Prepositions in Discourse

John Benjamins Publishing Company

This is a contribution from Functions of Languages 15:2

2008. John Benjamins Publishing Company
This electronic file may not be altered in any way.
The author(s) of this article is/are permitted to use this PDF file to generate printed copies to
be used by way of offprints, for their personal use only.
Permission is granted by the publishers to post this file on a closed server which is accessible
to members (students and staff) only of the authors/s institute.
For any other use of this material prior written permission should be obtained from the
publishers or through the Copyright Clearance Center (for USA: www.copyright.com).
Please contact [email protected] or consult our website: www.benjamins.com
Tables of Contents, abstracts and guidelines are available at www.benjamins.com
English prepositions in Functional

Discourse Grammar
Evelien Keizer
University of Amsterdam
Adpositions have always been problematic in terms of analysis and representation: should they be regarded as lexical elements, with an argument structure,
or as semantically empty grammatical elements, i.e. as operators or functions?
Or could it be that some adpositions are lexical and others grammatical, or even
that one and the same adposition can be either, dependent on its use in a particular context? In Functional Grammar (Dik 1997a,b) adpositions are analysed
as grammatical elements, represented as functions expressing relations between
terms (referring expressions). Various alternative treatments have been proposed
within FG, all of which, however, fail to solve all the problems, or address all the
relevant questions involved. This article offers an analysis of English prepositions
within the model of Functional Discourse Grammar (Hengeveld and Mackenzie
2006, 2008), based on the semantic, syntactic and morphological evidence available and fully exploiting the novel features of this model.
1. Introduction1
Most linguistic frameworks (formal or functional, formalistic or non-formalistic)
acknowledge the importance of differentiating between lexical (or content) elements and grammatical (or form) elements. Functional Grammar (Dik 1997a,b)
forms no exception, and distinguishes between the two classes of elements in underlying representation: whereas the former are analysed as predicates, the latter
are analysed as operators or functions (e.g. Dik 1997:159). Such an approach is unproblematic for those elements that are clearly lexical or grammatical. Thus, whatever their background, linguists seem to agree that full verbs, nouns and adjectives
are prototypical lexical elements, whereas affixes, auxiliaries and particles are generally regarded as grammatical elements. There are, however also elements in
some cases, classes of elements whose status as lexical or grammatical is less
straightforward. One of the most problematic classes of elements in this respect
Functions of Language 15:2 (2008), 216256. doi 10.1075/fol.15.2.03kei

issn 0929998X / e-issn 15699765 John Benjamins Publishing Company
English prepositions in Functional Discourse Grammar 217
is that of adpositions, which have been classified as either lexical or grammatical

(depending on the linguistic framework in question), and sometimes as both (with
some adpositions being classified as lexical and others as grammatical).
In all standard works on the theory on Functional Grammar (Dik 1978, 1989,
1997a,b) adpositional phrases are analysed as terms with a semantic function.
This corresponds to the view taken in most grammaticalisation studies, where adpositions are typically regarded as part of the grammatical inventory (e.g. Bybee
2003:145; Heine and Kuteva 2002a:34; Hopper and Traugott 1993:4). Such an
approach deviates from the position taken in both formal and cognitive semantics,
as well as in many more syntactically oriented frameworks, where adpositions are
often assumed to form a lexical category (alongside V, N and A). Unfortunately,
the classification of adpositions as either lexical or grammatical is rarely justified
by linguistic evidence (in the form of tests or a list of criteria) (see Keizer, 2007).
The controversial nature of adpositions is also evident from the fact that, in
the course of time, various alternative treatments of adpositions have been proposed within Functional Grammar. As will be demonstrated, however, these earlier proposals fail to solve, or often even address, the basic problems involved. The
aim of the present article is to propose an analysis and representation of English
prepositions and prepositional constructions within the newly developed theory
of Functional Discourse Grammar (Hengeveld and Mackenzie 2006, 2008) which
provides answers to the following questions:2
1. Is it plausible to assume that all prepositions in English are grammatical elements and that the wide range of meanings they express are represented in
underlying structure by a very limited number of highly abstract semantic
functions?
2. If, on the other hand, we choose to analyse prepositions as predicates, should
this be true of all prepositions, or is there evidence to suggest some prepositions are better regarded as more lexical and others as more grammatical (e.g.
Lehmann 2002:1,8)?
3. Is it justified and/or feasible to distinguish between lexical and grammatical
uses of one and the same preposition (e.g. on in the book on the table vs. his
dependence on his wife)?
4. What kind of evidence can be used to determine whether prepositions are
lexical or grammatical?
The article is organised as follows. Section2 presents an overview of the various
treatments of adpositions proposed in Functional Grammar so far. In Section3
the arguments and evidence provided in these earlier proposals are evaluated
and further evidence is brought in from semantics and cognition, morphology
and syntax. Together, this information forms the basis for a new proposal for the

All rights reserved
218 Evelien Keizer
treatment of English prepositions and prepositional constructions within Functional Discourse Grammar, made in Section4. Section5 offers some conclusions.
Finally, as it will be assumed that the answers to the questions addressed may differ
from language to language, this article will be concerned with English data only.
2. Adpositions in FG: An overview

In standard Functional Grammar (henceforth FG), adpositions, as grammatical
elements, are represented as semantic functions assigned to terms (i.e. referring
expressions); the expression the mouse under the table in (1a), for instance, is given
the representation in (1a), (cf. Dik 1997a:206207):
(1) a. the mouse under the table
a. (d1x1: mouseN: {(d1x2: tableN)Loc})
In (1a), under the table is represented as a term with the semantic function
Loc(ation); within the expression the mouse under the table, this locative term is
used in an attributive function, as the second restrictor on the term variable x1
(representing the referent of the entire term).3
This approach, however, turned out to be problematic in a number of respects.
The first problem concerns the idea that adpositional phrases are represented as
terms: whereas terms are defined as referring expressions, adpositional phrases
typically fulfil a non-referring (modifying or predicative) function (as is clear from
the representation in (1a), where the curly brackets indicate a predicative function). To account for the predicative function of terms with such semantic functions as location, direction, path and source, Dik (e.g. 1997a:206207) introduces
a term-predicate formation rule, converting such terms into one-place predicates.4
This does not, however, solve the problem: the input is still a term referring to
an entity. Take an example like the mouse under the table. The modifying phrase
under the table is represented as a term, the table, with the semantic function Location. However, the table referred to is not itself the location of the mouse; instead
the term the table refers to an entity in relation to which the location of the mouse
is indicated. Nor would such a view be logically acceptable: in a sentence like The
mouse under the table looked frightened, reference is made to a mouse and a table;
the place, or location, on the other hand, is not referred to.5 Instead the property of
being under the table is predicated of the mouse, a point which was convincingly
made by Jackendoff (1983:161163), and which, within FG, has resulted in the
introduction of the p-variable (Mackenzie 1992a; now the l-variable, Hengeveld
and Mackenzie 2006, 2008).

All rights reserved
Secondly, if adpositions are to be regarded as grammatical elements, this

means that they cannot have a meaning of their own: inserted by means of an expression rule, they can only derive their meaning from the semantic function assigned to the term as a whole. The number of semantic functions available is, however, limited. As a result, such an approach cannot do justice to (a) the differences
in meaning between different adpositions expressing the same semantic function
(e.g. at, in, by, above, below, beside, opposite, across, between, before, behind, beyond
and around, all of which express Location); (b) the large range of senses of one
single adposition (e.g. the many senses of over described in various studies from
cognitive grammar, such as Brugman 1988; Brugman and Lakoff 1988, Dewell
1994; Kreitzer 1997; Tyler and Evans 2001; see also evidence from psycholinguistic
research on the functional elements determining the use of spatial prepositions,
e.g. Coventry et al. 1994, Garrod et al. 1999, Coventry 2003).
Various attempts have been made to come up with a more satisfactory treatment of adpositions and adpositional phrases in FG. Thus it has been suggested
that, to tackle the problem of underspecification, more semantic functions need to
be distinguished to match the large number of adpositions. De Groot (1989:13),
for instance, introduces the semantic function Subessive to trigger the preposition
under. As pointed out by Mackenzie (1992b:4), however, this would lead to a proliferation of semantic functions, without, however adding to the explanatory value
of the theory (see also Samuelsdorff 1998:273).
For most authors challenging the standard FG view of adpositions, however,
the crux of the matter lies in the assumption that all adpositions are grammatical elements. Weigand (1990), for instance, proposes to represent prepositional
phrases through a combination of semantic functions (e.g. Loc, Path) and a limited number of prepositional predicates (e.g. on and under). A simplified example
is given in (2):
(2) a. The mouse ran from under the table.
a. e run (ag m mouse) ([path, from] p under (y table)))
where the preposition under is analysed as a one-place predicate heading a locative

expression p,6 which in turn is used as the argument with the semantic function
Path (expressed as from). The advantage of this approach is that we need only a
small set of semantic functions, as it is not necessary to have a separate semantic
function for each locative or directional preposition.
Taking the idea of lexical prepositions one step further, Mackenzie (1992b,
2001) proposes to represent most prepositions as predicates. In addition, Mackenzie distinguishes a small set of grammatical (spatial and temporal) prepositions, which he regards as direct realisations of the semantic functions Loc(ation),
So(urce), Path, All(ative) and App(roach):7

All rights reserved
220 Evelien Keizer
(3)

Loc:
So:
Path:
All:
App:
at (spatial and temporal)

from (spatial and temporal)
via (spatial), for (temporal)
to (spatial and temporal), until/till (temporal)
towards (spatial)
All other prepositions are assumed to be stored in the lexicon as one-place predicates. These lexical prepositions can then combine with one of the five semantic
functions to form a complex preposition (e.g. in + All = into). In addition, Mackenzies proposal can accommodate such embedded prepositional constructions as
in (4a) by analysing from as the direct realisation of the semantic function Source,
which is assigned to a place-denoting term headed by the prepositional predicate
under:
(4) a. from under the table
a. (d1pi: underP (d1xi: fj: tableN)Ref )So
Moreover, Mackenzie argues, such a system can account for the use of locative adverbs (as in I met him outside), as well as for the fact that some, but not all prepositions (i.e. only the lexical ones) allow adverbial modification (e.g. right behind the
door).8
Another interesting contribution to the debate is that by Bakker and Siewierska (2002), who advocate the view that, since the distinction between lexical and
grammatical elements (of any kind) is graded rather than strict, both ought to be
included in the lexicon. This means that all adpositions, whether predicates or
grammatical elements, will be assumed to have an entry in the lexicon, which will
include either a meaning definition (in the case of lexical adpositions) or an abstract predicate (in the case of a grammatical adposition). Thus, following Mackenzie (1992b), they consider above to be a lexical preposition, with a complex set
of semantic features, procuring it a full blown entry in the lexicon (Bakker and
Siewierska 2002:160). The meaning postulate for above will therefore consist of a
number of abstract locational predicates, as represented in (6a). The lexical entry
for the grammatical preposition at, on the other hand, will contain only the abstract predicate LOC, corresponding to the semantic function LOC. Whether or
not an adposition will appear as a predicate in the underlying representation of an
expression will therefore depend on the complexity of its lexical entry.

(6) a. LEX.ENTRY= above, [P], ((xi)ZERO (xj)REF, , [SUPERIOR (xi)ZERO

(xj)GOAL: (VERTICAL)DIR & NOT (CONTACT (xi)ZERO (xj)GOAL)]
b. LEX.ENTRY= at, [P], ((xi)ZERO (xj)REF, , [LOC (xi)ZERO (xj)REF]
Prez Quintero (2004) differs from all these proposals in that she regards all adpositions as lexical items, to be represented as one-place predicates. Her reasons for

All rights reserved
rejecting the standard approach are the same as those given by Mackenzie (1992b,
2001, 2002): firstly, even if it were possible in every utterance to trigger the correct
adposition by extending the set of semantic function, such an approach would
have very little explanatory value; secondly, since adpositions can be said to have
ascriptive value in the sense that they designate a relation between two entities,
and since, according to Mackenzie (2002:3), availability for the communicative
subact of ascription is the defining property of a predicate, adpositions qualify
for predicate status (Prez Quintero 2004:158). Prez Quintero sees no reason,
however, to distinguish a set of grammatical adpositions: the fact that some adpositions have a primitive meaning, she argues, does not necessarily mean that they
cannot function as predicates (Prez Quintero 2004:159). Prez Quintero does,
however, allow for a grammatical use of certain adpositions. In that case, the adposition performs a purely syntactic function that of indicating case and has no
independent meaning. In English only three prepositions can be used this way: to,
when used to express the semantic function of Recipient; by when used to express
the semantic function of Agent; and of in such constructions as a man of honour
(Prez Quintero 2004:163164).
Most of the arguments put forward in these proposals are semantic in nature
(primitive meaning, independent meaning). Sometimes syntactic criteria have
been used (combinatory properties, predicative use, modifiability), while some arguments are of a more general theoretical nature (lack of explanatory value). With
the notable exception of Mackenzie (1992b), however, no attempt has been made
to identify a set of criteria for systematically testing the lexical-grammatical status
of adpositions. The next section will be a first step in that direction.
3. The evidence
3.1 Grammatical vs. lexical adpositions
3.1.1 Semantics and cognition
One of the central questions in determining the lexical-grammatical status of adpositions is that of whether or not adpositions some or all can be said to
have lexical content. In an approach in which all adpositions are treated as grammatical items, the answer is clearly no. All of the alternative proposals mentioned,
however, argue in favour of a system in which there are grammatical as well as
lexical adpositions.9 Such a view does, of course, beg the question of which adpositions have meaning and which do not. Unfortunately, the various proposals,
even within FG, offer different classifications. As we have seen, Weigand (1990)
recognises only a limited number of lexical predicates (including on and under),

All rights reserved
222 Evelien Keizer
whereas Mackenzie classifies almost all English prepositions as lexical, reserving

grammatical status for a small set only (at, from, to, towards, via, for and until/till).
In other frameworks, we find other classifications. In Role and Reference Grammar (RRG; Van Valin and LaPolla 1997, based on Jolly 1993), be-at and be-via are
regarded as primitives; all other prepositions are predicative and are analysed in
terms of these primitives (e.g. from = BECOME NOT be-at). Dowty (1979), on the
other hand, qualifies be-at, be-in and be-on as primitives. Clearly, there is no consensus on which (kinds of) criteria are to be used to distinguish the two classes.
Nor is it just a problem of which adpositions belong to which class; matters
are further complicated by the fact that it is difficult, if at all possible, to determine
where one category ends and the other begins. As is well-known from grammaticalisation studies, grammatical elements typically develop out of lexical elements,
whereby the exact point of transition from one category to the other is difficult,
perhaps even impossible, to establish.10 This means that even within a particular
category of words, such as adpositions, some elements will be more lexical than
others, and that, even if we are able to provide a satisfactory classification of the
clearest cases, there may always be cases that defy straightforward categorisation.
Let us, however, for the purposes of this article, concentrate on the semantic evidence available for considering adpositions (some or all) as either grammatical or lexical. As pointed out before, the standard approach of analysing all
adpositions as semantically empty elements, expressing a limited number of semantic functions, is untenable because it leads to gross underspecification. This
underspecification is due to the fact that the different prepositions have different
meanings: in English, for instance, on, under, near and in do not merely indicate
location; they each have their own specific meaning, which determines their appropriateness in a particular context. Now it may be that, generally speaking, these
meanings are more abstract and schematic than those of most nouns or verbs; this
does not, however, mean that they lack semantic content. It might, for instance, be
argued that the primary meaning of a preposition like in (say, containment), is, in
fact, much more specific than the meaning of a noun like thing or a verb like do;
similarly, the meaning of under, even in its metaphorical use, seems to be more
tangible than that of a highly abstract noun like sort. In both cases, it can be argued
that the preposition makes a greater semantic contribution to the construction
they appear in than the particular noun or verb (cf. Langacker 1987:1819).11
If there is one preposition that is generally claimed to have little semantic content, it is the preposition of (e.g. Huddleston and Pullum 2002:658; Lindstromberg
1998:195; see also Prez Quintero 2004:164), partly because, unlike the other prepositions, it no longer has any concrete, literal sense, and partly because, possibly
as a result of this, it can be used to indicate a wide range of sometimes very disparate meanings (including possession, relative-of, part-of, property-of, type-of,

All rights reserved
quantity-of, as well as age, size, form, source, content, depiction; e.g. Huddleston
and Pullum 2002:442,477). Clearly, it will be difficult to find a definition which
would capture all of these meanings. The result would obviously be a very abstract
meaning definition, like the one offered by Langacker (1992b:487; cf. 1992a:296),
according to whom [i]t seems quite accurate to describe [all these relationships]
as designating some kind of intrinsic relationship between the two participants.
Interestingly enough, however, of is typically not included in the class of grammatical prepositions. So which semantic criteria are used to decide on the lexical
or grammatical status of a preposition? Mackenzie (1992b:10) justifies the recognition of a small set of grammatical prepositions on the basis that they are primitive
elements, which defy further etymological or syntactic analysis (with reference
to Kahr 1975:43) (for syntactic evidence, see Section3.1.2). But is this really true
of each of the prepositions classified as grammatical by Mackenzie? According to
Mackenzie (1992b:16), what the five spatial prepositions at, from, to, towards and
via have in common is the property that the Ref-argument of these prepositions
is always understood as a (zero-dimensional) point in space. Now, this has indeed been acknowledged to hold for at; Herskovits (1986:128140), for instance,
describes the functions of at as locating two entities at precisely the same point in
space and construing them as geometric points (cf. Lee 2001:23). In the case of
from, via, to and towards, however, some further meaning element is clearly present, which would seem to disqualify them as primitives. To confuse matters even
more, in RRG from and to are indeed analysed as decomposable lexical elements,
whereas via is treated as a primitive (like at).
One might, however, wonder whether qualifying the referent of via and from
as zero-dimensional points in space is, in fact, justified. Although both prepositions can certainly be used in combination with non-dimensional entities, this is
by no means a requirement. Lindstromberg (1998:3940) shows that the referent
of spatial from, indicating the starting point of some path, may be precisely named
(the non-dimensional argument in (7a)), or more or less vaguely located (by means
of the one-, two- or three-dimensional arguments in (7bd), respectively).
(7) a.
b.
c.
d.

Draw a line from point A to point B.

Draw a line from somewhere on line A to line B.12
The rocker slowly rose from the surface of the moon.
The noise came from the cave.
(Lindstromberg 1998:40)
Note also that neither the Oxford English Dictionary (OED) nor the Merriam-Webster Online Dictionary (Webster) restricts the meaning of from to non-dimensional
arguments only. OED describes the relevant meaning of from simply as indicating
the place, quarter, etc. whence something comes or is brought or fetched, while

All rights reserved
224 Evelien Keizer
Webster describes from as indicating the starting point of a physical movement.

The same can be said of the preposition via. Both dictionaries describe its meaning as by way of ; OED adds to this by the route which passes through or over
(a specified place). Once again this explains why, like from, via allows for a threedimensional argument. Lindstromberg (1998:128129) even regards this as the
central meaning of via:
via is vague about the exact relationship of the path in relation to the Landmark
[here: argument; MEK]. Thus, via can be used when the path goes through the
Landmark (which is probably the central meaning), or when the path only grazes
the Landmark, or even when the path goes near but does not touch the Landmark.
Of the five spatial prepositions, this leaves us with at as the last surviving serious
candidate for primitive (and as such grammatical) status. Unlike the other four
prepositions, the primary meaning of at does seem to be location at some nondimensional point in space. By not indicating the exact relationship between the
two entities, at is indeed less specific than, for instance, in (which indicates that
the argument is a container) or on (which indicates contact with a surface). Thus,
to say of a person that he or she is in the church is more specific than to say he or
she is at the church, which may also mean in the churchyard. The question is, however, whether this makes at less meaningful, less of a content word, than in or on.
Note also that it is not merely that at is more neutral or less specific in one or more
respects than other locational prepositions. In addition, the requirement that the
argument be conceptualised as a point in space places certain restrictions on its
use, making it more specific in this respect. Consider the following examples:
(8) a. John is at the supermarket.
b. John is in the supermarket.
As pointed out before, example (8a) does not actually require John to be inside the
building, thus making its use less specific than (8b). At the same time, however, the
requirement that the argument be conceived of as a point in space causes (8a) to
be appropriate only when the speaker is at some distance from the supermarket
which, as a result, loses its dimensions whereas (8b) is more likely to be uttered
when the speaker is standing just outside (or even inside) the supermarket (Lee
2001:23). The difference between the two prepositions is therefore not simply that
the one is more abstract than the other, but that they have different meanings.13 As
such, it is doubtful whether it is justified to analyse one of them as a lexical and the
other as a grammatical element.
Finally, note that an analysis of at as the direct expression of the semantic
function Loc suggests that at functions as a superordinate to all other locational

All rights reserved
prepositions, in the sense that it expresses an abstract feature shared by all other,
more specific, locational prepositions. In that case, it should, of course, be possible
to replace any of these more specific locational prepositions by at, thereby simply
making the expression less informative. This prediction is, of course, not correct,
precisely because at has a specific meaning of its own, at least one crucial feature of
which (geometric point in space) is not shared by other locational prepositions.
In sum, it looks as though there is no semantic justification for classifying the
five spatial prepositions at, from, via, to and towards as grammatical elements:
semantically they have little in common (apart from being locational, they each
have their own specific meaning); as such, there is nothing to set them apart, as a
group, from other prepositions.
3.1.2 Syntax
One of the most widely used arguments for regarding adpositions as a grammatical
category is that they form a closed class (e.g. Samuelsdorff 1998). In many cases,
however, this statement is weakened to allow for occasional additions. Thus Huddleston and Pullum (2002:603) describe prepositions as a relatively closed class,
while in Yule (1996:76), too, we read that prepositions form a closed class because
we almost never add new prepositions to the language. According to Langacker
(1987:19), on the other hand, such additions are in fact far from exceptional, and
the class of adpositions is, in fact, essentially open-ended.14
More convincing evidence for classifying adpositions, or at least some adpositions, as grammatical elements may be provided by their behaviour with regard to
modification. The underlying assumption is, of course, that grammatical elements,
lacking lexical content, cannot be modified (e.g. Bybee et al. 1994:7). However, at
least some adpositions, so it seems, do allow for (adverbial) modification. According to Halliday (1985:188189), for instance, some English prepositions can form
groups by modification, as in right behind (the door), not without (some misgivings)
and all along (the beach). Mackenzie (1992b) uses similar examples to support his
proposal for distinguishing two classes of prepositions: a small set of grammatical prepositions, which do not allow for adverbial modification, and a larger class
of lexical prepositions, which do. Thus, an expression like right behind the door
would be analysed as in (9), where the adverb right functions as a restrictor on
the prepositional predicate (and its argument), indicating that the spatial relation
holds with more than normal geometrical precision (Mackenzie 1992b:12):
(9) a. right behind the door
a. (f1: behindP (d1x1: f2: door)Ref : f3: rightAdv)
Mackenzie subsequently argues that his proposal correctly leads one to expect that
right cannot be combined with any of the five grammatical prepositions. Admittedly,

All rights reserved
226 Evelien Keizer
it is possible to say He came right from London, but in that case, the adverb right
should not, according to Mackenzie, be analysed as modifying the from-relation,
but rather as a modifier indicating manner (modifying not the preposition, but the
verbal predicate). In other words, in example (10a) right is analysed as a restrictor
on behind me, while in (10b) it modifies the verbal predicate come:
(10) a. They came from right behind me.
b. They came right from behind me.
Note, however, that in (10b) the adverb right is analysed as a restrictor not just on
the preposition, but on the preposition together with its argument, an interpretation which I think is correct (in most cases at least). This means that the restrictor has, in fact, an entire PP in its scope, and that the expression right behind the
door had perhaps better be represented as in (11) (notice the extra pair of brackets
around behind the door):15
(11) ((f1: behindP (d1x1: f2: door)Ref ): (f3: rightAdv))
If this is, indeed, the correct analysis of constructions of this kind, the presence of
an adverb tells us nothing about the status of the preposition; even if the preposition is believed to have a grammatical function, the PP is a lexical expression and
can, as such, be modified. This may then also account for the fact that right (or
similar adverbs) can, in fact, combine with such prepositions as at, towards and
via as well. Consider, for instance, the following examples, where just and utterly
occur in non-verbal predications headed by at:
(12) a. If you go up that road its just at the bottom of that <ICE-GB:S1A-071
#53:1:D>16

b. Im not utterly at the bottom of the road <ICE-GB:S1A-020 #281:1:B>
Now, it might be argued that the adverbs in these examples function as modifiers
of the entire predication rather than the PP; although the difference in interpretation may be negligible, what would be modified would be the assignment of the
property rather than the property itself. In other cases, however, such an explanation is less plausible. Consider in this respect the sentences in (13), where by far
the most likely interpretation is that in which the adverbs right and only function
to specify more precisely the locations at the back and at Tokyo University.
(13) a. Youll find Miss Jardine right at the back there <ICE-GB:W2F-011 #11>
b. Thank you for your letter proposing a branch of Dillons dealing with
customers orders only at or for Tokyo University. <ICE-GB:W1B-019
#15:2>
Similar examples can be found for the preposition toward, as shown in (14).

All rights reserved
(14) Their planned escape was cut short when they discovered that the only way
back was the roadway the vehicle had come from straight toward the
buildings and the mysterious alien ship, still performing its excavation.
Another type of syntactic evidence that can be used to distinguish between lexical and grammatical prepositions is that of mutual exclusivity: as pointed out by
Bybee et al. (1994:7), for instance, grammatical elements cannot co-occur with
members of the same class. This is indeed one of the criteria used by Mackenzie
to justify the distinction of a separate class of grammatical preposition. Thus, sequences of prepositions like from under or to behind are analysed as combinations
of a lexical preposition (under/behind) and a grammatical preposition (from/to)
(Mackenzie 1992b:11). For the sequence from under the table to behind the door
Mackenzie therefore proposes the following underlying representation:
(15) a. from under the table to behind the door
a. (d1p1: f1: underP (d1x1: f2: tableN)Ref )So (d1p2: f3: behindP (d1x2: f4:
doorN)Ref )All
The analysis in (15b) leads us to expect that from can combine freely with all lexical prepositions, but not with any of the other grammatical prepositions. This prediction does indeed seem to be borne out, since the sequences from at, from via,
from to and from towards seem to be impossible.
Nevertheless, there are weaknesses in this line of argumentation. First of all,
the analysis proposed would lead us to expect all grammatical prepositions to
occur in similar combinations; as Mackenzie himself admits, however, such sequences as in (15a) are basically limited to from (and occasionally to) + preposition. Similarly, not all lexical prepositions can co-occur with from either. Taken
together, these facts may lead one to suspect that there may be other reasons why
only certain combinations are allowed semantic reasons, for instance. One of
these reasons may very well be that (spatial) from requires a locative argument
(denoting typically, though not necessarily, a zero-dimensional point in space).
This would explain the absence of the sequences from via, from towards and from
to (where the second prepositions are analysed as grammatical elements indicating Path, Allative and Approach), as well as such sequences as *?from up, *?from
down, *?from across (where the second prepositions would be lexical prepositions
denoting Path).17 Given the kind of relations designated by the two prepositions, it
is difficult to find a plausible interpretation for these sequences.
This leaves us with the combination from at. The first question to be answered
is whether this combination is indeed entirely unacceptable. A quick Google
search shows that this seems not to be the case (although judgements may differ);
some of the attested examples are given in (16).18

All rights reserved
228 Evelien Keizer
(16) a. Removing page headers from at the bottom and top of the page while
printing the page.
(www.mail-archive.com/[email protected]/msg04647.html)
b. Make a right turn from at the United Way sign into the parking lot.
(www.californiateenhealth.org/events_060104.asp)
c. The Boat Crew Seamanship Manual can be downloaded from at the
Office of Auxiliary website (click on Publications). (www.cgaux.org/
cgauxweb/public/tbaskaux.shtml)
Nevertheless, it is undeniably true that the combination from at is relatively rare,

and that in many cases the presence of at may even seem superfluous. The same
seems to be true for certain other combinations of from with a locative preposition (from in, from on). So why is it that some combinations of prepositions are
frequently used and perfectly acceptable, while others are rarely used and at best
questionable? A number of factors may play a role here.19 One of these concerns
the availability of a default interpretation in the given context of the PP introduced by the second preposition. Consider the contrast between ??from on the
table, which is questionable (or at least exceptional), and from under the table,
which is perfectly ok. One way to account for this difference is to assume that with
an argument like table the on-relation is the default interpretation (the surface of
the table being its most prominent feature). In other words, the expression from
the table will be interpreted as meaning from on the table; as such, use of on would
be redundant. The same would apply to the combination from in: from the building will be interpreted as from in(side) the building; from his chair as from in
his chair. Any other relation, however, must be coded explicitly (e.g. from behind,
under, nearby the building/chair). The same explanation can be given for the rare
occurrence of the combination from at: a PP like from the station, for instance, will
always be interpreted as either from at the station or from in(side) the station.
A second, related, factor that seems to play a role here is that of locational opposition: the possibility of expressing a second preposition following from seems
to be determined by the availability of (inherent) contrast with some other locational preposition (or, perhaps more generally, some other locational concept).
Thus, the acceptability of from under the table can be accounted for by the presence
of the opposition on-under; the same can be said for such combinations of from
below, from behind, from inside etc. Note that this factor interacts with the availability of a default reading: where, given the nature of the object(s) referred to, one
of the two prepositions denotes the default relation, it is the non-default preposition that is more likely to be expressed (under rather than on, behind rather than in
front of etc.). Naturally, these default prepositions can also be explicitly expressed,
in which case not surprisingly a fuller, more emphatic, form of the preposi-

All rights reserved
tion tends to be preferred (hence from inside the building, rather than from in the
building; from on top of the table rather than from on the table).20
This second factor, too, helps to explain why the combination from at is rarely
used, as at does not have a locational counterpart: the opposite of at seems to be
not-at. Finally, such an account explains why from can quite easily be followed by
at in those cases where at (or the PP introduced by at) is modified, as the presence
of a modifier adds some the inherent opposition at lacks (hence from straight at
home, from almost at the middle).
So let us continue by considering such expressions combining the presence of
a modifier with that of a second preposition. By analysing the modifiers as having
scope over the PPs as a whole, and/or by analysing all prepositions as predicates,
we can also account for the following examples, where the grammatical spatial
prepositions distinguished by Mackenzie co-occur with both a modifier and another preposition:
(17) a. Subject: HOW TO MAKE MONEY FROM RIGHT AT HOME
(THOUSANDS) (http://vishnu.mth.uct.ac.za/omei/gr/BBS/Template/
messages/msgs31703.html)
b. Work a long-bladed knife into the sidewall, then saw a zig-zag, wavy, or
more detailed pattern, cut from almost at the center hole to right where
the tire begins to curve towards the sidewall. (www.suite101.com/print_
article.cfm/75/78343)
(18) a. Mostly, its light exiting the crown at oblique angles, as when the stone is
tipped appreciably away from straight toward the viewer. (www.ganoksin.
com/orchid/archive/200206/msg00067.htm)
b. But I hadnt prepared myself for an x-ray picture, showing heavy
calcium deposit from almost to the wrist to near my elbow. (www.
chesapeakestyle.com/celebrate/mar04.html)
In (17) the preposition at occurs in combination with the preposition from. In between the two prepositions we find an adverb, which can only be taken to modify
either the following PP or, if we assume at, from, via, to and toward(s) to be lexical
elements, just the second preposition. The examples in (18) show that this is also
is true for the prepositions to and towards.
A last source of syntactic evidence to be considered here is the predicative use
of locative adverbs, some examples of which are given in (19).
(19) a. John is in/out.
b. All the shades were down.
c. The kitchen is below.

All rights reserved
230 Evelien Keizer
Now clearly, there is a relation between these locative adverbs and prepositions, as
in many cases we are dealing with the same lexical element. To account for this,
Mackenzie (1992b:12) introduces the following predicate formation rule, which
changes the category of the input predicate from preposition to adverb, while reducing the number of arguments from one to zero:
(20) Locative Adverb Creation
Input: P (xi)Ref (e.g. below the deck)
Output: Adv (e.g. below)
What is interesting is that not all prepositions can serve as input to this rule; more
specifically, that the grammatical prepositions distinguished by Mackenzie do not
have an adverbial use (see (21a)). If we assume that locative adverbs are indeed
created through a predicate formation rule, this can easily be accounted for: since
these prepositions are not lexical elements, they cannot serve as input to the predicate formation rule. This does not explain, however, why the rule does not apply to
a large majority of lexical prepositions either (on the relevant spatial reading;
example (21b)):
(21) a. * John is at/from/via/to/towards.
b. * John is under/over/above/beneath/between/against/through/beside/
among/amid/beyond.
In Section4.2, I will come back to the use of Locative Adverbs and their relation
to prepositions. I will argue that there is no need for the Locative Adverb Creation
rule, for the simple reason that we are, in fact, dealing with one and the same lexical element on both the prepositional and the adverbial use.
3.1.3 Morphology
In English, prepositions do not easily combine with other elements to form complex
elements; nor do they often form the basis of derivation or conversion processes.
Occasionally, however, they do, in which case their morphological behaviour can
be taken as an indication of their status as lexical elements.21 In this section I will
first briefly consider compounds consisting of two prepositions, as well as combinations of prepositions with elements belonging to some other lexical category;
subsequently I will consider a few cases of derivation and conversion.
Unfortunately, morphological evidence is scarce. This is partly due to the
fact that the class of prepositions is relatively small since the number of spatial
and temporal relations that require linguistic coding tends to be limited. As far
as derivation is concerned, matters are further complicated by the fact that it is
not always clear which category an element belongs to: is the over- in overact or
overhang an adverb or a preposition? And what about in and out in the expression

All rights reserved
the ins and outs? More importantly, do we really have to assume different entries
for such forms as over, in and out in the lexicon, one for the preposition and one
for the adverb?
Nevertheless, it may be useful to look at the little evidence we have. English has
a number of preposition compounds, some spelt as one word (into, onto, upon),
some as two but clearly acting as one unit (out of, up to, off to). Mackenzie (1992b)
offers a very elegant analysis of the obvious analogy in the compounds into and
onto, which is argued to result from an interaction between semantic function
(Locative vs Allative) and lexical choice (in vs on) (Mackenzie 1992b:10). The
analogy can, however, equally insightfully be accounted for if to, like in and on,
is regarded as a predicate, in which case the complex prepositions into and onto
simply result from predicate formation. This, I feel, may even be a more attractive
analysis. After all, if into is derived from a combination of a lexical locative preposition in and the semantic function All, this suggests that an extra step is required
every time before the predicate into can be retrieved from the lexicon. This would
mean that the processing of into and onto is more complex than that of compound
(lexical) prepositions (e.g. upon, nearby). A more likely explanation may be to view
into and onto as compounds of two lexical elements, stored in and retrievable from
the lexicon. The same would hold for such fixed combinations as up to, down to, off
to, on to, in from, out of, as well as for any other prepositional compounds.
Prepositions can, however, also combine with other kinds of elements. A fairly
productive combination is (or was) that of the adverbs here, there and where and a
preposition. Thus we have, for instance, thereby, therefore, thereafter, therein; hereby, hereinafter; but we also have thereto (formal, legal); hereto (formal, legal, dialectal) and whereto, as well as hitherto, and even hereat, thereat, whereat, herefrom,
therefrom and wherefrom (formal and archaic/rare). Once again there is no reason
to assume that the prepositional elements by, in and after in these compounds differ in status from at and from: the most straightforward way to deal with these cases
is to regard them all as resulting from predicate formation. The same would hold
for the combinations toward(s), inward(s), outward(s), upward(s), downward(s),
forward(s) and backward(s).
Prepositions also play a modest role in derivational processes. The most obvious examples are those where the preposition functions as a prefix. Some of these
are, in fact, quite productive, such as, for instance, over, which, can be used in
combination with nouns (overcoat, overlord), verbs (overgrow, overlook),22 adjectives (overripe, overzealous) and adverbs (overmuch). The counterpart of over, under, is similarly productive (underdog, underestimate, underhand, underwater).23
Now obviously, not all prepositions can be used as prefixes. In addition to
over and under, prepositions like out (quite productive with verbs), up (upgrade,
update), on (on-base, on-beat), off (off-beat, off-hand) and by (with nouns, e.g. by-

All rights reserved
232 Evelien Keizer
line, bypass) can be found in this function. In Present Day English we do not,
however, find at, from, via and towards as prefixes, nor for that matter most other
prepositions (such as across, between, below, beneath, against, around or among).
In OE and ME, on the other hand, some of these prepositions do occur as prefixes.
The preposition at, for instance, was a frequent prefix in OE (meaning at, close
to, to); it can also be found in some words in ME, such as at-stand(en) to stand
close to, at-rech(en) to reach to, get at, at-fore(n) before and at-hind(en) behind.
Similarly, in ME, to used to occur as a prefix in combination with verbs, nouns,
adjectives and adverbs (resulting in such forms as to-come to happen/to arrive/to
come to, to-draught, following, retinue/resort, together and to-when until what
time/how long). All these facts seem to suggest that these prepositions must be
(or must have been) lexical elements which could function as the input of some
predicate formation rule.
The same conclusion can be drawn on the basis of the few cases of conversion
involving prepositions. Thus we have the nouns up(s) and down(s) and ins and
outs, as well as the nominal forms to(s) and fro(s). It is true that in the latter cases
we are dealing with fixed expressions (Mackenzie 1992b:12), but at some point
they must have been formed on the basis of the lexical elements in and out, and to
and from.
Finally, it is possible, though rare, for a preposition to function as the root of a
derivation. Examples are the adverb inly (inwardly, internally/intimately, closely,
fully) as well as the fairly recent (specialised) derived noun aboutness.
3.2 Grammatical vs. lexical use of adpositions
So far, we have been looking at arguments and evidence presented by linguists to
show that all or some adpositions should be seen as grammatical or lexical. There
is, however, another way of looking at the lexical/grammatical distinction, whereby this distinction applies to the use of adpositions rather than the adpositions
themselves. Thus in various theoretical frameworks it is assumed that modifying
PPs are introduced by lexical adpositions, whereas complements (of verbs, nouns
or adjectives) are introduced by grammatical adpositions (in this context also referred to as functional prepositions). Consider the examples in (22).
(22) a. the book on the table
b. Johns reliance on his own ingenuity
In (22a) the PP on the table functions as a modifier of the non-relational nominal

head table; in this construction the preposition on, indicating a spatial relation,
is often assumed to be a lexical preposition with its own argument structure. In
(22b), on the other hand, the PP on his own ingenuity is generally taken to be

All rights reserved
the complement of the relational (deverbal) head reliance; in this construction

the preposition on does not indicate location and is typically analysed as a purely
grammatical element, without argument structure. This is the position taken in
recent work within generative grammar (see Corver and Van Riemsdijk 2001).
A similar approach can be found in RRG, where a distinction is made between
predicative and non-predicative uses of adpositions:
Adpositions in the periphery of the clause are always predicative, while non-predicative adpositions normally mark oblique core arguments. A given adposition
may function either predicatively or non-predicatively, depending upon which
verb it appears with; for example, from is non-predicative when it occurs with a
verb like take, which licences a source argument, as in Sally took the book from the
boy, whereas it is predicative with a verb like die, as in She died from malaria. (Van
Valin and LaPolla 1997:52)
One important implication of such an approach is that not only does the lexical/
grammatical distinction not necessarily apply to an entire syntactic category (say
adpositions), neither does it apply to specific members of a category (say, the preposition on) in all their occurrences; instead, the status of the adposition depends
on the function of the PP it introduces (modifier or complement).
The distinction between modifiers and complements is, however, notoriously
difficult to make. This is particularly true for modifier and complement PPs within
the noun phrase, which appear in the same form and position. Various attempts
have been made to draw up lists of defining features, semantic and syntactic, to
distinguish between the two types of PP. The syntactic criteria suggested include
restrictions on the number of complements (but not modifiers), coordination
(possible between PPs with the same function, but not between complements and
modifiers), use of the pronoun one (combinable with modifiers but not with complements), and differences with regard to mobility within the clause (see e.g. Huddleston 1984; Radford 1988; Huddleston and Pullum 2002). As shown in Keizer
(2004, 2007), however, what are often given as rules are at best tendencies, and as
such make very unreliable tests.
Nor is it altogether clear to what extent the prepositions introducing complements and modifiers differ semantically. One of the semantic tests suggested to distinguish between the two types of preposition concerns their selection. According
to Huddleston and Pullum (2002: 440), for instance, complements must be licensed by the head noun; i.e. in the case of a complement PP, the head noun determines the choice of preposition (or the limited range of permitted choices).
The default preposition is of, and this is often the only possibility, as in the King of
France. In some cases, particularly with deverbal nouns, some other preposition is
selected (as in reliance on, collaboration with).

All rights reserved
234 Evelien Keizer
Although it is certainly true that in the case of complements the choice of

preposition is generally speaking more restricted, this choice can hardly be used
as a criterion for complement status of the PP. For one thing, we have seen that one
and the same preposition can be used to introduce both complements and modifiers. We can have the king of France, where of is claimed to introduce a complement,
as well as a king of wood or a king of considerable intelligence, where it introduces
PP-modifiers; we can have the book on physics as well as the book on the table. According to Huddleston (1984:264), however, we can still tell which prepositions
are grammatical and which are lexical, since lexical prepositions have their own
(typically locative) meaning, whereas grammatical prepositions, lacking semantic content, are determined by the head noun (see also Huddleston and Pullum,
2002:653ff). Now, this might be argued to apply to a preposition like on in the book
on physics and the book on the table: clearly the meaning of on is locative in the
latter construction, but not in the former. In many cases, however, the distinction
between lexical and grammatical prepositions is not so straightforward. This is
first of all due to the fact that it is not always possible to establish the lexical meaning of a preposition. The best example is, presumably, the preposition of, which
at an early stage lost its original locative meaning (close to away/from, indicating
source) and has since developed a host of meanings, including geographical origin, possession, part-of, attribute-of, type-of, quantity-of and depiction-of (Huddleston and Pullum 2002:658; see also 477; Lindstromberg 1998:195). Thus, in its
present-day use, of does not seem to have a literal (spatial) meaning; nor is it at
all clear which of its uses is to be regarded as primary. Nevertheless, it is generally
acknowledged that of can introduce both complements and modifiers. In the examples in (23), for instance, the of-PPs are usually regarded as complements; those
in (24), on the other hand, are generally taken to function as modifiers.
(23)

a.
b.
c.
d.
e.
f.
the sister of Mary

the house of her former husband
the spire of the cathedral
a glass of water
the death of the emperor
the conquest of Persia
(24)

a.
b.
c.
d
e.
f.
the wines of France

a man of honour
a girl of a sunny disposition
a boy of sixteen
a frame of steel
a matter of no importance

All rights reserved
In (24), Huddleston and Pullum (2002:659) argue, the preposition of is not selected by the head and thus makes an independent contribution to the meaning. This
claim may be justified for example (24a), where of, indicating geographical origin,
comes closest to having its original locative meaning. It may also be justifiable in
(24b), where, as pointed out by Huddleston (1984:264), of can be paraphrased by
means of the (lexical) preposition with and alternates with without (a man of/without honour). In examples (24cf), however, it is much more difficult to maintain
that of contributes independently to the meaning of the construction. In all these
cases the lexical meaning of of seems to be much more elusive than, for instance,
its meaning in examples (23b&c), where the semantic contribution of the preposition (possession, part-of) is much more tangible.
In other cases, the claim that the prepositions used to introduce complements
lack semantic content is even more doubtful. Huddleston and Pullum (2002:661),
for instance, acknowledge that on the basis of this criterion it is not always easy to
draw the distinction between with-complements and with-modifiers. The primary
meanings of with are generally assumed to be non-specific proximity (Lindstromberg 1998:208) and accompaniment (Huddleston and Pullum 2002:661). When
used to express this meaning, with is usually taken to introduce PP-modifiers (a
house with a garden, a mother with her child). Other closely related and very common uses of with involve cooperation (my work with John; collaboration with other
countries), while, as pointed out by Lindstromberg (1998:212), with has a similar
meaning when it accompanies communication and participation words, such as
communication, discussion, consultation, conspiracy, talk, negotiation and participation. In all these cases with clearly has a meaning related to the primary sense.
As such, there seems to be no justification for classifying these uses of with as
functional. Nevertheless, the PPs they introduce are typically analysed as complements.
The opposite is also true: modifiers are not always introduced by lexical prepositions. This, we have seen, is true of the preposition of, the meaning of which
proved to be difficult to establish in the first place. Another case in point is the
preposition by. Its primary sense is a locative one, indicating proximity in the horizontal plane (Lindstromberg 1998:141; Huddleston and Pullum 2002:655), which
we do, indeed, find in PP-modifiers (example (25a)), though most uses of by when
used to introduce PP-modifiers are metaphorical (e.g. example (25b)). When used
to mark the Agent (example (25c)), by is generally regarded as functional, despite
the fact that here, too, the PP it introduces is a modifier.
(25) a. The belt smashed into the table by his side and his whole body flinched.
<ICE-GB:W2F-001 #135:1>

All rights reserved
236 Evelien Keizer
b. Douglas Hurd a quiet man by nature a diplomat by training couldnt

resist pointing out on Guy Fawkes Day that hed had enough of the
explosions of the past few days <ICE-GB:S2B-007 #16:1:C>
c. We had a lecture by that guy Rene Weis over there <ICE-GB:S1A-006
#20:1:B>
In sum, distinguishing between lexical and functional prepositions or uses of

prepositions on the basis of their (lack of) semantic content is problematic because the notion of semantic content is obviously graded. Neither will it do simply
to say that complements are introduced by functional prepositions and modifiers
by lexical prepositions, first of all because the distinction between complements
and modifiers is similarly difficult to make (none of the tests suggested in the literature turns out to be reliable; see e.g. Keizer 2004, 2007), and secondly because,
as we have seen, there is no reason to assume a one-to-one relationship between
complement/modifier status and the distinction between functional and lexical
(uses of) prepositions.
4. English prepositions in Functional Discourse Grammar

4.1 Relevant features of Functional Discourse Grammar
4.1.1 A general outline of the model
Now that we have discussed several current approaches and reviewed the evidence,
it is time to consider the possibilities offered by Functional Discourse Grammar
(henceforth FDG) for a more unified and consistent treatment of adpositions and
adpositional phrases. Before we can do so, however, a brief summary of the relevant features of the model may be required.
FDG (Hengeveld 2004, 2005; Hengeveld and Mackenzie 2006, 2008) is a functional-typological grammar model that is intended as the successor to FG (Dik
1997a,b). Although it shares many of the basic assumptions and general features
of FG, it also deviates from FG in a number of important respects. One crucial
difference concerns the fact that, unlike FG, FDG has a top-down organisation:
the starting point of analysis is the Speakers intention, which, via the ordered
operations of (pragmatic and semantic) formulation and (morphosyntactic and
phonological) encoding, is gradually prepared for articulation. The operations
of formulation and encoding are FDGs main concern; they are part of the central component of the model, the grammatical component. This grammatical
component does not, however, operate in isolation, but interacts with three further components: a conceptual component, containing information concerning
the intention of the Speaker; a contextual component, containing linguistic and

All rights reserved
non-linguistic contextual information (including shared speaker-hearer longterm knowledge); and an output component, where phonetic, graphological, or
signed articulation takes place (see Fig.1).
Another feature that distinguishes FDG from FG is the internal organisation
of the grammatical component, where four modules interact to produce the appropriate linguistic forms. Each module deals with a different level of linguistic organisation. At the top of the model we find the pragmatic and semantic modules,
rendering the Interpersonal and Representational levels of analysis, respectively.
Both these levels are the output of the process of formulation. Information from
these levels then enters the phase of encoding: first that of morphosyntactic encoding, yielding a Morphosyntactic level, and finally that of phonological encoding, yielding a representation at the Phonological level.
Finally, these four levels not only receive their input from the other components and levels, but also from further information contained within the grammatical component; i.e. from the language users knowledge of his/her language.
This information, which takes the form of primitives, comes in three types. Firstly,
there is information about the patterns available in the language; at the Interpersonal and Representational levels, this information takes the form of frames (indicating acceptable valency structures), at the Morphosyntactic level of templates
(yielding the possible word order patterns), and at the phonological level of prosodic patterns. Secondly, there is information about the items available: lexemes
for formulation, auxiliaries for morphosyntactic encoding and (bound) morphemes for phonological encoding. Thirdly, each level has its own set of operators
and functions, i.e. abstract triggers of grammatical processes.
For the purposes of this article, we only need to consider two levels of analysis:
the Interpersonal level and, in particular, the Representational level. The Interpersonal level deals with all the formal aspects of a linguistic unit that reflect its role
in the interaction between Speaker and Addressee. In keeping with the overall
architecture of FDG, the units of discourse relevant at this level are hierarchically
organised. A simplified representation is given in (26):
(26) (MI: [(AI: [(FI: ILL) (PI) (PJ) (CI: [(TI) (RI) ])])])
At the highest level in this hierarchy we find the Move, which describes the entire segment of discourse relevant at this level. The Move consists of one or more
(temporally ordered) Acts (AI, AJ), which together form its (complex) Head. Each
Act in turn consists of an Illocution (FI), the Speech Participants (PI and PJ) and a
Communicated Content (CI). Finally, within the Communicated Content, one or
more Subacts of Reference (RI) and Ascription (TI) are evoked by the Speaker.

All rights reserved
238 Evelien Keizer
Conceptual component
Frames
Lexemes
Interpersonal and
representational
operators
Formulation
Interpersonal Level
Representational Level
Templates
Grammatical
morphemes
Morphosyntactic
operators
Morphosyntactic Encoding
Morphosyntactic level
Templates
Suppletive forms
Phonological
operators
c
o
n
t
e
x
t
u
a
l
c
o
m
p
o
n
e
n
t
Phonological Encoding
Phonological level
Output component
Figure1. FDG: general layout
The Representational level deals with the semantic aspects of a linguistic unit; it
is the level of denotation, where the appropriate lexemes are chosen to describe
entities in some (real or imagined) non-linguistic world. Since these entities are of
different orders, the linguistic units at this level differ with respect to the ontological category they denote. The hierarchical structure of the Representational level
is as follows:
(27) (pi: [(epi: [ (ei: [(fi)) (xi) (li) (ti) ])])])

All rights reserved
The most straightforward type of entity is the individual, a first-order entity (xi).
Entities of this type are concrete, can be seen and touched, can be located in time
and space, and can be evaluated in terms of their existence. State-of-affairs (SoAs)
are second-order entities (ei); they can be located in space and time, and evaluated in terms of their reality. Third-order entities, or propositions, are mental constructs (pi). They cannot be located in time or space; they can be evaluated in terms
of their truth. In addition to these three orders of entities (already distinguished
by Lyons 1977:442447), several more semantic categories can be distinguished.
Thus, there are separate units denoting location (li) and time (ti); since the concepts of time and space cannot be reduced to any of the other entity types, but instead rather specify dimensions of those entity types, they are seen as constituting
independent semantic categories (cf. Mackenzie 1992a for location and Olbertz
1998 for time). At the lowest level of the hierarchy, the entity denoted is a property
or relation (fi). Properties and relations do not have independent existence but can
only be evaluated in terms of their applicability. Finally, in between the proposition and the SoA we find another special semantic category, the episode (epi).
Episodes represent a combination of units of another semantic category; they may
be defined as a semantically coherent set of SoAs.
The primitives selected at the Representational level include the lexemes (predicates) used to describe the entity denoted and the appropriate frames, specifying
the number of arguments and their roles in the SoA. These roles are represented
by means of a restricted number of semantic functions (indicated by in (27), e.g.,
Actor, Undergoer, Location).
4.1.2 The representation of adpositions
At the Interpersonal level, the adpositional construction as a whole is analysed
either as a subact of reference (as in I put the book on the table, where on the
table is interpreted as referring to a place) or as a subact of ascription (when used
predicatively, as in The book is on the table). Note that by separating the referential
and ascriptive functions of adpositional phrases from their denotation, and by
analysing adpositional phrases as denoting places and times rather than objects,
one of the problems of the standard FG account has been solved: since adpositional phrases are no longer analysed as terms referring to first order entities, they
can be used either in a referential or in an ascriptive function as such the rule of
term-predicate formation is no longer required.
The crucial level of analysis for adpositions and adpositional phrases is, however, the Representational level: first, because it is at this level that a distinction
is made between different types of entity denoted (e.g., events, propositions, objects or locations); secondly, and even more importantly, because it is here that
the distinction between predicates and semantic functions is relevant. What we

All rights reserved
240 Evelien Keizer
know is that adpositional constructions are analysed as units denoting location or

time, symbolised by the variables l and t, respectively (Hengeveld and Mackenzie
2006). What still needs to be established, however, is whether to analyse adpositions (all or some) as predicates restricting these variables, as in (28a), or as semantic functions assigned to a term headed by a nominal predicate, as in (28a):24
(28) a. (1l1: atP (1l2: (stationN))Ref )
a. (1l1: stationN)Loc
In addition, we need to consider the question of which frames are needed for the
coding of adpositional constructions at the Representational level.
4.2 The proposal
4.2.1 Lexical prepositions and grammatical use
The proposal for the treatment of adpositions and adpositional phrases to follow
will start from the assumption that all English prepositions are lexical elements,
to be represented in the lexicon with a meaning definition. Such a treatment does
justice to the fact that all adpositions, in most of their uses, have some measure
of semantic content (Section3.1.1), as well as to the fact that adpositions share a
large number of syntactic and morphological properties with nouns, verbs and
adjectives (Sections3.1.2 and 3.1.3). In addition, it will be argued that a very limited number of adpositions can, in some of their uses, more plausibly be regarded
as grammatical elements, marking rather than describing the relation between a
head and some argument or modifier. In English the only prepositions to be analysed this way are of and by, and only in those cases where they are used to indicate
the semantic functions of Agent (or Positioner, Force, Processed or ) or Patient;
i.e. the semantic functions typically assigned to the first or second argument of a
verbal predicate.
This means, first of all, that adpositional phrases are analysed as expressions
with an adpositional first restrictor. Their prototypical denotation is a place, but
PPs may also denote a time or some other abstract entity. They typically function
as modifiers within a clause (e.g. I bought the book in Paris) or NP (e.g. the book
on the table). If, however, their presence is required to complete the meaning of
a verbal, nominal or adpositional first restrictor (head), they are to be analysed
as arguments (e.g. of verbal predicate like put, a nominal predicate like father, or
some other preposition, typically from).
However, before we turn to the representation of lexical adpositions, let us
first explore in some more detail the idea that some adpositions can be used grammatically, and that in English this use is limited to the prepositions of and by when
used to introduce terms with the semantic function of Agent, Positioner, Force,

All rights reserved
Processed, or Patient (i.e. the semantic functions that can be assigned to the first
and second arguments of verbal and adjectival predicates; see Dik 1997a:118).
Consider, for instance, the use of the preposition of in NPs with a deverbal nominal predicate. Some simple examples are given in (29):
(29) a. the fall of the city
b. the destruction of the city
Following Dik (1989, 1997a; see also Dik 1985a, 1985b), I will regard the nominal
predicates fall and destruction in these examples as derived from the corresponding verbal predicates to fall and to destroy, respectively, through a rule of predicate formation affecting the syntactic category and form of the derived predicate.
According to this rule, both the valency of the input predicate and the semantic
functions of the arguments are preserved in the process. The predicates fall and
destruction in (29) are formed as follows:
(30) Verbal Noun Formation:
a. Input: fallV (xi)Proc
Output: fallN (xi)Proc
b. Input: destroyV (xi)Ag (xj)Go
Output: destructionN (xi)Ag (xj)Go
However, the rule of verbal noun formation does not only affect the form of the input predicate, but also that of the arguments. This is accounted for by the Principle
of Formal Adjustment (Dik 1997b:157158):
(31) Principle of Formal Adjustment (PFA):25
Derived secondary constructions of type X are under pressure to adjust their
formal expression to the prototypical expression model of non-derived,
primary constructions of type X.
Modifiers within the term typically appear in the form of adjectives, genitives,
adpositions and relative clauses. The PFA accounts for the fact that the Agent and
Patient terms of a transitive input predicate (e.g. example (32a)) can appear as ofor by-PPs, or as genitival modifiers in the term (e.g. examples (32a-a)):
(32)

a. Caesar destroyed the city.

a. the destruction of the city by Caesar
a. Caesars destruction of the city
a. the citys destruction by Caesar
In other words, the grammatical use of adpositions is restricted to those cases where
the adposition expresses a semantic function which in the corresponding verbal
construction is not expressed by means of an adposition but by some (other) gram-

All rights reserved
242 Evelien Keizer
matical means, such as word order, agreement or case. It seems therefore reasonable
to assume that when they appear in the derived construction the prepositions of and
by do not have semantic content, but are indeed simply expressions of the semantic
relations in question. This approach is further corroborated by the fact that it is only
these semantic functions (i.e. Agent, Positioner, Force, Processed, or Patient in
FG; now Actor and Undergoer) which in the nominal domain can be expressed by
the likewise grammatical means of a genitive (e.g. examples (32a& a)).
Nor does the fact that in these cases the PPs are represented as terms form a
problem. Unlike in the case of locative PPs, the semantic functions of Agent, Patient etc. do apply to the term itself. Thus, whereas in a construction like the mouse
under the table, the term the table does not itself refer to the location, but rather as
the object in relation to which the location of the mouse is defined, in an expression
like the destruction of the city by Caesar, the terms the city and Caesar do refer to the
Patient (Undergoer) and Agent (Actor), respectively, of the predication denoted.
Observe furthermore that the current proposal is in line with the observations made earlier about the lack of a one-to-one relationship between the distinction between complements and modifiers and the distinction between lexical and
grammatical adpositions. Thus, as pointed out before, of does not always have a
grammatical function: when used in combination with a non-derived head (as in
the house of my parents, a man of honour, the spire of the cathedral, a girl of eighteen,
etc.), the preposition of will be analysed as a lexical element, indicating an intrinsic relation between the head and the argument of the preposition. Similarly, by
can be used both grammatically and lexically; in both cases, however, it typically
introduces modifiers.
Finally, we need to account for the fact that in most cases prepositions require
an NP-complement (we cannot say *I saw the book on), but that many of them also
can stand by themselves (e.g. John is in). With nominal heads, the opposite is true:
while in most cases omission of the PP is, syntactically and semantically at least,
unproblematic (we can refer to the book as well as to the book on the table), in other
cases the PP is more or less required to complete the meaning of the head (we do
not, as a rule, refer to a destruction). Note, once more, that this is not directly related to the lexical or grammatical status of the preposition (Sections3.1.2 and 3.2).
Clearly what we need is a flexible system, one which allows us to use one and
the same (adpositional or nominal) predicate with and without a complement.
FDG seems to allow for the required degree of flexibility: unlike in FG, where
predicates were stored in the lexicon along with a predicate frame, in the new
model predicates and frames are stored separately. By allowing predicates to combine with more than one frame, the system becomes both more flexible and more
efficient. What follows is a more detailed account of the some of the frames needed
and the ways in which they can be combined.

All rights reserved
4.2.2 Frames in FDG

4.2.2.1 Preposition frames. For the purposes of the present discussion, it will be
argued that two prepositional frames need to be distinguished:
a. Type 1: a frame for PPs headed by a transitive preposition (i.e. a one-place
prepositional predicate);
b. Type 2: a frame for PPs headed by an intransitive preposition (i.e. an avalent
prepositional predicate).
In combination with the various frames for nominal (and verbal) constructions
(some of which will be discussed below), these frames can account for the different
constructions and combinations prepositions and prepositional phrases appear in.
Type 1:
(33) PredP (i)Ref26
In this frame the preposition has its prototypical (relational) function: it indicates
a relation between some entity or SoA and the argument of the preposition. Every
preposition can be inserted in this frame. The argument is always assigned the semantic function Ref (as the entity with reference to which the relation indicated by
the preposition is asserted to hold) and may be of any entity type (i.e. the preposition phrase as a whole will denote a place, but the argument may denote any type
of entity, e.g. an event (during the party) or a proposition (in this proposal)). This
is the frame which is selected for PPs functioning as PP arguments or modifiers
within the term. Some examples are given in (34):
(34)

a.
a.
b.
b.
the book on the table

(l1: onP (1x1: tableN)Ref )
the noise from under the table
(l1: fromP (l2: underP (1x1: tableN)Ref )Ref )
Type 2:
(35) PredP
Here the preposition is used without an argument; this is the frame which is selected in the case of so-called prepositional adverbs. Consider the examples in (36):
(36) a. John is in.
b. All the shades are down.
c. The kitchen is below.
These constructions will be analysed as non-verbal predications the non-verbal

predicates of which consist of an intransitive preposition only (cf. Lee 1998). In
each case the argument is, to some extent, inferrable: in (36a) John is in the same

All rights reserved
244 Evelien Keizer
building as the speaker; in (36b) the shades are down in relation the vertical dimension of the window/room; and in (36c) the kitchen is below either in relation to the
position of the speaker, or in relation to the vertical dimension of the building.
The proposal has a number of advantages. First, it treats prepositions in the
same way as any other lexical category in allowing for a transitive (relational) and
an intransitive (non-relation) use. Secondly, by regarding the elements in, on and
below here as prepositions rather than adverbs, it accounts for the fact that in many
cases one and the same element is used in all cases. Recall that in Mackenzies
(1992b:12) proposal this parallelism was accounted for by the introduction of a
predicate formation rule converting prepositions into adverbs. In the present proposal there is no need for such a rule: rather than assuming that in constructions
such as (36) valency reduction has taken place, it will now be assumed that these
constructions result from a combination of the prepositional element in question
and the proper (intransitive) frame. Since prepositions are regarded as lexical elements, their functioning as non-verbal predicates poses no problem.27 Nor, as
pointed out before, do we need a separate rule of Locative Adverb Creation.
Finally, as mentioned before, not all prepositions allow the use of this frame
(e.g. *John is at/from, *The journey is via, *The mouse is under); and even if they do,
not all of them can occur in all syntactic environments. Thus, unlike PPs of Type
1, the intransitive frame is hardly ever selected for modification or complementation within the term (i.e. we do not have *the book on, *the mouse in or *the shades
down; note, however, that the kitchen below and (on a temporal use) the day after
are fine). As to why it is that not all prepositions can be used in this way, one can
only speculate; as pointed out before, there is little reason to assume that this has
anything to do with the status of the element as grammatical or lexical.
4.2.2.2 Noun frames. Let us finally consider how the approach advocated would
work in the case of N + PP constructions. As it will not be possible to discuss all
possible combinations of nouns and prepositional phrases, discussion will be confined to three common configurations, whose treatment may serve as a basis for
other, more complex, combinations.
(i) Type 1: a frame for NPs headed by non-relational nominal predicates and
modified by a prepositional phrase;
(ii) Type 2: a frame for NPs headed by relational (typically deverbal) nouns with
one (inherited) PP argument; the preposition is selected by the head (inherited from the verb along with the argument);
(iii) Type 3: a frame for NPs headed by relational (typically deverbal) nouns followed by arguments introduced by of or by; these prepositions (in this use at

All rights reserved
least) are grammaticalised and as such expressions of semantic functions (Actor/Undergoer).

Type 1:
(37) PredN: (li: PredP (i)Ref )
In (37) a non-relational nominal head is modified by a PP functioning as a (nonfirst) restrictor. As illustrated in (38a), this PP-modifier is headed by a prepositional predicate which itself takes an argument (typically a term, but it may also
be a PP or adverb; see e.g. Lee 1998:140); this argument is assigned the semantic
function Ref. The form of the prepositional predicate depends on the relation (locative, temporal, or otherwise) between the head and the argument of the preposition. Although the head may, in some cases, place (semantic) restrictions on the
choice of preposition, the preposition is not selected by the head.
(38) a. the book on the table
a. (1x1: [bookN: (l1: onP (1x2: tableN)Ref )])
In constructions such as (38a), the preposition is represented as a lexical element.

As such, it is assumed to have semantic content: it functions to indicate a relation
(locative, temporal, or otherwise) between two entities. Note that the preposition
is not necessarily used in its primary (literal) sense. Which interpretation is given
to the preposition depends on the nature of the two entities; thus in a construction
like the class on Tuesday the preposition will be given a temporal interpretation
on account of the predicational head class (an event) and the temporal argument
Tuesday.
Since the prepositions in these constructions are analysed as predicates, we
can easily account for the fact that they can be modified, irrespective of whether
the adverbs in question are analysed as modifying just the prepositional predicate
or the PP as a whole. Opting for the latter solution (see discussion of example (10)
above), we can (roughly) represent the NP in (39a) as in (39a):
(39) a. the girl right at the back (cf. (13a))
a. (1x1: [girlN: [(l1: atP (1x2: back)Ref ): (rightAdv)]])
Similarly, co-occurrence with other prepositions, whatever the combination, is unproblematic: the first preposition simply takes the following PP as its argument. As
can be seen in (40), this means that such constructions contain two Ref arguments:
one a regular term functioning as the argument of the second preposition, the
other a prepositional phrase functioning as the argument of the first preposition:
(40) a. the noise from behind the door
a. (1e1: [noiseN: [l1: fromP (l2: behindP (1x1: doorN)Ref )Ref ]])

All rights reserved
246 Evelien Keizer
Moreover, such an approach can deal effortlessly with such examples as (41), which
involve both modification and co-occurrence of prepositions:
(41) a. a close-up view from right at the edge
(www.philaprintshop.com/niagara.html)
a. (1e1: [viewN: close-upA: [(l1: fromP ((l2: atP (1x1: edgeN)Ref ): (rightAdv))
Ref )]])
b. heavy calcium deposit from almost to the wrist to near my elbow. (cf.
(18b))
b. (1x1: [calcium depositN: heavyA: [(l1: fromP ((l2: toP (1x2: wrist)Ref ):
(almostAdv))Ref ) (l1: toP (l2: nearP (1x2: elbow)Ref )Ref )]])
Finally, the fact that prepositions can combine with other prepositions, as well as
with other types of predicate, to form derived predicates (compounds as well as
derivations) can also easily be accounted for: since all prepositions are predicates
stored in the lexicon, they are available as input for predicate formation rules.
Type 2:
(42) PredN (fi: PredP (j)Ref )Ref
In Type-2 constructions, the PP functions not as a restrictor, but as a prepositional argument within the term. A possible instantiation of this frame would be
the term in (43a), which would be given the underlying structure given in (43a).
Here the nominal (deverbal) first restrictor reliance is analysed as taking an PP
argument, which it will be assumed to have inherited from the verbal predicate
rely. This PP argument (on his ingenuity) will be assigned the semantic function
Ref. It is headed by the prepositional predicate on, which, in turn, takes an argument, here in the form of a term (his ingenuity); this argument, too, is assigned the
semantic function Ref.
(43) a. his reliance on his ingenuity
a. (1e1 : [(relianceN (l1: onP (f1: [ingenuity: his])Ref )Ref ): his])
A final difference between Type-1 and Type-2 constructions is that in the latter the
nominal predicate selects (or has a strong preference for) a particular preposition
as the head of the prepositional argument (e.g. reliance on, appeal for, reference to,
interest in, conversation with, protest against). Although in these constructions the
preposition can still have its primary, spatial meaning (e.g. their emergence from
the room), in most cases the preposition is not used in its primary sense. Nevertheless, it is still regarded as a lexical element (i.e. as a metaphorical extension, but
still involving a spatial aspect). As in the case of non-prototypical uses of prepositions introducing modifiers (e.g. the class on Tuesday), the exact interpretation is
triggered by the context: in an example like (43a) it is the co-occurrence of the

All rights reserved
preposition on with the abstract noun ingenuity which leads to the correct nonprimary interpretation of on.
Syntactic evidence for analysing the prepositions in these constructions as
predicates is scarce. Semantically, however, an analysis of these prepositions as
predicates is far more plausible. First of all, we have seen that they are not meaningless elements; quite often the relation denoted is direct linked to the primary sense
of the preposition. At the same time it will be clear that where the prepositions
express more abstract relations, it will be difficult (and cumbersome) to have all
these relations represented by different semantic functions.
Observe finally that from a purely formal point of view, Type-1 and Type-2
frames yield identical constructions: in both cases the construction takes the form
of a nominal head and one or more PPs. Nevertheless, different analyses are called
for on account of the different status of the PPs. Although it is true that the criteria
for distinguishing between argument (complements) and modifiers are far from
clear-cut, the distinction is nevertheless a valid one, since the syntactic behaviour
of (typical) modifiers differs significantly from that of (typical) arguments. Despite
the many in-between cases, this difference therefore ought to be reflected in underlying representation.
Type 3:
(44) a. PredN (i)A/U
b. PredN (i)U
intransitive input verb

transitive input verb
In constructions of this type we are once again dealing with derived nominal predicates which function as one-place predicates. Where the input verb is an intransitive verb, the only argument of the verb (Actor, Undergoer) is inherited by the
derived nominal predicate. The frame for such a construction is given in (44a); an
example can be found in (45a).
(45) a. the arrival of the train
a. (1e1 : [arrivalN (1x1: train)U])
In the case of a transitive input predicate, it will be assumed that only the second
argument of the verb (the Undergoer) is inherited; any other arguments of the verb
(e.g. Actor, Recipient) will be regarded as (implied) modifiers in the derived nominal structure (for a discussion, see Mackenzie 1996). The basic frames for constructions of this kind is given in (44a); an instantiation of this frame can be found
in (46a). It is, of course, possible, to express both the Undergoer and the Actor. In
such constructions, the (optional) Actor PP will be given the status of a restrictor;
its semantic function inherited from the frame of the input verbal predicate
triggers the use of the grammatical element by. An example is given in (46b).

All rights reserved
248 Evelien Keizer
(46)

a.
a.
b.
b.
the treatment of the patient

(1e1: [treatmentN (1x1: patient)U])
the treatment of the patient by the doctor
(1e1: [treatmentN (1x1: patient)U]: (x2: doctor)A)
The reasons for assigning different underlying representations to these PPs will by
now be clear. As pointed out before, the argument-modifier distinction is based
on differences in syntactic behaviour (concerning its position in and outside of the
term, coordination, anaphora etc.) of the PPs in question. Only the first argument
of nominal predicates derived from one-place verbal predicates, and the second
argument of nominal predicates derived from two-place verbal predicates behave
consistently as arguments of the derived nominal head. The status of the preposition (lexical or grammatical) is not directly related to the argument-modifier
distinction. Instead, grammatical status is confined to those expressions which
appear as terms in the corresponding verbal constructions, i.e. the first and second
arguments. Within the nominal domain these arguments take the form of PPs.
However, since they fulfil the same function with regard to the head, it will be assumed that the prepositions in question (by and of) do not denote this function;
instead they will be taken to express the (inherited) semantic functions Actor and
Undergoer, their presence the result of the PFA.28 Analysing these expressions as
terms is further justified by the fact that, unlike in the case of their locative counterparts, it is the entities referred to by these terms that function as the Actor,
Undergoer etc.
5 Conclusion
In this article I have tried to show the need for a treatment of English prepositions
and prepositional phrases that differs radically from the way these are treated in
the standard FG approach, and which is also crucially different from any of the
alternatives offered since then. I started by identifying the various problems and
reviewing all the evidence and arguments used in previous proposals as well as additional evidence from semantics, syntax and morphology. Making use of the new
possibilities offered by the model of FDG, I have subsequently proposed an analysis which, as I hope to have demonstrated, can remedy most of the shortcomings
of previous attempts, while offering a unified and consistent treatment of prepositions and prepositional phrases. The most important features of this proposal are:
Prepositions are lexical elements; they are predicates stored in the lexicon,
provided with a meaning definition.

All rights reserved
Prepositions can be transitive (prototypical use) or intransitive. This means we

have two frames for prepositions, one in which the preposition functions as a
one-place predicate and one in which it is avalent. In this way, prepositions are
no different from the other syntactic categories (N, V and A).
The arguments of prepositions are typically terms; they may, however, also
take the form of prepositions or prepositional phrases, adverbs or clauses.
Some prepositions (English by and of) allow for a grammatical use, their
presence being a result of PFA. This use is restricted to those cases where the
preposition heads a PP which functions as the inherited argument of a deverbal/deadjectival nominal head with the semantic functions Actor or Undergoeror. In those cases, the PPs are analysed as terms with semantic expressions,
and the prepositions are grammatical elements expressing these semantic
functions.
Elements traditionally categorised as adverbs (in constructions like John is in)
are simply to be regarded as instances of intransitive prepositions.
The current proposal can account for many of the specific properties of prepositions and prepositional phrases. The fact, for instance, that prepositional phrases
can be used both referentially (as subjects or complements) and ascriptively (as
modifiers or non-verbal predicates) can easily be accounted for. Similarly, the fact
that prepositions and prepositional phrases can be modified is no longer a problem; the same is true for the occurrence of sequences of two or more prepositions.
In addition, the proposal can deal with those instances where the NP-arguments
of prepositions are not expressed; the same is true for unexpressed PP-arguments
of relational nominal predicates. Moreover, the proposal offers an explanation for
the fact that many prepositions are identical in form to what were traditionally
regarded as adverbs.
Finally, it has been shown that the analysis proposed provides a unified treatment for a number of constructions containing prepositions and prepositional
phrases, including constructions with locative adverbs (by distinguishing a frame
for the intransitive use of prepositions) and nominal constructions with PP arguments and modifiers (by means of different frames for NPs headed by relational
and non-relational nominal predicates).

Notes
Received 22 August 2006.

Revised version 21 April 2007.
1. I would like to thank Lachlan Mackenzie, as well as the two anonymous reviewers, for insightful comments on earlier versions of this article. All remaining errors and inconsistencies
are, of course, entirely my own.

All rights reserved
250 Evelien Keizer

2. It will be clear that for any discussion of the lexical/grammatical status of an element, one
needs to know what exactly is meant by these terms, and on the basis of which criteria the distinction can be made. In this article, a number of semantic, syntactic and morphological criteria
will be applied to prepositions to justify their qualification as grammatical or lexical. For a more
general discussion of the lexical-grammatical dichotomy in FDG, see Keizer (2007).
3. For an illustration of how such an approach might apply to English temporal modifiers (including prepositional phrases), see Connolly (1994, 1995).
4. For discussion of this rule, see also Mackenzie and Hannay (1982:47) and Mackenzie
(1983).
5. It is, of course, possible to refer to places, as in Under the table is a good place for a mouse to
hide. In that case, however, the PP functions not as a modifier, but as a participant in the SoA
(with the semantic function 0).
6. Note that in other approaches (e.g. Zeller 2001; also the treatment of predicative/adjunct
prepositions in RRG) prepositions are often regarded as two-place predicates, indicating a relation between two arguments. Like Weigand, I prefer to see them as one-place predicates, requiring only one complement. When this position is filled, the PP as a whole can function as a
modifier of some other predicate (the book on the table), as a predicate (the book is on the table)
or as an argument (I put the book on the table).
7. For a similar treatment of spatial and temporal prepositional phrases in French, see Franois
(1996).
8. Samuelsdorff (1998) also assumes there to be a small set of semantic functions that are directly realised by a preposition. He rejects the idea of representing prepositions as one-place predicates because predicates, he argues, express a property of their argument, which prepositions do
not: in the phrase under the table, under does not express a property of the table. I do not believe,
however, that this is a valid argument. Relational predicates do not express a property of their
argument; instead predicate and argument together express a property of some other entity (the
entity they restrict or modify).
9. A similar view can be found in the analyses of prepositional phrases developed in other
theoretical frameworks, e.g. Lexical Functional Grammar (Bresnan 1982, 2001), Generative
Grammar (Van Riemsdijk and Williams 1986; Radford 1988; Corver and Van Riemsdijk 2001),
Generalised Phrase Structure Grammar (Gazdar et al. 1985), Head-Drive Phrase Structure
Grammar (Pollard and Sag 1984), as well as in traditional, descriptive approaches (e.g. Huddleston and Pullum 2002).
10. Note that although grammaticalisation studies emphasise the fact that the change from lexical to grammatical is, indeed, a gradual one and that we are dealing with a continuum rather
than with two clearly distinguished sets (e.g. Bybee 2003; Bybee et al. 1994; Heine and Kuteva
2002a, 2002b; Hopper 1991; Hopper and Traugott 1993; Lehmann 1989, 1995), they do, at the
same time, continually classify elements as either lexical or grammatical, without, however, giving any indication of where and how to draw the boundary. For a more principled approach,
see Keizer (2007).

All rights reserved
11. Compare also Halliday and Hasan (1976:274), who describe general nouns, such as thing,
people, place or matter, as borderline case[s] between a lexical item (member of an open set)
and a grammatical item (member of a closed set).
12. Observe that such an analysis is corroborated by the fact that the second PP can be replaced
by the pronominal forms here and there.
13. Lindstrombergs choice of example here is unfortunate as it does not really prove his point;
after all, the head of argument of the preposition, somewhere, still designates a specific point
along the line. A better example would be He lives three miles from the border, where border does
not designate a point but a line, even though the hearer may choose to interpret the construction
as meaning He lives three miles from a particular point along the border. In that case, however,
the aspect of non-dimensionality is not part of the meaning of from, but inferred from a combination of lexical, contextual and general knowledge.
14. One of the reviewers offered an alternative explanation for the difference in meaning between (8a) and (8b). According to him/her, at does not indicate an exact geometric relationship between the two entities; instead its meaning is functional rather than locational. Thus a
sentence like John is at the piano is only appropriate if John is positioned canonically to use the
piano for its intended purpose. Similarly, the reviewer would prefer to explain the difference
between (8a) an (8b) as indicating a difference in the interaction between the entities involved:
whereas at seems to indicate that John is engaged in the usual activities associated with supermarkets (i.e. hes shopping), in appears neutral with respect to his activities (i.e. he might have
ducked in to get out of the rain).
I quite agree with the reviewers interpretation of the sentence John is at the piano. However,
this particular functional element of at seems to be present only in a restricted number of instances: e.g. only with [+human] subjects (note, for instance, its absence in such expression as
the house at the corner or the light at the end of the tunnel) and only in certain combinations and
contexts. I am not really convinced that it is this functional element that accounts for the difference between (8a) and (9b); to me the use of at still seems to be primarily spatial here. Note in
his respect also the following example (from Google): I was about three years old, and standing
at the piano, which my sister was playing, and my eyes were at the keyboard level). Nevertheless, it
will certainly be worthwhile to see how the various uses of at (its different functional elements)
relate to the restrictions on its distribution. Unfortunately, I will not be able to do so in the present article.
15. An example of a recently added member of the class of prepositions may be contra, which
in most dictionaries is not mentioned as a preposition (the OED gives a few examples, but the
prepositional status of contra in these examples is disputable), but which is nowadays used as a
transitive preposition (as in the unjust relationship of the US contra the world (Google) or Again,
contra Lawrence, it is obvious that for Eliot the idea that modern western society should adopt savage customs is seen as ludicrous and reprehensible (BNC)). Other examples of prepositions that
at some stage have entered the English lexicon are on top of and in front of (the latter becoming
a complex preposition, losing its definite article, in the 19th century). Bakker and Siewierska
(2002:151ff.) describe the birth of a new Dutch preposition (richting towards).
16. The examples followed by a code beginning with ICE-GB are from the ICE-GB Corpus, a
one-million-word corpus, tagged and parsed, of spoken and written English composed at the
Survey of English Usage, University College London.

All rights reserved
252 Evelien Keizer

17. As pointed out by one of the reviewers, it is possible to use such combinations in expressions
like He came from up North and The cat came from across the street. Note, however, that in these
cases the PPs up North and across the street do not describe a path (as in He walked up the road
or The cat ran across the street), but a location (the endpoint of a conceptualised path between
the position of the speaker and that of some other entity).
18. Even more surprising (or questionable) may be the following examples, both of which are
from the official websites of universities and organisations:
(ia) After completing his service in the Marine Corps, Brigham pursued graduate studies in
chemical engineering from at the University of Oklahoma from 1956 to 1962. (www.stanford.
edu/dept/news/pr/2004/brighamobit218.html)
(ib) Because of her expert knowledge, Dr Bennett is working close together with Radiologists
and Radiographers from at the Leicester Royal Infirmary to further develop scanning methods for NHS patients. (http://www.le.ac.uk/press/press/eyespecialists.html)
What seems to be happening in these examples, as well as in the examples in (16), is that the second preposition (at) is used as a place marker; in other words, it is used to indicate that the argument of the first preposition (from) is not an object (the default interpretation), but a location.
19. I am grateful to one of the reviewers for his/her comments on this part of the article, which
have forced me to reconsider some of the arguments used.
20. Note that on a truly contrastive reading, every place-denoting preposition can follow from,
e.g. I wasnt near the car park but at the car park.
21. In FG it has always been assumed that only lexical elements can be input to predicate formation rules. As one of the reviewers correctly points out, however, it is possible to for certain
derivational affixes to appear in conversion processes (e.g. isms and ologies), while a great many
of such affixes are semantically transparent (e.g. such neoclassical affixes as arch-, mono-, bio-,
etc., but also such more mundane affixes as -ful and -less, un- and -able). This seems to clash with
the FG treatment of such affixes as grammatical elements, triggered by the expression rules (e.g.
e.g. Dik 1997a:349; Dik 1997b:168).
The exact status of derivational elements is, however, far from unequivocal. In grammaticalisation studies, for instance, it is generally assumed that they have more lexical content than inflectional affixes (e.g. Halliday 1961: 249; Bybee 1985; Hammond and Noonan 1988; Lehmann
1989; Hopper and Traugott 1993; Bybee, Perkins and Pagliuca 1994: 22). As such they may be
seen as, at least, partially lexical; i.e. they can be considered to serve a role between contentive
and grammatical forms (Hopper and Traugott 1993: 4). This then may account for the fact that
occasionally they may be used independently, and may even undergo conversion processes. Alternatively, it may be argued that sometimes a grammatical (or highly grammaticalised) element
can occasionally be upgraded (degrammaticalised) for the occasion.
22. As pointed out by one of the reviewers, adding a prepositional prefix may increase the valency of a verbal predicate, as in look/overlook, live/outlive, go/undergo. This may be seen as supporting an analysis of prepositions as lexical elements.
23. It might be claimed that in some of these cases the prefix has adjectival or adverbial force.
Where the preposition is used in its primary, spatial sense, however, there is good reason to as-

All rights reserved
sume that we are dealing with the basic, prepositional form (overcoat, overtower). With regard
to under, OED specifically states that the prefix is always a preposition.
24. For practical reasons, the discussion to follow will concentrate on spatial prepositions, for
which representation by means of the l-variable is unproblematic. In those cases where the
preposition does not have a locational meaning, representation by means of an f-variable will
be more appropriate (indicating that the preposition or PP ascribes some property f (e.g. the
property for Mary in a present for Mary or with a handle in a cup with a handle) to an entity (the
present/cup)).
25. PFA, of course, also applies to deadjectival nouns like height. This means that also in a construction like the height of the building, the preposition is regarded as having a grammatical use.
26. The variable is used to denote any type of entity (generalising over the various variables
at the Representational level).
27. Note that in an expression like John is in, the intransitive preposition in must undergo nonverbal predicate formation, to supply it with an argument position. This is necessary to account
for the fact that the subject (John) is not the argument of the preposition (which is intransitive,
with an implied argument like the building/the office/the house), but of the prepositional phrase
as a whole (i.e. the intransitive preposition together with its implied argument).
28. Note that for the very same reason the agent PPs in passive constructions are best analysed
as modifiers introduced by the grammatical element by.
References
Bakker, D. and A. Siewierska (2002) Adpositions, the lexicon and expression rules. In R. Mairal
Usn and M.J. Prez Quintero (eds.) New perspectives on argument structure in Functional
Grammar. Berlin: Mouton de Gruyter. 125177.
Bresnan, J. (ed.) (1982) The mental representation of grammatical relations. Cambridge, MA:
MIT Press.
Bresnan, J. (2001) Lexical-functional syntax. Oxford: Blackwell.
Brugman, C. (1983) The story of over. Bloomington, Indiana: Indiana University Linguistics Club.
Brugman, C. and G. Lakoff (1988) Cognitive typology and lexical networks. In S. Small, G. Cottrell, and M. Tanenhaus (eds.) Lexical ambiguity resolution. Palo Alto, CA: Morgan Kaufman. 477507.
Bybee, J.L. (2003) Cognitive processes in grammaticalization. In M. Tomasello (ed.) The new
psychology of language. Cognitive and functional approaches to language structure: Vol. 2.
Mahwah, NJ: Erlbaum. 145167.
Bybee, J.L., R. Perkins and W. Pagliuca (1994) The evolution of grammar: tense, aspect, and modality in the languages of the world. Chicago: The University of Chicago Press.
Connolly, J.H. (199) On the generation of English temporal satellite terms. In E. Engberg-Pedsen, L. Falster Jakobsen and L. Schack Rasmussen (eds.) Function and expression in Functional Grammar. Berlin: Mouton de Gruyter. 303315.
Connolly, J.H (1995) On the generation of internally complex English temporal satellite terms.
Working Papers in Functional Grammar 57.

All rights reserved
254 Evelien Keizer

Corver, N. and H. van Riemsdijk (2001) Semi-lexical categories. In N. Corver and H. van Riemsdijk (eds.) Semi-lexical categories. Berlin: Mouton de Gruyter. 119.
Coventry, K.R. (2003) Spatial prepositions, spatial templates, and semantic versus pragmatic
visual representations. In E. van der Zee and J. Slack (eds.) Representing direction in language and space. Oxford: Oxford University Press. 255267.
Coventry, K.R., R. Carmichael and S.C. Garrod (1994) Spatial prepositions, object-specific function, and task requirements. Journal of Semantics 11: 289309.
Dewell, T. (1994) Over again: Image-schema, transformations in semantic analyses. Cognitive
Linguistics 5.4: 351380.
Dik, S.C. (1978) Functional Grammar. Dordrecht: Foris.
Dik, S.C. (1985a) Valentie en valentie operaties in Funktionele Grammatika. Tijdschrift voor
Taal- en Tekstwetenschap 5: 95114.
Dik, S.C. (1985b) Formal and semantic adjustment of derived constructions. In A.M. Bolkestein,
C. De Groot and J.L. Mackenzie (eds.) Predicates and terms in Functional Grammar. Dordrecht: Foris. 128.
Dik, S.C. (1989) The theory of Functional Grammar. Dordrecht: Foris.
Dik, S.C. (1997a) The theory of Functional Grammar: Vol. 1: The structure of the clause. Berlin:
Mouton de Gruyter.
Dik, S.C. (1997b) The theory of Functional Grammar: Vol. 2: Complex and derived constructions.
Berlin: Mouton de Gruyter.
Dowty, D.R. (1979) Word meaning and Montague grammar. Dordrecht: Reidel.
Franois, J. (1996) Reference to space and time through local adpositions and appositional cases:
beyond the proposals of C. de Groot, J.H. Connolly and J.L. Mackenzie. Working Papers in
Functional Grammar 62.
Garrod, S.C, G. Ferrier and S. Campbell (1999) In and on: investigating the functional geometry
of spatial prepositions. Cognition 72: 167189.
Gazdar, G., E. Klein, G. Pullum and I. Sag (1985) Generalised Phrase Structure Grammar. Oxford: Blackwell.
Groot, C. de (1989) Predicate structure in a Functional Grammar of Hungarian. Dordrecht: Foris.
Halliday, M.A.K. (1985) An introduction to functional grammar. London: Arnold.
Halliday, M.A.K. and R. Hasan (1976) Cohesion in English. London: Longman.
Heine, B. and T. Kuteva (2002a) The world lexicon of grammaticalization. Cambridge: Cambridge University Press.
Heine, B. and T. Kuteva (2002b) On the evolution of grammatical forms. In A. Wray (ed.) The
transition to language, Oxford: Oxford University Press. 376397.
Hengeveld, K. (2004) The architecture of a Functional Discourse Grammar. In J.L. Mackenzie and M.. Gmez Gonzlez (eds.), A New Architecture for Functional Grammar. Berlin:
Mouton de Gruyter. 121.
Hengeveld, K. (2005) Dynamic expression in Functional Discourse Grammar. In C. de Groot
and K. Hengeveld (eds.) Morphosyntactic expression in Functional Grammar. Berlin: Mouton de Gruyter. 5386.
Hengeveld, K. and J.L. Mackenzie (2006) Functional Discourse Grammar. In K. Brown (ed.).
Encyclopedia of language and linguistics. Vol. 4. Oxford: Elsevier. 668676.
Hengeveld, K. and J.L. Mackenzie (2008) Functional Discourse Grammar. Oxford: Oxford University Press.
Herskovits, A. (1986) Language and spatial cognition: An interdisciplinary study of the prepositions in English. Cambridge: Cambridge University Press.

All rights reserved
Hopper, P.J. (1991) On some principles of grammaticization. In E.C. Traugott and B. Heine
(eds.) Approaches to grammaticalization. Vols. 1 and 2. Amsterdam: Benjamins. 1735.
Hopper, P.J. and E.C. Traugott (1993) Grammaticalization. Cambridge: Cambridge University
Press.
Huddleston, R. (1984) Introduction to the grammar of English. Cambridge: Cambridge University Press.
Huddleston, R. and G.K. Pullum (2002) The Cambridge grammar of the English language. Cambridge: Cambridge University Press.
Jackendoff, R. (1983) Semantics and cognition. Cambridge, MA: MIT Press.
Jolly, J.A. (1993) Preposition assignment in English. In R.D. Van Valin (ed.) Advances in Role and
Reference Grammar. Amsterdam: Benjamins 275310.
Kahr, J.C. (1975) Adpositions and locationals: typology and diachronic development. Working
Papers on Language Universals 19: 2154.
Keizer, M.E. (2004) Postnominal PP complements and modifiers: a cognitive distinction. English Language and Linguistics 8/2: 323350.
Keizer, M.E. (2007) The English noun phrase: The nature of linguistic categorization. Cambridge:
Cambridge University Press.
Keizer, M.E. (2007) The lexical-grammatical dichotomy in Functional Discourse Grammar.
Alfa-Revista Lingstica 51/2 (Advances in Functional Discourse Grammar):3556.
Kreitzer, A. (1997) Multiple levels of schematization: a study in the conceptualization of space.
Cognitive Linguistics 8: 291325.
Lakoff, G. (1987) Women, fire and dangerous things: What categories reveal about the mind. Chicago: University of Chicago Press.
Langacker, R.W. (1987) Foundations of cognitive grammar: Vol. 1: Theoretical prerequisites. Stanford: Stanford University Press.
Langacker, R.W. (1992a) Prepositions as grammatical(izing) elements. Leuvense Bijdragen 81:
287309.
Langacker, R.W. (1992b) The symbolic nature of Cognitive Grammar: the meaning of of and
of-periphrasis. In M. Ptz (ed.) Thirty years of linguistic evolution: Studies in honour of Ren
Dirven on the occasion of his sixtieth birthday. Amsterdam: Benjamins. 483502.
Lee, D. (1998) Intransitive prepositions: are they viable? In P. Collins and D. Lee (eds.), The
clause in English. Amsterdam: Benjamins, 133147.
Lee, D. (2001) Cognitive linguistics: an introduction. Oxford: Oxford University Press.
Lehmann, Chr. (1989) Grammatikalisierung und Lexikalisierung. Zeitschrift fr Phonetik,
Sprachwissenschaft und Kommunikationsforschung 42: 1119.
Lehmann, Chr. (1995) Thoughts on grammaticalization. Mnchen: Lincom Europa.
Lehmann, Chr. (2002) New reflections on grammaticalization and lexicalization. In I. Wischer and
G. Diewald (eds.) New reflections on grammaticalization. Amsterdam: Benjamins. 118.
Lindstromberg, S. (1998) English prepositions explained. Amsterdam: Benjamins.
Lyons, J. 1977. Semantics. 2 vols. Cambridge: Cambridge University Press.
Mackenzie, J.L. (1983) Nominal predicates in a Functional Grammar. In S.C. Dik (ed.) Advances
in Functional Grammar. Dordrecht: Foris. 3151.
Mackenzie, J.L. (1992a) Places and things. In M. Fortescue, P. Harder and L. Kristoffersen (eds.)
Layered structure and reference in a functional perspective. Amsterdam: Benjamins. 253276.
Mackenzie, J.L. (1992b) English spatial prepositions in Functional Grammar. Working Papers in
Functional Grammar 46.

All rights reserved
256 Evelien Keizer

Mackenzie, J.L. (1996) English nominalizations in the layered model of the sentence. In B.
Devriendt, L. Goossens and J. Van der Auwera (eds.) Complex structures: A functionalist
perspective. Berlin: Mouton de Gruyter. 325353.
Mackenzie, J.L. (2001) Adverbs and adpositions: the Cinderella categories of Functional Grammar. In M.J. Prez Quintero (ed.) Challenges and developments in Functional Grammar.
(Monographic section of Revista Canaria de Estudios Ingleses 42.) 119135.
Mackenzie, J.L. (2002) The predicate in Function Grammar. In R. Mairal Usn and M.J. Prez
Quintero (eds.) New perspectives on argument structure in Functional Grammar. Berlin:
Mouton de Gruyter. 138.
Mackenzie, J.L. and M. Hannay (1982) Predicational predicates and focus constructions in a
Functional Grammar of English. Lingua 56: 4357.
Olbertz, H. (1998) Verbal periphrases in a Functional Grammar of Spanish. Berlin: Mouton de
Gruyter.
Prez Quintero, M.J. (2004) Adpositions in FG: has Cinderella been invited to the ball? In H.
Aertsen, M. Hannay and R. Lyall (eds.) Words in their places: Festschrift for J. Lachlan Mackenzie. Amsterdam: Faculty of Arts, Vrije Universiteit Amsterdam.
Pollard, C. and I. Sag (1984) Head-driven Phrase Structure Grammar. Chicago: Chicago University Press.
Radford, A. (1988) Transformational Grammar: A first course. New York: Cambridge University
Press.
Riemsdijk, H. van and E. Williams (1986) Introduction to the theory of grammar. Cambridge,
MA: MIT Press.
Samuelsdorff, P.O. (1998) Pronouns, adpositions, adverbs and the lexicon. In H. Olbertz, K.
Hengeveld and J. Snchez Garca (eds.) The structure of the lexicon in Functional Grammar.
Amsterdam: Benjamins. 267278.
Tyler, A. and V. Evans (2001) Reconsidering prepositional polysemy networks: the case of over.
Language 77(4): 724765.
Van Valin, R.D. and R.J. La Polla (1997) Syntax: structure, meaning and function. Cambridge:
Cambridge University Press.
Weigand, H. (1990) Linguistically motivated principles of knowledge base systems. Dordrecht:
Foris.
Yule, G. (1996) The study of language. Cambridge: Cambridge University Press.
Zeller, J. (2001) Lexical particles, semi-lexical postpositions. In N. Corver and H. van Riemsdijk
(eds.) Semi-lexical categories. Berlin: Mouton de Gruyter. 505549.
Authors address
Evelien Keizer
Spuistraat 210
1012 VT Amsterdam
The Netherlands
[email protected]

All rights reserved

Prepositions in Discourse

Uploaded by

Copyright:

Available Formats

Prepositions in Discourse

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Prepositions in Discourse

Uploaded by

Copyright:

Available Formats

John Benjamins Publishing Company

This is a contribution from Functions of Languages 15:2

English prepositions in Functional

Functions of Language 15:2 (2008), 216256. doi 10.1075/fol.15.2.03kei

English prepositions in Functional Discourse Grammar 217

is that of adpositions, which have been classified as either lexical or grammatical

2008. John Benjamins Publishing Company

218 Evelien Keizer

2. Adpositions in FG: An overview

2008. John Benjamins Publishing Company

English prepositions in Functional Discourse Grammar 219

Secondly, if adpositions are to be regarded as grammatical elements, this

where the preposition under is analysed as a one-place predicate heading a locative

2008. John Benjamins Publishing Company

220 Evelien Keizer

at (spatial and temporal)

(6) a. LEX.ENTRY= above, [P], ((xi)ZERO (xj)REF, , [SUPERIOR (xi)ZERO

2008. John Benjamins Publishing Company

English prepositions in Functional Discourse Grammar 221

2008. John Benjamins Publishing Company

222 Evelien Keizer

whereas Mackenzie classifies almost all English prepositions as lexical, reserving

2008. John Benjamins Publishing Company

English prepositions in Functional Discourse Grammar 223

Draw a line from point A to point B.

2008. John Benjamins Publishing Company

224 Evelien Keizer

Webster describes from as indicating the starting point of a physical movement.

2008. John Benjamins Publishing Company

English prepositions in Functional Discourse Grammar 225

2008. John Benjamins Publishing Company

226 Evelien Keizer

2008. John Benjamins Publishing Company

English prepositions in Functional Discourse Grammar 227

2008. John Benjamins Publishing Company

228 Evelien Keizer

Nevertheless, it is undeniably true that the combination from at is relatively rare,

2008. John Benjamins Publishing Company

English prepositions in Functional Discourse Grammar 229

2008. John Benjamins Publishing Company

230 Evelien Keizer

2008. John Benjamins Publishing Company

English prepositions in Functional Discourse Grammar 231

2008. John Benjamins Publishing Company

232 Evelien Keizer

In (22a) the PP on the table functions as a modifier of the non-relational nominal

2008. John Benjamins Publishing Company

English prepositions in Functional Discourse Grammar 233

the complement of the relational (deverbal) head reliance; in this construction

2008. John Benjamins Publishing Company

234 Evelien Keizer

Although it is certainly true that in the case of complements the choice of

the sister of Mary

the wines of France

2008. John Benjamins Publishing Company

English prepositions in Functional Discourse Grammar 235

2008. John Benjamins Publishing Company

236 Evelien Keizer

b. Douglas Hurd a quiet man by nature a diplomat by training couldnt

In sum, distinguishing between lexical and functional prepositions or uses of