461803

2012
SLR29110.1177/0267658312461803Second Language ResearchHopp

second
language
research

Second Language Research

Grammatical gender in adult

29(1) 33­–56
© The Author(s) 2012
Reprints and permission: sagepub.
L2 acquisition: Relations co.uk/journalsPermissions.nav
DOI: 10.1177/0267658312461803
between lexical and syntactic slr.sagepub.com

variability

Holger Hopp
University of Mannheim, Germany

Abstract
In order to identify the causes of inflectional variability in adult second-language (L2)
acquisition, this study investigates lexical and syntactic aspects of gender processing in real-
time L2 production and comprehension. Twenty advanced to near-native adult first language
(L1) English speakers of L2 German and 20 native controls were tested in a study comprising
two experiments. In elicited production, we probe accuracy in lexical gender assignment. In a
visual-world eye tracking task, we test the predictive processing of syntactic gender agreement
between determiners and nouns. The findings show clear contingencies (1) between overall
accuracy in lexical gender assignment in production and target predictive processing of
syntactic gender agreement in comprehension and (2) between the speed of lexical access
and predictive syntactic gender agreement. These findings support lexical and computational
accounts of L2 inflectional variability and argue against models positing representational
deficits in morphosyntax in late L2 acquisition and processing.

Keywords
grammatical gender, inflectional variability, adult L2 acquisition, L2 processing, morphosyntax,
production, comprehension

I Introduction
In contrast to child L1 and L2 learners who master the morphosyntax of grammatical
gender within a few years of exposure (e.g. Hopp, 2011; Unsworth, 2008), adult L2 learn-
ers frequently show variability in establishing gender agreement between nouns, deter-
miners and adjectives that persists to advanced proficiency levels and often characterizes

Corresponding author:
Holger Hopp, English Linguistics, University of Mannheim, Schloss EW 266, 68131 Mannheim, Germany.
Email: [email protected]
34 Second Language Research 29(1)

non-native-like ultimate attainment (Franceschina, 2005). Grammatical gender thus

constitutes a striking example of the general phenomenon of inflectional variability in
adult L2 acquisition (for overview, see White, 2007) and lends itself to studying how late
L2 acquisition and processing differ from native language processing.
From a grammatical perspective, two answers have been given to account for differ-
ences between natives and non-natives. One set of approaches appeals to age-related
grammatical impairments arguing that L2 grammars do not encode abstract syntactic
gender features necessary for computing agreement relations unless these features are
instantiated in the L1 (Hawkins, 2009; Hawkins and Casillas, 2008). Other approaches
hold that L2 grammars do not differ from native grammars in syntactic representations
and that variability in morphosyntactic agreement follows from lexical retrieval diffi-
culties in mapping the target morphophonological forms to syntactic features in situa-
tions of real-time processing pressure (Prévost and White, 2000).
This article reports two experiments that examine whether the L1 morphosyntax or
whether problems with lexical aspects of gender condition variability in the real-time
processing of syntactic gender agreement in comprehension. In an elicited production
task and a visual world eye tracking task, we consider the relation between problems
with lexical gender assignment in production and syntactic gender agreement in on-
line comprehension in a group of L1 English advanced to near-native adult L2 speak-
ers of German.
This article is structured as follows: Section II gives an overview of gender assign-
ment and agreement in late L2 acquisition. Section III presents the hypotheses, and
Section IV introduces the properties of gender marking inside the German DP. The study
is presented in Section V, and Section VI discusses the results.

II  Gender in late L2 acquisition

Many languages classify nouns by virtue of assigning them to one of several lexical
gender classes (Corbett, 1991). In the mental lexicon of native speakers, gender class
information is stored as a gender node at the lemma level as part of each noun entry
(Carroll, 1989) or as a gender node to which all nouns of this gender class are linked (e.g.
Schriefers and Jescheniak, 1999).
At the syntactic level, gender is expressed by means of grammatical agreement on
other constituents inside the nominal domain, e.g. determiners and adjectives, as well as
on predicative adjectives and coreferential pronouns outside the nominal domain.
According to, for example, Carstens (2000) gender agreement is realized by feature
checking between an interpretable gender feature on the noun and uninterpretable gender
features on determiners and adjectives. Many psycholinguistic models hold that speakers
consult these abstract gender features in real-time production and comprehension by a
process that matches features through agreement chains (e.g. Franck et al., 2008). Both
aspects of gender – (1) gender assignment, i.e. the classification of nouns to gender
classes in the mental lexicon and (2) gender agreement, i.e. the feature-based computa-
tion of gender across syntactically related constituents – prove to be variable in late L2
acquisition.
Hopp 35

For gender assignment, even very advanced L2ers exhibit residual variability in real-
time production, in particular L2ers whose L1 does not have grammatical gender
(Franceschina, 2005). For instance, advanced L1 English L2ers of Spanish only reach
levels from 75% to 90% accuracy in gender assignment in elicited production (Alarcón,
2011; Bruhn de Garavito and White, 2003; Franceschina, 2005; Grüter et al., 2012;
Montrul et al., 2008). Such residual variability suggests that L2ers do not consistently
establish or access gender nodes for nouns in the L2 mental lexicon.
However, gender errors are not random in production. First, L2ers resort to over-
generalizing a default gender (e.g. Bruhn de Garavito and White, 2003; Hawkins,
2009; McCarthy, 2008: White et al., 2004). Second, compared to real-time production,
L2ers perform significantly better, though not at ceiling, in untimed written produc-
tion. Third, L2ers have much lower error rates on gender agreement than on gender
assignment in production (e.g. Alarcón, 2011; Grüter et al., 2012). Irrespective of L1
differences, advanced L2ers have been shown to reach native-like levels on deter-
miner–noun and adjective–noun agreement in elicited production (e.g. White et al.,
2004). These findings indicate that errors with gender at advanced L2 proficiency lev-
els are predominantly restricted to lexical gender assignment and constrained by task
demands.
According to the Missing Surface Inflection Hypothesis (Prévost and White, 2000),
such a performance pattern implies that L2 inflectional variability stems from failures in
retrieving inflectional forms under real-time processing pressure, in particular in produc-
tion (White, 2011). In these situations, L2ers rely on default forms or other cues for
gender because they fall short of mapping the appropriate target lexical form into the
given grammatical context.
Comprehension experiments also show that processing constraints contribute to non-
target L2 performance on syntactic gender agreement. In reaction-time and eye tracking
studies, very advanced L2ers fail to compute non-adjacent determiner–noun agreement
at elevated processing speed (Hopp, 2010), and they show contingencies between accu-
racy on non-local adjective–noun gender agreement with proficiency (Sagarra and
Herschensohn, 2010) as well as with working memory capacity (Keating, 2010; see also
Foote, 2011).
For local determiner–noun agreement, however, behavioural and electrophysio-
logical studies report that even relatively low-proficient learners are sensitive to
gender agreement violations (Tokowicz and MacWhinney, 2005), and highly profi-
cient learners demonstrate native-like ERP (event related potential) responses to
gender agreement violations (Foucart and Frenck-Mestre, 2011; Gabriele et al.,
2012; Gillon Dowens et al., 2010). These effects are found for L2ers across L1s,
although learners whose L1s also mark grammatical gender show target-like ERP
signatures sooner (Sabourin and Stowe, 2008). For adjective–noun agreement, how-
ever, very advanced L2ers demonstrate attenuated (Gillon Dowens et al., 2010, 2011)
or no target ERP signatures of detecting gender violations, even if the L1 has gram-
matical gender (Foucart and Frenck-Mestre, 2011). Mastery of grammatical gender
agreement thus seems affected by whether gender marking is expressed in the same
contexts in the L1 and the L2, e.g. on postnominal adjectives (though see Gabriele
et al., 2012).
36 Second Language Research 29(1)

While the evidence available to date suggests that L2ers across L1s demonstrate on-
line sensitivity to gender agreement in some contexts, almost all behavioural and neuro-
physiological comprehension experiments on gender agreement test for sensitivity to
gender violations by recording whether L2ers identify mismatches between the gender
assigner, the noun, and its dependents in ungrammatical sentences. Such experiments
involving ungrammatical sentences suffer from the potential limitation that they do not
represent the typical course of language comprehension, because they essentially meas-
ure whether the language processor can successfully exercise a backward compatibility
check between morphological forms expressing gender agreement.
In normal language comprehension, however, morphological forms can act as inform-
ative cues for forward or anticipatory processing (Kamide, 2008). For instance, gender
marking on dependent constituents, e.g. determiners and adjectives, may facilitate the
predictive processing of the subsequent noun in grammatical sentences. As detailed in
the following, experiments addressing gender in its predictive function have found pro-
nounced effects of non-target performance in adult L2ers.
Native speakers and early bilinguals of French display shorter naming latencies of
nouns if they are preceded by gender-informative versus gender-neutral articles (e.g.
le/leur joli bateau – ‘theMASC/theirAMBIGUOUS nice ship’). In contrast, even highly advanced
late L1 English L2ers of French make no difference in naming speed according to pre-
ceding gender information (Guillelmon and Grosjean, 2001). In addition, adult L2ers do
not benefit from congruent gender agreement on pre-nominal adjectives when making
lexical decisions on the following noun (Scherag et al., 2004). For adult L2ers, then gen-
der realized on preceding elements in the noun phrase apparently does not lead to the
predictive activation of nouns by virtue of gender agreement.
Predictive effects of gender have also been studied in the visual-world eye tracking
paradigm. In visual-world eye tracking, participants look at a display of different objects
while listening to linguistic stimuli. Their relative proportion of looks to individual
objects is then interpreted as an index of how listeners predictively resolve reference on
the basis of partial input as a sentence is unfolding (for overview, see Huettig et al.,
2011). For instance, when native English speakers listen to a sentence like Look at the
candy, they show effects of cohort competition for words with phonologically similar
onsets (e.g. candy – candle) by looking equally often at these objects until the sound
stream disambiguates reference on the final syllable (e.g. Allopenna et al., 1998).
However, when words are preceded by informative gender cues on determiners, e.g. in
French (le bouton – ‘theMASC button’ versus la bouteille – ‘theFEM bottle’), native French
listeners do not consider phonologically similar nouns as competitors anymore (Dahan
et al., 2000). Instead, they solely look towards the object that is compatible with the
gender-marked form of the determiner by virtue of grammatical gender agreement.
Hence, gender marking on the determiner leads listeners to anticipate the upcoming noun
by computing a predictive agreement relation. Similarly, children as young as 28 months
make anticipatory use of grammatical gender in their looks to target objects (Johnson,
2005; Lew-Williams and Fernald, 2007). In identical experiments, however, intermediate-
level English-speaking L2 learners of Spanish fail to use grammatical gender as a guid-
ing cue to orient faster to target pictures, irrespective of whether existing nouns or novel
nouns are used (Lew-Williams and Fernald, 2010).
Hopp 37

Summarizing the findings on L2 gender, L2ers whose L1s do not have grammatical
gender can come to master syntactic gender agreement at advanced proficiency levels;
yet, they continue to struggle with gender assignment in production. In comprehension,
L2ers have protracted difficulty in using gender as a predictive agreement cue.
This pattern of L2 inflectional variability raises the question whether problems with
lexical gender assignment in production and variability in syntactic agreement in predic-
tive comprehension are related and can be traced to a single underlying cause. A recent
study by Grüter et al. (2012) addresses potential relations by comparing performance on
gender assignment and agreement in oral production as well as in comprehension for the
same group of 19 very advanced non-immersed adult L1 English learners of L2 Spanish.
The L2ers scored at ceiling for gender agreement in an untimed comprehension task. In
oral production, the L2ers performed to criterion on determiner–adjective agreement, yet
they only assigned nouns the correct gender at a mean rate of 80%. In a subsequent visual
world study adopted from Lew-Williams and Fernald (2010), the L2ers, like native
speakers of Spanish, showed predictive effects of grammatical gender for novel nouns
that they had first been exposed to in the course of the experiment; yet, unlike natives,
they did not use gender as a predictive cue as a predictive cue for familiar nouns.
Grüter et al. (2012) take the asymmetry between sensitivity to gender marking in predic-
tive processing for novel nouns and the null effect for familiar nouns as a starting point for
developing a lexical gender learning hypothesis of inflectional variability. The core tenet of
the hypothesis is that inflectional variability in production and comprehension is the con-
sequence of weaker links between nouns and gender nodes in the adult L2 mental lexicon.
Following Arnon and Ramscar (2012) and Lew-Williams and Fernald (2010), they argue
that problems with gender arise as the consequence of different learning environments
between children and adults. Children are exposed to non-segmented, aural input. In order
to map nouns to a conceptual representation, children first need to segment nouns from the
speech stream of continuous determiner–noun sequences. For this task, children exploit
high co-occurrence probabilities between determiners and nouns, ultimately mapping
determiner forms to abstract grammatical nodes expressing gender. Given the strong links
between determiners and nouns, the resulting lexical gender node is strong, and gender-
marked forms act as a powerful cue for the associated nouns in processing.
In contrast, adults come to the task of L2 acquisition with prior metalinguistic knowl-
edge about nouns and determiners as well as gender classes. In addition, they receive
written input, which is visually pre-segmented by gaps between words. Hence, L2 adults
can map nouns directly to a conceptual representation without having to rely on comput-
ing co-occurrence relations for extracting gender-marking determiners and nouns. In
consequence, the associations between determiners and nouns via gender nodes are
likely to be weaker in the adult L2 mental lexicon, such that (1) the appropriate gender
cannot always be retrieved in real-time production and (2) gender-marked forms do not
act as informative cues in on-line comprehension.
For novel nouns that the participants first encountered aurally in the context of the
experiment, however, they are likely to exploit the same distributional learning strategy
as children, thus establishing strong lexical links between gender-marked determiners
and nouns. Therefore, they can take advantage of gender marking on determiners to
anticipate upcoming nouns.
38 Second Language Research 29(1)

A similar relation between input learning type and gender processing is suggested by
studies that compare effects of different types of instruction for learning L2 gender process-
ing. For instance, adult L2ers show more target-like ERP components for gender agreement
violations in an artificial miniature language that they had learned implicitly compared to
when they had been explicitly taught an agreement rule (Morgan-Short et al., 2010; for
effects of feedback on gender in L2 processing, see Davidson and Indefrey, 2009).
The lexical gender learning hypothesis provides an intriguing post-hoc account of
parallels between gender misassignment in production and selective problems with gen-
der agreement in comprehension by tieing both to weaker lexical gender representations.
It thus complements the Missing Surface Inflection Hypothesis (Prévost and White,
2000) (1) by extending its scope from production to real-time comprehension and (2) by
specifying the causes of retrieval failures of inflectional forms as a likely result of the
particular learning history in adult L2 acquisition that leads to lexical gender nodes being
less active or accessible.
With a view to establishing whether problems with gender at the lexical level occa-
sion syntactic aspects of L2 inflectional variability, the present study directly tests for
contingencies between gender assignment in production and gender agreement in
comprehension within the same group of L2 speakers whose L1 does not have gram-
matical gender. Lexical accounts of gender variability will be contrasted to grammati-
cal impairment approaches that situate problems with gender at the syntactic level.
According to the latter approaches, L2 learners whose L1s do not encode grammati-
cal gender do not project syntactic gender features for agreement and hence need to rely
on non-grammatical cues for computing gender concord in the L2, i.e. phonological
similarity between determiners and nouns and/or probabilistic associative information.
The Representational Deficit Hypothesis by Hawkins (2009) suggests that adult L2ers
initially pair all nouns in the L2 with the most frequently occurring article form as a
default gender and then gradually list nouns that they encounter in the input with other
articles as exceptions to the default. If the activation level of the exceptions is too weak
for the non-default gender form to be retrieved in real-time production, the gender
default will appear in non-default contexts. By extension, a similar asymmetry arises in
on-line comprehension, where marked non-default forms could act as informative gen-
der cues because they tap into an associative memory trace of noun-gender co-occur-
rence. In contrast, the default gender marker will not act as a predictive cue as the
default is not specified for a restricted set of nouns. According to the Representational
Deficit Hypothesis, predictive effects of gender marking should be limited to the mem-
orized exceptions of listed gender–noun associations, because feature-based syntactic
agreement between determiners and nouns has become unavailable in L2 adults.
Hence, both the lexical gender learning and the Representational Deficit hypothesis argue
that lexical aspects of gender are implicated in L2 inflectional variability with syntactic gen-
der agreement. For the lexical gender learning hypothesis, strong lexical gender representa-
tions are the prerequisite for the successful recruitment of predictive grammatical gender
agreement through syntactic feature checking. For the Representational Deficit Hypothesis,
associative lexical learning acts as a compensatory strategy to memorize exceptional, i.e.
non-default, co-occurrence of gender-marked forms and nouns because the L1 morphosyntax
restricts the availability of syntactic feature checking in adult L2 acquisition.
Hopp 39

For adjudicating between the two approaches, it is imperative to make sure that one
genuinely probes syntactic agreement in comprehension. In their pioneering study on
predictive processing of grammatical gender in French native speakers, Dahan et al.
(2000) discuss two ways in which a gender-marked determiner can lead to the anticipa-
tion of the following noun in spoken word recognition. First, predictive processing of
gender could reflect the computation of syntactic agreement between determiner and
noun through the activation of a common abstract gender feature. Second, lexical selec-
tion could occur solely on the basis of formal correspondences between phonologically
similar word forms or endings. For instance, in Spanish, feminine articles end in -a (una,
la), and masculine articles are consonant final (un, el). Further, 99.8% of nouns ending
in -o are masculine, and 96.3% of nouns ending in -a are feminine (Teschner and Russell,
1984). For French, noun endings predict the gender at rates above 80% (Lyster, 2006)
Gender agreement in phonologically regular languages like Spanish, Italian and French
could thus largely be computed on the basis of phonological regularity. To avoid con-
founds of non-syntactic gender agreement cues like phonological regularity, this study
considers a target language that has phonologically and semantically opaque gender,
namely German.

III  Hypotheses and predictions

Grammatical impairment accounts of L2 inflectional variability hold that L2ers use asso-
ciative relations between constituents for establishing gender agreement. For an opaque
gender system like that of German, these accounts, first, predict variability in predictive
processing of gender agreement by L1 English learners. Second, they predict an asym-
metry between gender-marking forms, with the most frequent gendermarker being over-
generalized as the default in production. In turn, the default gender should not act as a
predictive cue in on-line comprehension, since it does not lead to the activation of a
restricted set of nouns.
Lexical accounts of L2 inflectional variability, such as the Missing Surface Inflection
and the lexical gender learning hypothesis, predict that adult L2ers across L1s can com-
pute syntactic agreement based on gender features across all gender forms. However,
performance on syntactic gender agreement should be modulated by lexical aspects of
gender. First, production and comprehension accuracy should be correlated in L2ers
because strong lexical gender representations should lead to target gender assignment in
production, and target-like predictive processing of gender agreement in comprehension.
Conversely, weak lexical gender representations should engender variability in produc-
tion and comprehension. Second, if inflectional variability in production and compre-
hension is conditioned by the retrieval of lexical (gender) representations, L2 performance
on gender should relate to measures of lexical access.

IV  Grammatical gender in German

German has a rich nominal paradigm with a three-way gender distinction, a two-way
number distinction, a four-way case distinction, and a two-way definiteness distinction.
Grammatical gender on nouns is phonologically and semantically opaque, and the few
40 Second Language Research 29(1)

phonological generalizations about gender there are – e.g. bisyllabic nouns ending in
schwa are feminine – have many high-frequency exceptions (e.g. Auge – ‘eyeNEUT’; Käse
– ‘cheeseMASC’). Given the lack of robust correspondences between morphophonological
form of the noun and gender class, gender information is only inferable from constituents
dependent on the noun by virtue of gender agreement, i.e. determiners and adjectives.
Due to a large degree of syncretism, gender is often marked ambiguously. However, in
nominative singular contexts, definite determiners provide three different gender-mark-
ing forms (der, die, das). Table 1 presents the determiner and adjectival paradigms in
nominative case for definite nouns.
In terms of relative frequency, about 50% of German nouns are masculine, 30% femi-
nine and 20% neuter (Bauch 1971, as cited in Hohlfeld, 2006). The learner input, how-
ever, provides somewhat different frequency patterns of gender-marking determiners. In
modern German usage, feminine (44%) outweighs masculine (33%) and neuter (23%)
among unambiguous gender markers in large-scale corpora (Chiarcos, 2009). Finally,
the overall occurrence of the prototypical definite determiner forms der, die, das is
around 43% for der and die, respectively, and 23% for das (Kornai, 2009), since die also
marks plural across genders. On any count, neuter represents the marked, non-default
gender or determiner form, while masculine and/or feminine could act as default gender
forms, depending on whether learners attend to global token frequency or more fine-
grained statistical properties of the input.

V Study
1 Materials
Based on the German paradigm in Table 1, we constructed stimuli in which gender (1) is
unambiguously realized on determiners only, with adjectives being ambiguous between
all three genders.

(1)  Wo   ist  der/die/das    gelbe [Noun]?

 Where is theMASC/FEM/NEUT  yellow [Noun]?

Twenty items in total were created for the gender condition in (1), i.e. five difference trials
for each of the three gender forms, i.e. masculine, feminine and neuter, in (1), and five same
trials. For each item, four-picture displays using coloured drawings or reduced photographs
of pictureable and easily identifiable inanimate objects were designed (Figure 1). All object

Table 1  Determiners and adjectives for definite nominative nouns in German

  Determiners Adjectives
Singular Masculine der gelb-e
Feminine die gelb-e
Neuter das gelb-e
Plural all die gelb-en
Hopp 41

Figure 1  Display for difference trials (1)

labels had a frequency of at least 1,200 tokens in the 5.4 billion COSMAS II corpus of con-
temporary written German (COSMAS II, 2008). Overall, there were no frequency differ-
ences of the nouns by gender form (masculine, feminine and neuter), as confirmed by a
one-way ANOVA (F(2,42) = 0.864, p = .429).
For the difference trials one object with a clearly identifiable colour was the target; there
were two identically coloured competitor objects that bore one of the other two genders
each, and the fourth object in the display was a differently coloured distracter of a gender
different from the target (Figure 1). For the same trials five ambiguous displays were
designed along the same lines as Figure 1; however, the three colour-matched objects all
bore the same gender and the differently coloured distractor bore a different gender. The
stimuli thus consisted of 15 difference trials, in which gender on the determiner was a poten-
tially informative predictive cue for the target object by making it the only referent that is
grammatically compatible with the determiner, and five same trials, in which the determiner
and the adjective were grammatically compatible with three objects in the display. In addi-
tion, 10 items with a numeral (Where do you see two [A] [N]?) and displays as in Figure 2
were included as items that tested for the use of lexical cues (two) as opposed to gender
cues. Finally, 40 trials involving number inflection on determiners and adjectives were
included as fillers.
The resulting 70 displays were assigned to one of four lists which counter-bal-
anced target noun and object position for the experimental items. Objects were
42 Second Language Research 29(1)

Figure 2  Display for lexical cue condition

equidistant from the central fixation cross. Each object/noun served as a target in
one list and as a competitor in other lists. In each list, the target object was in a dif-
ferent quadrant of the screen. All sentences were recorded at a moderate speech rate
by a male native speaker of German, with all elements ([carrier phrase], [D], [A(s)],
[N]) spliced and all word onsets temporally aligned. Mean length of the segments
was 1,103 ms for the carrier phrase, 405 ms (sd = 28) for determiners, and 640 ms
(sd = 46) for adjectives.

2 Participants
Twenty native English L2 speakers of German (M = 47.3 years, 16 female) and 20
native German speakers (M = 21.0 years, sd = 3.1, 16 female) took part in the
experiment.1 In order to test L2ers who had had enough exposure for acquiring German
gender, we recruited L2 participants who scored in the advanced to near-native profi-
ciency range (> 20) in a standardized 30-item written placement test (Goethe Institut,
2010). All participants had normal or corrected-to-normal vision and reported no his-
tory of learning difficulty or language deficits. At the time of testing, all participants
were residents in Germany, and the L2ers had been resident in Germany for 2 to 44
years (M = 23.6, Table 2).
Hopp 43

Table 2  L2 participant characteristics (n = 20)

Age (years) Age of onset Length of Length of Goethe score

(years) exposure (years) residence (years) (out of 30)
Mean 47.3 15.0 31.9 23.6 26.4
Range 19–70 11–24 5–59 2–44 20–30
sd 15.5 3.3 16.2 13.1 3.5

3  Apparatus, procedure and analysis

The study combined a production and a comprehension task. Participants sat in front of a
19-inch screen at a distance of approximately 70cm. Participants were instructed, and they
completed two practice trials before a calibration began. The calibration aimed for visual
acuity below 0.5 degrees and was repeated in the course of the experiment if necessary.
Throughout the experiment, an SMI RED eye tracker recorded gaze position at 60 Hz.

a  Production task  In the production task that immediately preceded the comprehension task,
participants saw still images of the four-picture displays and were asked to name the four
objects in each display, including their colour. This way, we could (1) ensure that participants
recognized the objects and knew their labels and (2) assess which gender L2 participants
assigned to the labels of the objects. In their descriptions of the objects, the participants used
noun phrases containing determiners, colour adjectives and nouns.2 In total, participants
named 60 different objects in the experimental condition (see Appendix 1). Participants’
responses were recorded and transcribed by the experimenter. Responses were coded for
gender on the basis of the gender form used on the determiner (der, die, das) for definite NPs
as in example (1) or the adjective (e.g. gelb-er, gelb-e, gelb-es – ‘yellowMASC/FEM/NEUT’) for
indefinite NPs. Overall accuracy was coded as percentage of target responses.

b  Comprehension task  After the production part, the screen changed immediately to
show the same displays including a fixation cross. After a preview of 3,000 ms, a sound
signal alerted participants to fixate the cross. The participants’ gaze was directed to the
fixation cross prior to each trial to avoid baseline effects of participants already look-
ing at the target item before the onset of the critical noun phrase (Barr et al., 2011).
1,500 ms after the sound signal, the auditory presentation of the sentences began. In the
analysis, gaze position after determiner onset was coded every 20 ms for 3,000 ms.
Looks to the target region within the first 200 ms after determiner onset were dis-
carded, because they could not reflect linguistically guided gaze shifts (e.g. Huettig et
al., 2011). In all, this affected less than 1% of the data. In total, the experiment com-
prising the two tasks took about 30 minutes for native speakers and approximately 40
minutes for the L2ers.

4  Production results
For gender assignment in production, the natives scored at ceiling, and the L2ers showed
variable performance with accuracy ranging from 53% to 100% (Table 3). Only one L2er
44 Second Language Research 29(1)

Table 3  Production results by group: L2 participants (n = 20)

Group Gender Goethe score Length of Length of

accuracy exposure residence
L2 Gender Mean 77% 24.0 23.8 17.7
variable (n = 11) range 53%–90% 20–28 5–37 2–39
sd 12.1 3.0 12.5 12.0
L2 Gender Mean 97% 29.2 41.7 30.1
consistent (n = 9) range 95%–100% 27–30 8–59 8–40
sd 0.2 1.1 15.1 11.2

Table 4   Production results: Gender assignment errors in the L2 groups

Non-target Masculine Feminine Neuter

gender

substituted for for for for for for for

Feminine Neuter Masculine Neuter Masculine Feminine
L2 Gender 27 (19%) 26 (19%) 22 (16%) 32 (23%) 29 (21%) 4 (3%)
variable 53 (38%) 54 (39%) 33 (24%)
(n = 11)
L2 Gender   5 (29%)   1 (6%)   2 (12%)   1 (6%)   7 (41%) 1 (6%)
consistent   6 (35%)   3 (18%)   5 (47%)
(n = 9)

did not make any gender assignment mistakes. For investigating effects of the lexical
representation of gender, the L2ers were divided into two lexical gender groups on the
basis of their overall gender assignment accuracy: A group with weak lexical gender, i.e.
variable gender assignment, that made at least six gender mistakes (10%) for the 60
objects (mean accuracy on gender = 77%), and a group with strong lexical gender, i.e.
consistent target gender assignment, that made at most three (5%) gender mistakes (mean
accuracy = 97%).
One-way ANOVAs yield significant differences between the lexical gender groups
for Gender Accuracy (F(1,19) = 23.051, p < .001), Goethe Score (F(1,19) = 24.908, p <
.001), Length of Exposure (F(1,19) = 8.349, p = .010) as well as for Length of Residence
(F(1,19) = 6.123, p = .024). Hence, accuracy in lexical gender assignment is also modu-
lated by the amount of target-language input L2ers have encountered which, in turn,
affects their levels of general proficiency. These factors are virtually necessarily associ-
ated in late L2 acquisition.
As for gender assignment errors, Table 4 presents the breakdown of substitution errors
by gender forms and groups. Table 4 illustrates that gender assignment errors are widely
distributed across gender forms. Thus, they are not characterized by the overgeneraliza-
tion of one (default) gender to other contexts. Instead, L2ers assign non-target genders to
nouns as a consequence of having misclassified nouns or wavering between gender
forms in production.
Hopp 45

5  Comprehension results
In the comprehension task, we analysed the trials based on the production accuracy of
gender assignment by the L2ers. We categorized the comprehension trials of the L2ers
into (1) difference trials, if the L2er had assigned the target lexical gender to the target
item and other (target) genders to the other items. Hence, these were trials in which
gender marking on determiners would give unambiguous cues to the target item as per
the L2ers’ subjective gender assignment. Of all the 20 gender-related trials per partici-
pant, the difference trials constituted 205 out of 400 (51%) for all L2ers, that is 81
(37%) for the gender-variable L2 group and 124 (69%) for the gender-consistent L2
group. Further, we classified trials as (2) same trials if the L2er had assigned the target
item and other items the same gender. These comprised 117 out of 400 (29%) of all
trials among the L2 group, 74 (34%) for the gender-variable L2 group and 43 (24%)
for the gender-consistent L2 group. In the same trials, gender marking could not act as
an unambiguous cue for the target item. Finally, trials were categorized as (3) unin-
formative trials, 77 (19%), if the L2er had assigned the target item a non-target gender.
In these cases, the target gender cue is incongruous with the subjectively assigned
gender and could not guide predictive looks to the target item. These items were
excluded from analysis.
In sum, we made sure that we only classified trials as difference trials for the L2ers in
which gender marking could serve as an informative cue for the upcoming noun as per
the subjective gender assignment by the L2 participants in the production task. We also
ensured that same trials would be trials in which gender marking is ambiguous, yet not
incompatible with regard to the subjective gender assignment of the participants.
Importantly, only these subjectively accurate difference and same trials were used in the
analyses of the comprehension trials.
We first analyse the mean reaction time of looking towards the target region, i.e. when
participants first look towards the target as a potential referent of the unfolding noun
phrase. Table 5 lists the mean reaction times in the target region and gives the difference
between the means in the same and the difference trials.
For mean reaction times, we find a significant main effect of Type (F(1,37) = 52.543,
p < .001) which, however, only interacts marginally with Group (F(2,37) = 2.519, p =
.094). In planned pairwise comparisons, the differences in mean reaction times are highly
significant for the natives (t(19) = 7.223, p < .001) and the gender-consistent L2 group

Table 5  Mean reaction times after determiner onset by group and type in ms (standard
deviation) and differences between same and difference sentences

Determiner Δ
Natives same 2231 (214)
456
difference 1775 (286)  
L2 Gender variable same 2449 (249)
210
difference 2239 (282)  
L2 Gender consistent same 2071 (324)
412
difference 1659 (341)  
46 Second Language Research 29(1)

Figure 3  Time course (in ms from determiner onset) of proportion of looks to the target
object in same and difference trials by group.Vertical lines indicate onset of adjective and noun,
respectively

(t(8) = 5.063, p = .001), while there is only a marginal effect in the gender-variable L2
group (t(10) = 1.968, p = .077).
Figure 3 graphs the proportions of looks to the target region over time for gender
marking on determiners for the natives and the two L2 groups. Broken lines represent
the difference conditions for the three groups, and solid lines represent the same condi-
tions for natives (grey) and all L2ers (black). For ease of illustration, the same condi-
tions are plotted for all L2ers, although all comparisons are based on separate group
means. Figure 3 shows that the natives and the gender-consistent L2 group look sooner
to the target region in the difference trials compared to the same trials after encountering
the determiner. In contrast, the curves of looking proportions to the target region diverge
only later and less strongly between the difference and the same trials for the gender-
variable L2 group.
In order to track effects of predictive gender over time, we conducted Repeated
Measures ANOVAs with the within-participant factors Type (same, different) and the
between-participant factor Group (native, gender-consistent, gender-variable) in a time-
course analysis of proportion fixations. We analyse mean proportion of looks to the tar-
get object in 300 ms time windows starting 300 ms after determiner onset (Table 6).
Hopp

Table 6  Time course analysis for proportion of looks to the target region by time windows after onset of determiner (F-values). Factors are Type
and Group

Time window 1: 0–300 ms 2: 300–600 ms 3: 600–900 ms 4: 900–1,200 ms 5: 1,200–1,500 ms

Type  1.335  1.636  5.509* 15.962*** 18.135***
Type*Group  1.341  3.563*  6.450**  2.047  1.417

Time window 6: 1,500–1,800 ms 7: 1,800–2,100 ms 8: 2,100–2,400 ms 9: 2,400–2,700 ms 10: 2,700–3,000 ms

Type 33.042*** 46.993*** 52.630***  1.818  2.173
Type*Group  3.615*  4.381*  5.423**  2.440  <1
Notes. (*) = p < .1, * = p < .05, ** = p < .01, *** = p < .001
47
48 Second Language Research 29(1)

In the time-course analysis (Table 6), the significant interactions of Type and Group
indicate that the gender-variable L2 group (dashed black line) differs from the other
groups in its use of grammatical gender. In subsequent pairwise comparisons (Table 7),
the gender-consistent L2 group shows significant differences between the difference
and the same trials starting in time window 2, whereas the gender-variable L2 group
makes significantly more fixations to the target in the difference compared to the same
trials in time windows 6 through 8 only. Accordingly, the gender-variable L2ers use
gender on determiners only relatively late and thus weakly in predictive processing,
while the gender-consistent L2 group and the natives demonstrate early and clear effects
of predictive gender processing. Note that this difference between groups obtains even
though only the items to which gender had been correctly assigned in production are
included in the analysis.
Next, we break down the difference trials according to gender form. A Repeated
Measures ANOVA on the mean reaction times with the factors Gender (masculine, femi-
nine, neuter) and Group unearths a main effect of Gender (F(2,68) = 6.449, p = .003), yet
no effect of Group (F(4,68) = 1.169, p = .332). Across groups, masculine forms trigger
more delayed looks to the target regions than feminine or neuter forms, with the latter
two being statistically indistinguishable from each other. Importantly, while the means
for each gender form in the difference trials is reliably different from the same trials for
the natives (all ps < .001) and the gender-consistent L2 group (all ps ≤ .014), the gender-
variable L2 group does not show any significantly earlier reaction times in difference
versus same trials for any gender form (all ps > .05). Whereas the natives and the gender-
consistent L2 group recruit gender as a robust agreement cue, irrespective of the particu-
lar gender marker, the findings of the gender variable L2 group illustrate that no gender
form acts as a reliable cue in predictive noun processing.
Finally, we investigate whether variability in the use of grammatical gender is related
to speed of lexical access. As a measure of lexical access, we use mean reaction times in
the lexical cue condition that includes the numeral ‘two’ (Where do you see two [A]
[N]?). Since this predictive cue does not involve inflection, reaction times in the lexical
cue condition offer insight in how quickly participants use predictive lexical, as opposed
to inflectional, cues, in real-time comprehension. For the lexical cue sentences, mean
reaction time is 655 ms after onset of the colour-disambiguating adjective for the natives;
for the L2ers, it is 620 ms. There is no significant difference between the groups (F(1,38)
= 0.357, p = .582).
To see whether problems of using gender as a predictive cue for agreement in real-
time comprehension is related to speed of lexical access, mean reaction times in the
lexical cue condition were correlated with the size of the predictive gender effect, i.e.
the mean difference in reaction times to the target picture in difference and same trials
for (1). For all participants, we find a significant contingency of speed of lexical access
and size of the gender effect (r = –.513, p = .001), which is also significant among
natives (r = –.455, p = .044) and the L2 group (r = –.634, p = .003) individually. These
correlations indicate that less automatic lexical access leads to lower levels of predic-
tive use of gender agreement in real-time comprehension across the whole group of
participants.
Hopp

Table 7  Pairwise comparisons by group for proportion of looks to target region between difference versus same condition for time windows after
onset of determiner (T-values and p-values)

Time window 1: 0–300 ms 2: 300–600 ms 3: 600–900 ms 4: 900–1,200 ms 5: 1,200–1,500 ms

Natives (n = 20) < 1,, p = 1 1.118, p = .277 .428, p = .673 2.675, p = .015 4.105, p = .001
L2 gender variable (n = 11) < 1, p = .341 < 1, p = .341 < 1, p = .449 1.074, p = .308 1.449, p = .178
L2 gender consistent (n = 9) 1.642, p = .139 2 360, p = .050 2.919, p = .019 3.167, p = .016 2.491, p = .037

Time window 6: 1,500–1,800 ms 7: 1,800–2,100 ms 8: 2,100–2,400 ms 9: 2,400–2,700 ms 10: 2,700–3,000 ms

Natives (n = 20) 5.812, p < .001 7.670, p < .001 8.920, p < .001 2.814, p = .011 1.365, p = .188
L2 gender variable (n = 11) 2.465, p = .033 3.059, p = .012 2.779, p = .019 < 1, p = .423 < 1, p = .851
L2 gender consistent (n = 9) 3.031, p = .016 2.424, p = .042 2.523, p = .036 1.217, p = .258 < 1, p = .754
49
50 Second Language Research 29(1)

VI Discussion
This study tested whether L1 English adult speakers can come to use gender in establishing
predictive agreement relations in L2 German. The results yield differences in anticipatory
use of inflection between natives and adult L2 speakers; yet, a subgroup of L2ers used
gender cues on determiners for establishing predictive syntactic agreement relations.
To our knowledge, this is the first study that establishes on-line sensitivity to gender
marking on determiners in predictive gender processing for L2 speakers whose L1 does
not encode grammatical gender. Before discussing the findings in detail, we consider
whether the particular procedure of the present study might account for target-like L2
performance on gender. First, participants named the objects immediately prior to the
comprehension trials, which may have boosted their performance in comprehension
because the production of the target gender primes its predictive use in subsequent com-
prehension. While effects of lexical production-to-comprehension priming (Tooley and
Traxler, 2010) cannot be definitively ruled out, they seem unlikely to account for target-
like L2 performance. Participants were free in how they named the objects, and most
chose the pragmatically most felicitous option of using indefinite determiners in deter-
miner–adjective–noun sequences. Hence, the descriptions of the objects did not exactly
match the stimulus participants heard immediately afterwards in the condition with gen-
der being marked on a definite determiner.
Second, target-like L2 performance may be taken as owing to the directive nature of
the task, in which participants followed explicit orders where to look. L2 processing is
generally found to be more native-like in tasks that have some metalinguistic or explicit
component compared to tasks that involve comprehension only (e.g. Roberts, to appear;
Williams, 2006). Note, though, that previous studies on L2 predictive gender processing
which did not unearth L2 sensitivity to gender cues also employed directive looking
tasks (Grüter et al., 2012; Lew-Williams and Fernald, 2010). We thus conclude that the
present findings are informative with respect to the question whether speakers establish
automatic predictive gender agreement relations in processing.
Importantly, the predictive effects of gender agreement observed in the present com-
prehension experiment could not reflect surface phonological or form regularities of
gender agreement. Given the phonologically non-transparent German gender paradigm,
L2ers cannot rely on regularity for creating associative agreement relations between
determiners and nouns. Moreover, predictive effects of gender marking in the gender-
consistent group hold across all gender-marking forms and are not limited to low-
frequent or non-default gender markers, like neuter in German. These findings present a
challenge to models of L2 acquisition positing that maturational constraints impair gram-
matical gender representations in late L2 acquisition, so that L2ers are limited to con-
structing surface-based or probabilistic associative agreement relations for features not
instantiated in the L1 (e.g. Hawkins, 2009; Hawkins and Casillas, 2008).
Although approaches positing representational deficits on the real-time integration of
L2 inflection may appear to adequately characterize problems with inflectional morphol-
ogy up to and including advanced proficiency levels in adult L2 acquisition, they cannot
account for target-like predictive processing of phonologically opaque gender across
different gender forms among the gender-consistent L2 group. In turn, however, such
Hopp 51

target performance goes to attest that the processing of syntactic gender agreement in late
L2 acquisition does not fundamentally differ from native processing of inflection.
Besides target-like performance in a subgroup of adult L2ers, the present data also
furnish evidence of variability in the predictive processing of inflection. In the following,
we consider the pattern of inflectional variability in more detail.
First, the group differences among L2 speakers point to a clear contingency between
inflectional variability in gender assignment in production and variability in using gen-
der as a cue for establishing predictive agreement relations in real-time comprehension.
Only the L2ers who have consistent target-like overall gender assignment in production
and whose lexical representations of gender are thus strong enough to be consistently
accessed perform target-like on gender agreement in comprehension.
In contrast, the group with residual variability in gender assignment shows weaker
and inconsistent predictive use of gender marking on determiners, even though only the
trials were analysed in which the target gender had been produced. Problems in the pre-
dictive use of gender in comprehension thus do not reflect that the L2ers assign the par-
ticular nouns a non-target gender or that they cannot access the target gender in
production. Although L2ers have been shown to rely on their subjective gender assign-
ment in detecting gender agreement violations in ERP studies (Lemhöfer et al., 2011),
subjectively accurate gender assignment is apparently not sufficient to ensure target use
of grammatical gender agreement in predictive comprehension. Further, problems with
gender do not differ across gender forms in production or comprehension, such that the
non-target performance of the gender-variable L2 group is neither specific to a particular
gender form nor the result of the overgeneralization of a default gender. Rather, the L2
group differences suggest that the ability to employ grammatical gender as a predictive
cue hinges on overall mastery of lexical gender assignment.
Overall mastery of the target gender system in production means (1) that speakers
have assigned nouns to their target gender classes, e.g. by linking them to the appropri-
ate abstract gender class features, and (2) that this link can be consistently and rapidly
called upon in real-time processing. For (1), we hypothesize that gender acts as an
informative cue in processing only once L2ers have converged on the overall target
system of lexical gender in the L2, i.e. they have classified (almost) all nouns to the
target gender classes. If, however, a speaker retains variability in lexical gender assign-
ment, the speaker will experience frequent mismatches in comprehension between
their subjective gender assignment of lexical items and the target gender of the item
they hear. Imagine a speaker assigns a proportion of L2 nouns a non-target gender (e.g.
[fem] to Hundmasc – ‘dog’ and [masc] to Katzefem – ‘cat’) or wavers between different
gender classes for them. Once he or she encounters gender marking on a determiner or
adjective in comprehension, using it predictively as a cue to restrict the set of potential
nouns would lead to the predictive activation of nouns other than the target noun (e.g.
[fem] activates dog). When the L2er hears the subsequent noun (cat), the prediction the
L2er has made is immediately defeated. As a consequence, the language processor is
unlikely to treat gender marking as an informative cue, because doing so would fre-
quently lead to incorrect anticipations in processing. We speculate that these experi-
ences undermine the reliability of gender marking as a predictive cue in comprehension
and prevent L2ers with non-target overall lexical gender assignment from using it
52 Second Language Research 29(1)

across the board, even though gender marking could potentially lead to facilitation in
processing for the subset of nouns they have assigned to the target gender classes.
Second, (2) the gender of a noun needs to be rapidly accessible in the mental lexicon in
order to be useable in comprehension. In order to test this potential relation directly, we
explored whether inflectional variability is related to the efficiency of lexical access in cor-
relational analyses. Using reaction times for the items that test predictive lexical cues as a
measure of lexical access speed, we find that they correlate significantly with the size of the
predictive gender effect in comprehension. This association suggests that inflectional vari-
ability with gender in adult L2 acquisition is indeed moderated by the efficiency of lexical
access. Moreover, the association between lexical access speed and performance on gender
agreement holds within and across the native and non-native groups. Such a contingency
may be interpreted as further evidence to suggest continuity in the processing of gender as
predictive processing of gender agreement is affected by individual differences in lexical
computation rather than by group differences between natives and non-natives.
This pattern of findings from, for instance this study accords well with accounts of
inflectional variability in adult L2 acquisition that locate problems with gender at the lexi-
cal level as well as in the retrieval of inflectional forms, for instance the Missing Surface
Inflection Hypothesis (Prévost and White, 2000) and the lexical gender learning hypoth-
esis (Grüter et al., 2012). These models predict the observed contingencies between the
production of lexical aspects of gender and comprehension of gender agreement. In stress-
ing lexical difficulties in grammatical processing, these hypotheses link up to models of
the bilingual mental lexicon which hold that L2ers create weaker lexical links of and
between L2 representations than natives in the L1 because they use the L2 comparatively
less frequently (e.g. Gollan et al., 2002). One consequence of such weaker links is that
L2ers are delayed in accessing grammatical and lexical information in L2 processing (e.g.
Runnqvist et al., 2011). In this respect, the present findings support capacity models of L2
inflectional variability that attribute problems with inflection in very advanced late L2
acquisition to lower processing resources in non-natives compared to natives which con-
strain L2ers in consistently accessing and computing the appropriate inflection in real
time (Hopp, 2010; McDonald, 2006). In positing weaker lexical links and concomitant
delays in accessing inflectional forms in adult L2 acquisition, lexical and capacity models
together provide a principled account of why predictive syntactic gender agreement is
selectively problematic in real-time L2 processing.
At any rate, the evidence from the current study that overall mastery of lexical gender
affects the processing of syntactic gender agreement may lead to the reinterpretation of find-
ings from previous research. By finding group differences among L2ers, the present findings
resonate with research investigating sensitivity to gender agreement within the violation para-
digm tracking reading times (Keating, 2009; Sagarra and Herschensohn, 2010) or ERP signa-
tures in natural (Gabriele et al., 2012; Gillon Dowens et al., 2010) and artificial (Morgan-Short
et al. 2010) languages. Across studies, it has been found that late L2ers at higher proficiency
perform more target-like on gender agreement. Although the small sample of L2ers in this
study does not allow for disentangling effects of general L2 proficiency and the specifics of
lexical gender, the disjunction between two groups in this study may be interpreted as sug-
gesting that the general proficiency differences reported for the processing of gender agree-
ment index more specific differences in overall mastery of lexical aspects of gender. However,
in view of the covariance of proficiency and accuracy in gender assignment in the present L2
Hopp 53

sample, future research needs to disentangle these factors in larger-scale studies of L2ers
matched in overall proficiency levels.
Currently, we are investigating how L1 differences impact on the relation between
gender assignment and predictive processing by studying L2 groups whose L1s differ
in the distributional and lexical realizations of gender. We also plan to explore the rela-
tions between lexical access and gender processing by using various measures of lexical
access and varying noun frequency to see whether the stability of gender categorization
of specific lexical items affects agreement processing. Experiments along these lines
will cast more light on the interrelations between lexical and syntactic aspects of L2
gender and reveal whether L2 inflectional variability can ultimately receive a unified
explanation by situating it in an integrated model of the L2 lexicon and grammar.

VII Conclusions
The present study finds that a group of late L2 learners whose L1 does not encode gram-
matical gender can use syntactic determiner–noun gender agreement in predictive pro-
cessing. Mastery of predictive gender agreement in the L2 appears to hinge on overall
mastery of lexical gender in the L2. The observed contingencies between lexical gender
assignment in production and syntactic agreement in comprehension suggest that varia-
bility in the use of gender agreement in real-time L2 comprehension reflects lower levels
of activation of and access to gender nodes in the adult bilingual mental lexicon rather
than representational deficits in L2 grammars.

Funding
This research received no specific grant from any funding agency in the public, commercial, or
not-for-profit sectors.

Acknowledgments
I would like to thank the audience at BUCLD 36, the editors of SLR and two anonymous reviewers
for helpful comments on previous versions of this paper.

Note
1 The native and the L2 group differ in mean age, and this age difference may have put the L2 group
at a disadvantage due to slower overall processing compared to the natives. As shown in the analy-
sis of the mean reaction times in the lexical cue condition in Section 5, however, there appears to
be no general group difference between the natives and the L2ers in terms of speed of processing.

2 Given the syncretism of German gender marking, it is unfortunately not feasible to clearly distinguish
between errors of lexical assignment and errors of syntactic agreement, since the gender of the nouns
is only recoverable from the determiner (for definites) or the adjective (for indefinites), yet not both at
the same time. However, genuine agreement errors between unambiguously marked determiners and
adjectives (e.g. eineFEM gelbesNEUT Haus – ‘a yellow house’) were virtually unattested in the data.

References
Alarcón IV (2011) Spanish gender agreement under complete and incomplete acquisition: Early
and late bilinguals’ linguistic behavior within the noun phrase. Bilingualism: Language and
Cognition 14: 332–50.
54 Second Language Research 29(1)

Allopenna PD, Magnuson JD, and Tanenhaus MK (1998) Tracking the time course of spoken
word recognition using eye movements: Evidence for continuous mapping models. Journal of
Memory and Language 38: 419–39.
Arnon I and Ramscar M (2012) Granularity and the acquisition of grammatical gender: How
order-of-acquisition affects what gets learned. Cognition 122: 292–305.
Barr DJ, Gann TM, and Pierce RS (2011) Anticipatory baseline effects and information integration
in visual world studies Acta Psychologica 137: 201–07.
Bruhn de Garavito J and White L (2003) The second language acquisition of Spanish DPs: The
status of grammatical features. In: Pérez-Leroux AT and Liceras JM (eds) The acquisition of
Spanish morphosyntax: The L1/L2 connection. Dordrecht: Kluwer, 135–78.
Carroll S (1989) Second-language acquisition and the computational paradigm. Language
Learning 39: 535–94.
Carstens V (2000) Concord in minimalist theory. Linguistic Inquiry 31: 319–55.
Chiarcos C (2009) Frequency of masc./fem/neut. in German. Corpora List. Retrieved from: http://
mailman.uib.no/public/corpora/2009-April/008723.html (October 2012).
COSMAS II (2008) Corpus Search, Management and Analysis System. Retrieved from: http://
www.ids-mannheim.de/cosmas2 (August 2012).
Corbett GG (1991) Gender. Cambridge: Cambridge University Press.
Dahan D, Swingley D, Tanenhaus MK, and Magnuson JS (2000) Linguistic gender and spoken-
word recognition in French. Journal of Memory and Language 42: 465–80.
Davidson DJ and Indefrey P (2009) An event-related potential study on changes of violation and
error responses during morphosyntactic learning. Journal of Cognitive Neuroscience 21: 433–46.
Foote R (2011) Integrated knowledge of agreement in early and late English–Spanish bilinguals.
Applied Psycholinguistics 32: 187–220.
Foucart A and Frenck-Mestre CA (2011) Grammatical gender processing in L2: Electrophysiological
evidence of the effect of L1–L2 syntactic similarity. Bilingualism: Language and Cognition
14: 379–99.
Franceschina F (2005) Fossilized second language grammars: The acquisition of grammatical
gender. Amsterdam: John Benjamins.
Franck J, Vigliocco G, Antón-Méndez I, Collina S, and Frauenfelder UH (2008) The interplay of
syntax and form in sentence production: A cross-linguistic study of form effects on agreement.
Language and Cognitive Processes 23: 329–74.
Gabriele A, Fiorentino R, and Alemán Bañón J (2012) Examining second language development
using event-related potentials: a cross-sectional study on the processing of gender and number
agreement. Bilingualism: Language and Cognition.
Gillon Dowens M, Vergara M, Barber HA, and Carreiras M (2010) Morphosyntactic processing in
late second-language learners. Journal of Cognitive Neuroscience 22: 1870–87.
Gillon Dowens M, Barber H, Guo Dowens M, et al. (2011) Gender and number processing in Chinese
learners of Spanish: Evidence from event related potentials. Neuropsychologia 49: 1651–59.
Goethe Institut (2010) German placement test. Retrieved from: http://www.goethe.de/cgi-bin/ein-
stufungstest/einstufungstest.pl (October 2012).
Gollan TH, Montoya RI, and Werner G (2002) Semantic and letter fluency in Spanish-English
bilinguals. Neuropsychology 16: 562–67.
Grüter T, Lew-Williams C, and Fernald A (2012) Grammatical gender in L2: A production or a
real-time processing problem? Second Language Research 28: 191–215.
Guillelmon D and Grosjean F (2001) The gender marking effect in spoken word recognition: The
case of bilinguals. Memory and Cognition 29: 503–11.
Hawkins R (2009) Statistical learning and innate knowledge in the development of second lan-
guage proficiency: Evidence from the acquisition of gender concord. In: Benati AG (ed.)
Issues in Second Language Proficiency. London: Continuum, 63–78.
Hopp 55

Hawkins R and Casillas G (2008) Explaining frequency of verb morphology in early L2 speech.
Lingua 118: 595–612.
Hohlfeld A (2006) Accessing grammatical gender in German: The impact of gender-marking reg-
ularities. Applied Psycholinguistics 27: 127–42.
Hopp H (2010) Ultimate attainment in L2 inflectional morphology: Performance similarities
between non-native and native speakers. Lingua 120: 901–31.
Hopp H (2011) Internal and external factors in the child L2 development of the German determiner
phrase. Linguistic Approaches to Bilingualism 1: 238–64.
Huettig F, Rommers J, and Meyer AS (2011) Using the visual world paradigm to study language
processing: A review and critical evaluation. Acta Psychologica 137: 151–71.
Johnson E (2005) Grammatical gender and early word recognition in Dutch. In: Brugos A, Clark-
Cotton MR and Ha S (eds) Proceedings of the 29th Annual Boston University Conference on
Language Development (BUCLD 29). Somerville: Cascadilla Press, 320–30.
Kamide Y (2008) Anticipatory processes in sentence processing. Language and Linguistics
Compass 2: 647–70.
Keating GD (2009) Sensitivity to violations of gender agreement in native and nonnative Spanish:
An eye-movement investigation. Language Learning 59: 503–35.
Keating GD (2010) The effects of linear distance and working memory on the processing of gen-
der agreement in Spanish. In: VanPatten B and Jegerski J (eds) Research in second language
processing and parsing. Amsterdam: Benjamins, 113–34.
Kornai A (2009) Frequency of masc/fem/neut in German. Corpora List. Retrieved from: http://
mailman.uib.no/public/corpora/2009-April/008721.html (October 2012).
Lemhöfer K, Schriefers H, and Indefrey P (2011) Cognitive mechanisms of L2 errors and ‘fos-
silization’: An ERP study. Poster presented at Architectures and Mechanisms for Language
Processing (AMLaP), September, Paris, France.
Lew-Williams C and Fernald A (2007) Young children learning Spanish make rapid use of gram-
matical gender in spoken word recognition. Psychological Science 18: 193–98.
Lew-Williams C and Fernald A (2010) Real-time processing of gender-marked articles by native
and non-native Spanish speakers. Journal of Memory and Language 63: 447–64.
Lyster R (2006) Predictability in French gender attribution: A corpus-based analysis. French
Language Studies 16: 69–92.
McCarthy C (2008) Morphological variability in the comprehension of agreement: An argument
for representation over computation. Second Language Research 24: 459–86.
McDonald JL (2006) Beyond the critical period: Processing-based explanations for poor gram-
maticality judgment performance by late second language learners. Journal of Memory and
Language 55: 381–401.
Montrul S, Foote R, and Perpiñán S (2008) Gender agreement in adult second language learn-
ers and Spanish heritage speakers: The effects of age and context of acquisition. Language
Learning 58: 503–53.
Morgan-Short K, Sanz C, Steinhauer K, and Ullman MT (2010) Second language acquisition of
gender agreement in explicit and implicit training conditions: An event-related potential study.
Language Learning 60: 154–93.
Prévost P and White L (2000) Missing surface inflection or impairment in second language acqui-
sition? Evidence from tense and agreement. Second Language Research 16: 103–33.
Roberts L (to appear) Sentence processing in bilinguals. In: van Gompel RPG (ed.) Sentence pro-
cessing. London: Psychology Press.
Runnqvist E, Strijkers K, Sadat J, and Costa A (2011) On the temporal and functional origin of L2
disadvantages in speech production: A critical review. Frontiers in Psychology 2: 1–8.
Sabourin L and Stowe LA (2008) Second language processing: When are first and second lan-
guages processed similarly? Second Language Research 24: 397–430.
56 Second Language Research 29(1)

Sagarra N and Herschensohn J (2010) The role of proficiency and working memory in gender and
number agreement processing in L1 and L2 Spanish. Lingua 120: 2022–39.
Scherag A, Demuth L, Rösler F, Neville HJ, and Röder B (2004) The effects of late acquisition of
L2 and the consequences of immigration on L1 for semantic and morpho-syntactic language
aspects. Cognition 9: B97–B108.
Schriefers H and Jescheniak JD (1999) Representation and processing of grammatical gender in
language production: A review. Journal of Psycholinguistic Research 28: 575–600.
Teschner RV and Russell WM (1984) The gender patterns of Spanish nouns: An inverse diction-
ary-based analysis. Hispanic Linguistics 1: 115–32.
Tokowicz N and MacWhinney B (2005) Implicit and explicit measures of sensitivity to viola-
tions in second language grammar: An event-related potential investigation. Studies in Second
Language Acquisition 27: 173–204.
Tooley KM and Traxler MJ (2010) Syntactic priming effects in comprehension: A critical review.
Language and Linguistics Compass 4: 925–37.
Unsworth S (2008) Age and input in the acquisition of grammatical gender in Dutch. Second
Language Research 24: 365–96.
White L (2007) Some puzzling features of L2 features. In: Liceras JM, Zobl H and Goodluck H (eds)
The role of features in second language acquisition. Mahwah, NJ: Lawrence Erlbaum, 305–30.
White L (2011) Second language acquisition at the interfaces. Lingua 121: 577–90.
White L, Valenzuela E, Kozlowska-MacGregor M, and Leung YKI (2004) Gender and number
agreement in nonnative Spanish. Applied Psycholinguistics 25: 105–33.
Williams JC (2006) Incremental interpretation in second language sentence processing.
Bilingualism: Language and Cognition 9: 71–88.

Appendix 1  Experimental items: Nouns

Masculine Feminine Neuter

Topf – pot Axt – axe Radio – radio
Herd – stove Tasse – cup Glas – glass
Fernseher – television Treppe – staircase Lenkrad – wheel
Knopf – button Gabel – fork Tor – gate
Bus – bus Banane – banana T-Shirt – t-shirt
Apfel – apple Lampe – lamp Brot – bread
Schrank – cupboard Torte – cake Hemd – shirt
Zahn – tooth Hand – hand Trampolin – trampoline
Käse – cheese Flasche – bottle Flugzeug – plane
Teller – plate Karte – card Kleid – dress
Schlauch – hose Blume – flower Bett – bed
Besen – broom Schüssel – bowl Auge – eye
Löffel – spoon Rose – rose Schiff – ship
Roboter – robot Zange – pliers Kissen – cushion
Trichter – funnel Kasse – till Zelt – tent

Experimental items: Adjectives

rote – red braune – brown
grüne – green blaue – blue
gelbe – yellow weisse – white
Copyright of Second Language Research is the property of Sage Publications, Ltd. and its content may not be
copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written
permission. However, users may print, download, or email articles for individual use.