ‘having a priestly gift ([Link] + [Link])
of a thousand (cows)’
c. Preverb + Noun:
/ádhi + rukmá/ → ádhi-rukma– ‘having bright ornaments upon oneself’ (upon + ornament)
/abhí + krátu/ → abhí-kratu– ‘whose will is set against’ (against + will)

The Vedic evidence in (16) is again corroborated by “recessively” accented Greek bahuvrīhi compounds, e.g. klutó-
toxos ‘famed for the bow’ (on Greek’s recessive accent, see Gunkel 2014). The equation of Greek and Vedic accen-
tuation suggests that this analysis of compound accent can be extended at least to PNIE, and that bahuvrīhis with
1M accent like *h2 ugró-bh eh2 g̑ h u— (> Ved. ugrá-bāhu–) can be reconstructed for this stage. The more complicated
case of bahuvrīhis with 2M accent is discussed further in §3.2 below.

§3.2 PIE lexical accent: Expanding the analysis It was shown in §3.1 that morphemes in PIE were lexically
specified for one of three accentual features: accented, unaccented, or pre-accenting. In addition, PIE accentu-
ation was governed by the BAP, which assigns the surface accent to the leftmost of several inherently accented
morphemes, or in their absence, assigns a default initial accent. These three accentual features as well as the BAP
have strong typological parallels in Japanese and other languages with lexical accent; however, it is all but certain
that the PIE accentual system was of a more complex type than (e.g.) Cupeño, where the interaction between
the same three accentual features and a BAP-like phonological principle is sufficient to account for (effectively)
all of the accentual contrasts in the language (cf. Yates 2017). To account for the accentual patterns attested in
the oldest IE daughter languages, it appears to be necessary to enrich the PIE system with additional properties,
although exactly how it must be enriched is very much open for debate at present. In the remainder of this section,
we lay out some of the data that complicates the analysis, and discuss a few recent proposals that may offer a way
forward.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 36

One accentual phenomenon that does not easily submit to the tools developed in §3.1 is the “intermediate”
behavior of several athematic suffixes, which appear to attract the surface accent in simplex forms, but yield the
accent in further derivation. Two suffixes with this property — both traditionally analyzed as “proterokinetic”
under paradigmatic approaches to IE accent and ablaut (cf. §3.3) — are the deverbal action/process noun-forming
suffix *–ti/tey– (cf. §2.4.1) and the qualitative adjective suffix *–u/ew– (§2.5). For instance, in (earliest) Vedic *ti-
stem nouns like jū-tí– ‘speed’ (to the root jū– ‘hasten’) or vr̥ ṣṭí– ‘rain’ (to vr̥ ṣ– ‘rain’) regularly show attraction of the
surface accent to the derivational suffix (cf. Lundquist 2015), thus non-default accent in their strong case-forms
(e.g. [Link]. jū-tí-m, vr̥ ṣ-ṭí-m); the suffix also retains the surface accent in weak case forms (e.g. [Link]. jū-táy-e;
[Link]. vr̥ ṣ-ṭí-bhis) in preference to the inherently accented inflectional endings to its right ([Link]. /–é/; [Link].
/–bhís/; cf. paḍ-bhís ‘with the feet’ to /pad/ in (9–10) above). At first glance, this accentual pattern recommends
analyzing the suffix as inherently accented (i.e. “*/–tí/téy–/”), in parallel to the thematic adjective suffixes (/–
nó–/, /–ró–/); fixed suffixal accent would then be correctly predicted, since the suffix would be the only accented
morpheme in strong case forms and preferred by the BAP in weak case forms (i.e. “leftmost wins”).
The problem with this analysis, however, is that it makes incorrect predictions about the accentuation of
derivationally related forms. Issues arise in (e.g.) adjectives derived from Vedic ti-stems by addition of the suffix
–mant– (/–mánt–/), which consistently attracts the accent away from these stems, thus (e.g.) jūtimánt– ‘swift’,
vr̥ ṣṭimánt– ‘rainy’. This pattern would be unexpected if the noun-forming suffix Ved. –ti/tay– were inherently ac-
cented; rather, like Ved. gó-mant– ‘possessing cattle’ in (10) above and similarly (e.g.) Ved. mánas-vant– ‘thought-
ful’ (to the neuter as-stem in (14) mánas–), a stem containing an inherently accented morpheme should receive
the surface accent in preference to an accented suffix to its right as a direct consequence of the BAP.
This issue is not unique to *ti-stems nor is it specific to the suffix(es) *–ment–/*–went–. The same kind of
accentual behavior is also observed in *u-stem qualitative adjectives, which show fixed accent on the ablauting
suffix *–u/ew– throughout their inflectional paradigm in both Vedic and Greek, e.g. Ved. svād-ú–, svād-áv– = Gk.
hēd-ú–, hēd-é(w)– ‘sweet’ (< PIE *sweh2 d-ú–, *sweh2 d-éw–); Ved. pr̥ thú–, pr̥ th-áv– = Gk. plat-ú–, plat-é(w)– ‘broad’
(< PIE *pl ̥th2 -ú–, *pl ̥th2 -éw–); Ved. āśú–, āś-áv– = Gk. ōkú–, ōk-éw– ‘swift’ (< PIE *h1 ōk̑-ú-/*h1 ōk̑-éw–. Once again,
the derivational suffix is superficially amenable to treatment as an inherently accented morpheme (“*/–ú/éw–/”),
but such an analysis is problematized by the accentual behavior of the suffix in further derivation — for instance,
in combination with the “devī”́ feminine suffix P(N)IE */–íh2 /yéh2 –/ (> Ved. /–ī/́ yā́–/).
The feminine suffix does not generally attract the surface accent when there is an inherently accented mor-
pheme to its left, as shown (e.g.) by its interaction with the accented PNIE perfect participle suffix *–wos/us–
(*/–wós/ús–/), whose Greek and Vedic masculine reflexes bear suffixal accent, e.g. [Link].m. Ved. vid-vā ́ṁ-s,
[Link]. vid-úṣ-as; Gk. eid-(w)ṓ-s, eid-ót-os ‘knowing’ (< PIE *w(e)id-wṓs, *w(e)id-ús-). Significantly, the correspond-
ing feminine forms exhibit persistent accent on the perfect participle suffix — e.g. [Link].f. Ved. vid-úṣ-ī, Gk.
eid-uîa (< PGk. *–ús-ya) — as expected under the BAP: PIE */–ús-íh2 / → *–ús-ih2 . However, when the same suffix
is used in Vedic to form feminine *u-stem adjectives, it unexpectedly attracts the surface accent, thus [Link].f.
Ved. svād-v-ī ́ ‘sweet’, pr̥ th-v-ī ́ ‘broad’. This pattern is corroborated by archaisms in Greek — in particular, feminine
plural forms in –eiaí, –aiaí with synchronically irregular oxytone accent; this class includes the Greek toponym
Plataiaí (< PGk. *pl ̥th2 -(e)w-yéh2 –), whose accent matches its cognate Ved. pr̥ th-v-ī ́ ‘broad’ and therefore likely re-
sisted the analogical leveling of suffixal accent that produced the synchronic feminine adjective Gk. plateîa ‘broad’
with the regular accent of its morphological class (cf. de Lamberterie 1990:644–5, 2002; contra Sihler 1995:349–50
et al.).
The exceptional “intermediate” accentual behavior of *u-stem adjectives in combination with the feminine
suffix recurs in other derivationally related forms. First, there are cases in which these *u-stems are further suffixed
by adjectival *–ment– (*/–mént–/) and — as in the *ti-stems — this suffix attracts the surface accent, e.g. Ved. āśu-
mánt– ‘speedy’. Moreover, Vedic bahuvrīhi compounds with 1M *u-stem adjectives generally have surface accent
on the accented syllable of their 2M, e.g. svādu-kṣádman– ‘(lit.) having a sweet carving knife (kṣádman–); serving
sweet food’; āśu-héṣas– ‘having swift missiles (héṣas–)’; pr̥ thu-pā ́jas– ‘whose form (pā ́jas–) is broad’. While such
compounds show some accentual variation — e.g. both pr̥ thu-budhná– and unexpected pr̥ thú-budhna– ‘having a
broad foundation (budhná–)’ are attested in the Rigveda — the dominant pattern in this class is 2M accent, which
contrasts with the 1M accent pattern observed in the structurally comparable bahuvrīhi compounds in (16b) above.
In each case, the *u-stem adjective is predicted by the BAP to receive the surface accent if it were an accented
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 37

morpheme, but these predictions are not borne out; rather, the systematic failure of the *–u/ew– suffix — and
similarly, *–ti/tey– — to attract surface accent in secondary derivatives suggests that these suffixes are in fact
underlyingly unaccented (i.e. PIE */–u/ew–/, */–ti/tey–/), and that their secondary derivatives can be analyzed as
in (17):

(17) PIE */g̑euhx – t(e)y – mént – s/ → *g̑ uhx -ti-mént-s > Ved. jū-ti-mā ́n ‘swift’ ([Link].)
PNIE */pleth2 – (e)w – íh2 – ∅/ → *pl ̥th2 -u-íh2 > Ved. pr̥ th-v-ī ́ ‘broad’ ([Link])
PIE */pleth2 – (e)w – peh2 g̑ – ˊes – s/ → *pl ̥th2 -u-páh2 g̑-ēs > Ved. pr̥ thu-pā ́jās ‘having a broad
foundation’ ([Link])
An important implication of this analysis is that the fixed suffixal accent observed in primary *ti-stem nouns
and *u-stem adjectives must arise as the result of some other grammatical process that does not apply in further
derivation. The exact nature of this process is controversial and a topic of ongoing research. According to Kiparsky
(2010a:144), it is the “Oxytone Rule,” which places a lexical accent on the rightmost syllable of a polysyllabic word’s
inflectional stem. Because it applies only to a fully formed inflectional stem, the Oxytone Rule assigns a lexical
accent to *–ti/tey– and *–u/ew– when immediately followed by inflectional endings, but does not target these
suffixes when there is intervening morphological material, since they do not stand at the right edge of the stem. The
suffix accented via the Oxytone Rule then attracts surface accent (in preference to accented weak case endings)
due to the BAP.
An alternative hypothesis is advanced by Sandell (2015:176–214), who argues that PIE affixes may be assigned
lexical accent by virtue of being a word’s morphological head — in effect, the part of the word that determines cer-
tain of its fundamental morphosyntactic properties (e.g., whether it is a noun or adjective; cf. Zwicky 1985; Dresher
and van der Hulst 1998). Thus a derivational suffix like *–ti/tey–, which selects a verbal root (e.g. *men– ‘think’)
and forms an abstract noun ([Link]. *mn̥ -tí-s ‘thought’), is the word’s head and would consequently receive a
lexical accent; however, in the (hypothetical) derived adjective *mn̥ ti-mént–, the head of the word is the adjectival
suffix *–ment–, so no lexical accent would be assigned to the *–ti/tey– suffix. This analysis would align PIE with a
range of other languages in which morphological structure plays a direct role in determining word accent; included
among these languages are two of PIE’s living descendants, Modern Greek and Russian (Revithiadou 1999), which
are arguably conservative in this respect. However, adjudicating between this account and Kiparsky’s (2010a) Oxy-
tone Rule requires further systemic analysis of Vedic word accent, and still more research in the other daughter
languages is needed to establish the accentual properties of the “intermediate” suffixes at the PIE level.
Another problem encountered by the basic analysis laid out in §3.1 is the accentual behavior of certain suffixes
which appear to “override” the accentual features of the stem to which they attach. The existence of such mor-
phemes with this property — termed dominance by Kiparsky and Halle (1977) — was established in Balto-Slavic
linguistics already in the 1970s (see, in particular, Garde 1976, and for a conceptual overview with reference to
Ancient Greek, Petit 2016:11–4). Such morphemes are also found in non-IE languages with lexical accent systems
like Tokyo Japanese (see Kawahara 2015 with references). Dominant morphemes flout the language’s phonological
accent resolution pattern (in PIE, the BAP), imposing their accentual properties on the stem to which they attach;
in the IE languages, this effect can be observed mostly clearly when a dominant accented morpheme is suffixed to
a stem that itself contains an inherently accented morpheme.
An example of a dominant morpheme in Vedic is the adjective-deriving suffix –in– (/–ín–/; cf. Kiparsky
2010a:170). When it combines with nouns that have fixed surface accent (due to their underlying accented stems),
the resulting derived forms systematically exhibit fixed surface accent on the –in-suffix; this pattern is shown in
(18) below, where the same accented (thematic) noun stems that retain their accent in combination with non-
dominant accented suffixes like Ved. –vant– (/–vánt–/) or as the 1M in bahuvrīhi compounds always cede the
surface to the dominant suffix –in–:

(18) Ved. áśva– ‘horse’ ⇒ aśvín– ‘horseman; Aśvin’ cf. áśva-vant– ‘possessing horses’
rátha– ‘chariot’ ⇒ rathín– ‘charioteer’ rátha-vant– ‘possessing chariots’
dyumná– ‘brilliance’ ⇒ dyumn-ín– ‘brilliant’ dyumná-vant– ‘possessing brilliance’
putrá– ‘son’ ⇒ putrín– ‘having a son’ putrá-kāma– ‘desirous of sons’
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 38

Dominance effects can also be found in the verbal system. In Vedic, verbal adjectives may be formed by suf-
fixing –ta– /–tá–/ (< PIE *–to–; cf. §2.5 above) directly to the verbal root. Whether the root is unaccented (the
majority type, e.g. /(g)han–/ ‘smash; kill’) or accented (/tákṣ–/ ‘fashion’), the suffix –ta– consistently attracts sur-
face accent (ha-tá– ‘smashed; killed’, taṣ-ṭá– ‘fashioned’). Dominant accented /–tá–/ thereby contrasts with the
non-dominant accented present participle suffix /–(a)nt–/, which receives surface accent when added to unac-
cented roots (e.g. ghn-ánt– ‘smashing’) but not to accented roots (tákṣant– ‘fashioning’).
The nature of accentual dominance in the PIE lexical accent system is a topic of ongoing research. Kiparsky
(2010a) treats dominance as an arbitrary lexical property of morphemes (i.e. [+/– dominant]), but observes that
there is a strong tendency for (prototypical) derivational suffixes to be dominant. In Greek, in fact, it appears that
all derivational suffixes are dominant (Steriade 1988, and cf. Probert 2006b:146, Gunkel 2014); several examples of
Greek’s inherently accented derivational suffixes are given (19), where their accentual dominance can be observed:
(19) /–ikó-/ hellád-os ‘Greece’ ([Link].) ⇒ hellad-ik-ós ‘Greek’
adelph -ós ‘brother’ ⇒ adelph -ik-ós ‘brotherly’
/–ísko–/ aspíd-os ‘shield’ ([Link].) ⇒ aspid-ísk-os ‘small shield’
kratḗr ‘mixing bowl’ ⇒ kratēr-ísk-os ‘small bowl’
/–éu–/ hípp-os ‘horse’ ⇒ hipp-eú-s ‘horseman’
kh alk-ós ‘copper’ ⇒ kh alk-eú-s ‘coppersmith’
Given that Vedic appears to have both dominant and non-dominant derivational suffixes, the Greek situation
likely reflects an innovation with respect to PIE. Nevertheless, the strong correlation in both languages between an
affix’s morphosyntactic properties and its (non-)dominant status suggests that accentual dominance effects are in
some way a consequence of morphological structure — i.e. the accent of the (last) derivational suffix is privileged
because it is the morphological head (as in the *ti-stems discussed above; see Sandell 2015:182–92 for a proposal
and formal implementation to this end). Yet how accentual dominance should be formally implemented in PIE
(and cross-linguistically) is far from a settled question; see generally Revithiadou (1999) and Alderete (2001a), and
for specific application to (pre-)PIE word accent, Frazier (2006), Keydana (2013b), and Kim (2002, 2013b).
The cross-linguistically well-established analytic tools introduced in §3.1 — i.e. the distinction between in-
herently accented, unaccented, and preaccenting morphemes together with the BAP — make empirically testable
predictions about PIE accentuation that correctly account for the distribution of surface accent in many securely
reconstructible PIE words and morphological categories. However, it is also clear that there are morphological
conditions under which these predictions are systematically violated — i.e. when a word contains an accentu-
ally “intermediate” or dominant morpheme. One possibility would be to take the behavior of these morphemes
as evidence that the analysis laid out in §3.1 — in particular, the BAP — is incorrect; yet in view of the the far-
reaching accentual generalizations that are correctly derived by the BAP, we have proposed instead that the theory
should be refined. Specifically, we have suggested that the PIE accentual system had additional morphophonolog-
ical properties relevant to the accentuation of words containing accentually “intermediate” and dominant mor-
phemes. We have also discussed several promising hypotheses about what these properties might be and how
they should be integrated into a general analysis of PIE word accent.
Under this view, the PIE lexical accent system is of a complex type similar to that of Thompson Salish, Tokyo
Japanese and Chamorro (cf. §3.1 above): surface accent is in some cases determined by a purely phonological
computation over the inherent accentual properties of morphemes (i.e. the BAP), but there is also an additional
“layer” associated primarily with derivational suffixes in which a word’s morphological structure may influence
the computation of the surface accent. Further research in this vein on the accentual systems of the ancient IE
daughter languages — in particular, Vedic, Greek, Balto-Slavic, and the Anatolian languages — will continue to
shed light on the synchronic principles governing the distribution of surface accent in PIE, on the reconstructible
accentual properties of individual morphemes, and in turn, on what forms constitute real archaisms already at
this stage of the proto-language — i.e. reconstructible words whose accent cannot be generated by productive
morphophonological processes, and so must have been learned on an item-by-item basis. A still broader issue is
the extent to which accent and ablaut are related at the PIE stage (and at the earlier pre-PIE stage), an issue we
take up immediately below (§3.3).
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 39

§3.3 Reconstructing PIE ablaut The relationship between accent and ablaut in PIE has been a major topic of
research since the beginning of IE studies. Accent and ablaut correspond only partially in the daughter languages
and so too at the stage of PIE that is accessible by the comparative method. In PIE, every kind of vowel may
surface with or without surface accent: *bh ér-e-ti ‘carries’ and *mn̥ -téy-es ‘thought’ ([Link].) surface with two
full-grades each (the [Link]. *–es- never has a reduced allomorph); *septḿ̥ ‘7’ (> Ved. saptá, Gk. heptá) bears an
accented zero-grade and an unaccented e-grade; *bh ór-o-s ‘burden’, *pód-s ‘foot’ and *k̑éy-(t)or ‘lies’ have accented
and unaccented *o-grades. These examples are easily multiplied. However, there are also strong indices to suggest
a relatively tight connection between surface accent and full-grade, as seen in (e.g.) verbal paradigms like *h1 éy-
ti ‘goes’, 3pl. *h1 y-énti or *h1 és-ti ‘is’, 3pl. *h1 s-énti. Accordingly, it is widely thought that these quantitive ablaut
alternations (i.e. *e : *∅) were once purely phonologically conditioned — in its strongest formulation, that an
*e vowel would surface only if it bore the surface accent, and all other morphemes would thus appear in their
zero-grade forms (see Szemerényi 1996:111–112, who traces this view back to the 1860s; cf. Weiss (2011:47) for a
recent, skeptical formulation). Viewed in generative terms, these alternations would reflect an accent-conditioned
syncope process deleting all unaccented */e/ vowels at the relevant stage of the proto-language. Similarly, a link
has long been suspected between surface accent and *o-grade, i.e. qualitative ablaut (e.g. Hirt 1900: esp. 156, but
see the doubts voiced earlier by de Saussure 1879:134, 235, et passim). For this view, however, one finds even less
consensus, since it has not yet been demonstrated just what that link would be (see Penney 1978 for an extensive
treatment and the concise overview by Weiss (2011:47); Kümmel (2012:307–20) gives one recent attempt to explain
the origin of *o-grades). Quantitative ablaut especially has often been treated as a shortcut to accent — i.e. if a
word contains an *e-grade morpheme, it should once have been accented, and a zero-grade morpheme should
have been unaccented — but at the PIE level such a shortcut is clearly not tenable.
A major program of research, developed principally in the 1960s and 1970s (but with older roots, esp. Pedersen
1926 and Kuiper 1942), has focused on reconstructing the formal patterns of athematic nominal formations at this
pre-PIE stage when the relationship between accent and ablaut would have been more transparent. For instance,
in a foundational paper Schindler (1975b:261) proposed that neuter *–es-stem nouns of the type PIE nom./acc.
*wékw -os, [Link]. *wékw -es-os (> Ved. vácas, vácasas etc.; cf. §3.1 above), looked substantially different at a pre-
PIE (“vorindogermanisch”) stage. He argued that, although no attested language exhibits synchronic accent shifts
or ablaut alternations of the root in this nominal class, it is nevertheless possible to reconstruct pre-PIE accentual
mobility between root and derivational suffix. In support of this hypothesis, Schindler cites lexicalized compounds
with 1M reflecting *mén-s- ‘thought’ (e.g. OAv. mazdā–) where the apparent zero-grade of the suffix would reflect
the predicted [Link]./[Link]. form (**men-s + dh eh1 –; cf. PIE *mén-os > Ved. mán-as, Gk. mén-os). At this pre-PIE
stage, all unaccented morphemes would surface in their zero-grade forms, since accent and full-grade would be
directly dependent on one another (“…die Ablautstufen im Wort akzentabhängig waren”, [Link]. 261). Provided that
this assumption is correct for pre-PIE, the PIE paradigm *wékw -os, *wékw -es-os would continue pre-PIE **wékw -s,
**ukw -és-s, whose accent was assigned morphologically and whose ablaut resulted predictably from the pre-PIE
syncope rule.
Under this approach, the hypothesized formal patterns are reified as a set of “paradigmatic” classes; all PIE ath-
ematic nominals of the structure R(oot) + S(uffix) + (E)nding would belong (historically) to one of these classes.
Thus pre-PIE **wékw -s, **ukw -és-s would instantiate the “proterokinetic” class, structurally R(é)-S(ø)-E(ø) in the
strong cases (e.g. [Link]./[Link]. **wékw -s) and R(ø)-S(é)-E(ø) in the weak ([Link]. **ukw -és-s). In the most widely
accepted model, developed in particular by Schindler (1972a, 1975b,a) and the “Erlangen School” (e.g. also Rix
1992:122–4), four or five “kinetic” (/“dynamic”) and “static” classes are posited. The “Leiden School” reduces the
model to three such classes (see Beekes 1985, Beekes and de Vaan 2011:190–1 et passim, Kloekhorst 2013), while
other scholars have posited additional accent and ablaut paradigms — for instance, Tichy (2004:75–81) and Neri
(2003:37–9) allow a “mesokinetic” paradigmatic class. This body of research has clarified especially which forms
could be relics already in PIE (such as the isolated *men-s- mentioned above) and offers a possible starting point
for analyzing the subsequent development of many PIE athematic nominal formations. Overviews of the paradig-
matic classes can be found in all recent IE handbooks: see Watkins (1998:61–2, skeptical), Clackson (2007:79–89),
Fortson (2010:119–23), Weiss (2011:257–62); Meier-Brügger (2010:336–53) offers the fullest history of research.
Despite its widespread acceptance, a rapidly growing body of scholarship has expressed dissatisfaction with
the conceptual and empirical limits of this theory (cf. in general Kiparsky 2010a, fthcm.; Keydana 2013b; Kümmel
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 40

2014 with reference to Indo-Iranian; and Yates 2016 on Anatolian); we outline some of these criticisms here. One
issue concerns the extent of the changes that separate reconstructible PIE forms from the pre-PIE paradigmatic
classes. Early work within the paradigmatic framework recognized that this approach, which relies extensively on
internal reconstruction, yields paradigms whose patterns of accentual mobility and ablaut grades display numer-
ous mismatches with the patterns observed in the daughter languages, some of which are directly reconstructible
for PIE by application of the comparative method (cf. Pedersen 1933:21). To obtain PIE morphophonology, further
diachronic assumptions are therefore required: the pre-PIE paradigmatic classes would be transformed by a se-
ries of analogical processes whose combined operations eliminated intraparadigmatic allomorphy by analogical
leveling of accent, ablaut, or both (sometimes referred to with the descriptive label “columnarization”). The mor-
phological upheavals here envisaged must have occurred in the internal history of the proto-language, i.e. prior to
PIE as accessible by the comparative method, since no daughter language organizes its morphology into produc-
tive paradigmatic classes (cf. the methodological discussion by Hale 2010, as well as Stüber 2002:esp. 211–6, both
with reference to *es-stems). Because the hypothesized changes are situated deep in prehistory, their plausibility
is difficult to evaluate, either within individual classes or collectively, at the systemic level.
Beyond these uncertainties, a problematic consequence of the focus on the internally reconstructed pre-proto-
language is that much of the morphophonology of PIE and its daughter languages is left unexplained, since the
theory was not designed to handle material at this chronological level. For instance, numerous bedrock formations
of PIE have no clear position in the paradigmatic classes. The classes refer only to athematic nominal formations
of the structure R(oot) + S(uffix) + E(nding), thus excluding thematic nouns and adjectives, athematic nominal
formations with multiple derivational suffixes (i.e. of the structure R + S + S (+ S…) + E), and even root nouns. The
fact that the paradigmatic approach does not address these PIE formations is not a criticism per se, since this is not
strictly the goal of the theory; however, it does mean that this theory, with its pre-PIE focus, sheds little light on the
distribution of the accent (discussed in §§3.1–3.2 above) or its synchronic relationship to ablaut at the “shallow”
chronological stage of PIE, which we are reconstructing here, and which was inherited directly into the daughter
languages.
A further criticism relates to the evidential basis for the paradigmatic reconstructions, which in a number
of cases has been called into question. For instance, in a widely followed thesis, Kuiper (1942:221) proposed that
the different accentuation of Vedic matí– ‘thought’ beside máti– ‘id.’, coupled with indirect evidence elsewhere,
showed a trace of erstwhile intraparadigmatic alternations in an accent and ablaut paradigm, i.e. **mén-ti–, **mn̥ -
téy–, so would be another proterokinetic paradigm (Rix 1992:146, Schaffner 2001:436–40). In this case, the zero-
grade ablaut of the root in the weak cases would have been leveled throughout the paradigm in Vedic, but with
a bifurcating accentual leveling: leveled accent of the strong cases would be preserved in some Vedic traditions
(i.e. *mń̥ -ti– > máti–), while the leveled accent of the weak cases would be preserved in others (i.e. leveled *mn̥ -
tí– > matí–). It has proven difficult to explain why the directions of leveling have taken the apparently arbitrary
courses they have; in this case, however, the quest to do so is in fact a red herring, since the two accentual patterns
stand in a clear chronological relationship: accented –tí-stems occur in the oldest textual layers, unaccented –ti–
in the younger. Thus early Ved. matí– and later Ved. máti– do not provide evidence for independently leveled bits
of a prehistoric paradigm, but instead reflect a Vedic-internal diachronic accentual change that can be otherwise
explained (Lundquist 2015; see further below). More generally, Kümmel (2014) has shown that the accent and
ablaut of “proterokinetic” nominals in Indo-Iranian is better explained without reference to paradigmatic class,
thereby undercutting an important source of evidence for the paradigmatic approach.
In assessing accentual change, it has become common practice to treat two attested accentual patterns as-
sociated with one suffix as reflecting independent analogical levelings of an alternating paradigm (as in the case
of Ved. matí– vs. máti–). However, it has now become clear that (pre-)PIE intraparadigmatic accentual mobility
is not a necessary condition for this situation to arise. This point has been conclusively demonstrated by Probert
(2006b,a), who investigates the diachronic development in Greek of two morphological categories that are by gen-
eral agreement reconstructed with fixed word-final surface accent, thematic adjectives (formed with the suffixes
*–ro–, *–no–, *–to–, and *–lo–; cf. §2.5) and feminine event/result nouns (formed with *–eh2 ; cf. §2.4.1). While
most attested reflexes of these categories show the historically expected pattern, some instead show “recessive” ac-
centuation, thereby arguably exhibiting an accentual change. Probert attributes this change to a process termed
“demorphologization” whereby morphologically complex words lose their compositionality due to semantic or
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 41

formal opacity and come to be treated as monomorphemic (“demorphologized”). As a further consequence, words
affected by this morphological change strongly tend to adopt the language’s default accentual pattern (whether
or not this occurs depends on word frequency and other factors; cf. Sandell 2015:192–214) — in Greek, recessive
accentuation, which ultimately reflects the BAP in modified form (i.e. leftmost within the accentable domain).
The differing surface accents of (e.g.) Gk. ekh th rós ‘enemy’ and Gk. gûros ‘circle’ thus do not reflect a fundamental
difference in the historical formation of each item; rather, the connection between reconstructible *gū-rós ‘circle’
(substantivized from the adj. gū-rós ‘round’) and other *–ro– adjectives became opaque and, as a result, the word
was eventually subject to default accentuation, whence *gūr-ós > gū ̂ ros (on this example see Probert 2006b:232–3).
Cases of this kind show definitively that two accentual patterns can emerge diachronically without an ear-
lier synchronic intraparadigmatic accentual alternation. Furthermore, such cases provide evidence for a type of
prosodically optimizing, non-proportional analogical change that can also be observed within the historical record
of English (cf. Kiparsky 2015:82–3). Within the ancient IE languages, the Greek evidence for this type of change
finds further support in Vedic, where a similar analysis can account for the development of Vedic *–ti-stems (like
Ved. matí– > máti–), as well in the Anatolian languages, where it can explain a variety of forms (such as PIE nasal-
infix presents; cf. §4.3.1) that unexpectedly exhibit initial surface accent (i.e. leftmost, in accordance with the PIE
default pattern; see Yates 2015). A broader implication of this finding is that the existence of more than one ac-
centual pattern associated with a single suffix is not a sufficient condition to reconstruct an alternating accentual
paradigm at any historical stage. To the extent that individual paradigmatic reconstructions are founded on this
premise (as in “proterokinetic” *–ti-stems), their (pre-)PIE status must be viewed as uncertain.
Finally, taking a still wider perspective, Kiparsky (2010a, fthcm.) in particular has also challenged the typolog-
ical naturalness of the paradigmatic classes. Although it is true that the typological pool of known morphophono-
logical properties is not comprehensive (see however van der Hulst (1999) on the word prosodic systems of the
languages of Europe, as well as the StressTyp2 database site ([Link] no clear parallel for the pre-
PIE system has yet been brought forward. Part of the uncertainty here is terminological: before comparing the
pre-PIE system to that of another language family, the linguistic claim needs to be formulated more precisely —
in what sense do paradigms “exist” in pre-PIE morphology? Are they prosodic templates associated with certain
derivational categories, and if so, which ones? Or are they intended to be the surface result of a pre-PIE lexical ac-
cent system, perhaps not dissimilar from the one we have reconstructed above? Given the real gaps in knowledge
currently facing researchers who reconstruct PIE morphophonology (as outlined above) — in particular, the fact
that it is not yet fully clear what determines the surface accent of derivationally complex forms — the amount
that can be said confidently about pre-PIE accent and its relation to ablaut is limited. Reconciling the results of
research on pre-PIE paradigms with the morphophonology of PIE and its daughter languages will likely remain a
major project for years to come.

§4 PIE verbal morphology

This section provides an overview of the reconstructed morphology of the PIE finite verb and associated non-finite
verbal categories such as participles and infinitives. The structure and early history of the PIE verb continues to
be one of the most hotly contested areas in IE studies today. While some consensus concerning the reconstruc-
tion of the PNIE verb was reached in the early 20th century, the advent of Anatolian and Tocharian called into
question many of the generally accepted features of this traditional reconstruction (see Jasanoff, this volume).
Consequently, much of our discussion focuses, first, on the reconstructible features of the PNIE verb, then we pro-
ceed to address the more controversial PIE verb, as well as the issues that problematize its reconstruction. While
we attempt to flag serious points of contention and offer critical discussion of the major competing views, non-
specialists in particular need to be aware that there is little unanimity in the field on these topics and that, due
to limitations of space, not all views can be considered here. Further discussion can be found in recent general
overviews of the IE verb, which include Clackson (2007:90–113), Fortson (2010:88–112), Weiss (2011:377–98) and
Meier-Brügger (2010:295–321). The standard reference work in the field is Rix and Kümmel (2001) (= LIV 2 ), a com-
prehensive collection of reconstructed PIE verbal roots and their verbal formations in the individual languages.
Jasanoff (2003a) re-examines the foundations of the IE verb, especially in light of the Anatolian (and to an extent
Tocharian) evidence (see too Jasanoff fthcm. b). The collection of papers in Melchert 2012b is representative of
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 42

recent research on the IE verb.

§4.1 Structure of the PIE verb As in the nominal domain, PIE verbal morphology was highly affixal. This prop-
erty is observed in PIE verb inflection where five grammatical categories were distinguished: person, number, voice,
tense, and mood (we treat aspect (below) as a derivational category). Fusional inflectional suffixes encoded gram-
matical agreement with the subject (nominative-accusative syntactic alignment; see Keydana, this volume) for
person (1st, 2nd, 3rd) and number (singular, dual, plural), as well as voice (or “diathesis”), either active or middle;
for example, *–m is an exponent of the features [1st person, singular, active], while *–o expresses [3rd person, sin-
gular, middle]. Separate segmentable suffixes are reconstructible as markers of tense (non-past; past is unmarked)
and mood (subjunctive; optative; imperative; indicative is unmarked). These inflectional categories are discussed
individually in §4.2 below.
Verbal inflectional suffixes were added to the verbal stem, which was specified with certain grammatical fea-
tures. In PNIE, verbal roots canonically formed three morphologically distinct verbal stems, traditionally and here
referred to as “present,” “aorist,” and “perfect” (see further §4.3 below); this tripartite distinction is maintained only
in Indo-Iranian and Greek. It is widely thought that the three stems expressed primarily differences of grammati-
cal aspect. A speaker could his or her view of the eventuality of the verb as internally complex, which was the work
of the present (or “imperfective”) stem; as a bounded, complete whole, using the aorist (or “perfective”) stem; or
as a resulting state, using the perfect stem. The three grammatical aspects interact with lexical aspect. By “lexical
aspect” (German Aktionsart) we mean the inherent semantics of a verb’s event structure, such as durativity or telic-
ity, which are inherent as opposed to chosen by a speaker to express a viewpoint. In the case of PNIE, it is generally
assumed that there was close agreement between grammatical and lexical aspect in the formation of tense-aspect
stems: verbal roots with telic lexical aspect had an underived aorist stem (i.e. root aorist), whereas verbs with atelic
lexical aspect had an underived present stem (i.e. root present). However, the agreement between lexical aspect
and stem formation is in practice not nearly so neat; rather, there are numerous mismatches in both directions,
relatively clear cases in which apparently telic roots form underived present stems, and apparently atelic roots
form underived aorist stems. We will return to some of the specific mismatches below (§4.3). Another real issue
with the PIE verbal system stems from the well-known difficulties associated with analyzing the “perfect” as an
aspectual category cross-linguistically (cf. Comrie 1976:52), to which may be added the challenge of establishing
the prototypical meaning of the PNIE perfect (see further §4.3.3 below). The question of grammatical aspect and
stem formation has been and continues to be a major locus of research in Indo-European linguistics.
The deeper prehistory of the PNIE verbal system is one of the most controversial topics in IE linguistics today.
In particular, two important structural features of the verbal system reconstructible for PNIE are absent in the Ana-
tolian languages: (i) a grammaticalized aspectual contrast between present and aorist stems; and (ii) the perfect as
a grammatical category. It is therefore a priori uncertain whether these verbal features — as well as certain others,
like the subjunctive and the optative (see §4.2.4 below) — should be reconstructed for PIE and their absence in
Anatolian attributed to historical loss, or whether they should instead be viewed as post-PIE innovations. These
issues are discussed in more detail below, but we lay out now the major assumptions that guide our presentation.
We adopt the position, shared by the majority of scholars, that PIE had an imperfective/perfective aspectual
contrast realized in the distinction between present and aorist stems. With respect to (ii), however, we follow
Jasanoff (2003a) in the view that a PIE verbal system was broadly Anatolian-like, in that all verbs belonged to
one of two formally distinct but — from a synchronic perspective — functionally undifferentiated conjugational
classes, the *m-conjugation or the *h2 e-conjugation. Furthermore, we assume with Jasanoff (fthcm. b) that an im-
portant innovation of PNIE — i.e. after the departure of the Anatolian branch — was the the grammaticalization
of the perfect, which developed out of a set of PIE verbs with the formal characteristics of PNIE perfects, includ-
ing reduplication and *h2 e-conjugation inflection (see further §4.2 and §4.3.3 below). Adopting these views has
significant implications for the PIE verbal system — for instance, on how the inflectional endings of the PIE verb
are reconstructed. This issue is addressed further in §4.2.5 and §4.2.6, where the evidence for the reconstruction
of PIE *m-conjugation endings and *h2 e-conjugation are separately assessed.

§4.2 PIE verbal inflection The PIE verb inflects for five grammatical categories, whose reconstructions are
discussed individually below: tense (§4.2.1), person and number (§4.2.2), voice (§4.2.3), and mood (§4.2.4). The ex-
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 43

ponents of person, number, and tense were fusional inflectional suffixes (“personal endings”), which were added
directly to a verbal aspectual stem. Two distinct sets of active voice inflectional endings are reconstructible for PIE,
one that became associated with the PNIE “perfect” stem, and another with the PNIE present and aorist stems; the
latter are sometimes referred to together as “eventive” active endings (and the present and aorist stems together as
the “eventive” system), a label that stems from the older view that verbs marked with these endings were seman-
tically opposed to a fundamentally stative perfect (now generally viewed as resultative-stative; see further §4.3.3
below).
These two sets of active endings have distinct cognates in the Anatolian languages, where all verbs belong to
one of two synchronically arbitrary inflectional categories, usually referred to as the mi- and ḫi-conjugations (after
their respective [Link]. endings in Hittite, –mi and –ḫ(ḫ)i). Active forms of Anatolian mi-conjugation verbs
have active personal endings clearly cognate with PNIE present/aorist active endings, and ḫi-conjugation verbs,
with PNIE perfect (active) endings. In what follows, we refer to PIE verbal endings that underlie the former as the
endings of the PIE *m-conjugation, and to the latter as the endings of PIE *h2 e-conjugation; the evidence for their
reconstruction is discussed in §4.2.5 and §4.2.6 respectively. In addition, PIE had a third set of verbal inflectional
endings associated with the middle voice. The distinction between verbs that select *h2 e-conjugation endings
in their active forms and those that select *m-conjugation endings is not realized in their corresponding middle
voice forms, both of which are marked the same set of middle endings; we assess the evidence for the formal
reconstruction of these endings in §4.2.7 below.

§4.2.1 Tense Tense is a grammatical category that relates the time of the event described to another point in
time, typically to the moment of the utterance (“absolute tense”), but in some cases, to the time of some other
discourse-relevant event (“relative tense”) (for the distinction, cf. Comrie 1976:2). Tense cuts asymmetrically across
the PNIE verbal aspectual categories. The imperfective stem shows a morphological contrast between non-past
and past tense forms (present vs. imperfect), and according to a majority of researchers, so does the “perfect” stem
(perfect vs. pluperfect), while the perfective stem has only forms that lack non-past tense marking (aorist); this
system is represented in (20):

aspectual stems
imperfective perfective “perfect”
(20)
non-past present — perfect
tense

past imperfect aorist pluperfect?

The imperfective and perfect stems differ in the way the tense contrast is encoded. Separate segmentable
markers of tense are clearly reconstructible in the imperfective stem, where non-past tense (i.e. present) verbal
forms are generally distinguished from past tense (i.e. imperfect) forms by the presence of an additional suffixal
element — in the active voice, by the “hic et nunc particle” *–i, and in the middle voice, by *–r (Yoshida 1990;
cf. Jasanoff, this volume). These morphemes may be viewed as markers of non-past tense (i.e. [– past]). Inflec-
tional endings characterized by these suffixal elements are traditionally referred to as “primary” endings, while
the unmarked endings of the past tense are called “secondary” (these labels, which confusingly appear to reverse
their morphological relationship, are due to their association with “sequences of tenses” in traditional grammars,
“primary” and “secondary” respectively). Thus (e.g.) the PNIE [Link]. was marked with the primary ending
*–m-i (vs. the imperfect “secondary” ending *–m), and the [Link]. form was marked with the primary ending
*–o-r/*–to-r (vs. imperfect *–o/–to); for the precise distribution of these tense markers and the evidence for their
reconstruction, see the detailed discussion of the reconstructible verbal “personal endings” in §4.2.5 and §4.2.7
below. The aorist employs the same secondary endings as the imperfect, and is thus formally indistinguishable
from the imperfect in certain stem classes (cf. §4.3).
Whether PNIE had a tense contrast in the “perfect” stem has long been debated (cf. Wackernagel 1926-8
[2009]:238 with references to older literature). It is now the majority view that the pluperfect, a past tense of
the perfect, should be reconstructed for this stage (see especially Jasanoff 2003a:34–43). The synchronic systems
of both Greek and Vedic include a separate pluperfect tense generally functioning as a past tense to the perfect, but
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 44

its PNIE status is complicated by serious difficulties in reconstructing the formal markers of this category — in par-
ticular, reconciling what appear to be significant discrepancies between the Greek and Vedic inflectional endings.
It is most likely, however, that the PNIE pluperfect was formed by addition of the secondary endings associated
with the present/aorist system to the perfect stem, as in Vedic, e.g. 1sg. ávedam ‘I knew’ (to the unreduplicated
perfect véda ‘knows’; cf. §4.3.2 below); 3sg. á-bi-bhe-t ‘feared’ (to the presential perfect bi-bhāy-a ‘fears’). For a
possible (albeit complicated) scenario by which the same endings underlie the markers of the Greek pluperfect,
see Katz (2008) and Jasanoff and Katz (2016).
The reconstruction of a future, i.e. as a morphologically distinct, inflectional category of the verb, is contro-
versial. Futurity could be expressed by the present indicative stem with or without an adverb expressly indicating
the future (on expressions of the future in ancient IE languages see Wackernagel 1926-8 [2009]:246–65 and refer-
ences in 247 n.14). Additionally, the subjunctive could refer to the future, with further modal meanings, in at least
PNIE. A desiderative suffix *–h1 se/o– meaning ‘wanting to do X’ comes to mark the future in a number of daughter
languages. The suffix has the shape *–h1 s-e/o– as reflected directly in Greek, indirectly elsewhere (for instance in
the Celtic futures descended from desideratives; cf. Stüber, this volume). Examples from Greek include tenéō, tenō ̂
‘I will stretch’ < *ten-h1 se/o- (cf. prs. teínō), or dérk-so-mai ‘I will see’ < *derk̑-h1 s-e/o–. What is very likely the same
suffix with a slight formal innovation, viz. *–h1 s-ye/o–, underlies the futures in Indo-Iranian and Baltic; e.g. Ved.
drak-ṣyá-ti ‘he will see’ < *derk̑-h1 s-ye-ti (on this morpheme cf. Jasanoff 2003a:134–5; note that others — e.g. Willi
(2011) — would derive this future instead from an *s-aorist subjunctive).
An additional prefix *(h1 )e–, the “augment,” marks past tenses in Indo-Iranian, Greek, Phrygian, and, in a
phonologically restricted way, Classical Armenian. Examples include Ved. á-han ‘he smashed’ < *e-gwh en-t (cf.
[Link]. *gwh én-ti ‘smashes’), Gk. é-ph er-e ‘he was carrying’ < *e-bh er-e-t (cf. [Link]. *bh er-e-ti ‘he carries’).
However, in the earliest Indo-Iranian and Greek texts past tense forms are not obligatory marked with the aug-
ment, which looks instead like an emerging, additional marker of [past]. Since no certain traces of the augment
have been found in other IE languages, augmented verbal forms are not reconstructible for PIE. The augment is
most often derived from a temporal deictic/anaphoric particle *h1 e ‘then’ (cf. e.g. Meier-Brügger 2010:315–6 with
references), although other etymological attempts have been made: Watkins (1963) (= 1994:3–51) derives the aug-
ment from a sentence connective seen in Anatolian (but cf. Melchert fthcm. a); as an alternative proposal, Willi
(2007) proposes to derive it from a reduplicating syllable, originally marking perfective aspect and only secondarily
past tense.

§4.2.2 Person & number There is general consensus that the PIE verb was morphologically marked for three
persons (1st, 2nd, 3rd) and three numbers (singular, dual, plural). Of these features, only the dual is somewhat un-
certain. As in the nominal system (cf. §2.1.2 above), the Anatolian languages synchronically lack dual number. It is
generally held that the 1du. marker (of the *m-conjugation) has ousted the 1pl. marker in the prehistory of Anato-
lian. The Proto-Anatolian 1pl. primary active ending may be uncontroversially reconstructed *–weni (based on e.g.
Hitt. –weni, Pal. –wini/–wani, CLuw. –unni < *–weni). Based on the resemblance of initial w in *–weni to the recon-
structed dual *–we-, it is thought that 1pl. *-weni is ultimately cognate with the ending of (primary/secondary) 1du.
in Indo-Iranian (Ved. –vaḥ/–va) and Balto-Slavic (Lith. –va, OCS –vě). The n-element would be presumably the
same as in the Gk. 1pl. –men (cf. Jasanoff 2003a:3, and cf. n.39, 47 n.98; more hesitantly, Kloekhorst 2008:1000–1).
Against this reconstruction, we note that the diachronic change whereby a dual ousts the plural is not typologically
trivial (see Corbett (2000:38–50, 268–71) for possible examples and discussion), and that no Anatolian language
shows any other trace of the dual in the verb or in pronouns (possible traces in the noun are discussed in §2.1.2
above). Although no alternative scenario has yet won acceptance, it may be the case that Proto-Anatolian *–weni
does not reflect an erstwhile dual marker (blended from dual *–wes and pl. *–meni). One attractive (if spec-
ulative) suggestion would reconstruct the cross-linguistically common category “inclusive” for the marker *-we,
which would then have become the Anatolian 1pl. *–weni and the PNIE 1du., thus constituting another significant
rift between the PIE and PNIE verb; for this reconstruction see Watkins (1969:46-8) (cf. Sihler 1993).

§4.2.3 Voice Two morphological voices are reconstructible for PIE, active and middle. This bivalent system is
maintained unaltered in Anatolian and Tocharian; the opposition between active and middle is also continued in
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 45

Indo-Iranian and in Greek, albeit with the later development of a separate (partially morphologically distinct) pas-
sive voice in these branches. This opposition is securely reconstructible only for the PNIE present/aorist system.
Indo-Iranian and Greek both synchronically make middle forms to the the perfect stem, but do so using the same
morphology as the present/aorist system (rather than distinctive PNIE “perfect” morphology); this lack of differ-
entiation suggests that the development of the perfect middle as a category was chronologically “late,” although
potentially already a feature of PNIE itself (cf. Jasanoff 2003a:44–5). Active and middle voices are characterized
by distinctive inflectional endings. The active and middle endings reconstructible for PNIE generally bear little
formal relationship to one another (e.g. [Link]. *–mi vs. mid. *–h2 er); rather, the middle endings closely re-
semble the endings of the PNIE perfect (active), a feature which has been argued to reflect a pre-PIE connection
between them (on which see §4.2.7 below).
Already by the PIE stage, however, the middle had become both formally and functionally differentiated from
the ancestor of the PNIE perfect. One core function of the PIE middle was to express subject affectedness, which
is clearly observed in transitive verbal stems that alternate between active and middle forms. In such opposi-
tional pairs, middle morphology marks verbs that are reflexive (e.g. mid. Gk. loúe-tai ‘washes him/herself’ vs. act.
loú-ei ‘washes’), reciprocal (Ved. yúdhy-ante, Hitt. zaḫḫiy-anta ‘they fight each other’ vs. Ved. yúdhy-anti, Hitt.
zaḫḫiy-anzi ‘they fight (someone)’), and self-benefactive (Ved. yája-te ‘sacrifices for his/her own benefit’ vs. yája-ti
‘sacrifices’). Middle morphology is also frequently used when the subject of a verb (transitive or intransitive) is
non-agentive. It therefore surfaces on anticausatives in “causative alternation” verbs (see, e.g., Haspelmath 1993)
— for instance, mid. Gk. ph úe-tai, Ved. várdha-te ‘grows (intr.)’ vs. act. Gk. ph ú-ei, Ved. várdh-ati ‘grows (tr.)’.
Many non-agentive verbs, however, are media tantum, i.e. take only middle morphology. The class of PNIE media
tantum — traditionally referred to in the IE literature as “deponents” (following Latin grammarians) — includes
many verbs belonging to semantic types that cross-linguistically tend to exhibit middle morphology in languages
where such dedicated morphology exists (see Kemmer 1993:41–94). These types include: verbs of cognition, e.g.
PNIE *mn̥ -yé-tor > Ved. mánya-te, ‘thinks’, OIr. –maine-thar ‘id.’, Gk. maíne-tai ‘rages’; non-translational motion
verbs, e.g. PNIE *sékw -e-tor ‘accompanies; follows’ > Ved. sáca-te, OAv. hacai-tē, Gk. hépe-tai, Lat. sequi-tur, OIr.
sechi-thir; PNIE *h3 ér-(t)o > Ved. (prá) ār-ta ‘set forth’, Gk. ō r̂ -to ‘arose’ (and from the same root, Lat. ori-tur ‘rises’);
and stative verbs, e.g. PIE wés-(t)or ‘wears’ > Ved. vás-te, OAv. vas-tē, Gk. heî-tai, Hitt. wēš-ta. The IE languages
also attest a number of agentive media tantum verbs, e.g. Ved. dáya-te, Gk. daíe-tai ‘distributes’; TA/B pāṣ-tär,
Hitt. paḫḫš-ari ‘protects’. Several verbs of this type — which notably exhibit a “mismatch” between semantics and
morphology — are reconstructible for the proto-language; for an assessment of the evidence, see Grestenberger
(2014a:225–53, 2016).
No separate passive can be reconstructed for PIE (or PNIE), its functions being expressed by middle mor-
phology (for which reason it is often referred to as “mediopassive”). The passive use of the middle is attested in
all of the oldest IE languages (cf. Hettrich 1990), including with expressed agent (in the instrumental case; see
Jamison 1979a,b, Melchert 2016a), although the rarity of examples within these languages suggests that this usage
was relatively uncommon. A separate passive voice with distinctive morphology arises in many of the daughter
languages (with or without loss of the middle). For instance, in the imperfective stem Vedic has an opposition
between middle and passive, adding to the root the (always accented) suffix –yá– (a specialization of PIE *–yé/ó–;
cf. §4.3.1) plus middle morphology to mark passive voice, e.g. ([Link].) Ved. kṣī-yá-te ‘is destroyed’ (cf. mid.
́
kṣī-ya-te ‘perishes’ with root accent) (see Kulikov 2012). Meanwhile in Greek a similar opposition developed in the
perfective stem, with the emergence of a distinct aorist passive formed by suffixation of *–(th )ē– plus secondary
active endings to the verbal root, e.g. ([Link].) Gk. e-gráph -ē ‘it was written’, e-lū ́ -th ē ‘was released’ (cf. mid.
e-gráp-sa-to ‘wrote for him/herself’, e-lū ́ -sa-to ‘released him/herself’). It is standardly assumed that the passive us-
age was an inner-Greek innovation, with the original core of the category formed by non-passive intransitive (i.e.
anticausative) aorists, e.g. e-mán-ē ‘went mad’, e-(w)ág-ē ‘broke’; on the historical origin of this category, see §4.3.1
below, and for discussion of the –ē–/–th ē– alternation in the suffix, Jasanoff 2003b:165–7 with references.
An older position — advanced by Oettinger (1976), influentially upheld by Rix (1988) and presupposed in LIV 2
— maintains that PIE had a third voice beside active and middle, the “stative” (Germ. Stativ). According to this
view, the “stative” is continued in Indo-Iranian verbal forms like [Link]. Ved. śáy-e ‘lies’, pl. sé-re (= YAv. sōi-re/saē-
re), and ipfc. á-śe-ran (< *k̑éy-o-i, –ro-i, –ro(n)), which semantically indicate a state, and are marked with endings
that share features with the regular endings of the middle ([Link]. –te < *–to-i) and the perfect active ([Link]. –a
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 46

< *–e; pl. –ur < *–r̥ s) but differ synchronically from both.
However, clear typological parallels for a trivalent voice system contrasting active, middle, and stative are
lacking, and the actual evidence in support of reconstructing a third voice is slim. Only in the third person (sg./pl.)
would distinctive “stative” endings be reconstructible; elsewhere in their paradigm, the relevant verbs use ordinary
middle morphology (e.g. [Link]. Ved. śé-ṣe ‘you lie’), and functionally equivalent forms are attested in later texts
marked with synchronically regular middle endings ([Link]. Ved. śé-te (= YAv. saē-te), śé-re ‘lie(s)’). Moreover,
in Anatolian, there is robust evidence for a [Link]. ending *–or (e.g. CLuw. ziy-ar ‘lies’; see further §4.2.7
below), from which the “stative” [Link]. ending *–oi can be derived straightforwardly by regular Indo-Iranian
replacement of the inherited *r-present tense marker of the middle with the *–i of the active (cf. [Link]. Ved.
–te < PIIr. –tai < < PIE *–tor); within Anatolian, the reflexes of *–or mark ordinary [Link]. forms, some which are
clearly non-stative, e.g. Hitt. ḫatt-ari ‘strikes’, paršiy-a ‘breaks’ (cf. Yoshida 2013:157).
In view of these issues, the “stative” is better treated as a transient effect of the renewal of middle morphology
(cf. Jasanoff 2003a:49–51). In the third singular the situation is clearest: two allomorphs of the [Link]. ending are
reconstructible for PIE, older unproductive *–o(r), and younger productive *–tor, the latter having been created
on the model of the corresponding *m-conjugation active ending *–t(i) in accordance with a pattern that is well-
established in IE languages (cf. §4.2.7 below). Archaic *–o(r) was gradually replaced by productive *–to(r) within
the IE languages, but was exceptionally retained under certain conditions — for instance, when forms marked by
*–or became semantically specialized, such as Ved. bruv-é, OAv. mruii-ē ‘is called’, whose passive sense contrasts
with that of the renewed middle forms Ved. brū-té, YAv. mrūi-te ‘calls to onself’. In other cases, retention of *–or
may have been due to high frequency, e.g. in a core vocabulary item like Ved. śáy-e ‘lies’; yet even such forms are
liable to renewal, and indeed, in chronologically later Vedic texts 3sg. forms of this same verb are attested with
identical semantics marked with the productive [Link]. ending –te (as noted above).

§4.2.4 Mood The following moods may be reconstructed for the PNIE verb: indicative, imperative, subjunctive,
optative. These are the moods of the verb in Greek and Indo-Iranian; inheritance in the other branches of PNIE
assures at least a PNIE age. Anatolian, however, deviates from this picture: the Anatolian languages distinguish
only indicative and imperative moods. Hittite, for example, expresses the potential, the unreal, the wished for
— notions associated with the subjunctive and optative (as well as the indicative) in PNIE languages — with the
particle man. Consequently, the reconstruction of the subjunctive and the optative for the stage of PIE including
Anatolian will depend on one’s evaluation of possible relic forms in Anatolian, together with one’s stance as regards
loss vs. non-inheritance in the prehistory of Anatolian.
The current understanding of moods in PIE is buttressed by centuries of fine-grained philological work. Repre-
sentative research in this vein includes the foundational study of Delbrück (1871), more recently e.g. Tichy (2006);
for an overview of the study of moods within Indo-European linguistics (with older bibliography) see Wackernagel
(1926-8 [2009]:266–323). Studies that take advantage of recent theoretical research on modality are thin on the
ground (for one example see Willmott 2007); continued incorporation of research on modality into the descrip-
tions of ancient languages will aid progress towards a more refined reconstruction of the meaning of the moods in
PIE (on modality see, e.g., Portner 2009 and the survey in Nuyts and Van der Auwera 2016).
We note here that many authorities include an “injunctive” mood in the PIE inventory. The injunctive is for-
mally the augment-less verbal stem with secondary endings (on the “augment”, cf. §4.2.1 above). Because its ex-
istence depends on the contrast with augmented verbal stems, and because we do not reconstruct the augment
for PIE, we do not reconstruct an injunctive for PIE; with Watkins (1969:45) we treat it as a category primarily
of Old Indic grammar. In the most influential account of the injunctive, that of Hoffmann (1967), it is proposed
that the augment designates past tense and, inversely, that the augment-less forms — the injunctives — cannot
designate the past. In mythological (arguably narrative/preterital) passages of the Rigveda the injunctive would
have the function of “mentioning” (“Erwähnung”), and its modality would be “memorative.” The textual and cross-
linguistic plausibility of this verbal structure (a “memorative” modality) is questionable, and has been critiqued
especially by Kiparsky (1968, 2005), whom we follow in treating the injunctive not as a mood but rather as a stem
underspecified for mood (as well as tense), taking on its values for tense and mood from context.
For reference, a table of the PNIE moods is provided in (21):
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 47

Formation of modal stems

Indicative/Imperative Subjunctive Optative
*h1 (e)s– ‘be’ (*h1 s-dh í) *h1 es-e/o– (*h1 és-e-t(i)) *h1 s-yeh1 –/*h1 s-ih1 – (*h1 s-yéh1 -t)
(21)
*li-n(e)-kw – ‘leave’ (*li-n-kw -dh í) *li-ne-kw -e/o– (*li-né-kw -e-t(i)) *li-n-kw -yeh1 –/*li-n-kw -ih1 - (*li-n-kw -yéh1 -t)
*bh er-e/o– ‘carry’ (*bh ére) *bh er-e-e/o– (*bh ér-ē-t(i)) *bh er-o-ih1 – (*bh ér-oi(h1 )-t)
*pr̥k-̑ sk̑e/o– ‘ask’ (*pr̥k-̑ sk̑é) *pr̥k-̑ sk̑e-e/o– (*pr̥ks̑ k̑-ḗ-t(i)) *pr̥k-̑ sk̑o-ih1 – (*pr̥ks̑ k̑-oí(h1 )-t)

§[Link] Imperative The imperative basically expressed orders and commands (more generally and more tech-
nically, “directives”). In the 2nd singular active of athematic verbs, the ending was either zero or *–dh i added to
the weak stem; e.g. Ved. [Link]. śru-dhí ‘listen!’ < *k̑lu-dh í (root *k̑lew– ‘listen’). Thematic verbs used the bare
stem, as in Gk. ph ére ‘carry!’ < *bh er-e. The 2nd singular middle imperative ending exhibits greater diversity across
the daughter languages: Lat. –re (< *–so), Gk. –o (< *–so), Ved. –sva, Hitt. –(ḫ)ḫut (*-h2 u-dh i), etc. Jasanoff (2006)
attempts to reconcile the forms under the reconstruction *–sh2 (u)wo (for which Barnes 2015 provides Old Irish
comparanda). The 2nd person plural and dual active imperatives were identical to the corresponding indicative
forms, thus (e.g.) 2pl. Ved. bhára-ta, Gk. ph ére-te ‘carry!’ (< *bh ére-te); 2du. (athematic) Ved. i-tám, Gk. í-ton ‘you
two go!’ (< *h1 i-tóm), (thematic) Gk. ph ére-ton, Ved. bhára-tam ‘you two carry!’ (< *bh ere-tom). Similarly, plural
and dual middle imperatives deployed the same endings as the indicative (cf. §4.2.5). What are traditionally called
third-person imperatives are modal forms expressing the speaker’s wish that a third person act in some way. Two
formations encoding these third-person imperatives may be reconstructed. The first formation is the suffix *–u
agglutinated to the endings of the third-person, *–t-u, *–nt-u (Hitt. eš-tu, Ved. ás-tu ‘let it be’ < *h1 es-t-u). The
second formation is a suffix *–ōd, also added to the secondary endings, as in the so-called “future imperative” in
Lat. –tōd (Cl. Lat. –tō), Ved. –tād, Gk. és-tō < *h1 es-t-ōd.

§[Link] Subjunctive The subjunctive encoded various modal readings, of which a prospective and hortative
are traditionally reconstructed. In athematic verbs, the subjunctive marker is added to the full-grade root; for in-
stance, from the root *h1 es– ‘be’ was formed *h1 es-e-ti (cf. prs. *h1 es-ti ‘is’). Thus in athematic subjunctive forms
looked identical to thematic indicative forms — compare (e.g.) athematic subjunctive ([Link].) *h1 es-e-ti with
thematic indicative *bh er-e-ti (to *bh er– ‘carry’). This formal identity may indicate a functional split; it has been
suggested that the subjunctive functions developed from a present indicative (Bozzone (2012) and Dahl (2013) pro-
vide possible diachronic pathways for the change). If the stem was thematic, the theme vowel and the subjunctive
suffix contracted to a long vowel. As far as inflectional endings go, there is conflicting evidence for whether pri-
mary or secondary endings were used with the subjunctive (on the Vedic evidence see García Ramón 2009); we
reconstruct primary endings here, but this reconstruction is not certain. It should be noted that numerous daugh-
ter languages have categories called “subjunctive” in their grammars, but these may or may not derive from the
PIE subjunctive. In Latin, for instance, what grammarians call the “subjunctive” reflects in large measure the PNIE
optative, while the PNIE subjunctive has become one ingredient of the Latin future. We provide below a chart of
stem formation for athematic and thematic indicatives and subjunctives in PNIE:

athematic [Link]. athematic subj. thematic [Link]. thematic prs. subjunctive

(22) *h1 es-ti ‘he is’ *h1 es-e-ti *bh er-e-ti ‘he carries’ *bh er-e/o-e-ti (*bh erēti)
*h1 s-enti ‘they are’ *h1 es-o-nti *bh er-o-nti ‘they carry’ *bh er-e/o-o-nti (*bh erōnti)

Whether the subjunctive is to be reconstructed for PIE will depend on one’s assessment of the Anatolian evi-
dence. No Anatolian language has a living subjunctive; whether any Anatolian language has a relic of the subjunc-
tive is disputed (for different viewpoints see Jasanoff, this volume and Oettinger, this volume). Jasanoff (2012a)
analyzes the Hittite [Link]. paḫši ‘protect!’, eši ‘settle, occupy!’, and ēšši ‘do, perform!’ as containing a PIE im-
perative ending *–si, which ultimately derives from 2sg. subjunctives built to a variety of sigmatic formations via
haplology, i.e. *–s-e-si > *–si; thus paḫši ‘protect!’ would derive from a subjunctive *peh2 -s-(e-s)i. There is evidence
from Indo-Iranian, Celtic, and Tocharian for reflexes of an imperative in *–s-e-si > *–si (see Jasanoff 2003a:182–3
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 48

with references); however, it should be emphasized that Jasanoff’s (2012a)’s proposed Anatolian reflex of *–s(es)i
would be the sole Anatolian outcome of the PIE subjunctive.

§[Link] Optative The PIE optative expressed at least wishes and potentialities (traditionally “cupitive” and
“potential” respectively). In a more nuanced reading of the moods in Homeric Greek, Willmott (2007:113–52, esp.
120–1) argues that the optative shows broadly “negative epistemic stance,” i.e. the optative indicates that the event
is not in line with the speaker’s view of the world. The mark of the PIE optative was an ablauting suffix *–yeh1 /ih1 –
added to athematic stems, non-ablauting *–ih1 – to thematic stems (*–o-ih1 –), plus the secondary endings. Thus to
the root *h1 es– ‘be’ would be formed the [Link]. *h1 s-yeh1 -t ‘he would be’, and to the thematic stem *bh er-e/o–
‘carry’ would be formed [Link]. *bh er-o-ih1 -t ‘he would carry’. We note here that the thematic vowel and the
optative suffix — *–o– + *–ih1 – — appear not to have contracted within PIE; evidence from the daughter languages
suggests that the two morphemes remained disyllabic (for possible reasons why, see Jasanoff 2009). The table in
(23) provides illustrative optative forms for athematic and thematic present stems:

athematic thematic
Ved. Gk. Lat. PIE Ved. Gk. PIE
(23) h
syā ́t eíē siēt *h1 s-yéh1 -t bháret p éroi *bh ér-o-ih1 -t
syā ́ma eîmen sīmus *h1 s-ih1 -me bhárema ph éroimen *bh ér-o-ih1 -me-

The optative is well-preserved in Greek and Indo-Iranian. In other branches, reflexes of the optative are clearly
inherited but go by different names. For instance, the Italic subjunctive reflects in part the optative; we have
used the verb siēd (Cl. Lat. sit), sīmus to illustrate the paradigm (fuller details in Vine, this volume). In Balto-
Slavic the optative develops into the synchronic imperative (standard Lithuanian “permissive”); in Tocharian, the
optative has become the optative of TA and TB, as well as the TB imperfect; etc. Once again, Anatolian presents a
divergent picture: there is no evidence for the optative in Anatolian. This absence could be interpreted as either
loss (the optative would be inherited into Proto-Anatolian, with subsequent evanescence) or non-inheritance (i.e.
Anatolian branched off before the category had developed). We think the latter option is likelier, but the matter is
still sub iudice.

§4.2.5 Verbal endings of the PIE *m-conjugation It was noted in §4.2 that PIE had two sets of reconstructible
active verbal endings, fusional exponents of person, number, and voice. One of these sets was the common source
of the active verbal endings of the PNIE present/aorist system and of the Anatolian mi-conjugation. We refer to
these endings as the PIE *m-conjugation endings.
Verbal stems selecting the PIE *m-conjugation endings can be further subdivided into two conjugational classes,
athematic and thematic, the latter characterized by a stem-final ablauting thematic vowel (*o/e). As in the noun
(cf. §2.1.1), the distinction between these classes was purely formal. With the notable exception of the [Link].
ending (and for some scholars also the [Link].; see below), thematic verbs have the same inflectional endings
as the athematic classes, being formally distinguished from the latter only by the presence of the thematic vowel,
which has *o-quality in 1sg./pl. and 3pl. paradigmatic forms, and *e-quality elsewhere — thus (e.g.) [Link].
athematic *–m vs. thematic *–o-m; 3sg. *–t vs. *–e-t; 3pl. *–(e)nt vs. *–o-nt. The exceptional 1(/3)[Link]. primary
thematic endings are discussed below together with their corresponding athematic endings.
The PIE athematic *m-conjugation inflectional endings that are securely reconstructible are given in (24). A
following dash (-) indicates the possibility that the PIE ending had additional segmental material, the reconstruc-
tion of which is problematized by conflicting evidence in the daughter languages. The evidence for these indi-
vidual reconstructions, as well as their problematic or controversial aspects, are discussed further immediately
below.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 49

PIE *m-Conjugation Active Endings

singular plural
1ry 2 ry 1ry 2ry
(24)
1st *–mi *–m *–me-?
2nd *–si *–s *–te-
3rd *–ti *–t *–(e)nti *–(e)nt

The reconstruction of the primary (athematic) singular active endings is wholly uncontroversial and sup-
ported by robust evidence across the daughter languages. The [Link]. ending *–mi is clearly attested in (e.g.)
Hitt. Gk. ei-mí, Ved. ás-mi, OAv. ah-mī, OCS jes-mǐ, Hitt. ēš-mi ‘I am’, and somewhat less transparently in VOLat.
es-om (Lat. s-um), Goth. i-m, OIr. a-m (< PIE *h1 és-mi).
The [Link]. ending *–si is continued in Ved. á-si and OAv. a-hī ‘you are’, as well as Gk. e-î (< PGk. *e-hi), Goth.
i-s (< PIE *h1 é-si with degemination of */s-s/; see Byrd, this volume). For this lexical item, some languages attest
only forms with root-final *s analogically restored (e.g OLat. es-s, Hitt. eš-ši; pace Kloekhorst 2016:238–41), or else
such forms coexist with the directly inherited ones (e.g. Hom. Gk. es-si).
The [Link]. ending *–ti is reflected in Gk. es-tí, Ved. ás-ti, OAv. as-tī, OLith. ẽs-ti, ORuss. jes-tĭ, CLuw. āš-ti ‘is’,
and additionally, in Lat. es-t, Goth. is-t, OIr. is (< PIE *h1 és-ti).
Thematic inflection differs substantially from athematic in the primary [Link]. ending, where the daughter
languages reflect an ending *–ō instead of expected x *–o-mi, e.g. Gk. ph ér-ō, Lat. fer-ō, Go. bair-a, OCS ber-ǫ ‘I bear’
(< PNIE *bh er-ō); this morphological irregularity was eliminated within some language branches, e.g. Indo-Iranian
(cf. Ved. bhár-āmi, OP bar-āmiy, YAv. bar-āmi ‘id.’). It is the majority view that thematic 1sg. *–ō historically con-
tains the same suffix *–h2 e that is found in the 1sg. endings of the PNIE perfect (active) and the middle voice (PNIE
[Link]. *–h2 e, [Link]. *–h2 e-r; see further below). Since Pedersen (1938:80–6), some scholars have suspected that
the simple thematic conjugation, the PNIE perfect, and the middle voice are historically related; pursuing this hy-
pothesis, Watkins (1969:66–9, 105–23, et passim) proposed that *–ō descends from a unitary pre-PIE type underly-
ing these three categories (later developed by Jasanoff (1978; 1998; 2003a, et seq.) as the “proto-middle;” see further
discussion in §4.2.7), whose verbal paradigm had a 1sg. ending in **–h2 e and 3sg. in **–e (like the *h2 e-conjugation;
see §4.2.6 below). Some members of this category were eventually thematized — according to Watkins (1969), via
reanalysis of 3sg. forms like **bh er-e as zero-marked **bh ere-ø, whence new 1sg. **bh ér(-)e/o-h2 e. Generally these
“pre-thematic” forms would then be re-characterized with ordinary PNIE *m-conjugation active endings (e.g. 2sg.
*bh ér-e-si), but 1sg. **–e/o-h2 e was exceptionally retained, developing into P(N)IE *–ō (probably via *–oh2 , with
apocope due to the same phonological process as in the thematic neuter dual ending; see §2.1.1).
Watkins (1969) further argued that t-less 3sg. forms like **bh er(-)e are directly reconstructible for PIE. Most of
Watkins’ comparative evidence for this reconstruction (from Tocharian, Celtic, and Balto-Slavic) can be explained
more straightforwardly as reflexes of *–e-ti (see Jasanoff 2003a:59–60). Somewhat more problematic is the evi-
dence from Greek, where it is maintained by some (e.g. Rau 2009b:186 n. 14) that thematic verbs like Gk. ph ér-ei
‘carries’ directly continue **bh er(-)e-i (with only the addition of the present tense marker *–i; cf. §4.2.1). However,
this analysis would imply a surprising divergence between Greek and other NIE languages with closely related ver-
bal morphology (esp. Indo-Iranian, e.g. 3sg. Ved. –a-ti; economy therefore recommends the alternative approach,
first proposed by Kiparsky (1967) and revised by Cowgill (1985) and Willi (2012), which derives the Greek thematic
3sg. ending –ei from *–e-ti via metathesis at word boundary followed by the regular loss of word-final stops in
Greek (i.e. *–eti# > *–ei-t# > –ei#). Thus only a single thematic 3sg. ending *–e-ti is securely reconstructible for
PIE, although Watkins’s (1969) t-less reconstruction may have obtained at an earlier, pre-PIE stage (cf. Jasanoff
2003a:148–9).
Similarly straightforward is the reconstruction of the secondary singular active endings. The [Link]. ending
–m is reflected in (aor.) Ved. á-sthā-m, Gk. é-sth ē-n ‘I stood’ (< PNIE *steh2 -m), as well as in the Latin (synchronic)
imperfect ending –bā-m (see Vine, this volume). Thematic verbs show the expected [Link]. ending *–om, e.g. Gk.
é-ph er-on, Ved. á-bhar-am, YAv. bar-əm, OP a-bhar-am ‘I was bearing’ (< PNIE *bhér-om).
The [Link]. ending *–s is directly continued in Ved. á-dhā-s, OAv. dā-s(-ca) ‘you placed’, Hitt. tē-s ‘you said’
(< *dh eh1 -s), as well as the Germanic weak preterite ending (e.g. Goth. –de-s, OIc. –ðe-r), which should likely be
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 50

traced back to the same PIE form (see Harðarson, this volume). Further reflexes include (Dor.) Gk. é-bā-s ‘you
went’, Ved. á-gā-s, (< *gw eh2 -s), and the Latin imperfect ending –bā-s.
The [Link]. ending *–t is evident in (aor.) Ved. á-dhā-t, (Boet.) Gk. (an)é-th ē ‘placed’, and (pst.) Hitt. tē-t ‘said’
(< aor. *dh eh1 -t ‘placed’, with semantic innovation in Hittite).
Somewhat more problematic is the reconstruction of the PIE [Link]. endings. Several of the attested primary
and secondary endings in the daughter languages continue *–me- (e.g. Ved. 1ry –mas(i) / 2ry –ma, OAv. –mahī /
–mā (< PIIr. *–mas(i)/–ma); Att.-Ion. Gk. –men, Dor. Gk. –mes), which is expected on structural grounds, but
Italic and Slavic both reflect an o-grade *–mo- (Lat. –mus, OCS –mŭ), and at least Lith. –me appears to require
a lengthened variant *–mē; it is uncertain whether these differences are due to independent innovations within
these languages or reflect phonologically-conditioned allomorphy already at the P(N)IE stage (cf. Weiss 2011:385–
6). There is also variation within and across language branches with respect to the post-vocalic segment: Latin
–mos, Dor. Gk. –mes, and PIIr. (1ry ) *–mas(i) contain an element *s, while Att-Ion. –men has *n in its place, thus
matching Anatolian (e.g. Hitt. 1ry –w(/m)eni / 2ry –w(/m)en; on the fluctuation of the ending’s initial consonant
and its possible dual origin, see further in §4.2.2). Furthermore, it is unclear whether the primary and secondary
endings were differentiated: past and present tense verbal forms in Greek, Italic, and Balto-Slavic reflexes of the
PNIE 1pl. are identical, but in Indo-Iranian the primary ending is distinguished by an additional post-vocalic *s
(plus the “hic et nunc” particle *–i in all Avestan and some Vedic forms, which similarly characterizes non-past
tense forms in Anatolian).
Similar issues arise in the reconstruction of the PIE [Link]. endings, which had the basic shape *–te-, e.g. Gk. –
te, Lat. –tis, OCS –te and Goth. –þ. It is unclear whether there was any distinction between primary and secondary
forms; the four languages cited above employ the same ending for both, but in Indo-Iranian, the primary ending
PIIr. *–th a (> Ved. –tha, OAv. –θā) contrasts with secondary *–ta (> Ved. –ta, OAv. –tā) (see Kümmel, this volume).
As in the 1pl. ending, Latin shows a post-vocalic segment *–s, while Anatolian has *–n (Hitt. 1ry –teni / 2ry –ten),
but neither has external comparative support from Greek or Indo-Iranian. Lith. *–te reflects a lengthened variant
*–tē just as in the 1pl. ending.
The PIE athematic primary 3pl. act. ending was *–enti, e.g. prs. Ved. s-ánti, Myc. Gk. e-e-si [eh-ensi], Hitt.
aš-anzi Osc. s-ent, Goth. sind, OIr. it [id] (< PIE *h1 s-énti ‘are’). The corresponding secondary ending was *–ent
(aor. PIE *gw (e)h2 -ent ‘went’ > Ved. á-gan, Gk. é-ban; cf. perhaps Pal. –Vnta [–nt]). Zero-grade allomorphs of
these endings 1ry *–n̥ ti / 2ry *–n̥ t are also attested in several NIE languages in athematic verbal formations that had
fixed accent on a syllable preceding the ending: “Narten presents” (e.g. Ved. tákṣ-ati ‘they fashion’ < *té-tk̑-n̥ ti);
reduplicated presents (Ved. dád-ati ‘give’ < *dé-dh3 -n̥ ti; simple thematic presents (Ved. bhár-a-nti) and *s-aorists
(Gk. é-deik-s-an ‘showed’ < *deik̑-s-n̥ t; OCS (po-)grĕ-s-ę ‘buried’ < *gh rébh -s-n̥ t). The zero-grade allomorph *–n̥ ti is
therefore standardly reconstructed for PIE in these categories.
The reconstruction of the PIE dual endings is more difficult, given the more limited evidence for this category
in the IE languages. However, the NIE languages agree that the basic shape of PNIE athematic [Link]. ending
was *–we- (> (1/2ry ) Ved. –vas/–va, OCS –vě, Lith. –va; for Germanic traces, cf. Prokosch 1939:212, Ringe 2006:136);
although dual number is absent as a grammatical category in the Anatolian languages, it is generally held that
the Anatolian [Link]. endings (Hitt. –w(/m)eni, CLuw. –unni, Pal. –wini/wani) derive from *–we- and thereby
support projecting this ending back to PIE (but cf. §4.2.2 above). There is also comparative NIE evidence for
reconstructing the [Link]. ending as *–to- (> 2ry Ved. –tam, Gk. –ton, OCS –ta). A secondary 3du. ending *–teh2 m
is perhaps reconstructible as well in view of agreement between Gk. –tēn and Ved. –tām, but the P(N)IE situation
is complicated by a mismatch in the corresponding primary ending between Gk. –ton and Ved. –tas.

§4.2.6 Verbal endings of the PIE *h2 e-conjugation In addition to the *m-conjugation endings (§4.2.5), PIE had
a second set of active verbal endings that developed, on the one hand, into the endings of the PNIE perfect active
and, on the other, into the endings of the Anatolian ḫi-conjugation. PIE reconstructions for these inflectional
endings — referred to here as the *h2 e-conjugation endings — are given in 25; note, however, that these recon-
structions — much more so than the *m-conjugation endings discussed in §4.2.5 above or the middle endings in
§4.2.7 below — are quite uncertain. In particular, reconstructing the distinction between primary and secondary
endings in the *h2 e-conjugation is highly problematic, in part because only Anatolian provides direct evidence
for the original morphological opposition (the PNIE pluperfect uses *m-conjugation secondary endings; see §4.2.3
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 51

above), and in part due to open (and much disputed) questions surrounding the prehistory of the PNIE perfect
— above all, whether the perfect endings stand in correspondence with (and so provide evidence for the recon-
struction of) PIE primary endings (as recently argued by (e.g.) Jasanoff 2003a, Oettinger 2006) or with secondary
endings (per Jasanoff fthcm. b). These issues are discussed below, together with the evidence for the formal re-
construction of the *h2 e-endings.

PIE *h2 e-Conjugation Endings

singular plural
1 ry 2 ry 1 ry 2ry
(25)
1st *–h2 ei? *–h2 e *–me-?
2nd *–th2 ei *–th2 e *–te? *–e? , *–s?
3rd *–ei? *–e? , *–s(t)? *–(e)nti *–(e)rs?
The PIE [Link]. primary and secondary endings of the *h2 e-conjugation were probably *–h2 ei and *–h2 e re-
spectively. Both are directly reflected in Anatolian, the former in Old Hitt. –ḫḫe (e.g. dā-ḫḫe ‘I take’; replaced by
–ḫḫi in younger texts), and the latter in CLuw. –(ḫ)ḫa, Lyc. –xa (CLuw. a-ḫa, Lyc. a-xa ‘I made’; Hitt. –(ḫ)ḫun
is remodeled on the basis of the corresponding mi-conjugation ending). The 1sg. ending of the PNIE perfect was
*–h2 e, which yields Gk. –a, PIIr. *–a, and Goth. –∅ (e.g. Gk. oîd-a, Ved. véd-a, OAv. vaēd-ā, Goth. wait ‘I know’
< PIE *woid-h2 e). In Italic and Slavic, the perfect ending was recharacterized with the present tense marker *–i
(i.e. *–h2 e-i), whence (e.g.) OCS věd-ě ‘I know’, Fal. pe:parai ‘I got’ (cf. Lat. –ī; see Weiss 2011:392). The formal
identity between the endings of the PNIE perfect and the secondary endings of the *h2 e-conjugation suggests that
the former historically descend from the latter, and thereby offers some support for Jasanoff’s (fthcm. b) recent
derivation of the perfect from a PIE reduplicated *h2 e-conjugation aorist (rather than a reduplicated present, as
per Jasanoff 2003a:168–9, Oettinger 2006, i.a.). There is some evidence to suggest that the presence of the present
tense marker *–i in the primary ending *–h2 ei was a relatively recent innovation in PIE: the synchronically irregular
m-conjugation [Link]. primary thematic ending *–ō should probably be traced back to *–e/o-h2 e (as discussed in
§4.2.5), which would contain a [Link]. ending *–h2 e unmarked for tense. Yet given the robust evidence across the
IE languages for a formal opposition in singular verbal endings between primary and secondary forms, it seems
more likely that pre-PIE **–h2 e was recharacterized as *–h2 ei already in PIE, and thus that the development of
*m-conjugation thematic 1sg. *–ō also occurred prior to PIE (cf. §4.2.5 above).
The PIE [Link]. primary and secondary endings were likely *–th2 ei and *–th2 e. The primary ending is indi-
rectly reflected in Hitt. –(t)ti (phonologically expected –te* having been analogically replaced already in the oldest
texts). Just as in the first singular, the inherited secondary ending appears to be continued not only in [Link].
Hitt. –tta (e.g. da-tta ‘you took’), but also in PNIE [Link]. *–th2 e, which yields the regular perfect ending in
Indo-Iranian (PIIr. *–th a), e.g. Ved. dadh ā-tha, OAv. dadā-θā ‘you have placed’; Ved. vét-tha, OAv. vōis-tā ‘you
know’. Cognate Gk. oîs-th a ‘id.’ exceptionally preserves the same ending, although elsewhere it has been replaced
by –as (e.g. téth ēk-as ‘you have placed’), with the –s of the *m-conjugation active and a-vocalism by analogy to the
1sg. –a (see above). The second element of the Latin perfect ending –is-tī (e.g. fec-istī ‘you made’) also continues
*–th2 e-i, with the inner-Italic addition of the tense marker *–i.
The reconstruction of the PIE [Link]. ending is a vexed question. The primary ending in PIE was likely *–ei,
which is marginally continued in Old Hitt. –e (e.g. waršš-e ‘wipes’; replaced by –i in younger texts); it is unlikely that
Gk. –ei derives from *–ei despite its superficial resemblance (cf. §4.2.5 above). Jasanoff (2003a:70–1, 2012c) argues
that the primary/secondary distinction was instead realized in the *h2 e-conjugation by an opposition between *–e
and *–et, with the latter recharacterized by *m-conjugation [Link]. *–t already in PIE; see however Kim (2005:195)
for the PIE primary ending as *–ei with regular tense marking. Still more problematic is the reconstruction of
the corresponding secondary ending, for which at least two forms are arguably reconstructible. One of these is
structurally expected *–e, which is reflected in PNIE [Link]. *–e (> Ved. –a, OAv. –ā, Gk. –e, Goth. –∅). The
other is *–s(t), which is reflected in Hitt. –š (e.g. dā-š ‘took’) and in the ending associated with Tocharian Class III
preterites, TB/A –sa/–äs (e.g. prek-sa/prak-äs ‘asked’; see Melchert 2015). Both forms have strong claim to antiquity.
Jasanoff’s (fthcm. b) derivation of the PNIE perfect from a reduplicated *h2 e-conjugation aorist requires that *–e
marked the 3sg. in this category at the stage prior to its post-PIE grammaticalization as the perfect. However, the
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 52

match between Hittite and Tocharian with respect to structurally unmotivated *–s(t) is prima facie evidence for
a morphological archaism, and on these grounds Melchert (2015) argues that it is the original marker of the 3sg.
*h2 e-conjugation aorists (cf. Watkins 1969:54, Yoshida 1993:33–4). The distribution of these endings in PIE remains
at present unresolved.
In PIE the *h2 e-conjugation and the *m-conjugation appear to have had the same [Link]. ending *–me-, which
marked both primary and secondary forms (cf. §4.2.5 above). The ending *–me is reflected in PNIE [Link]. *–
me- (> Gk. (w)íd-men, Ved. vid-má ‘we know’) and in both primary and secondary forms of ḫi-conjugation verbs
in Anatolian, which have [Link]. forms that are inflectionally identical to mi-verbs (i.e. Hitt. 1ry –w(/m)eni / 2ry
–w(/m)en). See however Jasanoff (2003a:32) for the possibility that Ved. –mā (e.g. vid-mā ́ ‘id.’) — synchronically,
a lengthened allomorph of the ending — derives rather from a PIE form with final laryngeal (e.g. *–mehx ) that
was once unique to the *h2 e-conjugation.
Similarly, the PIE [Link]. primary ending of the *h2 e-conjugation was most likely *–te, just as in the *m-
conjugation (cf. §4.2.5 above); it is continued in Anatolian, e.g. [Link]. Hitt. da-tteni ‘you (pl.) take’. Reflexes
of *–te are also attested in the PNIE perfect (Gk. (w)ís-te, Goth. wit-uþ ‘you (pl.) know’), but these have clearly
been analogically introduced from the *m-conjugation, since Ved. –a (e.g. vid-á ‘id.’) preserves the inherited PNIE
[Link]. ending *–e. More complicated is the reconstruction of the corresponding secondary ending, for which
two forms are potentially reconstructible for PIE: the *–e ending discussed above, and *–s, which according to
Melchert (2015) is reflected (with additional morphological material) in both Anatolian (Hitt. –šten, e.g. dai-šten
‘you (pl.) placed’; cf. Kloekhorst 2008:498) and Tocharian ([Link]. PT *–sV(s); see Malzahn 2010 for TB/A
outcomes, as well as references to alternative explanations). Agreement between Hittite and Tocharian would ar-
gue strongly that *–s is an archaic feature, a common retention of these branches. However, there are compelling
reasons to believe that *–e is also archaic: analogical explanation is not viable, since it bears no affinity to any
other PIE 2pl. ending, For this reason the ending is liable to diachronic renewal by the functionally transparent
*–te of the m-conjugation. Moreover, a remarkable feature of both *–e and *–s is that each is identical to one of the
two possible secondary endings reconstructible for the [Link]. of the *h2 e-conjugation (i.e. *–e, *–s; see above).
Just as in the [Link]., the exact PIE distribution of *–e and *–s is at present unsettled, and still less clear is the
broader significance of the structural symmetry between 3sg. and [Link]., which appears to be a unique feature
of the *h2 e-conjugation.
The PIE [Link]. primary ending of the *h2 e-conjugation was probably *–(e)nti, once again identical to the
*m-conjugation; it is directly reflected in Anatolian, e.g. Hitt. [Link]. akk-anzi ‘die’. The [Link]. secondary
ending has several reflexes in the daughter languages (cf. Jasanoff 2003a:32–4): *–ēr, which yields [Link]. Hitt.
–ēr and pfc. Lat. –ēre (via *–ēr-i with inner-Italic addition of the primary tense marker; cf. 1/[Link]. above); *–r̥ s,
which is continued in OAv. –ərəš (see further below) and pfc./opt. Ved. –ur; and *–r̥ , which is continued in pfc.
OAv. –arə̄ (YAv. –ērə) and pret. OIr. *–(a)tar (< *–ont-r̥ , a composite of thematic 3pl. *–ont + *–r̥ ). For arguments
that *–r̥ is a later analogical innovation, see Jasanoff (2003a:33). The remaining two endings *–ēr and *–r̥ s can be
reconciled as ablaut variants of *–ers (whose status as a PIE surface form is however dubious; see discussion of
thematic [Link]. *–oms in §2.1.1 above): its expected zero-grade form is *–r̥ s, while full-grade *–ers would develop
straightforwardly into *–ēr via Szemerényi’s Law. According to Jasanoff (1997, 2003a:39–43), PNIE originally had
the full-grade allomorph *–ēr in the perfect and *–r̥ s in the pluperfect, a distribution which — with the exception
of the replacement of *–ēr by analogical *–r̥ — is maintained in Avestan (*–r̥ /*–r̥ s > OAv. –arə̄/–ərəš).
There is insufficient evidence to reconstruct dual endings for the *h2 e-conjugation. Greek and Indo-Iranian
both make perfect dual forms, but their endings cannot be derived from a single pre-form; rather, the endings in
each language show clear effects of analogical re-shaping — for instance, [Link]. Ved. –atuḥ and YAv. –atarə
have evidently been influenced by the corresponding [Link]. endings (on which see above). Anatolian offers no
help, since there is no trace of the dual in the ḫi-conjugation.

§4.2.7 Verbal endings of the PIE middle In PIE, both verbs whose active forms inflected according to the *m-
conjugation and those whose inflected according to the *h2 e-conjugation had middle voice forms marked with the
same set of endings. This situation is still observed within the Anatolian languages, where these two conjugational
classes remain distinct, but verbs of both classes make use of the same middle endings, e.g. [Link]. Hitt.
ištamaš-zi ‘hears’, kānk-i ‘hangs (tr.)’ vs. mid. išdamaš-tari ‘is heard’, kank(a)-ttari ‘hangs (intr.)’. Similarly, the PNIE
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 53

perfect active is marked by endings that descend from the *h2 e-conjugation, but perfect middle forms generally
employ the same endings as in the present/aorist system (e.g. [Link]. Gk. lélu-tai ‘has been released’; cf. prs.
lúe-tai), whose active endings come from the *m-conjugation.
Reconstructions for the PIE athematic middle inflectional endings are given in (26). We discuss the evidence
that supports — or else problematizes — the reconstruction of each ending below.

PIE Middle Endings

singular plural
1ry 2ry 1 ry 2ry
(26)
1st *–h2 er *–h2 e? *–medh h2
2nd *–th2 er *–th2 e *–dh h2 we?
3rd *–or, *–tor *–o, *–to *–ror? , *–ntor *–ro, *–nto
The PIE primary [Link]. ending was *–h2 er, which is most clearly reflected in Hitt. –ḫa(ri) (e.g. ar-ḫari ‘I
stand’) and — with regular renewal of *–r by *–i as marker of present tense (as in the active endings) — in PIIr.
*–ai (e.g. Ved. bruv-é, OAv. mruii-ē ‘I speak’). The synchronic “passive” endings Lat. –or and OIr. –or (e.g. Lat.
ori-or ‘I rise’; OIr. –mol-or ‘I praise’) continue the corresponding thematic form *–o-h2 er. In Tocharian and Greek,
the initial *m of the *m-conjugation active has been analogically introduced, thus TB/A –mar/–mār (for details,
see Malzahn 2010:36 with references) and Gk. –mai (with the same renewal of presential *–r by *–i as in Indo-
Iranian); this kind of analogical remodeling — viz. assimilation of the characteristics of the corresponding *m-
conjugation active endings — is typical of the development of the middle endings in the IE languages, as will
become clear below. Hittite also attests an “iterated” (or “reduplicated”) allomorph of the ending –(ḫ)ḫaḫari (cf.
ar-ḫaḫari), which points to a preform *–h2 eh2 er, but the antiquity of this form is uncertain (see discussion of the
corresponding secondary ending below).
The PIE primary [Link]. ending *–th2 er is directly reflected in Hitt. –(t)ta(ri), TB/A –tar/tār, and (in media
tantum verbs) OIr. –ther. The other IE languages continue an ending *–soi, with initial *s taken from the [Link].
*m-conjugation ending and renewal of *–r by *–i, e.g. Ved. –se, OAv. –hē/–šē, Myc./Arc.-Cypr. Gk. –soi (in other
dialects, –sai with vocalism after 1sg. –mai), Goth. (pass.) –za.
Two primary [Link]. endings are securely reconstructible for PIE, *–or and *–tor (cf. §4.2.3 above). The
archaic *–or allomorph is preserved in CLuw. ziy-ar(i) ‘lies’, Hitt. paḫš-a(ri) ‘protects’, OIr. ber-air ‘is carried’ and
— with renewal of the tense marker in Indo-Iranian — Ved. śáy-e ‘lies’, OAv. sruii-ē ‘is heard’ (see further Jasanoff
2003a:49–51). The productive allomorph *–tor — with analogical *t from the *m-conjugation [Link]. ending —
is also attested in the same languages (e.g. CLuw. puppušša-tari ‘is crushed’, Hitt. ki-tta(ri) ‘lies’ (cf. Pal. kī-tar ‘id.’),
OIr. sechi-thir ‘follows’), in some cases, even in the same lexical items (late Ved. śé-te, YAv. saē-te ‘lies’); the latter, in
particular, show the strong tendency for *–or to be morphologically renewed by *–tor, a pattern that likely began
in PIE itself and led eventually to the complete elimination of *–or in other NIE languages, which have only *–tor:
Lat. sequi-tur ‘follows, TB wike-tär ‘disappears’, Cypr. Gk. ke-i-to-i [kei-toi] ‘lies’ (cf. Att.-Ion. Gk. keî-tai ‘id.’ with
vocalism after 1sg. –mai).
The PIE secondary athematic [Link]. ending was likely *–h2 e, which is directly reflected in Hitt. –(ḫ)ḫat(i),
e.g. Hitt. ēš-ḫa-t(i) ‘I sat down’ (with further addition of a reflexive particle *–di, on which see Yakubovich 2010:182–
205); it may also be maintained, as an archaism, as the ending of optative forms in Indo-Iranian (PIIr. *–a, e.g. Ved.
sac-ey-a ‘may I accompany’, OAv. vāur-aii-ā ‘may I cover’). Elsewhere, the Indo-Iranian languages show endings
(Ved. –i, OAv. –ī < PIIr. *–i), which have been argued to derive from a shorter ending *–h2 (e.g. Kortlandt 1981,
García Ramón 1985); however, it is more likely that PIIr. *–i should be explained analogically (see Kümmel, this
volume). The Tocharian preterite endings (TB –mai; TA –we/–e) probably also contain *–h2 e; see the discussions
of Malzahn (2010:44–5 with references) and Pinault, this volume.
Less certain is the PIE status of an “iterated” allomorph of the [Link]. ending –*h2 eh2 e, which appears to
be continued in both Hittite (e.g. ēš-ḫaḫat(i) ‘id.’) and Lycian (a-xagã ‘I became’; see Melchert 1992a). Potential
evidence for its deeper reconstruction comes from Greek, where it has been suggested that the same form underlies
(non-Attic-Ionic) Gk. –mān < *–m-h2 eh2 e-m with analogical remodeling after the [Link]. *m-conjugation ending
(Weiss 2011:388–9; but cf. the critique of Yoshida 2010, 2013).
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 54

The PIE secondary [Link]. ending *–th2 e, is continued — with different additional morphological material
in each language — in Hitt. –(t)tat(i) (+ reflexive *–di; cf. 1sg. above), Ved. –thās, and TB/A –tai/–te, as well as OIr.
–tha. Other IE languages have replaced *–th2 e with *–so, an analogical form with the initial *s of the m-conjugation
2s. active ending and the vocalism of [Link]. *–(t)o(r); *–so is reflected in Gk. –so, OLat. –re (on Cl. Lat. –ris, see
Weiss 2011:388–91), and in the Iranian languages (OAv. –šā, OP –šā; on the split within Indo-Iranian, see Kümmel,
this volume).
Just as in the corresponding primary form, two allomorphs of the athematic [Link]. secondary ending are
reconstructible for PIE, *–o and *–to. Archaic *–o is maintained in Hitt. ēš-at(i) ‘sat down’ (with reflexive *–di; cf.
1sg/2sg. above), and famously, in Ved. á-śay-at ‘was lying down’ (with analogical final *t; Wackernagel 1907:309–13
(= 1953a:498–502)). Once again, the same languages also reflect productive *–to, including in forms of the same
lexical items attested in chronologically younger texts: Hitt. ēš-tat; late Ved. (a)śe-ta (cf. YAv. sae-ta). In other NIE
languages, older *–o has been wholly ousted by younger *–to: Gk. –to, Iranian (OAv. –tā, OP –tā), TB/A –te/–t.
The PIE [Link]. ending was *–medh h2 or *–mesdh h2 ; it is possible that one of these forms was once specialized
as the primary ending and the other as secondary, but if so, the daughter languages provide no clear evidence for
the original distribution. Support for reconstructing *–medh h2 comes from Gk. –meth a, as well as Tocharian and
Indo-Iranian; in the latter two, a distinction has been introduced between primary (Ved. –mahe, OAv. –maidē <
PIIr. *–madh ai; TB/A –mtär) and secondary forms (Ved. –mahe, OAv. –maidī < PIIr. *–madh i; TB/A –mte/–mät;
see Kümmel, this volume and Malzahn 2010:37, 46). However, Greek also attests a variant *–mesth a, which points
to *–mesdh h2 ; an *s is also found in the same position in Hittite, which has — like Indo-Iranian — differentiated
primary –wašta(ri) and secondary –waštat(i) (using the same morphological material as in the singular; see above).
As in the [Link]. (cf. §4.2.5), the initial *w of the Hittite form is usually attributed to the influence of the dual (see
§4.2.2 above), either directly (i.e. < [Link]. *–we/o(s)- + –d(h) h2 of the [Link].; cf. §4.2.5 above) or else by analogy
with the [Link]. ending Hitt. –w(/m)en(i).
The reconstruction of the PIE [Link]. ending is problematic. The [Link]. endings attested in the NIE lan-
guages (with the possible exception of Tocharian) can be derived straightforwardly from an ending *–dh we, likely
undifferentiated for primary/secondary as in Gk. –sth e (with s generalized from coronal-final roots, where it is
phonologically regular via the “Double Dental” rule; see Byrd, this volume). As in the [Link]., Indo-Iranian has
introduced the primary/secondary distinction; furthermore, the attested endings appear to continue *–dh uwe, a
variant of the ending conditioned by Siever’s Law (on which see Barber 2013 and Byrd, this volume): 1ry /2ry Ved.
–dhve/–dhvam (with frequent disyllabic scansion); OAv. –duiiē/OAv. –dūm (YAv. –θβe/–θβəm). The same phono-
logically conditioned variant underlies Cl. Arm. –(a)rukc (Jasanoff 1979:144–5), synchronically the 2pl. aorist pas-
sive ending. More difficult is Tocharian, where there is a clear split between (1ry /2ry ) TB *–tär/–t and TA –cär/–c;
there is no consensus about whether either ending is the phonologically expected outcome of *–dh we, but most
scholars agree that both are ultimately based on *–dh we (see Malzahn 2010:37–8 with references).
The deeper PIE situation is problematized by the endings attested in the Anatolian languages: (1ry /2ry ) Hitt. –
ttuma(ri)/–dumat (on m < *w, see §4.2.2 above), CLuw. –(d)duwar(i). The principal issue is that the initial gemi-
nate (or “fortis”) stop (Hitt. –tt–) cannot be the outcome of *dh . Melchert’s (1984:26) alternative derivation of the
ending from PIE *–dh h2 we explains the geminate stop, and in addition, accounts more neatly (i.e. without appeal
to Siever’s Law) for the post-consonantal anaptyctic u vowel clearly observed in the Hittite form (cf. Melchert
1994a:57–8, 77–8); however, whether *–dh h2 we can be reconciled phonologically with the NIE evidence remains
to be systematically assessed. A different solution is proposed by Jasanoff (2003a), who suggests that Anatolian
replaced ending-initial *dh with *t by analogy to the [Link]. (m-conjugation) ending *–te.
For the PIE primary [Link]. ending — like the corresponding singular — two allomorphs are reconstructible,
likely *–ror and *–ntor. The older ending *–ror is not continued as such in any IE language, but is in all probability
the source of PIIr. *–rai (> Ved. –re, YAv. –re), which would be derived by the across the board replacement of the
inherited middle tense marker *r by active *i in that branch; PIIr. *–rai is selected by the same set of verbs that
take the archaic [Link]. ending *–ai (< PIE *–or; cf. §4.2.3 above), e.g. Ved. duh-ré ‘give milk’, śé-re ‘lie’ (= YAv.
sōi-re/saē-re), and in the [Link]., e.g. Ved. jajñi-re ‘are born’. The endings attested in the other IE languages
and elsewhere in Indo-Iranian all derive from productive *–ntor: Hitt. –anta(ri), Arc-Cyp. Gk. –ntoi (–ntai in other
dialects with analogical vocalism), Ved. –ate (< *–n̥ toi; cf. thematic –ante), TB/A –ntär, Goth. –nda, OIr. –tir, Lat.
–ntur (but see Weiss 2011:390–1 on the complicated Italic evidence; alternative view in Clackson and Horrocks
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 55

2007:33).
Similarly, the PIE secondary [Link]. ending has two reconstructible allomorphs, *–ro and *–nto. Archaic *–ro
is continued in Ved. –ran/–ram (with added final nasal), which marks imperfects corresponding to presents in –re,
e.g. Ved. á-śe-ran ‘were lying’, á-duh-ran ‘were giving milk’, as well as 3pl. forms of the aorist “passive,” e.g. á-dr̥ ś-ran
‘were seen’, á-budh-ram ‘woke up’ (on the development of this category in Indo-Iranian, cf. Kümmel 1996, Jasanoff
2003a:153–73, 206–10). Productive *–nto yields Hitt. –antat(i), Gk. –nto, Ved. –ata (< *–n̥ to; cf. thematic –anta),
and TB/A –nte/–nt.
No secure reconstruction of dual middle endings is possible. The IE languages that preserve the dual, above
all Vedic and Greek, have dual middle endings that cannot be traced back common pre-forms; rather, the attested
endings generally appear to be created by combining features of the m-conjugation active dual endings and inher-
ited middle plural endings — for instance, [Link]. Ved. –vahe amalgamates [Link]. –vas and [Link]. –mahe,
while [Link]. Gk. –sth on mixes [Link]. –ton and [Link]. –sth e.

§4.3 PNIE verbal stem formation A tripartite division of tense-aspect stems into “present” (imperfective as-
pect), “aorist” (perfective aspect), and “perfect” (resultative-stative) is reconstructible for PNIE. Only Greek and
Indo-Iranian exhibit this three-fold distinction directly, but it underlies other PNIE languages which have merged
the aorist and the perfect, e.g. Latin. The present stem could be inflected in present and past tenses (the latter
called the “imperfect”). For example, to the root *gwh en– ‘smash; slay’ could be formed the [Link]. *gwh én-
ti ‘smashes, slays’ (Ved. hán-ti ‘id.’), [Link]. *gwh en-t ‘was smashing’ (Ved. (á)-han). The aorist stem expressed
perfective aspect and could be used in the indicative only to refer to past tense, e.g. *weg̑ h – ‘convey, move’ ⇒
*wēg̑h -s-t ‘conveyed, moved’ (> Lat. vēxit). Thus the perfective/ imperfective distinction is overlaid with a past /
non-past distinction only in the imperfective stem. The perfect stood apart from the present and aorist on for-
mal and functional grounds in ways we will discuss below; an example of a perfect is Ved. ca-kár-a ‘I have made, I
made’ ([Link].) < PNIE *kw e-kw ór-h2 e. Bybee and Dahl (1989) survey tense-aspect stems cross-linguistically,
from which the tripartite system reconstructed for PNIE emerges as commonest in the languages of the world; see
further Wackernagel (1926-8 [2009]:195–268) for an overview of tense-aspect in several ancient IE languages with
copious examples and references.
It is important to distinguish between various uses of the term “aspect”. We will use the term “grammatical
aspect” for the grammatical means by which a speaker expresses views on the action of the verb (such as ongoing,
imperfective or as a complete whole, perfective). Grammatical aspect is conveyed by the morphology of the verb.
We will use the term “lexical aspect” for what is considered the inherent, unmodified lexical meaning of the verbal
root; often this notion goes under “Aktionsart” in IE studies (the term is fairly elastic and may refer to other phe-
nomena as well, cf. Napoli 2006:45–51). In PNIE the assignment of a verbal root to the present or aorist stem was
related to the verb’s lexical aspect (cf. Hoffmann 1970, Strunk 1994). Basically, if the root was telic or “punctual” it
would be assigned to the aorist stem, if atelic it would be assigned to the present stem. Thus lexically telic roots
like *deh3 – ‘give’, *dh eh1 – ‘put, place’, and *mer– ‘die’ all made root aorists as their basic formation (e.g. *deh3 -t
‘gave’ > Ved. (a)-dāt). Atelic roots like *bh eh2 – ‘speak’, *h1 ed– ‘eat’, *h1 ey– ‘go’ all made root presents as their basic
formation (e.g. *bh eh2 -ti ‘he speaks’ > Gk. ph ēsí). A root with telic lexical aspect could derive a stem with atelic
grammatical aspect (i.e. the “present” stem) via affixation — for instance, *deh3 – ‘give’ formed a reduplicated
present *de-deh3 -ti ‘gives, is giving’. A number of different derivational affixes may derive present stems, includ-
ing a thematic vowel added to the root (*bh ér-e-ti ‘bears’ > Ved. bhár-a-ti) and a nasal-infix inserted into the root
(*yeug– ‘yoke’ forms *yu-né-g-ti ‘yokes’ > Ved. yu-ná-k-ti). Vice-versa, a root with atelic lexical aspect could derive a
stem with perfective grammatical aspect (i.e. the “aorist” stem) via affixation — most commonly, by suffixing *–s–
to the root; for instance, *weg̑ h – ‘convey, move’ makes the aorist stem *wēg̑ h -s-t ‘conveyed, moved’ (> Lat. vēxit).
This neat picture is, however, disturbed by numerous mismatches between semantics and root formation. For
instance, one notorious example is the root *gwh en– ‘kill, slay’, of prominent use in the Indo-European dragon-
slaying myth (Watkins 1995). Given its meaning ‘slay, kill’ one might expect a root aorist, yet it forms a root present
in PIE, *gwh én-ti (e.g. Ved. hán-ti). García Ramón (1998) proposes that the root originally meant ‘(repeatedly)
strike’, thus bringing into better accord semantics and stem formation. Similarly, lexically atelic *peh3 – ‘drink’
forms a root aorist, not a root present as would be predicted; here too it is surmised that *peh3 – originally had
a more telic meaning in line with its root aorist formation, i.e. *‘take a gulp’. In the end, a number of stubborn
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 56

mismatches between lexical aspect and stem formation remain.

Two further divisions of aspect must be mentioned. The first is “predicational aspect,” where aspect interacts
with syntax. For instance, lexical aspect may be changed in the presence or absence of additional arguments (e.g.
imperfective John reads a lot vs. perfective John reads a book). This domain has proven fruitful for understanding
the individual daughter languages (cf. e.g. Napoli 2006:85-128 on Homeric Greek) and future research will likely
cast light on the PIE verb; it is, however, situated more in the syntax, so we will leave off further discussion of it here.
Secondly, the more developed notion of “state-of-affairs” (or “actionality”) is sometimes used in Indo-European
studies to describe the types of situation a verb may express (following the seminal work by Vendler 1967). To
illustrate using Ancient Greek, where the PIE situation is often thought best preserved, many studies depart from
a first order distinction between verbs expressing states vs. dynamic situations (cf. Napoli (2006, 2015), and the
overview by George 2014, both with references). States include (e.g.) eînai ‘to be’, ékh ein ‘to have’, keîsthai ‘to lie’.
Dynamic verbs may be either telic or atelic. If the verbal eventuality is durative (i.e. persists through time), the telic
verb is called an “accomplishment” (e.g. manthánein ‘to learn’, poieîn ‘to create, make’); if it occurs instantaneously,
the telic verb is called an “achievement” (e.g. apokteínein ‘kill’). Atelic verbs are called “activities” if durative, as
with e.g. verbs of motion (phérein ‘carry’). Here too further research may shed light on the structure of the PIE
verb (cf. e.g. Dahl 2010 on Vedic; Weiss 2011:377–98 gives an overview on PIE).
Whether and to what extent the PNIE system also underlies Anatolian (and is thus of PIE age) is debated,
since the Anatolian verbal system shows no obvious trace of grammatical aspect. In the Anatolian languages all
finite and non-finite verbal forms are based on a single stem. Many of these stems are formed by suffixes that derive
imperfective stems in the PNIE languages — for instance, the suffix *–sk̑é/ó– makes stems in various NIE languages
with the aspectual value [imperfective] (e.g. Ved. gáchati ‘goes’ << *gw m̥ -sk̑é-ti), but its Hittite reflex –ške-(z)zi
modifies the lexical meaning of the verbal stem, indicating that iteration, pluractionality, or related notion is a
property of the event. The mere fact that PNIE has so many affixes all deriving the same functions ([imperfective,
perfective]) suggests a merger of categories; at an earlier stage the suffixes would have marked varieties of lexical
aspect, and it has been proposed that this stage underlies and is reflected by Anatolian (cf. Cowgill 2006:37–68 [=
1974b, 1979b]; Strunk 1994). Melchert (1997) contests this finding, arguing that Anatolian might have inherited a
prehistoric contrast in grammatical aspect. He points to Hittite and Luwian verbs reflecting the suffix *–ye/o– (see
4.3.1 below) exclusively in the present stem vs. a stem lacking *–ye/o– in the preterite (e.g. Hitt. npst. karp(i)ye–
‘lift’ beside pst. karp–). Thus Anatolian would have inherited PIE *–ye/o– as an imperfective formant confined to
the present system beside a perfective stem (i.e. a root aorist). But the question remains an open one due to the
paucity of evidence (see Melchert fthcm. a for a recent assessment of the Anatolian data).
One further means of instantiating the imperfective vs. perfective contrast should be noted here: stem sup-
pletion. The notion of “suppletion” is a fraught one, since what defines suppletion cross-linguistically has been
disputed (Veselinova (2003, 2013) is helpful for orientation and discussion). For present purposes, by “suppletion”
we mean the process whereby regular semantic relations are encoded by unpredictable formal means. In terms
of verbal suppletion according to tense and aspect, this will mean that one root is used for one tense-aspect stem
(e.g. present), a separate root is used to form another stem (e.g. aorist). For example, in numerous IE languages
reflexes of the present stem *bh éreti ‘bears’ have only a suppletive perfective, giving well known pairs like Gk. ph érō
: ḗnegkon, Lat. ferō : tulī, TB/A pär– : kām–. On suppletion in PIE, see García Ramón 2002 and Kölligan’s (2007b)
recent study of the Greek evidence (with particular attention to diachrony).
The basic architecture of the PNIE system of present and aorist stems is exemplified in (27):

present/imperfective stem aorist/perfective stem

present imperfect aorist
*de-deh3 -ti ‘gives’ *de-deh3 -t ‘was giving’ *deh3 -t ‘gave’
(27)
*dh e-dh eh1 -ti ‘places’ *dh e-dh eh1 -t ‘was placing’ *dh eh1 -t ‘placed’
*bh er-e-ti ‘bears’ *bh er-e-t ‘was bearing’ [suppletive]
*weg̑ h -e-ti ‘conveys’ *weg̑ h -e-t ‘was conveying’ *wēg̑h -s-t ‘conveyed’
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 57

§4.3.1 Imperfective stem formation Present (imperfective) stems show a wide variety of formations and we of-
fer here an abbreviated catalogue of verbal stem types, formally divided between athematic and thematic, therein
divided between “primary” formations (made to verbal roots) and “secondary” formations (derived verbal stems).
Our catalogue aims to be a descriptive overview of some present types reconstructible from the IE daughter lan-
guages, with the caveat expressed about the role of these suffixes in Anatolian. We will list the reckoning from LIV 2
for how many roots build a given formation, often followed by how many examples are considered “secure” by the
authors. We do not accept the analysis of LIV 2 in every instance: the numbers are provided merely as a rough
guide to current thinking in the field. Fuller inventories of verbal stem types may be found in Jasanoff (fthcm. a),
Meier-Brügger (2010:297–311) (based squarely on LIV 2 ), and Beekes and de Vaan (2011:251–86) (representative of
Leiden views of the verb, which differ in many ways from those presented here).
Root presents are formed by adding the endings to the root without overt affixation; in LIV 2 such a present
is listed for about 200 roots. Examples include 3sg. *h1 és-ti ‘is’, 3pl. *h1 s-énti ‘are’ (Hitt. ēš-zi, aš-anzi, Ved. ás-ti,
s-ánti, etc.); 3sg. *h1 éy-ti ‘goes, walks’, 3pl. *h1 y-énti (CLuw. ī-ti, Ved. é-ti, etc.). A number of prominent media
tantum are root presents, e.g. PIE *k̑éy-or ‘lies’ (CLuw. ziyar(i), Ved. śáye), and *wés-(t)or ‘clothes oneself, wears’
(Hitt. wešta, Ved. váste, Gk. héstai) (cf. §4.2). A controversial subtype is the “Narten” (or lengthened-grade)
present, named in honor of Johanna Narten’s work from the 1960s. This type showed *ḗ-grade in the singular,
*é-grade plural, for which the prime example is 3sg. *stḗw-ti, 3pl. *stéw-n̥ ti ‘praises’ (> Ved. stáuti, etc.). The
lengthened-grade in these root presents reflects a derived present type. Some examples form imperfective stems
to root aorists: Kümmel (1998) gives (e.g.) *dḗk̑- / dék̑- ‘to expect, accept’ (Ved. root dāś-, 3sg. dāṣ-ṭi ‘serve religiously’
via a semantic development of Vedic) beside the root aorist *dék̑- (Gk. [Link]. dék-to ‘received’). Other examples
are arguably formed to root presents: Melchert (2014b) gives (e.g.) *h1 ḗs-ti, *h1 és-n̥ ti ‘sits’ (OHitt. ēš-zi ‘is sitting’)
to the aforementioned root present *h1 és-ti ‘is’. The formation likely had an earlier aspectual nuance; Melchert
suggests iterative-durative.
molō-presents: Another kind of PIE root present had *o/e-ablaut in the root and — according to a still contro-
versial proposal by Jasanoff (2003a:64-90) — inflected with the perfect-like endings of the *h2 e-conjugation (on
which see §4.2.6 above). The verbs constituting this class are typically those of vigorous activity, such as PIE 3sg.
*bh ódh h1 -ei ‘digs’ (e.g. OCS bodǫ ‘I stab’, Lith. bedù ‘I poke’ beside Hitt. paddai ‘digs’). Jasanoff names the class
“molō-presents” after the Lat. outcome molō ‘I grind’, whose cognates give evidence for both *o-grade vocalism of
the root (e.g. Go. malan ‘to grind’, Lith. malù ‘I grind’, both with a < *o) and *e-grade (e.g. OIr. melid ‘grinds’, OCS
meljǫ ‘I grind’). As in the noun, these diverse ablaut grades suggest bifurcated levelings of a once unitary paradigm
*mólh2 –/*mélh2 –. Hittite arguably provides direct evidence for such a unitary paradigm in the ḫi-conjugation, a
class that includes the cognate verb (3sg.) Hitt. mall-(a)i ‘grinds’; although the original weak stem root vocalism
of this verb is obscured by sound change (3pl. mall-anzi), Hittite preserves *ó/é-ablaut in a recessive sub-class
of ā/e-ablauting ḫi-verbs, e.g. k(a)rāp-/k(a)rep– ‘devour’ (< PIE *gh róbh –/*gh rébh – ‘seize’), š(a)rāp–/š(a)rep– ‘sip’
(< *sróbh –/*srébh – ‘id.’). Kloekhorst (2012, 2014) disputes this evidence, arguing that the ḫi-conjugation in Hittite
reflects only *o/∅-ablaut, but his alternative inner-Hittite derivation of the weak stem e-vocalism of this class can-
not be maintained — for instance, the root e-vowel in the verbs cited above cannot be epenthetic, since there is
no plausible phonological or morphological motivation for epenthesis in this environment (see Melchert 2013; cf.
Yates 2015:154–5, 166 n. 43).
Reduplicated athematic presents: Partial copy reduplication is another major device for forming present
stems to root aorists. Two types of reduplicated presents may be formally distinguished, an athematic and a the-
matic (treated below). The athematic type is well attested in Greek and Indo-Iranian, but with formal differences
— in particular, in the vocalism of the reduplicant — that problematize its reconstruction. In Greek, all redupli-
cated presents have fixed i-segmentism in the reduplicant, e.g. WGk. hí-stā-mi ‘I stand’ (beside root aorist stem
WGk. stā–), tí-th ē-mi ‘I place’ (beside root aorist stem thē–). In contrast, Indo-Iranian has reduplicated presents
with i-, a- (< *e) and even u-vocalism of the reduplicant, e.g. Ved. í-yar-ti ‘moves’ (beside root aorist stem ar–
); Ved. dá-dhā-ti (= OAv. da-dāi-tī) ‘places’ (beside root aorist stem dhā–); Ved. ju-hó-ti ‘pours’. While the last
type, which occurs only when the verbal root contains u, is generally regarded as an innovation, both *e- and
*i-reduplicated forms are usually viewed as inherited — for instance, LIV 2 reconstructs two distinct athematic
reduplicated presents for PIE, one with *e-fixed segmentism, another with *i. Yet while this “maximal” reconstruc-
tion is possible, it still does not straightforwardly account for the mismatch between Vedic and Greek in cognate
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 58

lexical items (e.g. Gk. tí-th ē-mi, Ved. dá-dhā-mi < PIE *dV́ -dh eh1 -mi ‘I place’), or for the fact that a few roots are
attested in Indo-Iranian with both *e- and *i-reduplicated forms, e.g. Ved. 3sg. sí-ṣak-ti (= YAv. hišhaxti) vs. 3pl.
sá-śc-ati (: sac– ‘accompany’), Ved. 3sg. jí-gāt-i vs. fossilized [Link]. já-g-at– ‘(moving) world’ (: gā– ‘go’).
Various other interpretations of this evidence have been advanced. Jasanoff (2003a:128–32) contends that PIE had
only *e-reduplicated presents in the *m-conjugation, arguing that *i-reduplicated athematic presents in Greek
and Vedic are due to the analogical influence of PNIE thematic *i-reduplicated presents, which would ultimately
derive from PIE *h2 e-conjugation *i-reduplicated forms (see below). Another possibility — proposed already by
Hirt (1900:190–3) and further developed in recent scholarship (Sandell 2011; Hill and Frotscher 2012) — is that
all athematic presents descend from a single PIE paradigm in which the reduplicant had two allomorphs, one
with *e-vocalism and one with *i-vocalism; this intraparadigmatic allomorphy would then have been leveled out
separately in the individual languages. Dempsey (2015:339–41) suggests that this hypothesis better explains the
situation in Anatolian, where reduplicated *h2 e-conjugation verbs may have either fixed *e- or *i-segmentism in
the reduplicant (with no corresponding functional difference) — e.g. Hitt. we-wakk-i (: wek– ‘demand’) vs. Hitt.
li-lḫuwa-i (: laḫ(ḫ)u– ‘pour’). However, there is not yet scholarly consensus on this issue.
Nasal-infix presents: An ablauting nasal-infix *–ne/n– is one of the commonest means for making present
stems to root aorists: in LIV 2 it is reconstructed for 248 roots (168 secure). An example is the root *yeug– ‘yoke’:
the infix is inserted after the first syllable of the (zero-grade) root to derive a present 3sg. *yu-né-g-ti, 3pl.*yu-n-g-
énti ‘yokes’ (> Ved. yu-ná-k-ti, yu-ñ-j-ánti), beside the root aorist *yeug-t (> OAv. yaogət; cf. 1sg. Ved. yójam). The
formation is well attested across a number of branches and is traditionally divided into three varieties based on
the consonantal quality of the final segment of the root into which *–né/n– was inserted: (i) a final obstruent, e.g.
aforementioned *yu-né-g-ti; (ii) final laryngeal, *kw reyh2 - ‘buy’ > *kw ri-né-h2 -ti ‘buys’ (Ved. krī-ṇā-ti, TB [Link].
kärn-ās-tär); or (iii) glide *–w–, e.g. *k̑lew- ‘hear’ > *k̑l ̥-né-w-ti, *k̑l ̥-n-w-énti ‘hears’ (Ved. śr̥-ṇó-ti). The sequence
*–n(e)w– was reinterpreted as a suffix already in PIE and added suffixally (not infixally) to roots, e.g. *str̥-néw-ti
‘strews’ (Ved. str̥-ṇó-ti). Although in the NIE languages it is mainly attested as a present stem formant beside root
aorists (cf. Strunk 1967), there is some evidence to suggest that the infix may have earlier had a valency-increasing
role. The infix is clearly transitivizing in pairs like (transitive) Hitt. ḫar-ni(n)-k– ‘kill’ (also ḫarg(a)nu– ‘id.’) beside
(unaccusative) ḫark– ‘die’. In at least one case there is comparative evidence for a transitive/causative nasal-infix
verb derived from an adjective: Hitt. tep-nu-zi ‘belittles’ and Ved. dabh-nó-ti ‘deceives’ (cf. 2pl. OAv. dəbənaotā)
directly reflect PIE *dh ebh -né-u-ti ‘belittles’ (from *dh ebh -ú– ‘little, small’). Moreover, the related nasal suffix PIE
*–n(é)w– is highly productive in valency-increasing derivation in the Anatolian languages, e.g. Hitt. link– ‘swear’ :
ling(a)nu– ‘make swear’; HLuw. ta– ‘stand’ : tanu(wa)– ‘make stand’ (cf. Luraghi 2012). Accordingly Meiser (1993)
has argued that the nasal infix was originally valency-increasing and only secondarily used as a means for deriving
present stems. It is, however, noteworthy that higher transitivity aligns cross-linguistically with perfective, not
imperfective, aspect (see Hopper and Thompson 1980); the nasal-infix should thus be expected to derive a PIE
aorist, not present, stem (see too Clackson 2007:151–5).
*–eh1 -stative/fientives: Presents formed with *–eh1 -ye/o– make stative as well as change of state verbs across
a wide swath of IE languages. Such presents are sometimes made to a verbal root (e.g. Lat. hab-ē-re ‘to have’, OCS
bǔd-ě-ti ‘to be awake’, Lith. bud-ė-́ ti ‘to be awake’) and are sometimes deadjectival (e.g. Hitt. marš-e-zzi ‘be false’ to
marš-a–, Lith. sen-ė-́ ti ‘to grow old’ to sẽn-as). The deadjectival forms have been derived from “Caland” adjectives
since Watkins (1971). Greek has present forms reflecting the *–eh1 -stative (type tharséō ‘am bold’; cf. Tucker 1990)
but additionally the intransitive (“passive”) aorist is formed with *–eh1 – (e.g. e-mán-ē ‘went mad’), which is hard
to square with the evidence from the other languages. Harðarson (1998) posits that *–eh1 – formations were at
home in the aorist (privileging the Greek evidence) and calls the type “fientive” (i.e. change of state) meaning ‘to
become X’; presents to the fientive would be derived via further suffixation as *–h1 -ye/o–, named “essives,” which
some languages reformed as *–eh1 -ye/o–. This account was taken over wholesale by the influential LIV 2 . The
categories “essive” and “fientive” are both rejected by Jasanoff (2003b), in part on the phonological grounds that
*–h1 -ye/o– would infringe “Pinault’s Law” (cf. Byrd, this volume; note, though, that Byrd suggests restricting the
law to *h2 , *h3 ). Jasanoff reconstructs instead a suffix *–eh1 -ye/o–, which he derives from the predicatively used
[Link]. of a root noun in *–eh1 , e.g. *h1 rudh -éh1 ‘with redness’ > *h1 rudh -eh1 -yé/ó– ‘be(come) with redness, blush’
( > Lat. rub-ē-re ‘to be red, ruddy’). On the basis of the reanalyzed stative stem the daughter languages created or
extended other formations including: change of state verbs in *–eh1 -s– in Hittite; verbal abstracts (infinitives) in *–
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 59

eh1 -ti– in Balto-Slavic; and intransitive aorists in bare *–eh1 – in Greek. The matter has not been settled: Yakubovich
(2013b) presents an overview of the problem; Bozzone (2016) builds on Jasanoff’s scenario, with further typological
considerations.
*–h2 -factitives (the “newaḫḫi”-type): When added to thematic adjectives, the factitive suffix *–h2 – derives
transitive verbs. Examples include the class’s eponymous Hitt. newa-ḫḫ-i ‘make something new’ (< *newe-h2 -ei
‘new’; cf. Hitt. nēwa– ‘new’). Other languages probably reflect the *–h2 -suffix only in its extended form *–h2 -
ye/o–; for instance, the extra-Anatolian comparanda for newaḫḫi include Lat. nou-ā-re ‘make something new’ and
the rare Gk. verb neáō ‘plough up (fallow land)’ (both from extended *new-e-h2 -ye/o–). The derivation remains
productive in Italic, e.g. Lat. sānus ‘healthy’ ⇒ sānāre ‘heal’, etc.; see further Watkins (1971:61, 85–6) and Jasanoff
(2003a:139–41).
“Simple” thematic presents: Roots with an affixed thematic vowel *–e/o– are a bedrock formation of PNIE;
LIV 2 lists 426 roots (224 secure) that make simple thematic presents. Examples include *bh ér-e-ti ‘bears, carries’
(e.g. Ved. bhárati; cf. §4.2.5), *h2 ég̑-e-ti ‘leads, drives’ (Ved. ájati, Lat. agit, Arm. 1sg. acem, etc.). Simple thematic
presents are often found beside other present types in the daughter languages (e.g. athematic [Link]. Lat. fer-t
‘bears, carries’ < *bh ēr-ti). Jasanoff (1998) has argued that simple thematic presents in the IE languages come from
at least two historically distinct sources, as indicated by their relationships to other present formations, and the
kinds of aorists they co-occur with. The *bh ér-e-ti type occurs beside other present formations (e.g.*bh ēr-ti > Lat.
fer-t) and makes a suppletive aorist (both *bh ér-e-ti and *h2 ég̑e-ti make suppletive aorists). A second type, whose
present is formally identical, is represented by (e.g.) *wég̑h -e-ti ‘conveys’ (Ved. váhati, Lat. vehit etc.), which does
not have competing present formations, and makes its aorist stem with the *s-aorist (*wēg̑h -s-t > Lat. vēxit). This
evidence would indicate that the two thematic present types derive from historically distinct origins, a conclusion
bolstered by their “fit” within the chronology of IE dialects. That is, the *wég̑h eti type does not occur in Anatolian,
whether the *bh éreti type does is disputed. Many researchers find an isolated example of the *bh éreti type in
HLuw. [tammari]* ‘builds’ (in transcription: AEDIFICARE+MI-ra/i + i), which could derive from PIE *dém(h2 )-e-ti
with thematic cognates in Gk. dém-ō ‘build’ and Go. ga-timan ‘fit’ (but cf. Lehrman (1998) for a dissenting view).
The rarity — and possibly complete absence — of both present types in Anatolian is striking and suggests that
both types could represent post-Anatolian innovations. Tocharian knows thematic presents of the *bh éreti type
(Toch. class II presents and subjunctives) but in reduced numbers; arguably the *wég̑ h -e-ti type does not occur
in Tocharian, and therefore represents a PNIE innovation. Ringe (2000) leverages the dearth of such presents
in Anatolian and Tocharian to suggest an early branching off of these languages, a view Malzahn (2010:363-366)
disputes. Fitting the simple thematic type of PNIE into the picture of the earlier PIE verb is an ongoing project.
tudáti-presents: Zero-grade presents with accented thematic vowel — known as “tudáti”-presents after the
canonical class VI present of Sanskrit grammar tudáti ‘strikes’ — are considerably less well-represented than sim-
ple thematic presents; in LIV 2 it is reconstructed for 52 roots (20 secure). Significantly, at least one example of
this class is found in Anatolian: Hitt. šuwe-zzi ‘pushes away, shoves’ forms an equation with Ved. suv-á-ti ‘impels’
and OIr. soïd ‘turns’ < *suhx -é-ti ‘push’ (with Oettinger 1979:279, pace Kloekhorst 2008:797–8). It has often been
thought that this present class, with its preference for markedly telic activities in Vedic, might have developed
from aspectually shifted thematic aorists; the imperfect of the zero-grade present and the thematic aorist are for-
mally identical (e.g. imperfect *suhx -é-t ‘pushed’ and aor. *wid-é-t ‘found’; on the aorist type see below). Because
these presents are held to have their origins in aorists, the class sometimes goes by the unfortunate name “aorist
presents.” The early diachronic development of the tudáti-presents is in need of further investigation (on the Vedic
material see Hill 2007 and now Malzahn 2016). A number of tudáti-presents are made to roots in final -i- in Old
Indic (e.g. sy-á-ti ‘binds’ to root say–/si–, cf. Kulikov 2000); Jasanoff (2003a:105-107) argues that these represent
part of a wider class of presents with an *–i– suffix in the protolanguage.
Thematic reduplicated presents: A thematic reduplicated type is also found beside the athematic type dis-
cussed above. An example is *g̑i-g̑n(h1 )-e-ti > Lat. gi-gn-i-t, Gk. gí-gn-e-tai (deponent mid. beside root aorist
*g̑enh1 -to > Gk. e-géneto). In some cases, thematic reduplicated presents have athematic reduplicated cognates
(see above) in other NIE languages, e.g. Lat. si-st-ō, Ved. tí-ṣṭha-ti vs. WGk. hí-stā-mi (cf. root aorist PIE *stéh2 -t
‘stood’). The etymological equation between thematic reduplicated present Gk. mí-mn-ō ‘I stand fast’ and *h2 e-
conjugation i-reduplicated verb Hitt. mimma-i ‘refuses’ points to a diachronic connection between these cate-
gories, and it has been argued, specifically, that some (if not all) PNIE reduplicated thematic presents arise via
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 60

“thematization” of PIE *h2 e-conjugation *i-reduplicated presents (see esp. Jasanoff 2003a:128–132, García Ramón
2010; cf. §4.2.5).
*–yé/ó-presents: The suffix *–ye/o– is a thematic present formation only (i.e. there is no aorist *–ye/o–). A
prominent type has accented suffix and zero-grade root, many examples of which are deponent, including the
roots of birth and death: *mr̥-yé-tor ‘dies’ > Ved. mri-yá-te, Lat. mor-i-tur and *g̑n̥h1 -yé-tor ‘is born’ > OIr. gain-
i-thir, Ved. jā ́-ya-te, and *mn̥ -yé-tor ‘thinks’ > Ved. mán-ya-te, Gk. maíne-tai ‘rages’. In Indo-Iranian, this suffix,
accented and with middle inflection, becomes specialized as a present passive marker (e.g. 3sg. –yá-te; cf. §4.2);
Kulikov (2012) is an extensive treatment of the Vedic evidence. *–ye/o– is also the normal denominative suffix
forming verbs that mean ‘be, become, act like X’. Examples include Ved. vr̥ ṣā-yá-te ‘acting the bull (vŕ̥ṣan–)’, Gk.
poimaínō ‘I am a herdsman (a poimḗn)’ < *poh2 i-mn̥ -yō (cf. Tucker 1988). A number of primary *–ye/o– presents
give evidence for an accented full grade of the root, such as *(s)pék̑-ye-ti ‘sees, looks at’ ( > Ved. páś-ya-ti); in LIV 2
this full-grade formation is considered a distinct type made to 50 roots (19 secure).
*–sk̑é/ó-presents: The suffix *–sk̑é/ó– with the zero-grade of the root formed thematic presents in PNIE. Ex-
amples include *gw m̥ -sk̑é/ó– ‘be walking’ (Ved. gáchati ‘goes’, [Link]. Gk. báske ‘go!’, TA kumnäṣtär ‘comes’),
and the widespread item *pr̥k-̑ sk̑é– ‘ask’ (Ved. pr̥cháti, Lat. poscit, OIr. –airc). In PNIE the suffix derives present
stems especially to root aorists, with further innovations and extensions defining the daughter languages (see
Zerdin 1999, 2002 on this issue with special reference to Greek). There are, however, sufficient indications to re-
construct its earlier aspectual functions. In Hittite the suffix –ške– derives an aspectual stem whose function can
be iterative, habitual, and pluractional (cf. Hoffner and Melchert 2008:318–22). In Tocharian B, reflexes of the suf-
fix *–sk̑é/ó–, viz. –ṣṣə–/–ske–, form class IX presents (e.g. we-skau, we-ṣṣäṃ ‘say’), but the suffix is mostly used
in the present (and subjunctive) to form the causative — e.g., to the root wik- ‘disappear’ is formed a causative
present 3sg. wikäṣṣäṃ ‘drives away, removes’. Peyrot (2013:515–24) has recently presented new arguments that
the Tocharian A class VIII presents in –s–/–ṣ– of TA (“s-transitives” in his terminology) — traditionally held to
reflect presents in *–s-e/o– – derive via inner-Tocharian changes from the *–sk̑é/ó– suffix as well. This causative
feature is usually understood as an inner Tocharian development (recently Adams 2014 with references). Li and
Whaley (fthcm.) argue on cross-linguistic grounds that there is a grammaticalization cline of intensive > causative
> reciprocal; Tocharian would perhaps fit into this schema. One intriguing detail is that the suffix makes iterative
and durative stems not only in Anatolian but also an iterative preterite in –(e)skon in the Ionic dialect of Greek;
Puhvel (1991:13–20) and Watkins (2001:58–9 [= Watkins 2008:954–5]) plausibly attribute the spread (or rebirth) of
the iterative functions of this suffix to diffusion from Anatolian to the Greek speakers of the Ionic coast.
*–éye/o-causative-iteratives: A thematic formation in *R(o)-éye/o–, making transitive and causative verbs, is
widespread across the languages; in LIV 2 it is reconstructed to 400 roots (237 secure). Examples include *men–
‘think’ > *mon-éye– ‘call to mind’ (> Lat. monēre ‘warn’) and *sed–‘sit’ > *sod-éye- ‘set something’ (> Go. satjan
‘to set, plant’). Two etymological equations set the date of this formation back to PIE antiquity: Hitt. lukke-zzi
‘lights up, sets ablaze’ was taken by Watkins (1971:69) to derive from a causative *louk-éye/o– seen also in e.g. Ved.
rocáyati ‘make shine’, Lat. lūceō, -ēre ‘ignite, light’; and Hitt. waššezzi ‘clothes (someone)’ continues *wos-éye/o-,
to be equated with Ved. vāsáyati, Goth. wasjiþ (PGmc. *waz-jan, also Eng. wear), Alb. vesh, as demonstrated by
Eichner (1969). The formation knows a particularly rich development in its Old Indic avatar the –áya–presents
(extensively studied by Jamison 1983). In certain languages there are also verbs formed with the suffix that have
iterative meaning. Kölligan (2007a) argues that in the case of Latin the distinction depends on the agentivity
of the base verb: if the base is agentive, the derived verb is iterative-intensive; if the base verb is non-agentive,
the derived verb is transitive-causative. It is possible that both meanings of iterativity and transitivity-increasing
were available in the proto-language (see also Kölligan 2004). In some languages the reflexes of *R(o)-éye/o- have
merged with denominal verbs made to *o-grade nominals; Greek is a case in point (discussed in detail by Tucker
1990:123–84).

§4.3.2 Perfective stem formation There were fewer types of aorists — we reconstruct four — but still diversity
is found. As in the present system, the redundancy of four formal markers expressing one functional category
suggests that early mergers define the prehistoric development of the aorist.
Athematic root aorists: Like the athematic root presents, the (secondary) endings are added directly to the
root. Thus *dh eh1 – ‘to place’ formed a root aorist *dh eh1 -t ‘placed, put down’, reflected in Ved. dhā-t, Gk. é-th ē-k-e
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 61

(whose older k-less form is preserved in Boeot. Gk. (an)-é-th ē). Root aorists typically form their present stems by
further affixation; Gk. é-th ēke is the root aorist to the reduplicated present títh ēmi ‘I place, set something’. PNIE
root aorists show up in Anatolian as stems that make presential forms; thus beside the inherited root aorist *dh eh1 -t
‘placed, put down’ (> Hitt. tēt ‘said’) are attested Hitt. tē-zzi ‘says’ and Lyc. ta-di ‘puts’, and beside the root aorist
*kw er-t (> Ved. (a)kar ‘made’) is found Hitt. kuer-zi, kuranzi ‘cut(s)’ and CLuw. kuwarti, kur– ‘id.’. The Anatolian
forms are usually explained as innovations, when old aorists were retrofitted with new primary endings, in this case
*dh éh1 -t-i ‘places’; Malzahn (2010:267–8, et passim) calls this process of morphological renewal the “tēzzi-principle.”
*s-aorists: Athematic *s-suffixed aorists (“sigmatic aorists”) are a widespread aorist type in PNIE. The *s-aorist
and its offshoots make up the most productive aorist type in Greek, Indo-Iranian, and Slavic (although it is no-
tably absent from Baltic); furthermore, relics are uncontroversially found in Latin, Celtic, and elsewhere. From
the PNIE languages, a formation with lengthened grade root and secondary endings may be reconstructed; e.g.
the root *weg̑ h – ‘convey, move’ forms an *s-aorist *wḗg̑ h -s-t (Lat. vēxit ‘conveyed’, Ved. ávāṭ, etc.). Despite this
agreement between the NIE languages, reconstructing the *s-aorist for PIE — including Tocharian and Anatolian
— is beset with difficulties. Some connection of the Tocharian s-preterite (pret. class III) with the PNIE *s-aorist
is universally accepted; the nature of that connection, however, remains elusive. Essentially the following three
positions have been advanced: (i) the Tocharian s-preterite derives wholly from the *s-aorist; (ii) it represents in-
stead a conflation to some extent with the PIE perfect; or (iii) it reflects an ancestor of the PNIE aorist, namely
a “pre-sigmatic aorist” (see the review of literature in Malzahn 2010:208–14). No proposal has yet won universal
accord; recent investigations of this problem may be found in the volume edited by Malzahn et al (2015), espe-
cially the contributions therein by Kim, Melchert, and Oettinger (all against the pre-sigmatic aorist). Even more
difficult to pin down is the prehistory of Anatolian. There is widespread agreement that the Hittite preterite third
singulars of the ḫi-conjugation like nai-š ‘turned’ and the *s-aorist (cf. to the same root Ved. á-nāi-ṣ-am ‘I led’) are
historically related (from very different viewpoints see Oettinger 1979:405 and Jasanoff 2003a:174–214), but there
is no agreement on what that relationship is. Jasanoff’s innovative proposal (for which see already Jasanoff 1988b,
and also his account in this volume) has not won general acceptance (as witnessed by the critical remarks of Kim
2005:194 and Oettinger 2006:43–4, i.a.) and the issue remains unsettled at present. Further studies on the devel-
opments of the *s-aorist in the ancient Indo-European languages include Drinka (1995), Narten (1964) on Vedic,
Schumacher (2004) on Celtic, and Ackermann (2014) on Slavic.
Reduplicated thematic aorist: The reduplicated thematic aorist is not widely attested but the examples look
old; LIV 2 reconstructs it for only 18 roots (5 secure). Examples include the root *wekw – ‘say’, which makes a redu-
plicated aorist *we-ukw -e-t ‘said’ ( > Ved. voc-a-t, Av. -vaocaṯ, Gk. (w)eîp-e), and *werh1 – ‘find’ > *we-wr(h1 )-e/o–
‘found’ (> Gk. heûr-e, OIr. fo-fuair). Willi (2007) argues that the reduplication seen in the reduplicated aorist was a
marker of aspectual perfectivity in PIE. Besides Indo-Iranian examples like Ved. vocat ‘said’ (< *we-wkw -e-t), there
is also attested in Vedic a reduplicated preterite regularly aligned with the -áya- transitives (discussed above under
*–éye/o-presents), e.g. Ved. darś-áya-ti ‘shows, makes see’ beside the aorist a-dī-dr̥ ś-a-t. The fact that the redupli-
cant in this class regularly contains the vowels -i-, -u- (not -a-) leads Jamison (1983:216–9) to argue that it derives
from a different historical source than the PIE reduplicated aorist, viz. imperfects to the reduplicated present.
In a number of daughter languages, the reduplicated aorist is valency increasing; Ancient Greek is a case in
point (Duhoux 2000:79–80). Bendahman (1993:61–100, 140–70) finds in Greek about 30 reduplicated aorist stems,
which fall into two types: (i) roots referring to prototypically transitive events with an agentive subject form tran-
sitive reduplicated aorists, *gwh én-ti ‘strikes’ ⇒ *gwh e-gwh n-e/o– ‘struck’ (> Gk. péph n-e ‘slew’ = YAv. -jaγnaṯ); (ii)
roots referring to prototypically intransitive events form transitive reduplicated aorists, e.g. *h2 er– ‘fit’ ⇒ *h2 e-h2 r-
e/o– (>> Gk. arareîn ‘ to make fit (tr.), to adapt). Similarly the reduplicated aorist underlies the productive preterite
to “causative” formations in Tocharian A, viz. its class II preterite (e.g. ca-cäl ‘lifted’ to the root täl(ā) ‘lift’ < *telh2 –;
cf. Malzahn 2010:172–3 on the function of this preterite). Whether the TB preterite II can also be derived from the
reduplicated aorist is not certain; see Malzahn (2010:184–189) for an overview of the question and, in addition, the
recent analysis of Jasanoff (2012b), who books the TB forms under “long-vowel preterites,” a class which he derives
from the imperfects of “Narten presents” (see above under root presents). It is possible that the cross-linguistically
common alignment of high transitivity and telicity (cf. Hopper and Thompson 1980:270–6, Wagner 2006) feeds the
development of transitivity in this class of aorists, though the fact that not all types of aorists become transitivizing
implies a more complicated evolution.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 62

Thematic aorists: Aorists with zero-grade root and accented thematic vowel are known from at least two
equations: PNIE *wid-é-t ‘saw, found out’ ( > Ved. 3sg. á-vid-a-t, Gk. é-(w)id-e, Arm. e-git), and *h1 ludh -é-t ‘went
out’ ( > Gk. ḗluth -e ‘came’, OIr. luid ‘went’, TA läc, TB lac ‘went out’). The latter example in particular demonstrates
the PIE antiquity of the thematic aorist, since it is continued in languages where the category was by no means
productive (Old Irish and Tocharian). Cardona (1960) analyzes most thematic aorists in Greek and Indo-Iranian as
thematized root aorists, and considers only two examples cited above to be of PIE antiquity, although the fact that
we have these two examples suggests that more existed in the protolanguage. LIV 2 unaccountably fails to reckon
with a thematic aorist; for one account of the type’s origins (ultimately a type of imperfect reanalyzed as an aorist)
see Jasanoff (fthcm. b).

§4.3.3 Perfect stem formation “Perfect” stems exhibit far less formal diversity than present and aorist stems;
there is effectively one type of perfect, which is set off from the system of present and aorist stems in several
formal and functional ways. The perfect is formed by partial copy reduplication (with fixed e-segmentism in the
reduplicant) and *o/ø-ablaut in the root. The inflectional endings of the perfect (active) are distinct from the
present/aorist active endings (cf. §4.2). Examples include *men– ‘think’ ⇒ 3sg. *me-món-e ‘has in mind’, 3pl.
*me-mn-ḗr (> 3sg. Gk. mémone ‘intends’, Lat. meminit ‘remembers’); *gwh en– ‘smash’ ⇒ 3sg. *gwh e-gwh ón-e ‘has
slain’, 3pl. *gwh e-gwh n-ḗr (> 3sg. Ved. ja-ghā ́n-a). Another formal peculiarity of the perfect is its distinctive active
participle suffix *–wós- (contrast the eventive’s *–nt–). There is one certain example of a PIE root that makes an
unreduplicated perfect: *woíd-e ‘knows’ (Gk. (w)oîd-e, 1pl. (w)íd-men, Ved. véd-a 1pl. vid-mā ́, Go. wait, witum, etc.).
It has long been disputed whether this form represents an archaism (i.e. reflecting a pre-stage when perfect stems
were formed without reduplication), an innovation, or is something else entirely (for one account see Jasanoff
2003a:234–46 with references, but compare now Jasanoff fthcm. b).
Beyond these formal differences, it is notable that the perfect’s semantic value is resultative-stative, again
setting it apart from the eventive system. The three-way split between present, aorist, and perfect stems survives
only in Greek and Indo-Iranian, and it is therefore only in these two branches that semantic distinctions between
these categories can be investigated. Early Greek is thought to be most conservative in reflecting the value of
the PNIE perfect: Wackernagel (1904) established that in Homeric Greek a perfect can have the meanings of a
present state and/or a resulting state (cf. further Wackernagel 1926-8 [2009]:215–20 with the editor’s notes, and
Chantraine 1926). The value of the perfect in Indo-Iranian is broadly harmonious with that of Greek; in a thorough
investigation of the category Kümmel (2000:65–78) shows that the Indo-Iranian perfect divides into a stative-
like perfect and a past perfect, which refers to a greater or lesser extent to the present value relevance of a past
action. However, on the particulars of the perfect in Vedic a number of questions remain. Dahl (2010:343–424),
for instance, argues that the primary meaning is anteriority, a result critically reviewed by Jamison (2014), who
disputes that any overarching function of the perfect can be established for the Rigveda due to the heterogeneous
nature of the text. The diversity of functions in earliest Vedic would reflect ongoing diachronic change from the
resultative-stative value of PIE, found in earliest Vedic, to the anterior meaning found more consistently in its use
as a preterital narrative perfect in later Vedic, regularly in Epic and Classical Sanskrit. The precise functional value
of the perfect in Old Indic is thus a topic still undergoing investigation (see now also Jamison 2016). For further
analysis of the PIE perfect, see the three volume study by Di Giovine (1990-1996).
The status of the perfect in Anatolian is unsettled and inextricably bound up with one’s views on the founda-
tional question of the prehistory of the ḫi-conjugation (a helpful introduction to this complex problem is given by
Clackson 2007:129–56). Deriving the ḫi-conjugation as a whole from the perfect is simply not viable in the wake of
Jasanoff’s (2003a:1–27) criticism (following esp. Cowgill 1975, 1979). Whether any Anatolian items reflect the per-
fect is disputed. Jasanoff (2003a:11, 37, 117–8) claims that Hitt. wewakk– ‘demand’ and mēm(a)i– ‘speak’ descend
from PIE perfects, and Forssman (1994) argues that Hitt. šipand– ‘libate’ continues a perfect *spe-spónd–; how-
ever, Jasanoff (fthcm. b) now derives wewakk– ‘demand’ and mēm(a)i- ‘speak’ (and other apparent non-resultative
perfects like Gk. mémēke ‘bleats’) from reduplicated *h2 e-presents with a strong stem *Cé-CoC-ei, while deriving
the PNIE resultative-stative perfect from reduplicated *h2 e-aorists with a strong stem *Ce-CóC-e (cf. §4.2.6 above).
If Hitt. šipand– ‘libate’ reflects a reduplicated stem at all, its attested telic sense argues that it represents a redupli-
cated *h2 e-aorist *se-spónd– (Melchert 2016b).
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 63

§4.4 Non-finite formations PNIE made participles to each tense-aspect stem and for the two voices of active
and middle. Yet again, Anatolian does not conform to this model, and we address below the specific points at
which Anatolian problematizes the deeper PIE reconstruction. No single marker for the category infinitive can
be reconstructed for the protolanguage since the daughter languages disagree too greatly on how the category is
marked, although the fact that numerous daughter branches build infinitives with case-forms of abstract nouns
strongly suggests that the proto-language did so as well.

§4.4.1 Participles Morphologically participles attach to tense-aspect stems (present, aorist, perfect), making
verbal formations with adjectival agreement features. No recent work devoted entirely to participles in PIE exists;
Lowe (2015) is a thoroughgoing account of participles in the Rigveda, with diachronic material throughout. Lowe
(2015:5–6, 226–94) proposes to define participles along the cline of an adjective’s status as an inflectional part of
the verb system. Thus participles are defined as non-finite, inflectional forms of verbs, which are morphologically
adjectival. As inflectional forms, participles convey adjectival agreement of case, number, and gender with their
head noun; morphologically participles mark the verbal categories of voice (active and middle) and tense-aspect.
The participle is defined in distinction to verbal adjectives, which are lexical adjectives that display some verbal
properties. A deciding criterion between participle and verbal adjective is whether the adjective obligatorily in-
herits the argument structure of the base verb from which it is derived; participles in Vedic always inherit the
argument structure of the base, but with deverbal adjectives this may, but need not, be the case.
Present/aorist active participle: In PNIE the active participle to present and aorist stems is formed by an
ablauting suffix *–ont/nt– (fem. *–nt-ih2 –). Thematic forms were *–o-nt–, while in athematic verbs the suffix was
added to the weak stem — for instance, *h1 es– ‘be’ makes a [Link]. *h1 s-ónt– ‘being’ (cf. Lat. sōns ‘guilty’ and
in-sōns ‘innocent’, relics from *h1 s-ónt-s; Watkins 1967). In thematic verbs the zero-grade suffix was added to the
thematic vowel, as in *bh ér-o-nt– ‘bearing’ (Gk. ph ér-o-nt–). Similarly *–nt– could be added to aorist stems.
The Anatolian cognate of *–nt– presents several serious discrepancies. The Hittite cognate of the participial
suffix *–nt–, viz. –ā ̆nt–, regularly expresses a resultant state: Hitt. kunant– means ‘killed, having been killed’ (not
‘killing’), a meaning matched residually in Luwian and Lycian relics, as in CLuw. walant(i)–/ulant(i)– ‘dead’, Lyc.
lãta– ‘dead’. In the case of transitive verbs the Anatolian participles show usually a passive, but sometimes an
active sense, e.g. Hitt. šekkant– ‘knowing/known’, appānt– ‘taken, seized’. This state of affairs contrasts with other
IE languages, as illustrated by Hitt. kunant– ‘killed’ beside its cognate in Vedic ghnánt– ‘smashing, killing’ (though
see Watkins 1969:142–4 for possible relics of passive meaning of the *–nt-participle). In general, then, the Hitt.
–nt-participle in functional terms most closely resembles PNIE *–to–/*–no– adjectives. Precisely how to derive
the Anatolian or non-Anatolian attested function from the other remains an unsolved problem (Melchert fthcm.
b and Fellner and Grestenberger fthcm propose possible step-by-step diachronic scenarios).
It may be noted that a formally identical suffix *–nt– is also used outside the verbal system to build adjectives
to property concept roots (within the “Caland system,” Rau 2009a:176–7 et passim). For instance, Ved. br̥hánt–
‘high’, Av. bərəzaṇt– ‘id.’, TA kom-pärkānt ‘sunrise’, etc. all derive from the root *bh erg̑ h – ‘high’, whose meaning is
typical of property concept roots, and which builds adjectival stems (this example was identified already by Caland
1892:267). Verbal stems formed to this root are sporadically attested (see further Lowe 2014a:283–94).
Middle participle: The middle participle (present, aorist, perfect) is reconstructible as athematic *–mh1 no–,
thematic *–o-mh1 no–. The comparative method requires the reconstruction of this peculiar suffix shape, as shown
by Klingenschmitt (1975:161–3); the suffix is certainly composite in diachronic terms, although its internal structure
is opaque. The suffix is found as a productive participle marker in Indo-Iranian (Ved. [athem.] –āná–, [them.] –
a-māná–), Greek ([pfc.] –ménos, [prs.] –menos), and Tocharian (TA -māṃ, TB -mane). In other languages mere
vestiges remain, such as Arm. anasown ‘animal’ < *n̥ -h2 eg̑-omno– lit. ‘non-speaking’, and Latin relics include fēmina
‘woman’ and alumnus ‘nursling’ (cf. Weiss 2011:437). There is no trace of this participle in Anatolian; for arguments
against Luwian –Vmma– as a reflex of *–mh1 no– see Melchert (2014a:206–7).
Perfect Participle: The perfect participle active was formed with the ablauting suffix *–wos/us– (f. *–us-ih2 –)
added to the perfect stem. The formation is clearly continued into a number of daughter languages, as in Myc.
Gk. a-ra-ru-wo-a [arar(u)-woh-a] ‘fitted’ ([Link].), Ved. ca-kr̥-vā ́ṃs-am ([Link].) / ca-kr-úṣ-ī ([Link].) to the
root kr̥ /kar– ‘make’. Forms of the perfect participle active are continued in languages where the perfect has been
lost as a finite category; it is found in Tocharian’s preterite participle, e.g. TB kekamu/kekamoṣ (root käm– ‘come’ +
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 64

–u < *–wos–), and remade in Balto-Slavic (details in Olander 2015:94–5). A curious trace of the formation survives
in Go. berusjos ‘parents’ (reflecting the feminine *–us-yeh2 –). Possible vestiges remain in Italic (see Vine, this
volume §[Link]); no trace has been found in Armenian or Albanian. With greater consequences for PIE, the perfect
participle active is absent from Anatolian; it is highly likely that this absence is due to the category’s nascence after
Anatolian’s departure from the common ancestor of the NIE languages.

§4.4.2 Infinitives The infinitives in the IE languages are usually frozen case-forms of deverbal nominalizers (cf.
§2.4.1 above); it is very likely that infinitives were formed in this way in PIE too. However, the significant formal
diversity attested in the marking of infinitives in the daughter languages seriously problematizes efforts to recon-
struct the PIE exponent(s) of this category — that is, precisely which case form or forms of which nominalizers
marked the category of infinitive remains unclear (for one overview of the problem, see García Ramón 1997). Key-
dana (2013a) proposes criteria for segregating event nominalizations from true infinitives in Vedic. His strongest
proposed criterion for nounhood is that event nominalizers do not inherit argument structure from the verb, and
therefore cannot govern a transitive object (they instead take a genitive complement). True infinitives do inherit
verbal argument structure, which includes transitivity (and potentially tense, aspect, and voice), and thus will
govern accusative case (Keydana 2013a:25-58). It is not yet clear whether the Vedic texts always conform to the
proposed criteria (cf. Lowe 2014b); see also the extensive discussion of Old Irish verbal nouns and infinitives by
Stüber (2015).
The following infinitives are representative of the forms attested in the daughter languages. The suffix *–tu–
(forming abstract nouns) makes infinitives in various cases, for instance ([Link].) Ved. dā ́-tum ‘to give’, ([Link].)
Ved. pā ́-tave ‘for drinking’, also in Old Prussian da-twei ‘to give’. Likewise the suffix (forming abstract nouns of
feminine gender) *–ti– in various cases: Ved. pī-táye ([Link].) ‘for drinking’, Lith. bū ́ -ti ‘to be’ (from [Link]. *–
tēi). The suffix *–men– furnishes infinitives in various cases, such as Ved. vid-mán-e ([Link].) ‘to know’, Hom. Gk.
(w)íd-men-ai ‘to know’; comparable is *–wen–, which underlies the Anatolian infinitives, Hitt. –wanzi (< abl.-instr.
*–wen-ti), Palaic and Luvian –una (< Proto-Anatolian allative *–un-eh2 ). The suffix *–dh ye/o– (cf. Fortson 2012,
2013) makes infinitives across a number of branches: Indo-Iranian *–dh yāy (e.g. Ved. píba-dhyai ‘to drink’) can be
equated with Italic infinitives, viz. Osc. –fír, Umb. –f(e)i, Lat. [Link]. –rier, as well as the Tocharian infinitive in
–tsi (e.g. TB lkā-tsi ‘to see’).

§5 Conclusions

Our survey of PIE morphology, written in the first quarter of the 21st century, builds directly on the great foun-
dations of the field laid in the 19th and 20th centuries. However, the picture of PIE morphology it presents differs
radically in many respects from the one presented by our predecessors; as one adage has it, “no language changes
so fast as Proto-Indo-European.” We have attempted here to survey where there is consensus in the field and to
flag points of interest for future research. We have aimed to present a state-of-the-art view on PIE morphology, in
full knowledge that this picture will change in coming years. The continued integration of Hittite and Tocharian
into our understanding of PIE will undoubtedly play a major role in the 21st century, much as it has done in the
20th ; philological work on the daughter branches will continue apace, challenging and revising our understanding
of the proto-language; and advances in theoretical linguistics and in synchronic and diachronic language typology
will continue to shed new light on old problems.

Acknowledgments

For invaluable comments, criticisms, and suggestions, we are deeply indebted to colleagues at our home institu-
tion, UCLA, especially David Goldstein, Stephanie Jamison, Craig Melchert, Teigo Onishi, Ryan Sandell, and Brent
Vine. Our chapter also improved immeasurably from the help of Joe Eska, Ben Fortson, Mark Hale, and Jay Jasanoff.
We also thank José Luis García Ramón and Andreas Willi, who made preprint versions of their forthcoming work
available to us. Finally, it is our special pleasure to thank Jared Klein, without whose support this project could
not have been completed.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 65

References

Ackermann, Katsiaryna. 2014. Die Vorgeschichte des slavischen Aoristsystems: mit der kommentierten Belegsamm-
lung der Aoristformen und Formen des präteritalen passiven Partizipiums im Altkirchenslavischen. Leiden /
Boston: Brill.

Adams, Douglas Q. 2014. The polyvalent present-formative –äsk– in Tocharian B. Tocharian and Indo-European
Studies 15.1–33.

Aikhenvald, Alexandra Y. 2011. Word-Class Changing Derivations in Typological Perspective. In Alexandra Y.

Aikhenvald and Robert M. W. Dixon (eds.), Language at Large: Essays on Syntax and Semantics, 221–289. Leiden
/ Boston: Brill.

Alderete, John D. 2001a. Dominance Effects as Transderivational Anti-Faithfulness. Phonology 18(2).201–253.

———. 2001b. Morphologically Governed Accent in Optimality Theory. New York: Routledge.

———. 2001c. Root-Controlled Accent in Cupeño. Natural Language & Linguistic Theory 19(3).455–502.

Alexiadou, Artemis. 2014. Nominal Derivation. In Rochelle Lieber and Pavol Štekauer (eds.), Oxford Handbook of
Derivational Morphology, 235–256. Oxford ; New York: Oxford University Press.

Balles, Irene. 2004. Zur Rekonstruction des früh-urindogermanischen Nominalklassensystems. In Adam Hyllested,
Anders R. Jørgensen, Jenny H. Larsson and Thomas Olander (eds.), Per aspera ad asteriscos: Studia Indogerman-
ica in honorem Jens Elmegaard Rasmussen, 43–57. Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft.

———. 2006. Die altindische Cvi-Konstruktion: Form, Funktion, Ursprung. Bremen: Hempen.

———. 2008. The Old Indic cvi-construction, the Caland system, and the PIE adjective. In Jens E. Rasmussen
and Thomas Olander (eds.), Internal Reconstruction in Indo-European: Results, Methods, and Problems. Section
Papers from the XVI International Conference on Historical Linguistics, University of Copenhagen, 11th-15th August
2003. Copenhagen: Museum Tusculanum Press.

Bamyacı, Elif, Jana Häussler , and Barış Kabak. 2014. The interaction of animacy and number agreement: An
experimental investigation. Lingua 148.254–277.

Barber, Peter J. 2013. Sievers’ Law and the History of Semivowel Syllabicity in Indo-European and Ancient Greek.
Oxford / New York: Oxford University Press.

Barnes, Timothy. 2015. Old Irish cuire, its congeners, and the ending of the 2nd sg. middle imperative. Ériu 65.49–
56.

Bauer, Laurie, and Salvador Valera. 2005. Approaches to Conversion/Zero-Derivation. Münster: Waxmann.

Beekes, Robert S.P. 1985. The Origins of the Indo-European Nominal Inflection. Innsbruck: Innsbrucker Beiträge zur
Sprachwissenschaft.

Beekes, Robert S.P., and Michiel A.C. de Vaan. 2011. Comparative Indo-European Linguistics: an Introduction, 2nd
edn. Amsterdam/Philadelphia: John Benjamins Publishing.

Bendahman, Jadwiga. 1993. Der reduplizierte Aorist in den indogermanischen Sprachen. Egelsbach / Köln / New
York: Hänsel-Hohenhausen.

de Bernardo Stempel, Patrizia. 1999. Nominale Wortbildung des älteren Irischen: Stammbildung und Derivation.
Tübingen: Niemeyer.

Booij, Geert. 2012. The Grammar of Words: an Introduction to Morphology, 3rd edn. Oxford/New York: Oxford
University Press.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 66

Bopp, Franz. 1854. Vergleichendes Accentuationssystem. Berlin: Dümmler.

Bozzone, Chiara. 2012. The PIE Subjunctive: Function and Development. In Craig Melchert (ed.), The Indo-
European Verb: Proceedings of the Conference of the Society for Indo-European Studies, Los Angeles 13-15 September
2010, 7–18. Wiesbaden: Reichert.

———. 2016. The Origin of the Caland System and the Typology of Adjectives. Indo-European Linguistics 4(1).15–
52. doi: 10.1163/22125892-00401003.

Brosch, Cyril. 2010. Nominalkomposita und komponierende Ableitungen im Hethitischen. Altorientalische

Forschungen 37(2).263–301.

Brugmann, Karl. 1891. Zur Frage der Entstehung des grammatischen Geschlechtes. Beiträge zur Geschichte der
deutschen Sprache und Literatur 15.523–531.

Brugmann, Karl, and Berthold Delbrück. 1906. Vergleichende Laut-, Stammbildungs- und Flexionslehre, nebst Lehre
vom Gebrauch der Wortformen der indogermanischen Sprachen. Zweiter Band: Lehre von den Wortformen und
ihrem Gebrauch, 2nd edn. Grundriss der vergleichenden Grammatik der indogermanischen Sprachen. Strass-
burg: Trübner.

Bybee, Joan L., and Östen Dahl. 1989. The Creation of Tense and Aspect Systems in the Languages of the World.
Studies in Language 13(1).51–103.

Byrd, Andrew M. 2015. The Indo-European Syllable. Leiden / Boston: Brill.

Caland, Willem. 1892. Beiträge zur kenntniss des Avesta, no. 19. KZ 31.266–268.

———. 1893. Beiträge zur kenntniss des Avesta, no. 26. KZ 32.592.

Cardona, George. 1960. The Indo-European Thematic Aorists. Ph.D. diss., Yale University, New Haven, CT.

Casaretto, Antje. 2004. Nominale Wortbildung der gotischen Sprache: Die Derivation der Substantive. Heidelberg:
Winter.

Chantraine, Pierre. 1926. Histoire du parfait grec. Paris: Champion.

———. 1933. La formation des noms en grec ancien. Paris: Klincksieck.

Chung, Sandra. 1983. Transderivational Relationships in Chamorro. Language 59.35–66.

Clackson, James. 2007. Indo-European Linguistics : an Introduction. Cambridge / New York: Cambridge University
Press.

Clackson, James, and Geoffrey Horrocks. 2007. The Blackwell History of the Latin Language. Oxford / Malden, MA:
Blackwell.

Comrie, Bernard. 1976. Aspect. Cambridge / New York: Cambridge University Press.

———. 1989. Language Universals and Linguistic Typology: Syntax and Morphology, 2nd edn. Chicago, IL: Uni-
versity of Chicago Press.

Comrie, Bernard, and Sarah Thompson. 2007. Lexical Nominalization. In Timothy Shopen (ed.), Language
Typology and Syntactic Description, 334–381. Cambridge / New York: Cambridge University Press, 2nd edn.

Corbett, Greville. 2000. Number. Cambridge University Press: Cambridge / New York.

Corbett, Greville G. 1991. Gender. Cambridge / New York: Cambridge University Press.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 67

———. 2013. Gender. In Matthew S. Dryer and Martin Haspelmath (eds.), The World Atlas of Language Structures
Online (WALS). Leipzig: Max Planck Institute for Evolutionary Anthropology.

Cowgill, Warren. 1975. More Evidence for Indo-Hittite: The Tense-Aspect Systems. In Proceedings of the Eleventh
International Congress of Linguists, 2.557–70. Bologna: Società Editrice il Mulino.

———. 1979. Anatolian ḫi-Conjugation and Indo-European Perfect: Installment II. In Erich Neu and Wolfgang
Meid (eds.), Hethitisch und Indogermanisch: Vergleichende Studien zur historischen Grammatik und zur dialek-
tgeographischen Stellung der indogermanischen Sprachgruppe Altkleinasiens, 25–39. Innsbruck: Innsbrucker
Beiträge zur Sprachwissenschaft.

———. 1985. The Personal Endings of Thematic Verbs in Indo-European. In Bernfried Schlerath (ed.), Gramma-
tische Kategorien: Funktion und Geschichte. Akten der VII. Fachtagung der Indogermanischen Gesellschaft, Berlin,
20.–25. Februar 1983, 99–108. Wiesbaden: Reichert.

———. 2006. The Collected Writings of Warren Cowgill. Edited by Jared S. Klein. Ann Arbor, MI / New York: Beech
Stave Press.

Cuzzolin, Pierluigi, and Christian Lehmann. 2004. Comparison and Gradation. In Geert Booij, Christian Lehmann
and Joachim Mugdan (eds.), Morphologie. Ein internationales Handbuch zur Flexion und Wortbildung. 2. Halb-
band, 1212–1220. Berlin / New York: de Gruyter.

Dahl, Eystein. 2010. Time, Tense and Aspect in Early Vedic Grammar. Leiden / Boston: Brill.

———. 2013. The Indo-European Subjunctive and its Origin as a Present. Journal of Indo-European Studies
41(3/4).392–430.

Debrunner, Albert. 1917. Griechische Wortbildungslehre. Heidelberg: Winter.

Delbrück, Berthold. 1871. Der Gebrauch des Conjunctivs und Optativs im Sanskrit und Griechischen. Halle: Waisen-
haus.

———. 1888. Altindische Syntax. Syntaktische Forschungen. Halle: Waisenhaus.

Dell’Oro, Francesca. 2015. Leggi, sistemi e leghe suffissali “di Caland”: storia della questione “Caland” come problema
teorico della linguistica indoeuropea. Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft.

Dempsey, Timothy R. 2015. Verbal Reduplication in Anatolian. Ph.D. diss., University of California, Los Angeles.

Derksen, Rick. 2014. Etymological Dictionary of the Baltic Inherited Lexicon. Leiden / Boston: Brill.

Di Giovine, Paolo. 1990-1996. Studio sul perfetto indoeuropeo (3 vols.). Rome: Editrice ‘Il Calamo’.

Dieu, Éric. 2011. Le supplétisme dans les formes de gradation en grec ancien et dans les langues indo-européennes.
Geneva: Droz.

Dixon, Robert M. W. 2004. Adjective Classes in Typological Perspective. In Robert M. W. Dixon and Alexandra I.
Aikhenvald (eds.), Adjective Classes: A Cross-Linguistic Typology, 1–49. Oxford / New York: Oxford University
Press.

Dresher, B. Elan, and Harry van der Hulst. 1998. Head-dependent asymmetries in phonology: Complexity and
visibility. Phonology 15.317–352.

Drinka, Bridget. 1995. The Sigmatic Aorist in Indo-European: Evidence for the Space-Time-Hypothesis. JIES Mono-
graphs. Washington, D.C.: Institute for the Study of Man.

Duhoux, Yves. 2000. Le verbe grec ancien : éléments de morphologie et de syntaxe historiques, 2 edn. Louvain-La-
Neuve: Peeters.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 68

Dunkel, George E. 2014. Lexikon der indogermanischen Partikeln und Pronominalstämme. vol. I–II. Heidelberg:
Winter.

Eichner, Heiner. 1969. Hethitisch wešš-/waššiya- “(Gewänder) tragen; anziehen; bekleiden”’. Münchener Studien
zur Sprachwissenschaft 27.5–44.

———. 1985. Das Problem des Ansatzes eines urindogermanischen Numerus ‘Kollektiv’ (‘Komprehensiv’). In
Bernfried Schlerath (ed.), Grammatische Kategorien: Funktion und Geschichte. Akten der VII. Fachtagung der
Indogermanischen Gesellschaft, Berlin, 20.–25. Februar 1983, 135–169. Wiesbaden: Reichert.

Fábregas, Antonio. 2014. Adjectival and Adverbial Derivation. In Rochelle Lieber and Pavol Štekauer (eds.), The
Oxford Handbook of Derivational Morphology, 276–295. Oxford: Oxford University Press.

Fellner, Hannes. 2014. PIE feminine *–eh2 – in Tocharian. In Sergio Neri and Roland Schuhmann (eds.), Studies on
the Collective and Feminine in Indo-European from a Diachronic and Typological Perspective, 7–22. Leiden: Brill.

Fellner, Hannes, and Laura Grestenberger. fthcm. Gemeinsame Unterschiede und unterschiedliche Gemein-
samkeiten der hethitischen und tocharischen Nominalmorphologie: Die Reflexe der *-nt- und *-mh1 no-
Partizipien. Forthcoming in the proceedings of the Indo-European Society, Marburg: “100 Jahre Entzifferung
des Hethitischen” (21-23 September 2015).

Forssman, Bernhard. 1994. Zu hethitisch šipand- und išpant-. In Jens Rasmussen (ed.), In honorem Holger Pedersen.
Kolloquium der indogermanischen Gesellschaft vom 25. bis 28. März 1993., 93–106. Wiesbaden: Reichert.

Fortson, Benjamin W. 2010. Indo-European Language and Culture: an Introduction, 2nd edn. Oxford, U.K. and
Malden, MA: Wiley-Blackwell.

———. 2012. Latin -rier and its Indo-Iranian Congeners. Indogermanische Forschungen 117.118.

———. 2013. Pre-Italic *dhye (*-dhyeh1 ) versus Pre-Indo-Iranian textit*-dhyōy: Bridging the Gap. In Adam
Cooper et al. (ed.), Multi Nominis Grammaticus: Studies in Classical and Indo-European linguistics in honor of
Alan J. Nussbaum on the occasion of his sixty-fifth birthday, 50–60. Ann Arbor: Beech Stave.

Frazier, Melissa. 2006. Accent in Proto-Indo-European Athematic Nouns: Antifaithfulness in Inflectional

Paradigms. Ph.D. diss., University of North Carolina at Chapel Hill, Chapel Hill, NC.

García Ramón, José Luis. 1985. Die Sekundärendung der 1. Sg. Medii im Indogermanischen. In Bernfried Schlerath
(ed.), Grammatische Kategorien: Funktion und Geschichte. Akten der VII. Fachtagung der Indogermanischen
Gesellschaft, Berlin, 20.–25. Februar 1983, 202–217. Wiesbaden: Reichert.

———. 1997. Infinitive im Indogermanischen? Zur Typologie der Infinitivbildungen und zu ihrer Entwicklung
in den älteren indogermanischen Sprachen. Incontri Linguistici 20.45–69.

———. 1998. Indogermanisch *gwh en- ‘(wiederholt) schlagen’, ‘töten’. In Jay Jasanoff et al. (ed.), Mír Curad: Studies
in Honor of Calvert Watkins, 139–154. Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft.

———. 2002. Zu Verbalcharakter, morphologischer Aktionsart und Aspekt in der indogermanischen Rekonstruk-
tion. In Heinrich Hettrich (ed.), Indogermanische Syntax: Fragen und Perspektiven. Akten der Arbeitstagung der
Indogermanischen Gesellschaft., 105–136. Wiesbaden: Reichert.

———. 2009. Primär- und Sekundärendungen im Konjunktiv im Vedischen: Deixis und Sprechakt. In Elisabeth
Rieken and Paul Widmer (eds.), Pragmatische Kategorien: Form, Funktion und Diachronie. Akten der Arbeitsta-
gung der Indogermanischen Gesellschaft, Marburg 2007, 79–91. Wiesbaden: Reichert.

———. 2010. On Hittite Verbs of the Type mimma-hhi ‘refuse’: Aktionsart and Aspect in Indo-European Recon-
struction. In Ronald Kim et al. (ed.), Ex Anatolia Lux: Anatolian and Indo-European Studies in honor of H. Craig
Melchert on the Occasion of his Sixty-Fifth Birthday, 40–54. Beech Stave.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 69

Garde, Paul. 1976. Histoire de l’accentuation slave. Paris: Institut d’études slaves.

Garrett, Andrew. 1990. The Syntax of Anatolian Pronominal Clitics. Ph.D. diss., Harvard University.

———. 1996. Wackernagel’s Law and Unaccusativity in Hittite. In Aaron Halpern and Arnold Zwicky (eds.),
Approaching Second: Second-Position Clitics and Related Phenomena, 85–133. Chicago: University of Chicago
Press.

George, Coulter. 2014. Lexical Aspect (Aktionsart). In Georgios Giannakis et al. (ed.), Encyclopedia of Ancient
Greek Language and Linguistics. Leiden / Boston: Brill.

Gil, David. 1996. Maltese ‘collective’ nouns. Rivista di Linguistica 8.53–87.

Goedegebuure, Petra. 2006. A new proposal for the reading of the Hittite numeral ‘1’, šia–. In Theo van den Hout
(ed.), The Life and Times of Hattušili III and Tudhaliya IV: Proceedings of a Symposium Held in Honor of J. de Roos,
165–188. Leiden: NINO.

———. 2013. Split-ergativity in Hittite. Zeitschrift für Assyriologie und vorderasiatische Archäologie 102(2).270–
303.

Gotō, Toshifumi. 2013. Old Indo-Aryan Morphology and its Indo-Iranian Background. Vienna: Verlag der Österre-
ichischen Akademie der Wissenschaften.

Greenberg, Joseph H. 1963. Some Universals of Grammar with Particular Reference to the Order of Meaningful
Elements. In Joseph H. Greenberg (ed.), Universals of language, 73–113. Cambridge MA / London: MIT Press.

Grestenberger, Laura. 2009. The Vedic i-Stems and Internal Derivation. Master’s thesis, Universität Wien.

———. 2014a. Feature Mismatch: Deponency in Indo-European Languages. Ph.D. diss., Harvard University.

———. 2014b. Zur Funktion des Nominalsuffixes *-i- im Vedischen und Urindogermanischen. In Norbert Oet-
tinger and Thomas Steer (eds.), Akten der Arbeitstagung: Das indogermanische Nomen., 88–102. Wiesbaden:
Reichert.

———. 2016. Reconstructing Indo-European Deponents. Indo-European Linguistics 4.98–149.

Griepentrog, Wolfgang. 1995. Die Wurzelnomina des Germanischen und ihre Vorgeschicte. Innsbruck: Institut für
Sprachwissenschaft der Universität Innsbruck.

Gunkel, Dieter. 2014. Accentuation. In Georgios Giannakis et al. (ed.), Encyclopedia of Ancient Greek Language
and Linguistics. Vol. 1: A-F, 7–12. Leiden (longer version available online at [Link]
[Link]): Brill.

Gvozdanović, Jadranka (ed.). 1992. Indo-European Numerals. Berlin: de Gruyter.

Hackstein, Olav. 2011. Collective and Feminine in Tocharian. In Olav Hackstein and Ronald I. Kim (eds.), Linguis-
tic Developments Along the Silkroad: Archaism and Innovation in Tocharian, 143–177. Wien: Österreichischen
Akademie der Wissenschaften.

Hajnal, Ivo. 1994. Die lykischen a-Stämme: Zum Werdegang einer Nominalklasse. In Jens E. Rasmussen (ed.),
In honorem Holger Pedersen. Kolloquium der indogermanischen Gesellschaft vom 25. bis 28. März 1993., 135–171.
Wiesbaden: Reichert.

Hale, Mark. 2007. Historical Linguistics: Theory and Method. Oxford: Blackwell.

———. 2010. Návyasā vácaḥ: To Praise with a Really Old Word. In Ronald Kim, Norbert Oettinger, Elisabeth
Rieken and Michael Weiss (eds.), Ex Anatolia Lux: Anatolian and Indo-European Studies in honor of H. Craig
Melchert on the Occasion of his Sixty-Fifth Birthday, 85–97. Ann Arbor, MI / New York: Beech Stave Press.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 70

Harðarson, Jon Axel. 1987. Zum urindogermanischen Kollektivum. Münchener Studien für Sprachwissenschaft
48.71–113.

Harðarson, Jón Axel. 1998. Mit dem Suffix *-eh1 - bzw. *-(e)h1 -ie/o- gebildete Verbalstämme im Indogermanischen.
In Wolfgang Meid (ed.), Sprache und Kultur der Indogermanen : Akten der X. Fachtagung der Indogermanis-
chen Gesellschaft Innsbruck, 22.-28. September 1996, 323–339. Innsbruck: Innsbrucker Beiträge zur Sprachwis-
senschaft.

Harðarson, Jon Axel. 2015. The Proto-Indo-European Collective and the Hittite Neuter Plural. Paper presented at
Hrozný and Hittite: The First 100 Years, Prague, 12 November 2015.

Harðarson, Jón Axeln Axel. 2014. Das andere Wort für ‘Frau’ im Urindogermanischen. In Sergio Neri and Roland
Schuhmann (eds.), Studies on the Collective and Feminine in Indo-European from a Diachronic and Typological
Perspective, 23–56. Leiden: Brill.

Haspelmath, Martin. 1993. More on the typology of inchoative/causative verb alternations. In Bernard Comrie
and Maria Polinsky (eds.), Causatives and Transitivity, 87–120. Amsterdam / Philadelphia: John Benjamins.

Heidermanns, Frank. 2005. Bibliographie zur indogermanischen Wortforschung: Wortbildung, Etymologie, Onoma-
siologie und Lehnwortschichten der alten und modernen indogermanischen Sprachen in systematischen Publika-
tionen ab 1800 (3 vols.). Tübingen: Max Niemeyer.

Hettrich, Heinrich. 1990. Der Agens in passivischen Sätzen altindergmanischer Sprachen. Göttingen: Vandenhoeck
& Ruprecht.

Hill, Eugen. 2007. Die Aorist-Präsentien des Indoiranischen: Untersuchungen zur Morphologie und Semantik einer
Präsensklasse. Bremen: Hempen.

Hill, Eugen, and Michael Frotscher. 2012. The Accentuation of Old Indic Reduplicated (3rd Class) Presents. In
H. Craig Melchert (ed.), The Indo-European Verb: Proceedings of the Conference of the Society for Indo-European
Studies, Los Angeles 13-15 September 2010, 105–114. Wiesbaden: Reichert.

Hill, Jane H. 2005. A Grammar of Cupeño. Los Angeles / Berkeley, CA: University of California Press.

Hirt, Hermann. 1900. Der indogermanische Ablaut: Vornehmlich in seinem Verhältnis zur Betonung. Strassburg:
Trübner.

Hoffmann, Karl. 1955. Ein grundsprachliches Possessivsuffix. Münchener Studien zur Sprachwissenschaft 6.35–40.

———. 1967. Der Injunktiv im Veda. Heidelberg: Winter.

———. 1970. Das Kategoriensystem des indogermanischen Verbums. Münchener Studien zur Sprachwissenschaft
28.19–41.

———. 1976. Aufsätze zur Indoiranistik. vol. 2. Wiesbaden: Reichert.

Hoffner, Harry A., and H. Craig Melchert. 2008. A Grammar of the Hittite Language. Vol. I: Reference Grammar.
Winona Lake, IN: Eisenbrauns.

Hopper, Paul, and Sandra Thompson. 1980. Transitivity in Grammar and Discourse. Language 56(2).251–299.

van den Hout, Theo. 2001. Neuter Plural Subjects and Nominal Predicates in Anatolian. In Onofrio Car-
ruba and Wolfgang Meid (eds.), Anatolisch und Indogermanisch: Akten des Kolloquiums der indogermanischen
Gesellschaft, Pavia, 22-25 September 1998, 167–192. Innsbruck: Institut für Sprachwissenschaft der Universität
Innsbruck.

Huggard, Mattyas. 2015. Wh-words in Hittite: A Study in Syntax-Semantics and Syntax-Phonology Interfaces. Ph.D.
diss., University of Californa, Los Angeles.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 71

van der Hulst, Harry. 1999. Word Prosodic Systems in the Languages of Europe. Berlin / New York: de Gruyter.

———. 2014. The study of word accent and stress: Past, present, and future. In Harry Van der Hulst (ed.), Word
Stress: Theoretical and Typological Issues, 3–55. Cambridge, UK / New York: Cambridge University Press.

Hyman, Larry. 2006. Word prosodic typology. Phonology 2.225–257.

Jamison, Stephanie. 1983. Function and form in the -áya-formations of the Rig Veda and the Atharva Veda. Göttingen:
Vandenhoeck & Ruprecht.

———. 2014. Review of: Eystein Dahl “Tense, Time and Aspect in Early Vedic Grammar” Leiden: Brill, 2010.
Indo-Iranian Journal 57(1-2).151–160. doi: 10.1163/15728536-05701014.

———. 2016. The Vedic Perfect Subjunctive and the Value of Modal Forms to Tense/Aspect Stems. In Ivo Hajnal,
Daniel Kölligan and Katharina Zipser (eds.), Miscellanea Indogermanica. Festschrift für José Luis García Ramón
zum 65. Geburtstag, 323–334. Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft.

Jamison, Stephanie W. 1979a. The Case of the Agent in Indo-European. Die Sprache 15.129–143.

———. 1979b. Remarks on the expression of agency with the passive in Vedic and Indo-European. Zeitschrift für
vergleichende Sprachforschung 93(2).196–219.

Jasanoff, Jay H. 1976. Gr. ámphō, lat. ambo et le mot indo-européen pour ‘l’un et l’autre’. Bulletin de la Société de
Linguistique 71.125–131.

———. 1978. Stative and Middle in Indo-European. Innsbruck: Institut für Sprachwissenschaft der Universität
Innsbruck.

———. 1979. Notes on the Armenian Personal Endings. Kuhn’s Zeitschrift/Historische Sprachforschung 93.133–149.

———. 1983. A rule of final syllables in Slavic. Journal of Indo-European Studies 11.139–149.

———. 1988a. The sigmatic aorist in Tocharian and Indo-European. Tocharian and Indo-European Studies 2.52–
76.

———. 1988b. The s-aorist in Hittite and Tocharian. Tocharian and Indo-European Studies 2.52–76.

———. 1989. Language and Gender in the Tarim Basin: The Tocharian 1 sg. Pronoun. Tocharian and Indo-
European Studies 3.125–148.

———. 1997. Gathic Avestan cikōitərəš. In Alexander Lubotsky (ed.), Sound Law and Analogy: Papers in honor of
Robert S.P. Beekes on the Occasion of his 60th Birthday, 119–130. Amsterdam / Atlanta: Rodopi.

———. 1998. The Thematic Conjugation Revisited. In Jay Jasanoff et al. (ed.), Mir Curad: Studies in Honor of
Calvert Watkins, 301–316. Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft.

———. 2003a. Hittite and the Indo-European Verb. Oxford: Oxford University Press.

———. 2003b. “Stative” *-ē- Revisited. Die Sprache 43.127–70.

———. 2006. The Ending of the PIE 2 Sg. Middle Imperative. Die Sprache 46(2).203–212.

———. 2008. *–bhi, *–bhis, *–ōis: Following the trail of the PIE instrumental plural. In Jens E. Rasmussen
and Thomas Olander (eds.), Internal Reconstruction in Indo-European: Results, Methods, and Problems. Section
Papers from the XVI International Conference on Historical Linguistics, University of Copenhagen, 11th-15th August
2003, 137–150. Copenhagen: Museum Tusculanum Press.

———. 2009. Notes on the internal history of the PIE optative. In Kazuhiko Yoshida and Brent Vine (eds.), East
Meets West: Papers in Indo-European Studies, 47–68. Bremen: Hempen.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 72

———. 2011. Balto-Slavic mobility as an Indo-European problem. In Roman Sukač (ed.), From Present to Past and
Back: Papers on Baltic and Slavic Accentology, 52–74. Frankfurt: Lang.

———. 2012a. Did Hittite have si-imperatives? In Roman Sukač and Ondřej Šefčík (eds.), The Sound of Indo-
European 2: Papers on Indo-European Phonetics, Phonemics and Morphophonemics, 116–132. Munich: LINCOM.

———. 2012b. Long-vowel preterites in Indo-European. In Craig Melchert (ed.), The Indo-European Verb. Proceed-
ings of the Conference of the Society for Indo-European Studies, Los Angeles, Oct. 13-15, 2010, 127‒135. Wiesbaden:
Reichert.

———. 2012c. The Prehistory of the Thematic Aorist. Paper presented at the 31st Annual East Coast Indo-
European Conference, Berkeley, CA, 18 May 2012.

———. fthcm. a. Verbal Stem Formation. In Andrew Garrett and Michael Weiss (eds.), Handbook of Indo-European
Studies. Oxford / New York: Oxford University Press.

———. fthcm. b. The Prehistory of the IE perfect: A contribution to the theory of the h2 -conjugation. In 100 Jahre
Entzifferung des Hethitischen: Morphosyntaktische Kategorien in Sprachgeschichte und Forschung. Proceedings of
Arbeitstagung of the Indogermanische Gesellschaft, Marburg, 21. bis 23. September 2015.

Jasanoff, Jay H., and Joshua Katz. 2016. A Revised History of the Greek Pluperfect. In Ivo Hajnal, Daniel Kölligan and
Katharina Zipser (eds.), Miscellanea Indogermanica. Festschrift für José Luis García Ramón zum 65. Geburtstag,
347–361. Innsbruck: Institut für Sprachen und Literaturen der Universität Innsbruck.

Katz, Joshua T. 1998. Topics in Indo-European Personal Pronouns. Ph.D. diss., Harvard University, Cambridge, MA.

———. 2008. On the Origin of the Greek Pluperfect. Die Sprache 46 [2006].1–37.

Kawahara, Shigeto. 2015. The Phonology of Japanese Accent. In Haruo Kubozono (ed.), The Handbook of Japanese
Phonetics and Phonology, 445–492. Berlin: de Gruyter.

Kellens, Jean. 1974. Les noms-racines de l’Avesta. Wiesbaden: Reichert.

Kemmer, Suzanne. 1993. The Middle Voice. Amsterdam / Philadelphia: John Benjamins.

Keydana, Götz. 2013a. Infinitive im Rgveda: Formen, Funktion, Diachronie. Leiden / Boston: Brill.

———. 2013b. Proterokinetische Stämme. In Götz Keydana, Thomas Olander and Paul Widmer (eds.), Indo-
European Accent and Ablaut, Copenhagen Studies in Indo-European, vol. 5, 31–62. Copenhagen: Museum Tuscu-
lanum Press.

Kim, Ronald. 2002. Topics in the Reconstruction and Development of Indo-European Accent. Ph.D. diss., Univer-
sity of Pennsylvania.

———. 2005. Review: J. Jasanoff, Hittite and the Indo-European Verb. Diachronica 22(1).191–200.

———. 2008. The Celtic Feminine Numerals ‘3’ and ‘4’ Revisited. Keltische Forschungen 3.143–167.

———. 2013a. The Indo-European, Anatolian, and Tocharian “Secondary” Cases in Typological Perspective. In
Adam I. Cooper, Jeremy Rau and Michael Weiss (eds.), Multi Nominis Grammaticus: Studies in Classical and
Indo-European linguistics in honor of Alan J. Nussbaum on the occasion of his sixty-fifth birthday, 121–142. Ann
Arbor, MI / New York: Beech Stave Press.

———. 2013b. Metrical Grid Theory, Internal Derivation, and the Reconstruction of PIE Nominal Accent
Paradigms. In Götz Keydana, Thomas Olander and Paul Widmer (eds.), Indo-European Accent and Ablaut, 63–
106. Copenhagen: Museum Tusculanum Press.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 73

Kim, Ronald I. 2009. The feminine gender in Tocharian and Indo-European. In Kazuhiko Yoshida and Brent Vine
(eds.), East Meets West: Papers in Indo-European Studies, 69–87. Bremen: Hempen.

Kiparsky, Paul. 1967. A Phonological Rule of Greek. Glotta 44(3/4).109–132.

———. 1968. Tense and mood in Indo-European syntax. Foundations of language 4(1).30–57.

———. 2005. The Vedic Injunctive: Historical and Synchronic Implications. The Yearbook of South Asian Lan-
guages and Linguistics 35.219–235.

———. 2010a. Compositional vs. Paradigmatic Approaches to Accent and Ablaut. In Stephanie W. Jamison,
H. Craig Melchert and Brent Vine (eds.), Proceedings of the 21st Annual UCLA Indo-European Conference, 137–181.
Bremen: Hempen.

———. 2010b. Dvandvas, blocking, and the associative: The bumpy ride from phrase to word. Language
86(2).302–331.

———. 2011. Did Indo-European have Reflexive Pronouns? Paper presented at the 23rd Annual UCLA Indo-
European Conference, Los Angeles, 28–29 October 2011.

———. 2015. New Perspectives in Historical Linguistics. In Claire Bowern and Bethwyn Evans (eds.), The Rout-
ledge Handbook of Historical Linguistics, 64–102. London / New York: Routledge.

———. 2016. The Agent Suffixes as a Window into Vedic Grammar. In Dieter Gunkel, Joshua T. Katz, Brent Vine
and Michael Weiss (eds.), Sahasram ati srajas: Indo-European and Indo-Iranian Studies in Honor of Stephanie W.
Jamison, 170–192. Ann Arbor: Beech Stave Press.

———. fthcm. Accent and Ablaut. In Andrew Garrett and Michael Weiss (eds.), Handbook of Indo-European
Studies. Oxford / New York: Oxford University Press.

Kiparsky, Paul, and Morris Halle. 1977. Towards a Reconstruction of the Indo-European Accent. In Larry Hyman
(ed.), Studies in Stress and Accent, 209–238. Los Angeles, CA: USC Press.

Klein, Jared S. 2003. Āmreḍitas and Related Constellations in the Rigveda. Journal of the American Oriental Society
123(4).773–802.

Klingenschmitt, Gert. 1975. Tocharisch und Urindogermanisch. In Helmut Rix (ed.), Flexion und Wortbildung:
Akten der V. Fachtagung der Indogermanischen Gesellschaft, Regensburg, 9. bis. 14. September 1975, 148–163. Wies-
baden: Reichert.

Kloekhorst, Alwin. 2008. Etymological Dictionary of the Hittite Inherited Lexicon. Leiden / Boston: Brill.

———. 2012. Hittite “ā/e”-ablauting Verbs. In H. Craig Melchert (ed.), The Indo-European Verb: Proceedings
of the Conference of the Society for Indo-Eurpean Studies, Los Angeles 13-15 September 2010, 151–160. Wiesbaden:
Reichert.

———. 2013. Indo-European nominal ablaut patterns: The Anatolian evidence. In Götz Keydana, Paul Wid-
mer and Thomas Olander (eds.), Indo-European Accent and Ablaut, 107–28. Copenhagen: Museum Tusculanum
Press.

———. 2014. Once more on Hittite ā/e-ablauting hi-verbs. Indogermanische Forschungen 119.55–77.

———. 2016. The Anatolian Stop System and the Indo-Hittite Hypothesis. Indogermanische Forschungen 121.213–
247.

Kölligan, Daniel. 2004. Wenn zwei dasselbe tun: Iterativa und Kausativa. In Maria Kozianka, Rosemarie Lühr and
Susanne Zeilfelder (eds.), Indogermanistik–Germanistik–Linguistik. Akten der Arbeitstagung der Indogermanis-
chen Gesellschaft, Jena 2002, 193–247. Hamburg: Kovač.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 74

———. 2007a. Iteratives and Causatives in Latin: A Unified Approach. In Coulter George, Matthew McCulloch,
Benedicte Nielsen, Antonia Ruppel and Olga Tribulato (eds.), Greek and Latin from an Indo-European Perspective,
49–64. Cambridge: Cambridge Philological Society.

———. 2007b. Suppletion und Defektivität im griechischen Verbum. Bremen: Hempen.

Koontz-Garboden, Andrew. 2006. On the typology of state/change of state alternations. In Geert Booij and Jaap
van Maarle (eds.), Yearbook of Morphology 2005, 83–117. Dordrecht: Springer.

Kortlandt, Frederik. 1981. 1st Sg. Middle *–H2 . Indogermanische Forschungen 86.123–136.

Kubozono, Haruo. 2011. Japanese Pitch Accent. In Marc Van Oostendorp, Colin J. Ewen, Elizabeth Hume and
Keren Rice (eds.), The Blackwell Companion to Phonology, 2879–2907. Oxford / Malden, MA: Blackwell.

Kuiper, Franciscus Bernardus Jacobus. 1942. Notes on Vedic Noun-Inflexion. Amsterdam: Noord-hollandsche
uitgevers maatschappij.

Kulikov, Leonid. 2000. The Vedic Type syáti Revisited. In Bernhard Forssman and Robert Plath (eds.), Indoarisch,
Iranisch und die Indogermanistik: Arbeitstagung der Indogermanischen Gesellschaft vom 2. bis 5. Oktober 1997 in
Erlangen, 267–283. Wiesbaden: Reichert.

———. 2009. Evolution of Case Systems. In Andrej Malchukov and Andrew Spencer (eds.), Oxford Handbook of
Case, 439–457. Oxford / New York: Oxford University Press.

———. 2012. The Vedic –ya-Presents: Passives and Intransitivity in Old Indo-Aryan. Amsterdam: Rodopi.

Kümmel, Martin. 1996. Stativ und Passivaorist im Indoiranischen. Göttingen: Vandenhoeck & Ruprecht.

———. 1998. Wurzelpräsens neben Wurzelaorist im Indogermanischen. Historische Sprachforschung 111.191–208.

Kümmel, Martin. 2000. Das Perfekt im Indoiranischen. Wiesbaden: Reichert.

———. 2007. Konsonantenwandel: Bausteine zu einer Typologie des Lautwandels und ihre Konsequenzen für die
vergleichende Rekonstruktion. Wiesbaden: Reichert.

Kümmel, Martin. 2012. Typology and Reconstruction: The Consonants and Vowels of Proto-Indo-European.
In Benedicte Whitehead, Thomas Olander, Birgit A. Olsen and Jens E. Rasmussen (eds.), The Sound of Indo-
European. Phonetics, Phonemics, and Morphophonemics, 291–329. Copenhagen: Museum Tusculanum Press.

———. 2014. Zum “proterokinetischen” Ablaut. In Norbert Oettinger and Thomas Steer (eds.), Das indogerman-
ische Nomen: Akten der Arbeitstagung der Indogermanischen Gesellschaft, Erlangen, 2011, 164–179. Wiesbaden:
Reichert.

———. 2016. *syá– im Indoiranischen: Zahlwort und Demonstrativum? In Andrew Byrd, Jessica DeLisi and Mark
Wenthe (eds.), Tavet Tat Satyam: Studies in Honor of Jared S. Klein on the Occasion of his Seventieth Birthday, 179–
190. Ann Arbor / New York: Beech Stave.

de Lamberterie, Charles. 1990. Les adjectifs grecs en -υς: sémantique et comparaison. Louvain-la-Neuve: Peeters.

———. 2002. Review of M. Meier-Brügger, Indogermanische Sprachwissenschaft (8th edn.), 2002. Bulletin de la
Société de Linguistique de Paris 97(2).103–114.

Ledo-Lemos, Francisco J. 2000. Femininum Genus: A Study on the Origins of the Indo-European Feminine Gram-
matical Gender. Munich: Lincom Europa.

Lehrman, Alexander. 1998. Indo-Hittite Redux. Moscow: Paleograph.

The Morphology of PIE J. Lundquist & A.D. Yates ǀ 75

Li, Fengxiang, and Lindsay Whaley. fthcm. A Cross-Linguistic Examination of Morphological Causatives, Inten-
sives, and Reciprocals: A Grammaticization Perspective. MS Dartmouth College.

Lieber, Rochelle, and Pavol Štekauer. 2009. The Oxford Handbook of Compounding. Oxford / New York: Oxford
University Press.

———. 2014. The Oxford Handbook of Derivational Morphology. Oxford: Oxford University Press.

Lindner, Thomas. 2011-2016. Komposition. Indogermanische Grammatik IV.1 (4 vols.). Heidelberg: Winter.

Litscher, Roland. 2014. Voraussetzungen für ein feminines Genus und Implikationen für das Kategoriensystem
des frühindogermanischen Nomens. In Sergio Neri and Roland Schuhmann (eds.), Studies on the Collective and
Feminine in Indo-European from a Diachronic and Typological Perspective, 137–166. Leiden / Boston: Brill.

Lohmann, Johannes. 1932. Genus und Sexus. Eine morphologische Studie zum Ursprung der indogermanischen
nominalen Genus-Unterscheidung. Göttingen: Vandenhoeck & Ruprecht.

Lowe, John. 2014a. Indo-European Caland Adjectives in *-nt- and Participles in Sanskrit. Historische Sprach-
forschung 127(1).166–195.

———. 2014b. Review of Götz Keydana, “Infinitive im R̥ gveda: Formen, Funktion, Diachronie”, Leiden: Brill, 2013.
Indo-Iranian Journal 57(3).261–270.

———. 2015. Participles in Rigvedic Sanskrit. Oxford: Oxford University Press.

Lühr, Rosemarie, and Irene Balles. 2008. Nominale Wortbildung des Indogermanischen in Grundzügen: die wortbil-
dungsmunster ausgewählter indogermanischer Einzelsprachen. Band I: Latein, Altgriechisch. Hamburg: Kovač.

Lühr, Rosemarie, and Joachim Matzinger. 2008. Nominale Wortbildung des Indogermanischen in Grundzügen:
die wortbildungsmunster ausgewählter indogermanischer Einzelsprachen. Band II: Hethitisch, Altindisch, Altar-
menisch. Hamburg: Kovač.

Lundquist, Jesse. 2015. On the Accentuation of Vedic -ti-Abstracts. Indo-European Linguistics 3(1).42–72.

———. 2016. On the Accentuation of Compound s-Stem Adjectives in Greek and Vedic. In David M. Goldstein,
Stephanie W. Jamison and Brent Vine (eds.), Proceedings of the 27th Annual UCLA Indo-European Conference,
97–114. Bremen: Hempen.

Luraghi, Silvia. 2009a. Indo-European Nominal Classification: From Abstract to Feminine. In Stephanie W. Jami-
son, H. Craig Melchert and Brent Vine (eds.), Proceedings of the 20th Annual UCLA Indo-European Conference,
115–131. Bremen: Hempen.

———. 2009b. The origin of the feminine gender in PIE. In Vit Bubenik, John Hewson and Sarah Rose (eds.),
Grammatical Change in the Indo-European Languages: Papers presented at the Workshop on Indo-European Lin-
guistics at the 18th International Conference on Historical Linguistics, Montreal, 2007, 3–13. Amsterdam / Philadel-
phia: John Benjamins.

———. 2011. The origin of the Proto-Indo-European gender system: Typological considerations. Folia Linguistica
45(2).435–464.

———. 2012. Basic valency orientation and the middle voice in Hittite. Studies in Language 36(1).1–32.

———. 2014. Gender and word formation: The PIE gender system in cross-linguistic perspective. In Sergio Neri
and Roland Schuhmann (eds.), Studies on the Collective and Feminine in Indo-European from a Diachronic and
Typological Perspective, 199–231. Leiden: Brill.

Malzahn, Melanie. 2010. The Tocharian Verbal System. Leiden / Boston: Brill.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 76

———. 2013. Cutting Around “temós”: Evidence from Tocharian. In Adam Cooper et al. (ed.), Multi Nominis
Grammaticus: Studies in Classical and Indo-European linguistics in honor of Alan J. Nussbaum on the occasion of
his sixty-fifth birthday, 165–174. Beech Stave Press.

———. 2016. Tudati-presents and the tēzzi Principle. In Dieter Gunkel et al. (ed.), Sahasram Ati Srajas: Indo-
European and Indo-Iranian Studies in Honor of Stephanie W. Jamison, 227–238. Ann Arbor, MI / New York: Beech
Stave.

Malzahn, Melanie, Michaël Peyrot, Hannes Fellner , and Theresa-Susanna Illés (eds.). 2015. Tocharian Texts in Con-
text. International Conference on Tocharian Manuscripts and Silk Road Culture, Vienna, June 25-29, 2013. Bremen:
Hempen.

Matasović, Ranko. 2004. Gender in Indo-European. Heidelberg: Winter.

———. 2014. Slavic Nominal Word-Formation: Proto-Indo-European Origins and Historical Development. Heidel-
berg: Winter.

Mayrhofer, Manfred. 1986-2001. Etymologisches Wörterbuch des Altindoarischen. vol. 1-3. Heidelberg: Winter.

Meier-Brügger, Michael. 2010. Indogermanische Sprachwissenschaft, 9th edn. Berlin: De Gruyter.

Meillet, Antoine. 1922. La forme du génitif pluriel en ombrien. Mémoires de la Société de Linguistique de Paris
22.258–259.

———. 1925. La méthode comparative en linguistique historique. Paris: Champion.

———. 1931. Essai de chronologie des langues indo-européennes. Bulletin de la Société de Linguistique de Paris
32.1–28.

Meiser, Gerhard. 1993. Zur Funktion des Naslapräsens im Urindogermanischen. In Gerhard Meiser (ed.), Indoger-
manica et Italica: Festschrift für Helmut Rix zum 65. Geburtstag, 280–313. Innsbruck: Innsbrucker Beiträge zur
Sprachwissenschaft.

Meissner, Torsten. 1998. Das “Calandsche Gesetz” und das Griechische — nach 100 Jahren. In Wolfgang Meid (ed.),
Sprache und Kultur der Indogermanen : Akten der X. Fachtagung der Indogermanischen Gesellschaft Innsbruck,
22.-28. September 1996, 237–254. Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck.

———. 2006. S-stem Nouns and Adjectives in Greek and Proto-Indo-European. Oxford / New York: Oxford Univer-
sity Press.

Melchert, H. Craig. 1984. Studies in Hittite Historical Phonology. Göttingen: Vandenhoeck & Ruprecht.

———. 1992a. The Middle Voice in Lycian. Historische Sprachforschung 105.189–199.

———. 1992b. Relative Chronology and Anatolian: The Vowel System. In Robert S.P. Beekes and Alexan-
der Lubotsky (eds.), Rekonstruction und relative Chronologie. Akten der VIII. Fachtagung der Indogermanischen
Gesellschaft, Leiden, 31 August–4 September 1987, 41–53. Innsbruck: J. Weitenberg.

———. 1994a. Anatolian Historical Phonology. Amsterdam/Atlanta, GA: Rodopi.

———. 1994b. The Feminine Gender in Anatolian. In George E. Dunkel (ed.), Fruh-, Mittel-, Spatindogerman-
isch. Akten der IX. Fachtagung der Indogermanischen Gesellschaft vom 5. bis 9. Oktober 1992 im Zürich, 231–244.
Wiesbaden: Reichert.

———. 1997. Traces of a PIE Aspectual Contrast in Anatolian? Incontri linguistici 20.83–92.

———. 2000. Tocharian Plurals in –nt– and Related Phenomena. Tocharian and Indo-European Studies 9.530–575.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 77

———. 2002. Sibilants in Carian. In Matthias Fritz and Susanne Zeilfelder (eds.), Novalis Indogermanica:
Festschrift für Günter Neumann zum 80. Geburtstag, 305–313. Graz: Leykam.

———. 2003. Language. In H. Craig Melchert (ed.), The Luwians, 170–211. Leiden ; Boston: Brill.

———. 2011a. The PIE Collective Plural and the “τὰ ζῶα τρέχει rule”. In Thomas Krisch and Thomas Lindner
(eds.), Indogermanistik und Linguistik im Dialog: Akten der XIII. Fachtagung der Indogermanischen Gesellschaft
vom 21. bis 27. September 2008 in Salzburg, 395–400. Wiesbaden: Reichert.

———. 2011b. The Problem of the Ergative Case in Hittite. In Michèle Fruyt, Michel Mazoyer and Dennis Pardee
(eds.), Acts of the International Colloqium Variations, Concurrence, et Evolution des Cas dans Divers Domains Lin-
guistiques, Paris, 2–4 April 2007, 161–167. Chicago: Oriental Institute of the University of Chicago.

———. 2012a. Genitive Case and Possessive Adjective in Anatolian. In Vincenzo Orioles (ed.), Per Roberto
Gusmani: Linguistica Storica e Teorica, vol. 1, 273–286. Udine: Forum.

Melchert, H. Craig (ed.). 2012b. The Indo-European Verb: Proceedings of the Conference of the Society for Indo-
European Studies, Los Angeles 13-15 September 2010. Reichert.

Melchert, H. Craig. 2013. Ablaut Patterns in the Hittite ḫi-Conjugation. In Stephanie W. Jamison, H. Craig Melchert
and Brent Vine (eds.), Proceedings of the 24th Annual UCLA Indo-European Conference, Los Angeles, 26-27 October
2012, 137–150. Bremen: Hempen.

———. 2014a. Anatolian Nominal Stems in *-(C)o-. In Norbert Oettinger and Thomas Steer (eds.), Das Nomen im
Indogermanischen. Morphologie, Substantiv versus Adjektiv, Kollektivum. Akten der Arbeitstagung der Indoger-
manischen Gesellschaft vom 14. bis 16. September 2011 in Erlangen, 205–14. Wiesbaden: Reichert.

———. 2014b. ”Narten formations” versus ”Narten roots”. Indogermanische Forschungen 119.251–258.

———. 2014c. PIE *–eh2 as an “Individualizing” Suffix and the Feminine Gender. In Sergio Neri and Roland
Schuhmann (eds.), Studies on the Collective and Feminine in Indo-European from a Diachronic and Typological
Perspective, 257–271. Leiden: Brill.

———. 2015. The Tocharian s-Preterite. In Melanie Malzahn, Michaël Peyrot, Hannes Fellner and Theresa-
Susanna Illés (eds.), Tocharian Texts in Context: International Conference on Tocharian Manuscripts and Silk Road
Culture, Vienna, June 25–29, 2013, 127–135. Bremen: Hempen.

———. 2016a. The Case of the Agent in Anatolian and Proto-Indo-European. In Dieter Gunkel, Joshua T. Katz,
Brent Vine and Michael Weiss (eds.), Sahasram Ati Srajas: Indo-European and Indo-Iranian Studies in Honor of
Stephanie W. Jamison, 239–249. Ann Arbor, MI / New York: Beech Stave Press.

———. 2016b. Initial *sp- in Hittite and šip(p)and- ‘to libate’. Journal of Language Relationship 14(3).187–195.

———. 2016c. Relative Clauses in Anatolian. In Sergio Neri, Roland Schuhmann and Susanne Zeilfelder (eds.), dat
ih dir it nu bi huldi gibu: Linguistische, germanistische und indogermanistische Studien Rosemarie Lühr gewidmet,
287–295. Wiesbaden: Reichert.

———. fthcm. a. The Position of Anatolian. In Andrew Garrett and Michael Weiss (eds.),
Handbook of Indo-European Studies. Oxford / New York: Oxford University Press (Ms. available at
[Link] (accessed 3/1/2016)).

———. fthcm. b. The Source(s) of Indo-European Participles in *–e/ont–. In Adjectifs verbaux et participes dans
les langues indo-européennes. Proceedings of Arbeitstagung of the Indo-European Society, Paris, 24–26 September
2014. Proceedings of the Indo-European Society, Paris: “” (24-26 September 2014).

———. fthcm. c. An Allative Case in Proto-Indo-European? In To appear in a forthcoming Festschrift.

The Morphology of PIE J. Lundquist & A.D. Yates ǀ 78

Melchert, H. Craig, and Norbert Oettinger. 2009. Ablativ und Instrumental im Hethitischen und Indogermanis-
chen. Incontri Linguistici 32.53–73.

Morpurgo Davies, Anna. 1987. Folk Linguistics and the Greek Word. In George Cardona and Norman H. Zide
(eds.), Festschrift for Henry Hoenigswald on the occasion of his seventieth birthday, 263–280. Tübingen: G. Narr.

Napoli, Maria. 2006. Aspect and Actionality in Homeric Greek. A Contrastive Analysis. Milan: Angeli.

———. 2015. Telicity as a parameter of aspect in Homeric Greek. Activity and accomplishment verbs. Indoger-
manische Forschungen 112.124–169.

Narten, Johanna. 1964. Die sigmatischen Aoriste im Veda. Wiesbaden: Reichert.

Neri, Sergio. 2003. I sostantivi in -u del gotico: morfologia e preistoria. Institut für Sprachen und Literaturen,
Abteilung Sprachwissenschaft.

Neri, Sergio, and Roland Schuhmann (eds.). 2014. Studies on the Collective and Feminine in Indo-European from a
Diachronic and Typological Perspective. Leiden: Brill.

Nikolaev, Alexander. 2009. The Germanic word for ‘sword’ and delocatival derivation in Proto-Indo-European.
The Journal of Indo-European Studies 37(3).462–488.

Nussbaum, Alan J. 1975. Studies in Latin noun formation and derivation: ī in Latin denominative derivation. In
Calvert Watkins (ed.), Indo-European Studies II, 116–161. Cambridge, MA: Department of Linguistics, Harvard
University.

———. 1976. Caland’s “Law” and the Caland System. Ph.D. diss., Harvard University, Cambridge, MA.

———. 1986. Head and Horn in Indo-European. Berlin/New York: de Gruyter.

———. 1999. *Jocidus: An account of the Latin adjectives in -idus. In Heiner Eichner and Hans C. Luschütsky
(eds.), Compositiones Indogermanicae in Memoriam Joachem Schindler. Prague: Enigma Corporation.

———. 2014a. Feminine, Abstract, Collective, Neuter Plural: Some Remarks on Each (Expanded Handout).
In Sergio Neri and Roland Schuhmann (eds.), Studies on the Collective and Feminine in Indo-European from a
Diachronic and Typological Perspective, 273–306. Leiden: Brill.

———. 2014b. Greek τέκμαρ ‘sign’ and τέκμωρ ‘sign’: Why both? In Norbert Oettinger and Thomas Steer (eds.),
Das Nomen im Indogermanischen, 215–260. Wiesbaden: Reichert.

Nuyts, Jan, and Johan Van der Auwera. 2016. The Oxford Handbook of Modality and Mood. Oxford: Oxford University
Press.

Oettinger, Norbert. 1976. Der indogermanische Stativ. Münchener Studien zur Sprachwissenschaft 34.109–149.

———. 1979. Die Stammbildung des hethitischen Verbums. Nürnberg: Carl.

———. 1987. Bemerkungen zur anatolischen i-Motion und Genusfrage. Zeitschrift für vergleichende Sprach-
forschung 92.74–92.

———. 2006. Review of Jasanoff 2003. Kratylos 51.34–45.

Olander, Thomas. 2015. Proto-Slavic Inflectional Morphology. Leiden / Boston: Brill.

Olsen, Susan. 2001. Copulative Compounds: A closer look at the interface between syntax and morphology.
Yearbook of Morphology 2000 279–320.

———. 2014. Delineating Derivation and Compounding. In Rochelle Lieber and Pavol Štekauer (eds.), Oxford
Handbook of Derivational Morphology, 26–48. Oxford / New York: Oxford University Press.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 79

Ostrowski, Manfred. 1985. Zur Enstehung und Entwicklung des indogermanischen Nomens. In Bernfried Schlerath
(ed.), Grammatische Kategorien: Funktion und Geschichte. Akten der VII. Fachtagung der Indogermanischen
Gesellschaft, Berlin, 20.–25. Februar 1983, 313–323. Wiesbaden: Reichert.

Patri, Sylvain. 2007. L’alignement syntaxique dans les langues indo-européennes d’Anatolie. Wiesbaden: Reichert.

Pedersen, Holgar. 1938. Hittitisch und die anderen indoeuropäischen Sprachen. Copenhagen: Levin and Munks-
gaard.

Pedersen, Holger. 1926. La cinquième déclinaison latine. Copenhagen: A.F. Høst.

———. 1933. Études lituaniennes. Copenhagen: Levin & Munksgaard.

Penney, John H.W. 1978. Problems of Greek and Indo-European Ablaut: The Origin and Development of the
O-grade. Ph.D. diss., University of Oxford.

Petit, Daniel. 1999. *SUE- en grec ancien: la famille du pronom réfléchi. Linguistique grecque et comparaison indo-
européenne. Louvain-Paris: Peeters.

———. 2016. Force et dominance accentuelle en morphologie derivationelle grecque. In Alain Blanc and Daniel
Petit (eds.), Nouveaux acquis sur la formation des noms en grec ancien: Actes du Colloque international, Université
de Rouen, ERIAC, 17-18 octobre 2013, 5–35. Louvain: Peeters.

Peyrot, Michaël. 2013. The Tocharian Subjunctive: A Study in Syntax and Verbal Stem Formation. Leiden / Boston:
Brill.

Pike, Moss. 2011. Latin –tās and Related Forms. Ph.D. diss., UCLA Program in Indo-European Studies, Los Angeles.

Pinault, Georges-Jean. 2006a. Analyse étymologique d’un nom de parenté indo-européen. In Günter Schweiger
(ed.), Indogermanica: Festschrift Gert Klingenschmitt, 465–486. Taimering: Schweiger.

———. 2006b. Retour sur le numéral “un” en tokharien. Indogermanische Forschungen 111.71–97.

Portner, Paul. 2009. Modality. Oxford: Oxford University Press.

Poser, William John. 1984. The Phonetics and Phonology of Tone and Intonation in Japanese. Ph.D. diss., Mas-
sachusetts Institute of Technology, Cambridge, Mass.

Probert, Philomen. 2006a. Accentuation in Ancient Greek Deverbative ā-stems. In Daniel Kölligan and Ranjan
Sen (eds.), Oxford University Working Papers in Linguistics, Philology & Phonetics, vol. 11, 122–142. Oxford: Oxford
University.

———. 2006b. Ancient Greek Accentuation: Synchronic Patterns, Frequency Effects, and Prehistory. Oxford Clas-
sical Monographs. Oxford / New York: Oxford University Press.

———. 2006c. Clause Boundaries in Old Hittite Relative Sentences. Transactions of the Philological Society
104(1).17–83.

———. 2014. Relative Clauses, Indo-Hittite, and Standard Average European. In Stephanie W. Jamison, H. Craig
Melchert and Brent Vine (eds.), Proceedings of the 25th Annual UCLA Indo-European Conference, 2013, 137–164.
Bremen: Hempen.

———. 2015. Early Greek Relative Clauses. Oxford / New York: Oxford University Press.

Prokosch, Eduard. 1939. A Comparative Germanic Grammar. Philadelphia: University of Philadelphia Press.

Puhvel, Jaan. 1991. Homer and Hittite. Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 80

Rau, Jeremy. 2009a. Indo-European Nominal Morphology: The Decads and the Caland System. Innsbrucker Beiträge
zur Sprachwissenschaft. Innsbruck: Institut für Sprachen und Literaturen der Universität Innsbruck.

———. 2009b. Myc. te-re-ja and the Athematic Inflection of the Greek Contract Verbs. In Brent Vine and Kazuhiko
Yoshida (eds.), East and West: Papers in Indo-European Studies, 181–188. Bremen: Hempen.

———. 2013. Notes on Stative Verbal Roots, the Caland System, and Primary Verbal Morphology in Indo-Iranian
and Indo-European. In Multi Nominis Grammaticus: Studies in Classical and Indo-European linguistics in honor
of Alan J. Nussbaum on the occasion of his sixty-fifth birthday, 255–273. Ann Arbor, MI / New York: Beech Stave
Press.

———. 2014. The History of the Indo-European Primary Comparative. In Norbert Oettinger and Thomas Steer
(eds.), Das Nomen im Indogermanischen, 327–341. Wiesbaden: Reichert.

Revithiadou, Anthoula. 1999. Headmost Accent Wins: Head Dominance and Ideal Prosodic Form in Lexical Accent
Systems. Ph.D. diss., Leiden.

Rieken, Elisabeth. 1994. Der Wechsel –a–/–i– in der Stammbildung des hethitischen Nomens. Historische Sprach-
forschung 107(1).42–53.

———. 1999. Untersuchungen zur nominalen Stammbildung des Hethitischen. StBoT 44. Wiesbaden: Harras-
sowitz.

———. 2005. Neues zum Ursprung der anatolischen i-Mutation. Historische Sprachforschung 118.48–74.

Ringe, Donald. 2000. Tocharian Class II Presents and Subjunctives and the Reconstruction of the Proto-Indo-
European Verb. Tocharian and Indo-European Studies 9.121–142.

———. 2006. From Proto-Indo-European to Proto-Germanic. Oxford / New York: Oxford University Press.

Risch, Ernst. 1974. Wortbildung der homerischen Sprache, 2nd edn. Berlin: De Gruyter.

Rix, Helmut. 1988. The Proto-Indo-European Middle: Content, Forms, and Origin. Münchener Studien zur Sprach-
wissenschaft 49.101–119.

———. 1992. Historische Grammatik des Griechischen: Laut- und Formenlehre, 2nd edn. Darmstadt: Wis-
senschaftliche Buchgesellschaft.

Rix, Helmut, and Martin Kümmel. 2001. Lexikon der indogermanischen Verben: die Wurzeln und ihre Primärstamm-
bildungen (addenda and corrigenda maintained at [Link] 2nd edn. Wies-
baden: Reichert.

Sandell, Ryan. 2011. The Morphophonology of Reduplicated Presents in Vedic and Indo-European. In Stephanie W.
Jamison, H. Craig Melchert and Brent Vine (eds.), Proceedings of the 22nd Annual UCLA Indo-European Confer-
ence, 223–254. Bremen: Hempen.

———. 2015. Productivity in Historical Linguistics: Computational Perspectives on Word-Formation in Ancient

Greek and Sanskrit. Ph.D. diss., UCLA Indo-European Studies.

Sandell, Ryan, and Andrew M. Byrd. 2014. In Defense of Szemerényi’s Law. Paper presented at the 33rd Annual
East Coast Indo-European Conference, Blacksburg, VA, 5–7 June 2014.

Sasseville, David. 2014/15. Luwian and Lycian Agent Nouns in *-é-leh2 . Die Sprache 51(1).105–124.

de Saussure, Ferdinand. 1879. Mémoire sur le système primitif des voyelles dans les langues indo-européennes. Leip-
sick: Teubner.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 81

Scalise, Sergio, and Antonietta Bisetto. 2009. The Classification of Compounds. In Rochelle Lieber and Pavol
Štekauer (eds.), Oxford Handbook of Compounding, 34–53. Oxford: Oxford University Press.

Scalise, Sergio A., Antonio Fábregas , and Francesca Forza. 2009. Exocentricity in compounding. Gengo Kenkyu
135.49–84.

Scarlata, Salvatore. 1999. Die Wurzelkomposita im R̥ g-Veda. Wiesbaden: Reichert.

Schaffner, Stefan. 2001. Das Vernersche Gesetz und der innerparadigmatische grammatische Wechsel des Urgeman-
ischen im Nominalbereich. Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft.

Schindler, Joachim. 1980. Zur Herkunft der altindischen cvi-Bildungen. In Manfred Mayrhofer, Martin Peters and
Oskar E. Pfeiffer (eds.), Lautgeschichte und Etymologie, vol. 6. Wiesbaden: Reichert.

Schindler, Jochem. 1967. Zu hethitisch nekuz. Zeitschrift für vergleichende Sprachforschung 81.290–303.

———. 1972a. Das Wurzelnomen im Arischen und Griechischen. Ph.D. diss., Würzburg.

———. 1972b. L’apophonie des noms-racines indo-européens. BSL 67.31–38.

———. 1975a. L’apophonie des thèmes indo-européens en -r/n. BSL 70.1–10.

———. 1975b. Zum Ablaut der neutralen s-Stämme des Indogermanischen. In Helmut Rix (ed.), Flexion und
Wortbildung: Akten der V. Fachtagung der Indogermanischen Gesellschaft, Regensburg, 9. bis. 14. September 1975,
259–267. Wiesbaden: Reichert.

———. 1986. Zu den homerischen ῥοδοδάκτυλος-Komposita. In Annemarie Etter (ed.), o-o-pe-ro-si: Festschrift
für Ernst Risch zum 75. Geburtstag, 394–401. Berlin / New York: De Gruyter.

———. 1994. Alte und neue Fragen zum indogermanischen Nomen. In Jens E. Rasmussen (ed.), In honorem
Holger Pedersen. Kolloquium der indogermanischen Gesellschaft vom 25. bis 28. März 1993., 397–400. Wiesbaden:
Reichert.

Schmidt, Johaness. 1889. Die Pluralbildungen der indogermanischen Neutra. Weimar: Hermann Böhlau.

Schumacher, Stefan. 2004. Die keltischen Primärverben: Ein vergleichendes, etymologisches und morphologisches
Lexikon. Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft.

Sihler, Andrew. 1993. The Anatolian and Indo-European First Person Plural. In Bela Brogyanyi and Reiner Lipp
(eds.), Comparative-Historical Linguistics, Indo-European and Finno-Ugric. Papers in Honor of Oswald Szemerényi
III, 171–186. Amsterdam / Philadelphia: John Benjamins.

———. 1995. New Comparative Grammar of Latin and Greek. Oxford / New York: Oxford University Press.

Smith-Stark, T. Cedric. 1974. The Plurality Split. In Michael W. La Galy, Robert A. Fox and Anthony Bruck (eds.),
Papers from the Tenth Regional Meeting of the Chicago Linguistics Society, April 19–21, 1974, 657–671. Chicago:
Chicago Linguistic Society.

Starke, Frank. 1990. Untersuchung zur Stammbildung des keilschrift-luwischen Nomens. Wiesbaden: Harrasowitz.

Stassen, Leon. 2013. Comparative Constructions. In Matthew S. Dryer and Martin Haspelmath (eds.), The World
Atlas of Language Structures Online (WALS). Leipzig: Max Planck Institute for Evolutionary Anthropology. URL
[Link]

Steriade, Donca. 1988. Greek Accent: A Case for Preserving Structure. Linguistic Inquiry 19(2).271–314.

Stolz, Thomas, Cornelia Stroh , and Aina Urdze. 2011. Total Reduplication: The Areal Linguistics of a Potential
Universal. Bremen: Akademie.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 82

Strunk, Klaus. 1967. Nasalpräsentien und Aoriste: Ein Beitrag zur Morphologie des Verbums in Indo-Iranischen und
Griechischen. Heidelberg: Winter.

———. 1994. Relative Chronology and Indo-European Verb-System: The Case of Present-and Aorist-Stems. Jour-
nal of Indo-European Studies 22(3-4).417–434.

Stüber, Karen. 2002. Die primären s-Stämme des Indogermanischen. Wiesbaden: Reichert.

———. 2015. Die Verbalabstrakta des Altirischen (2 vols.). Bremen: Hempen.

Sturtevant, Edgar H. 1929. The relationship of Hittite to Indo-European. Transactions and Proceedings of the
American Philological Association 60.25–37.

———. 1933. Archaism in Hittite. Language 9.1–11.

Szemerényi, Oswald. 1996. Introduction to Indo-European Linguistics, 4th edn. Oxford: Oxford University Press.

Tichy, Eva. 1993. Kollektiva, Genus femininum und relative Chronologie im Indogermanischen. Historische Sprach-
forschung 106(1).1–19.

———. 1995. Die Nomina agentis auf –tar- im Vedischen. Heidelberg: Winter.

———. 2004. Indogermanistiches Grundwissen, 2nd edn. Bremen: Hempen.

———. 2006. Der Konjunktiv und seine Nachbarkategorien: Studien zum indogermanischen Verbum, ausgehend
von der älteren vedischen Prosa. Bremen: Hempen.

Tremblay, Xavier. 2003. La déclinaison des noms de parenté indo-européens en -ter-. Innsbruck: Innsbrucker Beiträge
zur Sprachwissenschaft.

Tribulato, Olga. 2015. Ancient Greek Verb-Initial Compounds: Their Diachronic Development within the Greek Com-
pound System. Cambridge / New York: Cambridge University Press.

Tucker, Elizabeth. 1988. Some innovations in the system of denominative verbs in early Indic. Transactions of the
Philological Society 86(2).93–114.

———. 1990. The Creation of Morphological Regularity: Early Greek Verbs in -éō, -áō, -óō, -úō and -íō. Göttingen:
Vandenhoeck & Ruprecht.

———. 2013. Old Persian asabāra– ’horse-borne’, RV kṣirapaká– ’cooked in milk’, and the Restructuring of Vowel
Quantities in Indo-Iranian Thematic Verbal Nouns. In Stephanie W. Jamison, H. Craig Melchert and Brent Vine
(eds.), Proceedings of the 24th Annual UCLA Indo-European Conference 2012, 229–242. Bremen: Hempen.

Uhlich, Jürgen. 2002. Verbal Governing Compounds (Synthetics) in Early Irish and Other Celtic Languages. Trans-
actions of the Philological Society 100(3).403–433.

de Vaan, Michiel. 2003. The Avestan Vowels. Leiden / Boston: Brill.

Vendler, Zeno. 1967. Verbs and Times. In Zeno Vendler (ed.), Linguistics in Philosophy, 97–121. (Revision of 1957,
”Verbs and Times”, Philosophical Review, 66.2, pp.143-160) Ithaca: Cornell University Press.

Veselinova, Ljuba. 2003. Suppletion in Verb Paradigms: Bits and Pieces of a Puzzle. Amsterdam / Philadelphia: John
Benjamins.

Veselinova, Ljuba N. 2013. Suppletion According to Tense and Aspect. In Matthew S. Dryer and Martin Haspelmath
(eds.), The World Atlas of Language Structures Online (WALS). Leipzig: Max Planck Institute for Evolutionary
Anthropology.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 83

Vijūnas, Aurelijus. 2009. The Indo-Europan Primary T-Stems. Innsbruck: Innsbrucker Beiträge zur Sprachwis-
senschaft.

Villanueva Svensson, Miguel. 2012. The ablaut of the middle root athematic presents in Indo-European. In H. Craig
Melchert (ed.), The Indo-European Verb: Proceedings of the Conference of the Society for Indo-European Studies,
Los Angeles 13-15 September 2010, 333–342. Wiesbaden: Reichert.

Vine, Brent. 1998a. Aeolic ὄρπετον and Deverbative -etó- in Greek and Indo-European. Innsbruck: Institut für Sprach-
wissenschaft der Universität Innsbruck.

———. 1998b. The Etymology of Greek κώμη and Related Problems. In Jay Jasanoff, Craig Melchert and Lisi Oliver
(eds.), Mír Curad: Studies in Honor of Calvert Watkins, 685–702. Innsbruck: Innsbrucker Beiträge zur Sprachwis-
senschaft.

———. 2002. On Full-Grade *-ro- in Greek and Indo-European. In Mark Southern (ed.), Indo-European Perspec-
tives, 329–50. Washington, D.C.: Journal of Indo-European Studies Monograph (43).

———. 2004. On PIE Full Grades in Some Zero-Grade Contexts: *-tí-, *-tó-. In James Clackson and Birgit A.
Olsen (eds.), Indo-European Word Formation: Proceedings of the Conference held at the University of Copenhagen
October 20th-22nd 2000, 357–80. Copenhagen: Museum Tusculanum Press.

———. 2013. A Hoarse of a Different Color (Plautus, Poen. 778 rāviō). In Adam Cooper, Jeremy Rau and Michael
Weiss (eds.), Multi Nominis Grammaticus: Studies in Classical and Indo-European Linguistics in Honor of Alan J.
Nussbaum, 315–328. Ann Arbor: Beech Stave.

Wackernagel, Jacob. 1877. Zum homerischen Dual. Zeitschrift für vergleichende Sprachforschung 23.302–310.

———. 1897. Vermischte Beiträge zur griechischen Sprachkunde. Programm zur Rektoratsfeier der Universität
Basel 3–62.

———. 1904. Studien zum griechischen Perfectum. Programm zur akademischen Preisverteilung (=Kl. Schr.
II.1000-21) 3–24.

———. 1905. Altindische Grammatik, Band II.1: Einleitung zur Wortlehre, Nominalkomposition. Göttingen: Van-
denhoeck & Ruprecht.

———. 1907. Indisches und Italisches. Kuhn’s Zeitschrift/Historische Sprachforschung 41.305–319.

———. 1926-8 [2009]. Lectures on Syntax: with Special Reference to Greek, Latin, and Germanic. Translated and
edited with notes and bibliography by David Langslow. Oxford / New York: Oxford University Press.

———. 1953a. Kleine Schriften: bd. 1. Ed. by Kurt Latte. Göttingen: Vandenhöck & Ruprecht.

———. 1953b. Kleine Schriften: bd. 2. Ed. by Kurt Latte. Göttingen: Vandenhoeck & Ruprecht.

Wackernagel, Jacob, and Albert Debrunner. 1954. Altindische Grammatik, Band II.2 Die Nominalsuffixe. Göttingen:
Vandenhoeck & Ruprecht.

Wagner, Laura. 2006. Aspectual bootstrapping in language acquisition: Telicity and transitivity. Language Learn-
ing and Development 2(1).51–76.

Watkins, Calvert. 1963. Preliminaries to a Historical and Comparative Analysis of the Syntax of the Old Irish Verb.
Celtic 6.1–49 (Repr. in Watkins 1994: i.3–51).

———. 1967. Latin sōns. In W.W. Arndt (ed.), Studies in Historical Linguistics in Honor of George Sherman Lane,
186–194. (= Selected Writings 2, pp.405- 413). Chapel Hill, N.C.: University of North Carolina Press.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 84

———. 1969. Geschichte der indogermanischen Verbalflexion (Indogermanische Grammatik: III/1, Formenlehre).
Heidelberg: Winter.

———. 1971. Hittite and Indo-European Studies: The Denominative Statives in *-ē-. Transactions of the Philolog-
ical Society 70(1).51–93.

———. 1994. Selected Writings. Volume I: Language and Linguistics. Lisi Oliver (ed.). Innsbruck: Institut für
Sprachwissenschaft der Universität Innsbruck.

———. 1995. How to Kill a Dragon: Aspects of Indo-European Poetics. Oxford / New York: Oxford University Press.

———. 1998. Proto-Indo-European: Comparison and Reconstruction. In Anna Giacalone Ramat and Paolo
Ramat (eds.), The Indo-European Languages, 25–73. London/New York: Routledge.

———. 2001. An Indo-European Linguistic Area and its Characteristics: Ancient Anatolia. Areal Diffusion as a
Challenge to the Comparative Method? In Alexandra Y. Aikhenvald and R.M.W. Dixon (eds.), Areal Diffusion and
Genetic Inheritance: Problems in Comparative Linguistics, 44–63. Oxford / New York: Oxford University Press.

———. 2008. Selected Writings. Volume III: Publications 1992-2008. Lisi Oliver (ed.). Innsbruck.: Innsbrucker
Beiträge zur Sprachwissenschaft.

———. 2011. American Heritage Dictionary of Indo-European Roots, 3rd edn. Boston MA: Houghton Mifflin Har-
court.

———. 2013. Aspects of the “expressive dimension” in Indo-European. Toward a comparative grammar of speech
registers. In Stephanie W. Jamison, H. Craig Melchert and Brent Vine (eds.), Proceedings of the 24th Annual UCLA
Indo-European Conference, 243–253. Bremen: Hempen.

Weiss, Michael. 2011. Outline of the Historical and Comparative Grammar of Latin, 2nd (corr. repr.) edn. Ann Arbor,
MI / New York: Beech Stave Press.

———. 2014. The Comparative Method. In Claire Bowern and Bethwyn Evans (eds.), Routledge Handbook of
Historical Linguistics, 127–145. New York: Routledge.

Whitney, William Dwight. 1889. A Sanskrit Grammar, 2nd edn. Cambridge MA: Harvard University Press.

Widmer, Paul. 2004. Das Korn des weiten Feldes: Interne Derivation, Derivationskette und Flexionsklassenhierarchie.
Aspekte der nominalen Wortbildung im Urindogermanischen. Innsbruck: Institut für Sprachen und Literaturen
der Universität Innsbruck.

———. 2005. Der altindische vr̥kī-Typus and hethitisch nakkī: Der indo-germanische Instrumental zwischen
Syntax und Morphologie. Die Sprache 45.190–208.

Willi, Andreas. 2007. Of aspects, augments, aorists–or how to say to have killed a dragon. In George Coulter (ed.),
Greek and Latin from an Indo-European Perspective, 34–48. Cambridge: Cambridge Philological Society.

———. 2011. Morphosyntaktische Überlegungen zum Ursprung des griechischen Futurs. In Thomas Krisch and
Thomas Lindner (eds.), Indogermanistik und Linguistik im Dialog, 605–615. Wiesbaden: Reichert.

———. 2012. Kiparsky’s Rule, thematic nasal presents, and athematic verba vocalia in Greek. In Philomen Probert
and Andreas Willi (eds.), Laws and Rules in Indo-European, 260–278. Oxford: Oxford University Press.

Willmott, Jo. 2007. The Moods of Homeric Greek. Cambridge: Cambridge University Press.

Wodtko, Dagmar S., Britta Sofie Irslinger , and Carolin Schneider (eds.). 2008. Nomina im Indogermanischen
Lexikon. Heidelberg: Winter.

Yakubovich, Ilya. 2010. Sociolinguistics of the Luvian Language. Leiden / Boston: Brill.
The Morphology of PIE J. Lundquist & A.D. Yates ǀ 85

———. 2013a. The degree of comparison in Luwian. Indogermanische Forschungen 118.155–168.

———. 2013b. Reflexes Of Indo-European ‘ē -Statives’ In Old Indic. Transactions of the Philological Society 1–23.

Yates, Anthony D. 2015. Anatolian Default Accentuation and its Diachronic Consequences. Indo-European Lin-
guistics 3.145–187.

———. 2016. Stress assignment in Hittite and Proto-Indo-European. Proceedings of the Linguistic Society of
America 1.25/1–15.

———. 2017. Against Root Faithfulness in Cupeño Stress. In Supplemental Proceedings of the 2016 Annual Meeting
on Phonology.

Yoshida, Kazuhiko. 1990. The Hittite Mediopassive Endings in –ri. Berlin / New York: de Gruyter.

———. 1993. Notes on the Prehistory of Preterite Verbal Endings in Anatolian. Historische Sprachforschung
106(1).26–35.

———. 2010. 1st Singular Iterated Mediopassive Endings in Anatolian. In Stephanie W. Jamison, H. Craig
Melchert and Brent Vine (eds.), Proceedings of the 21st Annual UCLA Indo-European Conference, 231–243. Bre-
men: Hempen.

———. 2013. The Mirage of Apparent Morphological Correspondence: A Case from Indo-European. In Ritsuko
Kikusawa and Lawrence A. Reid (eds.), Historical Linguistics 2011: Selected Papers from the 20th International
Conference on Historical Linguistics, Osaka, 25–30 July 2011, 153–172. Amsterdam / Philadelphia: John Benjamins.

Zerdin, Jason. 1999. Studies in the Ancient Greek Verbs in -SKŌ. Ph.D. diss., Oxford University.

———. 2002. The “Iterative-Intensives” in -σκον. Oxford University Working Papers in Linguistics, Philology &
Phonetics 7.103–130.

Zwicky, Arnold M. 1985. Heads. Journal of Linguistics 21.1–29.