Você está na página 1de 22

Language Acquisition

ISSN: 1048-9223 (Print) 1532-7817 (Online) Journal homepage: http://www.tandfonline.com/loi/hlac20

Age of Acquisition and Sensitivity to Gender in


Spanish Word Recognition

Rebecca Foote

To cite this article: Rebecca Foote (2014) Age of Acquisition and Sensitivity to
Gender in Spanish Word Recognition, Language Acquisition, 21:4, 365-385, DOI:
10.1080/10489223.2014.892948

To link to this article: http://dx.doi.org/10.1080/10489223.2014.892948

Accepted author version posted online: 28


Feb 2014.
Published online: 28 Feb 2014.

Submit your article to this journal

Article views: 237

View related articles

View Crossmark data

Citing articles: 1 View citing articles

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=hlac20

Download by: [Radcliffe Infirmary] Date: 15 July 2016, At: 03:44


Language Acquisition, 21: 365–385, 2014
Copyright © Taylor & Francis Group, LLC
ISSN: 1048-9223 print / 1532-7817 online
DOI: 10.1080/10489223.2014.892948

Age of Acquisition and Sensitivity to Gender in Spanish


Word Recognition

Rebecca Foote
University of Illinois
Downloaded by [Radcliffe Infirmary] at 03:44 15 July 2016

Speakers of gender-agreement languages use gender-marked elements of the noun phrase in


spoken-word recognition: A congruent marking on a determiner or adjective facilitates the recog-
nition of a subsequent noun, while an incongruent marking inhibits its recognition. However, while
monolinguals and early language learners evidence this gender-marking effect, late learners do not
(Guillelmon & Grosjean 2001). The goals of this study were to determine whether early learners
who are not dominant in the gender-marking language (Spanish) can use gender cues in spoken-word
recognition and whether the ability of both early and late learners to do so is a function of the noun’s
gender-marking transparency. Results of a word-repetition task, completed by 32 native Spanish
speakers and 64 English-dominant early and late learners of Spanish indicate that both types of learn-
ers make use of gender cues during spoken-word recognition, and that gender-marking transparency
may influence this process.

1. INTRODUCTION

One of the most difficult tasks for second-language (L2) learners of languages that mark gender
in the noun phrase (NP) is to fluently and accurately employ gender during the comprehension
or production of the L2. A relatively large number of studies suggests that the correct assignment
of gender to nouns and the implementation of gender agreement can cause lasting difficulties
for L2 learners (e.g., Carroll 1989; Dewaele & Veronique 2001; Franceschina 2001; McCarthy
2008; Grüter, Lew-Williams & Fernald 2012), particularly those whose first language (L1) does
not mark noun gender (e.g., Franceschina 2005; Sabourin, Stowe & de Haan 2006). And while
there is some evidence that gender is eventually acquirable by L2 learners (e.g., White et al. 2004;
Sagarra & Herschensohn 2011), the majority of empirical studies, along with anecdotal evidence,
suggests that most learners never acquire gender to nativelike levels.
One factor that may play a role in whether learners are able to make use of gender in a
nativelike way is age of language acquisition. Many researchers who have found that learners
experience persistent difficulties with gender in the L2 attribute these difficulties to a critical
period for the acquisition of language in general, and more specifically, for the acquisition of

Correspondence should be sent to Rebecca Foote, Department of Spanish, Italian, and Portuguese, University
of Illinois, 4080 Foreign Languages Building, MC-176, 707 South Mathews Avenue, Urbana, IL 61801. E-mail:
rfoote@illinois.edu
366 FOOTE

gender if it is not instantiated in the learner’s L1 (e.g., Hawkins & Franceschina 2004). However,
only a few studies have directly compared gender use in early and late learners in order to
determine whether early acquisition of gender is necessary (and sufficient) for nativelike usage.
In one such study, Montrul, Foote & Perpiñán (2008) compared gender knowledge in proficiency-
matched early and late learners of Spanish and found that early learners performed more like
native speakers than did late learners on a task that required spontaneous production, though both
groups differed from the native speakers. These results suggest that there is an advantage for early
acquisition, but they also imply that this advantage may not be enough to confer nativelikeness.
Foote (2011) investigated whether both early and late learners were sensitive to gender-agreement
errors while reading sentences in Spanish for comprehension and found that both types of learners
were sensitive; age of acquisition did not appear to play a role.
In a different type of task, however, Guillelmon & Grosjean (2001) found that age of acqui-
Downloaded by [Radcliffe Infirmary] at 03:44 15 July 2016

sition did matter. Research with native speakers of gender-agreement languages has shown that
gender-marked elements of the NP are used during spoken-word recognition: Congruent gender
marking on a determiner or adjective facilitates the recognition of a subsequent noun, while an
incongruent marking inhibits its recognition (Grosjeanet al. 1994; Bates et al. 1996). For example,
if the Spanish noun bolsa ‘bag,’ which is feminine, is preceded by a determiner marked for femi-
nine gender, such as la ‘the,’ native speakers of Spanish are faster to recognize the noun than they
are if bolsa is not preceded by a determiner marked for gender. In contrast, if bolsa is preceded
by a determiner that is marked for masculine gender, such as el ‘the,’ native Spanish speakers are
slower to recognize bolsa than they are if the noun is not preceded by a gender-marked deter-
miner. Based on these findings, Guillelmon & Grosjean compared the ability of early and late
learners to utilize gender cues in spoken-word recognition in French and found that early learners
made use of these cues in a nativelike way, but that late learners were insensitive to them.
More recently, Montrul et al. (2014) replicated the finding of differences between early and
late learners in the same type of task, again comparing proficiency-matched early and late learn-
ers of Spanish. In two tasks that were relatively metalinguistic in nature, they found that early and
late learners performed like native speakers in showing sensitivity to incongruent gender mark-
ing, but in a word-recognition task, late learners patterned differently from both native speakers
and early learners. However, only effects of incongruent gender marking were reported for the
native speakers and early learners in this study, in contrast to Guillelmon & Grosjean (2001),
who showed that native speakers and early learners of French were sensitive to both congruent
and incongruent gender cues in spoken-word recognition.
In addition to dissimilarities between these studies in terms of tasks used to test gender, dif-
ferences in the nature of the stimuli, in early learner language dominance, and in late learner
proficiency levels may explain the conflicting results obtained. Specifically, the majority of the
nouns used in Foote (2011) were transparently marked for gender, while those that were used in
Montrul, Foote & Perpiñán (2008) and Guillelmon & Grosjean (2001) were not. Montrul et al.
(2014) manipulated gender-marking transparency and found that all groups were affected by
whether the noun was clearly marked as masculine or feminine, but both learner groups were
more affected than native speakers. In addition, the early learners in Foote, Montrul, Foote
& Perpiñán (2008), and Montrul et al. (2014) were not dominant in the gender-marking lan-
guage under investigation. In contrast, those in Guillelmon & Grosjean were. And finally, while
the Guillelmon & Grosjean late learners were likely highly proficient in the gender-marking
AGE OF ACQUISITION AND SENSITIVITY TO GENDER 367

language, as were the late learners in Foote (2011), the late learners in Montrul, Foote & Perpiñán
(2008) and Montrul et al. (2014) showed a range of proficiency levels, from intermediate to
advanced. The goals of the current study are therefore first, to determine whether early learn-
ers who are not dominant in the gender-marking language are able to make use of gender cues
(specifically, determiners) during spoken-word recognition in the same way as native speakers,
showing sensitivity to both congruent and incongruent gender cues, and second, to investigate
whether the ability of highly proficient early and/or late learners to utilize these cues may depend
on the transparency of the noun’s gender marking. If age of acquisition is what determines one’s
ability to make use of gender cues during word recognition, then early learners should be able
to do so regardless of whether the noun is transparently marked for gender. If highly proficient
late learners are able to make use of these cues to some degree in spite of their late acquisition
of the gender-marking language, it may be that they can only do so when nouns are transpar-
Downloaded by [Radcliffe Infirmary] at 03:44 15 July 2016

ently marked for gender. In order to examine these issues, native speakers of Spanish, along with
advanced-proficiency early and late learners, participated in a partial replication of Guillelmon
& Grosjean. Results suggest that, like native speakers, both early and late learners make use of
gender cues during spoken-word recognition in Spanish. They also indicate that gender marking
transparency may affect this process.

2. GENDER IN SPANISH

All nouns in Spanish are classified according to gender as masculine or feminine. The classi-
fication of animate nouns is primarily based on biological gender, while the classification of
inanimate nouns is semantically arbitrary but tends to follow trends based on the noun’s morpho-
logical and phonological characteristics. Nouns that are considered to be morphophonologically
canonical, or transparently marked for gender, include masculine nouns that end with the suffix
–o (e.g., libro ‘book’), and feminine nouns that end with the suffix –a (e.g., bolsa ‘bag’). Nouns
that end in a consonant or –e, and nouns that end in other vowels are opaque with respect to gen-
der; they may be either masculine (e.g., lápiz ‘pencil’) or feminine (e.g., tribu ‘tribe’). Another
group of nouns end in –o or –a, but their gender is the opposite of what would be expected based
on the principal morphophonological trends (e.g., mano–fem. ‘hand’; mapa–masc. ‘map’). There
are approximately twice as many transparently marked nouns in Spanish as there are opaquely
marked ones (Harris 1991).
The gender of a noun has consequences for other elements in the noun phrase and beyond.
Specifically, any determiners and adjectives that accompany or modify a noun must agree with
the noun in gender (and in number). Both definite and indefinite articles in Spanish have mas-
culine and feminine forms that are different in the singular and the plural. Most adjectives have
forms that pattern with the canonical suffixes, so that the masculine form ends in –o, and the
feminine form in –a. However, there are many adjectives that do not vary in form depending on
the gender of the noun that they agree with, but end in –e, a consonant, or even another vowel
when accompanying both masculine and feminine nouns (e.g., verde ‘green,’ or mejor ‘better’).
In summary, while the assignment of gender to inanimate nouns in Spanish is relatively arbitrary,
there are morphophonological patterns or trends in gender assignment that are largely echoed in
the morphophonology of modifying elements in the NP.
368 FOOTE

3. THE PROCESSING OF GENDER IN WORD RECOGNITION

In terms of the formal representation of gender, proficient native speakers of a language are
thought to store gender as an inherent property of each noun (Schriefers & Jescheniak 1999).
From a psycholinguistic perspective, there are two main types of models that address how
this stored gender information may be processed during language comprehension (Friederici
& Jacobsen 1999). According to interactive models of lexical access, word recognition can be
affected not only by the form of the word as it is being heard or read but also by contextual
information, such as information about what word would plausibly fit next in the phrase (seman-
tics), or more pertinent to the present study, information about what word would fit syntactically
(Grosjean et al. 1994; Bates et al. 1996). In this view, gender information specified on preceding
elements in an NP can affect the recognition of a subsequent noun either by allowing the compre-
Downloaded by [Radcliffe Infirmary] at 03:44 15 July 2016

hender to narrow the possible word candidates to only those nouns with gender that is congruent
with (the same as) the preceding gender information, or by increasing the activation of nouns that
share the gender of the preceding element. This means that when the gender of the preceding
element is congruent with the gender of the noun, word recognition should be speeded up (facil-
itation). When the gender of the preceding element is incongruent, word recognition should be
slowed (inhibition).
The other main type of model that addresses how gender may be processed is modular in
nature, claiming that lexical access is independent of contextual information about semantics and
syntax (Tanenhaus & Lucas 1987). Modular models of word recognition assume that gender can-
not affect the process of recognizing a word but may affect processes that occur after recognition
takes place (postlexical as opposed to prelexical processes). Under this type of model, one way
that gender information specified on elements of an NP that precede the noun may affect the pro-
cesses that occur after the recognition of the noun is via a syntactic congruency check. According
to Friederici & Jacobsen (1999), gender information specified on NP elements may remain active
until it can be checked against the noun’s gender. When the gender information on the noun is
congruent with the gender information provided by the preceding context, this check is faster in
comparison to when this information is incongruent. If there is no gender information provided by
the preceding context, the check takes the same amount of time as when there is congruent infor-
mation. This implies that there should be no facilitation with gender-congruent information but
rather only inhibition of the syntactic congruency check when gender-incongruent information is
present.
To summarize, two distinct predictions about how gender is processed during language com-
prehension emerge from these two types of models. According to interactive views of word recog-
nition, gender information from the preceding context affects the recognition of a subsequent
noun; there is a prelexical effect that can be facilitatory or inhibitory. According to modular views,
gender information is only taken into account postlexically and so can only cause inhibitory
effects. In the literature on the processing of gender during word recognition, the presence of
both facilitatory and inhibitory effects has therefore been interpreted as support for interactive
views of lexical access, while the finding of only inhibitory effects has been seen as supporting
a modular view of how lexical access proceeds. Researchers have found both facilitatory and
inhibitory effects, though the inhibitory effects seem to be stronger and more reliably present.
In one of the first studies on gender priming, Carello, Lukatela & Turvey (1988) found pri-
marily inhibitory effects of incongruently marked possessive adjectives on lexical decisions to
AGE OF ACQUISITION AND SENSITIVITY TO GENDER 369

subsequent nouns in Serbo-Croatian. Jakubowicz & Faussart (1998) found that participants were
faster to make lexical decisions to target French nouns when a congruent gender marking was
present on a preceding determiner than when incongruent gender was present; this effect cannot
be specified as facilitatory or inhibitory, since there was no neutral baseline condition in their
experiments. Akhutina et al. (1999) investigated gender priming in Russian via word-repetition
tasks and found inhibitory effects for masculine, feminine, and neuter nouns, but facilitatory
effects for feminine nouns only. Bölte & Connine (2004) found facilitatory effects in auditory
lexical decisions to German nouns that were either preceded by an article that was congruently
marked for gender or preceded by noise. Results from Grosjean et al.’s (1994) investigation of
gender processing in French indicated that participants were faster to identify or make lexical
decisions to nouns when those nouns were preceded by a congruently marked article than when
they were not preceded by an article.
Downloaded by [Radcliffe Infirmary] at 03:44 15 July 2016

Bates et al. (1996) made use of three different tasks to explore gender processing during
spoken-word recognition in Italian. The first task was a word-repetition task in which participants
repeated nouns that were preceded by adjectives that were either congruently or incongruently
marked for gender or that were not marked for gender. In the second task, participants heard the
same type of stimuli but were asked to indicate the gender of the noun as quickly as possible by
pressing a button. The third task was a grammaticality-judgment task, in which participants had to
press a button to indicate whether the adjective-noun combination they heard was grammatical or
not. Results differed across the three tasks; both facilitatory and inhibitory effects were found in
the word-repetition task, although the inhibitory effect was larger. In the gender-monitoring task,
there was only an inhibitory effect, and in the grammaticality-judgment task, participants were
slower to make judgments with gender-incongruent adjectives in comparison to gender-congruent
adjectives. Bates et al. concluded that at least part of the gender-priming effect is prelexical in
nature, based on their characterization of pre- and postlexical processes. According to what they
call the “standard two-stage model of lexical access” (p. 994), during word recognition, prelexi-
cal processes include the activation of lexical items by both the form of the incoming word and
by spreading activation from within the lexicon that can include form and meaning information
from words that precede the current lexical item. These processes affect the recognition of the
word and are thought to be automatic in nature, meaning that they occur very quickly and are
not available for conscious inspection. Postlexical processes include the selection of candidates
that fit within the context, the inhibition of candidates that do not fit, and the integration of the
selected candidate within the larger context of the utterance. These processes do not affect the
recognition of the word but come into play soon after recognition. They are thought to be more
controlled or strategic processes, and as such, are slower than prelexical processes and more
related to conscious attentional mechanisms. Bates et al. argued that the three tasks that they
employed in their study varied on the degree of explicit attention to gender required to complete
the task. The word-repetition task was a more automatic task that required no attention to gender,
since participants only had to repeat the noun that they heard, while the gender-monitoring task
and the grammaticality-judgment task both required explicit attention to gender and thus were
more likely to elicit controlled, conscious processing. According to Bates et al., evidence for a
prelexical component of gender-priming effects was provided by the finding of both facilitatory
and inhibitory priming effects in the word-repetition task, the most automatic of the three tasks.
Most relevant to the current study is research carried out by Guillelmon & Grosjean (2001) to
investigate the nature of gender processing during spoken-word recognition in bilinguals. In their
370 FOOTE

study, they had native speakers of French along with early and late learners of French complete
a word-repetition task in which half of the participants in each group were asked to repeat nouns
that were preceded by either a congruently marked determiner and an adjective without gender
marking or a nonmarked determiner and a nonmarked adjective. The other half of the participants
repeated nouns that were preceded with an incongruently marked determiner and a nonmarked
adjective or a nonmarked determiner and a nonmarked adjective. The native speakers of French
and the early learners showed both facilitatory and inhibitory effects; however, the late learners
did not show evidence of sensitivity to either congruent or incongruent gender marking on the
determiner. Guillelmon & Grosjean concluded based on these results that early age of acquisition
is vital for nativelike processing of gender. A relatively large body of research on the acquisition
of gender in an L2 suggests that this may be true, though it is not clear if early age of acquisition
is sufficient for nativelikeness.
Downloaded by [Radcliffe Infirmary] at 03:44 15 July 2016

4. AGE OF ACQUISITION AND GENDER IN THE L2

In the literature on the adult L2 acquisition of gender-marking languages, there is evidence that
gender assignment and agreement can cause difficulties that persist even at high levels of profi-
ciency (Dewaele & Véronique 2001; Franceschina 2001, 2005). The existence of such difficulties,
however, may interact with the learner’s L1. Representational and processing deficit views of
L2 acquisition claim that if gender is not present in a learner’s L1, then he or she will be unable
either to acquire it to nativelike levels in the L2 (e.g., Hawkins & Franceschina 2004) or to process
it like native speakers do (e.g., Clahsen & Felser 2006). This hypothesis is supported by findings
of persistent gender errors in production and nonnativelike processing of gender in comprehen-
sion on the part of learners of a gender-marking L2 whose L1 does not mark gender (Franceschina
2001; McCarthy 2008; Grüter, Lew-Williams & Fernald 2012) and by studies showing an advan-
tage in gender production and processing for late learners whose L1 marks gender over those
whose L1 does not (Franceschina 2005; Sabourin, Stowe & DeHaan 2006). Evidence against the
role of the L1 in L2 gender acquisition comes from studies that suggest nativelike performance in
gender processing and production by late learners whose L1 lacks gender (Alarcón 2006; Keating
2009; Foote 2011; Sagarra & Herschensohn 2010, 2011) and by findings of no performance dif-
ferences between learners whose L1 differs on whether gender is instantiated or not (White et al.
2004).
The role of age of language acquisition in gender processing and production has not been
widely investigated except in an indirect manner (by comparing late learners to native speakers
rather than to early learners). Nevertheless, researchers typically attribute the gender difficulties
experienced by late learners to some sort of critical-period effect (Carroll 1989; Guillelmon &
Grosjean 2001; Hawkins & Franceschina 2004). Native speakers of gender-marking languages,
including Spanish, generally master gender by age 3 or 4 (Montrul 2004) and tend to make very
few gender errors (Carroll 1989). However, the picture is not quite so clear in early learners
who are acquiring two languages at approximately the same time. Montrul & Potowski (2007)
compared monolingual Mexican children to children from Spanish- and English-speaking homes
enrolled in a two-way Spanish-English immersion school in the U.S. on their gender production.
Results indicated that the bilingual children from the Spanish-speaking homes generally outper-
formed those from English-speaking homes but that both groups differed from monolinguals. The
AGE OF ACQUISITION AND SENSITIVITY TO GENDER 371

bilingual children from Spanish-speaking homes in the Montrul & Potowski study represent a par-
ticular group of early language learners, sometimes called heritage speakers. Heritage speakers
typically learn a minority language at home until beginning schooling in the majority language
around age 5. These speakers then often shift their language dominance so that they become
dominant in the majority language (their L2) rather than their L1 (Kohnert, Bates & Hernández
1999). The L1 is either not completely acquired or it is attrited due to reduced input.
Studies that investigate the linguistic knowledge of these speakers tend to show that their
knowledge of the L1 is not the same as that of monolingual L1 speakers (see Montrul 2008 for
an overview). In one study relevant to the current research, Polinsky (2008) examined knowledge
of gender in Russian heritage speakers and found that while her participants did have knowl-
edge of gender assignment, their knowledge differed from native speakers of Russian who grew
up as monolinguals. Montrul, Foote & Perpiñán (2008) directly compared knowledge of gender
Downloaded by [Radcliffe Infirmary] at 03:44 15 July 2016

in Spanish in proficiency-matched heritage speakers and late language learners with three tasks,
two written and one oral. They found that the heritage speakers outperformed the late learners
on the oral task (though the late learners outperformed the heritage speakers on the written tasks,
indicating a role for modality in the performance of the two groups). Based on their results, the
authors concluded that there is an advantage for early acquisition of a gender-marking language,
particularly when it comes to the oral, spontaneous production and processing of gender; this
is similar to what Guillelmon & Grosjean (2001) concluded, based on their study with early
and late French learners. However, Montrul et al. also noted that the heritage speakers in their
study differed from the native-speaker control group on all tasks, suggesting that the advantage
of learning a gender-marking language early may not be sufficient to guarantee nativelike pro-
cessing and production, though a strict view of the Critical Period Hypothesis implies that this
should not be the case. One important difference between the participants in the Montrul et al.
study and the Guillelmon & Grosjean study is that the Guillelmon & Grosjean early learners
were dominant in French (the gender-marking language), while the Montrul et al. learners were
dominant in English. A more recent study directly comparing highly proficient heritage speakers
and late learners on their sensitivity to gender-agreement errors while reading Spanish sentences
for meaning found results that differed from both Montrul et al. and Guillelmon & Grosjean
(Foote 2011). In this study, both heritage speakers and late learners were sensitive to errors in
gender, performing similarly to native speakers. One possible reason for the conflicting findings
is that the three studies employed different tasks: In Montrul et al. and Guillelmon & Grosjean,
the advantage for the early learners was present in an aural/oral task, while Foote employed a
reading task. However, one other factor that could have contributed to the divergent findings is
a difference in the relative transparency of gender markers on the noun stimuli across the three
studies. The nouns in Guillelmon & Grosjean and Montrul et al. varied in their gender-marking
transparency, while almost all of the nouns in Foote were transparently marked for gender.
According to Carroll (1989; see also Grüter, Lew-Williams & Fernald 2012), the central differ-
ence between the acquisition of gender by native speakers of gender-marking languages and the
acquisition of gender by late learners is that children acquiring a gender-marking language from
birth initially analyze determiners as part of the noun and only later reanalyze them as separate
words. This reanalysis “serves as a trigger for the activation of the universally specified gender
feature” (Carroll, p. 572). Late learners do not go through these stages of analysis and reanal-
ysis and thus do not develop gender representations in the same way as native speakers. They
may therefore be forced to rely on “mnemonic strategies” (p. 579) or “rules” in order to produce
372 FOOTE

gender accurately rather than simply retrieving the gender directly from the lexical representa-
tion as native speakers do. These strategies can be based on phonological patterns that may help
learners recognize which nouns are feminine and which are masculine. In French, these patterns
are somewhat complex and may involve positing a large number of rules with many exceptions
(Franck et al. 2008), possibly explaining why Guillelmon & Grosjean’s (2001) late learners did
not appear to be sensitive to gender cues in spoken-word recognition. In contrast, as described
previously, the majority of Spanish nouns fall into a canonically marked group of words in which
masculine nouns end in –o and feminine nouns end in –a, thus providing learners of Spanish with
an easy rule to follow in assigning gender, at least for this set of nouns. If Carroll’s hypothesis
is correct, and late learners are forced to rely on such rules, it could be the case that late learn-
ers of Spanish may be able to perform in a more nativelike manner than late learners of French
in general, and more specifically, they may be more accurate in Spanish when processing and
Downloaded by [Radcliffe Infirmary] at 03:44 15 July 2016

producing transparently marked nouns in comparison to nouns that are opaque.


Building on this idea, and as a follow-up to Montrul, Foote & Perpiñán (2008), Montrul et al.
(2014) conducted a partial replication of Bates et al. (1996) with native speakers of Spanish,
heritage speakers of Spanish, and late L2 learners of Spanish. In addition to varying the degree
of explicit attention to gender required to complete the three experimental tasks (as in Bates
et al.), Montrul et al. (2014) also varied the transparency of the gender-marking on the nouns
used in the task. Results indicated that all groups showed effects of gender-incongruency in the
two more explicit tasks (the gender-monitoring and the grammaticality-judgment tasks), but in the
word-repetition task, only the native speakers and heritage speakers were sensitive to incongruent
gender marking on preceding elements of the NP and only with nouns that were transparently
marked for gender (e.g., libro, bolsa). While the L2 learners patterned in the right direction
with transparently marked nouns, they showed the opposite pattern with nontransparently marked
nouns, repeating these nouns more slowly in the gender-congruent condition than in the gender-
incongruent condition. Montrul et al. concluded based on these results that early age of acqui-
sition provides an advantage in the processing of gender, at least when the processing task taps
into implicit rather than explicit gender knowledge. This finding generally replicates Guillelmon
& Grosjean’s (2001) results, except that no effects of gender congruency were reported in any
of the Montrul et al. participant groups. In other words, there was no evidence in their study
for a prelexical component of gender priming, as there was in both Guillelmon & Grosjean and
Bates et al. (1996). In addition, the late learners in the Montrul et al. (2014) study were not all at
advanced levels of proficiency but scored in the “intermediate and advanced” range on a written
proficiency test (p. 7); those tested in Guillelmon & Grosjean’s study (and in the current study)
were high-proficiency learners who functioned daily in both their L1 and L2. Two important
questions are therefore left open by the results of Montrul et al. (2014): (1) whether early learners
who are not dominant in the gender-marking language can make use of congruent gender cues in
spoken word recognition; and (2) whether late learners at more advanced proficiency levels are
able to make use of either type of gender cue, congruent or incongruent, during word recognition.

5. THE CURRENT STUDY

In order to further investigate the role of age of language acquisition in the processing of gender,
and to continue to explore the role of noun transparency in gender sensitivity while also adding to
AGE OF ACQUISITION AND SENSITIVITY TO GENDER 373

previous findings on the nature of lexical access, a partial replication of Guillelmon and Grosjean
(2001) was carried out. The following research questions and corresponding hypotheses were
posed:

RQ1: Can early language learners who are not dominant in a gender-marking language make use
of congruent and incongruent gender cues during the recognition of spoken words in that
language?

If early age of acquisition ensures nativelike processing of gender, then even early learners
who are not dominant in the gender-marking language should be able to use both congruent and
incongruent gender cues during spoken word recognition, as native speakers have been shown to
do.
Downloaded by [Radcliffe Infirmary] at 03:44 15 July 2016

RQ2: Does the ability of highly proficient early and/or late learners to utilize gender cues depend
on the transparency of noun gender markers?

Based on Carroll’s (1989) claim that late learners are forced to rely on general rules in order
to process and produce gender, late learners, but not early learners, should show differences
in sensitivity to gender cues as a function of gender marker transparency; however, based on
Montrul et al.’s (2014) results, both groups may show effects of transparency. Moreover, since
morphophonological information can likely only be taken into account during postlexical pro-
cesses that may be more available to conscious attention, rather than prelexical processes that
occur before word recognition, it is hypothesized that if either group of learners does show dif-
ferences in sensitivity to gender cues as a function of transparency, these differences will only be
observable in gender-incongruent conditions.

6. METHOD

6.1. Participants

Thirty-two early and 32 late learners participated in the experiment, along with 32 native speakers
of Spanish. All participants were either faculty or students at one of two large Midwestern uni-
versities in the United States. The early learners were heritage speakers of Spanish who learned
Spanish at home from birth and began to acquire English either during the early preschool years
or soon after. The late learners were native speakers of English who began learning Spanish in a
classroom environment at age 10 or later (except for two who had some classroom Spanish dur-
ing elementary school) and were primarily students working toward graduate degrees in Spanish.
Critically, none of the late learners had any prepuberty immersion experience in a Spanish-
speaking environment. The native speakers who participated in the study had only been in the
U.S. as adults after growing up in either Spain or Latin America.
The early and late learners completed a proficiency test in Spanish that included the
vocabulary section of an MLA test (30 multiple-choice items) and the cloze section (20 multiple-
choice items) of one of the tests to obtain the advanced Diplomas de Español como Lengua
374 FOOTE

Extranjera. This is the same test that has been used in several other studies to measure L2 pro-
ficiency in Spanish (e.g., White et al. 2004; McCarthy 2008; Montrul, Foote & Perpiñán
2008; Montrul et al. 2014). The mean score of both groups was around 83%, indicat-
ing advanced proficiency in Spanish. As an additional measure, self-ratings in Spanish and
English were collected from the participants. The ratings were based on a scale from 1 (can
barely speak/understand/read/write Spanish/English) to 5 (can speak/understand/read/write
Spanish/English like a native speaker). Self-ratings echoed the results of the proficiency test
and indicated that the participants in both bilingual groups were dominant in English. Table 1
summarizes participant characteristics.

6.2. Materials
Downloaded by [Radcliffe Infirmary] at 03:44 15 July 2016

6.2.1. Word-Repetition Task

A word-repetition task similar to the one used by Guillelmon & Grosjean (2001) was employed
as the main experimental task. Sixty-four inanimate Spanish nouns were included in the task, half
masculine and half feminine. Half of the nouns in each gender group were transparently marked
for gender, and half were opaque. Nouns were matched across the four conditions (masculine
transparent, masculine opaque, feminine transparent, feminine opaque) on frequency (Léxico
informatizado del español, Sebastián Gallés et al. 2000), length in syllables, and uniqueness point
(the point in the word when its spoken form becomes unique to that word; this was determined by
consulting the María Moliner Diccionario de Uso del Español and then measured in milliseconds
[ms] in the sound files used in the experiment). Table 2 lists stimuli characteristics, along with
the results of three one-way ANOVAs conducted to verify no significant differences between the
noun groups on any of these characteristics. See the Appendix for the full list of stimuli.

TABLE 1
Participant Characteristics (Standard Deviations in Parentheses)

Native speakers Early learners Late learners

Mean age 29 (6.2) 22 (8.1) 27 (7.1)


Mean Span proficiency (%a ) − 83 (10.6) 84 (10.9)
Mean self-ratings-Span (max 5)
Speaking − 4.44 (0.62) 3.90 (0.54)
Understanding − 4.75 (0.57) 3.81 (0.65)
Reading − 4.25 (0.88) 4.02 (0.69)
Writing − 3.66 (1.07) 3.77 (0.68)
Mean self-ratings-Eng (max 5)
Speaking − 4.75 (0.51) 5.00 (0.00)
Understanding − 4.84 (0.37) 5.00 (0.00)
Reading − 4.88 (0.34) 5.00 (0.00)
Writing − 4.84 (0.45) 5.00 (0.00)
Mean age of L2 acquisition − 3.6 (2.8) 13.1 (3.2)
a Fourteen of the early learners only completed the vocabulary portion of the proficiency test due to time
constraints. For those who did not complete the entire test, scores were calculated based on the completed
portion.
AGE OF ACQUISITION AND SENSITIVITY TO GENDER 375

TABLE 2
Characteristics (Means) of the Noun Stimuli Used in the Word-Repetition Task

Frequency (per million) Length (syllables) Uniqueness Point (ms)

Masc–Trans 112.4 2.1 443.1


Masc–Opaq 102.3 2.2 466.8
Fem–Trans 107.0 2.2 431.4
Fem–Opaq 111.0 1.9 418.8
ANOVAs F(3, 60) = .02, p = .996 F(3, 60) = 1.5, p = .212 F(3, 60) = .44, p = .726

All stimuli were recorded by a female native speaker of Mexican Spanish and were spliced into
three carrier phrases: (1) el mejor [NOUN] (themasc bestmasc/fem NOUN), (2) la mejor [NOUN]
Downloaded by [Radcliffe Infirmary] at 03:44 15 July 2016

(thefem bestmasc/fem NOUN), and (3) su mejor [NOUN] (his/hermasc/fem bestmasc/fem NOUN). The
adjective mejor ‘best’ was chosen as an adjective with no gender marking that would plausibly
fit with all of the nouns.
Following Guillelmon & Grosjean (2001) in order to allow for a direct comparison between
studies, four experimental lists were created. In two of the lists, each noun was presented
in either a gender-congruent or a gender-neutral phrase. In the gender-congruent phrases, the
gender of the determiner agreed with the gender of the noun (e.g., el mejor libro [themasc
bestmasc/fem bookmasc ]); in the gender-neutral phrases, the determiner was not marked for gen-
der (e.g., su mejor libro [his/hermasc/fem bestmasc/fem bookmasc ]). In the other two experimental
lists, each noun was presented in either a gender-incongruent or a gender-neutral phrase. In the
gender-incongruent phrases, the gender of the determiner disagreed with the gender of the
noun (e.g., la mejor libro [thefem bestmasc/fem bookmasc ]). The presentation of nouns was coun-
terbalanced across experimental lists, so that each noun appeared once in each of the four
lists. Noun phrases were presented in a pseudorandomized order that was the same for all
participants.

6.2.2. Gender-Assignment Task

In order to verify knowledge of the noun stimuli’s gender, participants completed a writ-
ten gender-assignment task in which they had to indicate whether each noun was masculine or
feminine.

6.3. Procedure

Participants were tested individually in a quiet room. They first completed a language-history
questionnaire with information about age of language acquisition and experience with English
and Spanish, and then the self-ratings of proficiency and the written proficiency test. Next, they
completed the word-repetition task on a PC running E-Prime. For this task, they were instructed to
listen to the phrases they heard via headphones and repeat the third word of each phrase as quickly
and accurately as possible. Half of the participants in each group were randomly assigned to one
of the two gender-congruent/gender-neutral lists, and the other half were randomly assigned to
one of the two gender-incongruent/gender-neutral lists, following the design of Guillelmon &
376 FOOTE

Grosjean (2001).1 In each list, the 64 experimental stimuli were preceded by four practice items.
After the word-repetition task, participants completed the gender assignment task.

6.4. Scoring

Responses on the word-repetition task were digitally recorded and checked for repetition accu-
racy. The time it took participants to begin repeating each noun (starting from the onset of the
noun in each NP) was recorded via the E-Prime serial response box. Known response box errors
and incorrectly repeated words were removed from the data before analysis, along with any
RTs above 3000 ms. In addition, any words that a participant incorrectly marked for gender
in the gender-assignment task were removed from that participant’s data if the words appeared in
congruent or incongruent conditions, since the actual gender for the word in that participant’s lin-
Downloaded by [Radcliffe Infirmary] at 03:44 15 July 2016

guistic system can be assumed to be the opposite of the word’s standard gender. A total of 9.9% of
the data in the gender-congruent/gender-neutral lists and 9.3% in the gender-incongruent/gender-
neutral lists were removed. In order to determine whether participants made use of the gender
cues on the determiner in the recognition of the subsequent noun, response times (RTs) in the
gender-congruent condition were compared to RTs in the gender-neutral condition in the gender-
congruent/gender-neutral lists, and RTs in the gender-incongruent condition were compared to
RTs in the gender-neutral condition in the gender-incongruent/gender-neutral lists.

7. RESULTS

7.1. Gender-Assignment Task

Both groups scored highly on the gender-assignment task, though scores differed according
to condition. Table 3 lists the mean number of errors made by each group in each condition
(out of 16).
As shown in Table 3, there was an average of less than one error in each condition, except
for the feminine opaque condition, in which approximately 25% of the nouns were assigned the
wrong gender in both learner groups. A mixed ANOVA with group as a between-participants

TABLE 3
Mean Number of Errors on the Gender Assignment Task by Group and by Condition (Standard
Deviations in Parentheses)

Masculine Feminine

Transparent Opaque Transparent Opaque

Early learners 0.19 (0.59) 0.50 (1.22) 0.09 (0.39) 3.91 (3.28)
Late learners 0.00 (0.00) 0.50 (0.72) 0.09 (0.30) 4.22 (3.58)

1 According to Guillelmon & Grosjean, this design was used in their study because results of a pilot had indicated that
the strength of the congruency effect depends in part on the grouping of the experimental conditions.
AGE OF ACQUISITION AND SENSITIVITY TO GENDER 377

variable of two levels (early learners/late learners) and gender and transparency as within-
participants variables of two levels each (masculine/feminine, transparent/opaque) confirmed
this pattern statistically, with a main effect of gender, F(1, 62) = 85.39, p < .001; a main effect
of transparency, F(1, 62) = 89.04, p < .001; and an interaction between gender and transparency,
F(1, 62) = 83.17, p < .001. There was no main effect of group and no interactions with group;
both the early and the late learners performed similarly on this task.

7.2. Word-Repetition Task

Participants repeated nouns at high levels of accuracy; the mean accuracy rate was 98.1% (SD =
1.8) in the early learner group and 99.1% (SD = 1.5) in the late learner group. In what follows,
I will first present the RT results for the gender-congruent/gender-neutral lists and then for the
Downloaded by [Radcliffe Infirmary] at 03:44 15 July 2016

gender-incongruent/gender-neutral lists.

7.2.1. Gender-Congruent/Gender-Neutral Lists

A mixed ANOVA with group as a between-participants variable of three levels


(native speakers/early learners/late learners) and congruency, gender, and transparency as
within-participants variables of two levels each (congruent/neutral, masculine/feminine,
transparent/opaque) was conducted on the RTs. There was no main effect of group, F(2, 45) =
0.82, p = .445, or of congruency, F(1, 45) = 1.81, p = .185, but there was an effect of gender, F(1,
45) = 12.08, p < .01, and of transparency, F(1, 45) = 48.73, p < .001, and an interaction between
gender and transparency, F(1, 45) = 25.01, p < .001. The gender and transparency effects along
with the interaction reflect the fact that participants responded more quickly to transparent nouns
than to opaque nouns but only when those nouns were masculine (transparent: 937 ms, opaque:
1011 ms). Feminine transparent nouns were responded to at approximately the same speed as
feminine opaque nouns (957 vs. 960 ms). This pattern of results is depicted in Figure 1.
Although there was no main effect of congruency, the native-speaker and early-learner groups
patterned in the right direction, with repetition onsets faster to nouns in gender-congruent contexts
than in gender-neutral contexts. The native speakers repeated nouns in the gender-congruent con-
dition 13 ms more quickly than in the gender-neutral condition, and the early learners performed
similarly with a 14 ms difference between the two conditions. The difference between the two
conditions in late learners was minimal (<1 ms). Figure 2 illustrates congruency effects in each
group.

7.2.2. Gender-Incongruent/Gender-Neutral Lists

The mixed ANOVA conducted on RTs in the gender-incongruent/gender-neutral lists again


included group as a between-participants variable of three levels and incongruency, gender, and
transparency as within-participants variables of two levels each. There was no main effect of
group, F(2, 45) = 0.34, p = .715, but there was a main effect of gender, F(1, 45) = 13.75, p <
.01, and of transparency, F(1, 45) = 31.57, p < .001. The main effects of gender and transparency
were qualified by an interaction between the two, F(1, 45) = 71.89, p < .001, reflecting the fact
that, as in the previous analysis, speed was only affected by transparency on masculine nouns
378 FOOTE

1050
1011
1000
957 960
Mean response time in ms

950 937

900

Transparent
850
Opaque

800

750
Downloaded by [Radcliffe Infirmary] at 03:44 15 July 2016

700
Masculine Feminine
Noun gender

FIGURE 1 Gender by transparency interaction (gender-congruent/


gender-neutral analysis).

1050
1006
1000 992
964
Mean response time in ms

951 942 941


950

900

Congruent
850
Neutral
800

750

700
Native speakers Early learners Late learners
Group

FIGURE 2 Effects of gender congruency by group.

(transparent: 983 ms, opaque: 1063 ms). Feminine transparent nouns were responded to at the
same speed as feminine opaque nouns (1005 vs. 1004 ms). Figure 3 illustrates this interaction.
In this analysis there was also a main effect of incongruency, F(1, 45) = 9.14, p < .01, along
with an interaction between incongruency and transparency that approached significance, F(1,
45) = 3.74, p = .059. Overall RTs to nouns in the gender-incongruent condition (1026 ms)
were slower than in the gender-neutral condition (1002 ms). However, the incongruency by trans-
parency interaction that approached significance was due to a larger difference between RTs to
AGE OF ACQUISITION AND SENSITIVITY TO GENDER 379

1100
1063
1050
1005 1004
Mean response time in ms

1000 983

950

Transparent
900
Opaque

850
Downloaded by [Radcliffe Infirmary] at 03:44 15 July 2016

800

750
Masculine Feminine
Noun gender

FIGURE 3 Gender by transparency interaction (gender-incongruent/


gender-neutral analysis).

1100
1051

1048
1045
1030

1026
1024

1021

1050
1012
1011
Mean response time in ms

982

979

1000
937

950

900 Incongruent
Neutral
850

800

750
Trans Opaq Trans Opaq Trans Opaq

Native speakers Early learners Late learners


Group

FIGURE 4 Effects of gender incongruency by group as a function of


transparency.
380 FOOTE

nouns in the gender-incongruent versus the gender-neutral condition with transparent nouns than
with opaque nouns (35 vs. 13 ms). Incongruency effects in each group are illustrated in Figure 4
as a function of transparency.

8. DISCUSSION

The goals of this study were first, to determine whether early learners who are not dominant in
a gender-marking language can make use of congruent and incongruent gender cues during spo-
ken word recognition in that language, and second, whether the ability of highly proficient early
and/or late learners to process gender cues depends on the morphophonological transparency
of noun gender markers. Although only incongruency effects reached significance in the current
Downloaded by [Radcliffe Infirmary] at 03:44 15 July 2016

study, based on the patterning of both gender-congruency and gender-incongruency results, it


appears that early learners who are not dominant in Spanish are able to make use of both con-
gruent and incongruent gender cues available in the NP when recognizing words in Spanish. The
size of the congruency effect in the native-speaker and early-learner groups in the present study
(13 and 14 ms respectively) was similar to the size of the effect found in both Bates et al. (1996)
and Guillelmon & Grosjean (2001): 19 ms in each study. This finding suggests that language
dominance does not play a role in the ability to make use of both congruent and incongruent
gender cues (though proficiency, which was not manipulated in this study, certainly may, par-
ticularly given the results of Montrul et al. 2014 with less-proficient late learners). In contrast
to Guillelmon & Grosjean’s (2001) findings, it also seems that the late learners in the present
study were able to make use of gender cues in word recognition, though only when those cues
conflicted with the gender of the noun; the difference in this group between gender-congruent
and gender-neutral RTs was less than 1 ms, while the difference between gender-incongruent and
gender-neutral RTs was 25 ms. In terms of the effects of the transparency of noun gender mark-
ers, there was no interaction between congruency and transparency, but there was an interaction
between incongruency and transparency that approached significance, such that incongruency
effects were larger with transparent than with opaque nouns. However, though it was predicted
based on Carroll (1989) that transparency would only affect gender-cue usage in late learn-
ers, there were larger differences between RTs to nouns in the gender-incongruent versus the
gender-neutral condition with transparent nouns than with opaque nouns across all three partic-
ipant groups. In the learner groups, this difference in incongruency effect size as a function of
transparency could be related to a decreased certainty about the gender of opaque nouns, even
though all nouns included in data analysis had been marked correctly for gender on the written
task. However, this does not explain why transparency would play a role in native speakers’ use
of gender cues.
There is some evidence in the literature that the gender of the noun as well as the trans-
parency of its ending can affect word recognition in native speakers. In the present study,
feminine nouns were recognized significantly faster than masculine nouns overall, and nouns
transparently marked for gender were recognized more quickly than nouns not transparently
marked. However, the effects of gender and of transparency were modulated by an interac-
tion between the two, such that feminine nouns were only recognized faster than masculine
nouns in the opaque condition. By the same token, transparency only affected response times
when participants responded to masculine nouns. There are other studies that have found that
AGE OF ACQUISITION AND SENSITIVITY TO GENDER 381

native speakers of Romance languages recognize feminine nouns faster than masculine nouns
and that they recognize transparently marked nouns faster than nouns that are not transparently
marked. In the word-repetition and the gender-monitoring tasks of Bates et al. (1996), feminine
nouns were responded to significantly faster than masculine nouns, and in the gender-monitoring
and grammaticality-judgment tasks, responses were faster to transparently marked nouns than
nontransparently marked nouns. However, there were no interactions between gender and noun
transparency, except in the grammaticality-judgment task, due to a larger effect of transparency
in feminine than in masculine nouns. In Bates et al. (1995), native speakers of Italian were faster
to respond to transparently marked nouns in comparison to nontransparently marked nouns in a
gender-monitoring task, though there was no such effect in a word-recognition task in that study.
Gender had no significant effect on response times, and there were no interactions between gender
and noun transparency. Colé, Pynte & Andriamamonjy (2003) examined the effects of gender-
Downloaded by [Radcliffe Infirmary] at 03:44 15 July 2016

marking transparency on lexical decisions in native speakers of French and found that feminine
nouns were recognized faster than masculine nouns, that transparently marked nouns were rec-
ognized faster than nontransparently marked nouns (when the nouns were of low frequency), and
that transparency played a role only in masculine nouns, similar to current findings.
The interpretation of gender effects varies across studies, with Bates et al. (1996) attribut-
ing faster RTs to feminine nouns to nonsignificant differences in stimulus characteristics, and
Colé, Pynte & Andriamamonjy (2003) claiming that orthographic differences in feminine and
masculine nouns in French played a role. As for transparency effects, they have generally been
interpreted as postlexical in nature (Bates et al. 1995, 1996; Colé, Pynte & Andriamamonjy
2003). Specifically, Bates et al. (1995) proposed that morphophonological cues may be used
in a postlexical “check” before responding to the noun, although in their study, these cues only
affected responses in a task in which participants had to specify the gender of the noun. The find-
ing that these cues only played a role with masculine nouns in Colé, Pynte & Andriamamonjy was
attributed to the predictability of the –e ending to indicate feminine gender in French being lower
than the predictability of the endings to indicate masculine gender in their stimuli. This explana-
tion does not make sense in Spanish, however, as –a is a clear predictor of feminine gender.
One possible way to account for transparency only affecting responses to masculine nouns in
the current study could lie in the status of the masculine as the default, or unmarked gender in
Spanish. While proposals regarding the representation of default gender vary, some scholars have
claimed that masculine gender is unspecified or featureless (e.g., Harris 1991; McCarthy 2008).
In other words, there is no gender information specified in the lexical entry of masculine nouns in
Spanish (Harris 1991:44). If this is true, then it could be the case that morphophonological cues
are only utilized during the postlexical check of grammatical agreement with masculine nouns
because there is no gender information in the lexical entry of these nouns that the processor can
utilize to carry out the check. For feminine nouns, on the other hand, since they are lexically
specified as feminine, the processor does not need to rely on morphophonological cues to carry
out the agreement check, and therefore the presence or absence of these cues does not affect
response times. This account can also explain why response times to masculine opaque nouns
were longer than to all other types of nouns in the present study (see Figures 1 and 3). The extra
time needed to respond to these nouns may be due to the processor “timing out” in its search
for gender information that it can use to conduct the check, whether from the lexical entry of
the noun, or morphophonological cues present on the form of the noun. If the processor does not
382 FOOTE

find this information, then it may assume that the default gender is in play and conduct the check
based on that assumption.
One question that arises regarding this account, however, is why it does not seem to apply
to Italian, which was the language studied by Bates et al. (1995, 1996). In these studies, there
was no evidence that participants took significantly longer to respond to opaque masculine nouns
than to other types; transparency did play a role in response times in some tasks, but with both
feminine and masculine nouns. It is also unclear why there was no interaction between gender,
transparency, and incongruency in the current study if incongruency effects come about, at least
in part, as a result of the slowing down of the postlexical check when the processor encounters
elements in the NP that do not agree. That is, if the processor only uses morphophonological
cues to gender to carry out the check with masculine nouns, then it would be expected that these
cues would not affect checks involving feminine nouns. However, there was no difference in the
Downloaded by [Radcliffe Infirmary] at 03:44 15 July 2016

effect of these cues based on gender; incongruent transparent cues caused a relatively greater
slowdown than incongruent opaque cues, whether the nouns containing the cues were masculine
or feminine. A possible explanation for this lack of gender effect is that the transparency of the
gender cue on both masculine and feminine nouns may more effectively draw the participant’s
conscious attention to the incongruency, thus causing an exaggerated slowdown. In other words,
the transparency effect is not related to the automatic postlexical check that is carried out by the
processor but rather by the noticing of the incongruency on the part of the participant. However,
this is very speculative. Further research is called for to investigate these interactions, both to
determine the conditions under which they occur and to establish whether the account proposed
here can explain these effects.
Most importantly for the goals of the current study, both early learners who are not dominant
in Spanish and highly proficient late learners who acquired Spanish during or after adolescence
were able to make use of gender cues during spoken word recognition. This suggests that neither
late age of language acquisition nor language dominance plays a role in the ability to use gen-
der cues, at least when those cues are incongruent. Although congruency effects did not reach
statistical significance in this study, according to the patterning of the three participant groups, it
seems to be the case that native speakers and early learners may have been able to utilize these
cues while late learners were not able to. If this is true (and further research with larger partic-
ipant groups may confirm this statistically), this may indicate that while gender cues in the NP
are used at both prelexical and postlexical stages of word recognition in native speakers and early
learners, late learners are only able to make use of these cues after word recognition has taken
place (recall that facilitatory or congruency effects are thought to be prelexical while inhibitory
or incongruency effects may be either pre- or postlexical). This is consistent with the idea that
grammatical knowledge tends to be less automatized or implicit in late learners (DeKeyser 2000;
Ullman 2001; Paradis, 2004). If postlexical stages of word recognition are more open to con-
scious control, then late learners may be able to employ explicit knowledge of gender during
those stages, while they cannot do so at earlier stages. This coincides in part with the finding
by Montrul et al. (2014) of differences in late learner versus early learner and native speaker
performance depending on task type and potential for making use of explicit knowledge of gen-
der. As described previously, Montrul et al.’s native speakers, early learners, and late learners
of Spanish all showed sensitivity to gender in the more explicit grammaticality-judgment and
gender-monitoring tasks. In contrast with the present study, however, the late learners of Montrul
et al. (who were less proficient than those who participated in the current study) did not show
AGE OF ACQUISITION AND SENSITIVITY TO GENDER 383

significant incongruency effects in the word-repetition task, while both the native speaker and the
early learner groups did.
The reason that present results differ from those of Guillelmon & Grosjean (2001), whose
participants were at very high levels of proficiency in French, is unclear; though late learners
were able to make use of incongruent gender cues in the current study, they were not able to in
Guillelmon & Grosjean. It is possible that the fact that gender in Spanish is more transparently
marked than it is in French could have contributed to this finding. It may also be the case that the
late learner participants of the current study were more sensitive to gender errors than those of
Guillelmon and Grosjean, since almost all of the late learners in the current study were teaching
Spanish at the university level at the time of testing. Because they are consistently forced to pay
attention to and correct gender errors in their students’ language, this type of learner may develop
a sensitivity to such errors that the late learners of Guillelmon & Grosjean did not have.
Downloaded by [Radcliffe Infirmary] at 03:44 15 July 2016

In terms of models of lexical access, the results of the current study may provide very tentative
support for interactive views, given that native speakers and early learners both showed evidence
of (nonsignificant) facilitatory effects of gender context on word recognition. This coincides with
findings in Bates et al. (1996) and Guillelmon & Grosjean (2001). Just as in those studies, how-
ever, inhibitory effects were larger in the present study than facilitatory effects, suggesting that
the role that congruent gender marking plays in word recognition is small, though not trivial, as
it provides evidence that syntactic context may, in fact, influence the recognition of words, either
by restricting the possible candidates considered by the word-recognition system or by boosting
activation of the candidates that would fit best in the context.
In conclusion, although gender may cause lasting difficulties for late learners, it appears that
at least some late learners are able to use incongruent gender cues during word recognition in
a nativelike manner. It remains for further research to determine, then, exactly how the late
learner gender system diverges from the native speaker gender system and what the causes and
consequences of this divergence are.

ACKNOWLEDGMENTS

I would like to thank Carmen Albaladejo along with Luis García and the College Assistance
Migrant Program Scholars Initiative at Michigan State University for their help with participant
recruitment. Thanks also go to Junkyu Lee and Soo Kim, who served as research assistants for
this project.

REFERENCES

Akhutina, Tatiana, Andrei Kurgansky, Maria Polinsky & Elizabeth Bates. 1999. Processing of grammatical gender in a
three-gender system: Experimental evidence from Russian. Journal of Psycholinguistic Research 28(6). 695–713.
Alarcón, Irma. 2006. The second language acquisition of Spanish gender agreement: The effects of linguistic variables
on accuracy. Munich, Germany: Lincom Europa.
Bates, Elizabeth, Antonella Devescovi, Arturo Hernández & Luigi Pizzamiglio. 1996. Gender priming in Italian.
Perception and Psychophysics 58(7). 992–1004.
Bates, Elizabeth, Antonella Devescovi, Luigi Pizzamiglio, Simona D’Amico & Arturo Hernández. 1995. Gender and
lexical access in Italian. Perception and Psychophysics 57(6). 847–862.
384 FOOTE

Bölte, Jens & Cynthia M. Connine. 2004. Grammatical gender in spoken word recognition in German. Perception and
Psychophysics 66(6). 1018–1032.
Carello, Claudia, G. Lukatela & M. T. Turvey. 1988. Rapid naming is affected by association but not by syntax. Memory
and Cognition 16(3). 187–195.
Carroll, Susanne. 1989. Second language acquisition and the computational paradigm. Language Learning 39(4).
535–594.
Clahsen, Harald & Claudia Felser. 2006. Grammatical processing in language learners. Applied Psycholinguistics 27(1).
3–42.
Colé, Pascale, Joel Pynte & Pascale Andriamamonjy. 2003. Effect of grammatical gender on visual word recognition:
Evidence from lexical decision and eye movement experiments. Perception and Psychophysics 65(3). 407–419.
DeKeyser, Robert M. 2000. The robustness of critical period effects in second language acquisition. Studies in Second
Language Acquisition 22(4). 499–534.
Dewaele, Jean-Marc & Daniel Véronique. 2001. Gender assignment and gender agreement in advanced French
interlanguage: A cross-sectional study. Bilingualism: Language and Cognition 4(3). 275–296.
Foote, Rebecca. 2011. Integrated knowledge of agreement in early and late English-Spanish bilinguals. Applied
Downloaded by [Radcliffe Infirmary] at 03:44 15 July 2016

Psycholinguistics 32(1). 187–220.


Franceschina, Florencia. 2001. Morphological or syntactic deficits in near-native speakers? An assessment of some current
proposals. Second Language Research 17(3). 213–247.
Franceschina, Florencia. 2005. Fossilized second language grammars. The acquisition of grammatical gender.
Amsterdam: John Benjamins.
Franck, Julie, Gabriella Vigliocco, Inés Antón-Méndez, Simona Collina & Ulrich H. Frauenfelder. 2008. The interplay
of syntax and form in sentence production: A cross-linguistic study of form effects on agreement. Language and
Cognitive Processes 23(3). 329–374.
Friederici, Angela D. & Thomas Jacobsen. 1999. Processing grammatical gender during language comprehension.
Journal of Psycholinguistic Research 28(5). 467–484.
Grosjean, François, Jean-Yves Dommergues, Etienne Cornu, Delphine Guillelmon & Carole Besson. 1994. The gender-
marking effect in spoken word recognition. Perception and Psychophysics 56(5). 590–598.
Grüter, Theres, Casey Lew-Williams & Anne Fernald. 2012. Grammatical gender in L2: A production or a real-time
processing problem? Second Language Research 28(2). 191–215.
Guillelmon, Delphine & François Grosjean. 2001. The gender marking effect in spoken word recognition: The case of
bilinguals. Memory and Cognition 29(3). 503–511.
Harris, James W. 1991. The exponence of gender in Spanish. Linguistic Inquiry 22(1). 27–62.
Hawkins, Roger & Florencia Franceschina. 2004. Explaining the acquisition and non-acquisition of determiner-noun
gender concord in French and Spanish. In Philippe Prévost and Johanne Paradis (eds.), The acquisition of French in
different contexts, 175–206. Amsterdam: John Benjamins.
Jakubowicz, Celia & Christelle Faussart. 1998. Gender agreement in the processing of spoken French. Journal of
Psycholinguistic Research 27(6). 597–617.
Keating, Gregory D. 2009. Sensitivity to violations of gender agreement in native and nonnative Spanish: An
eye-movement investigation. Language Learning 59(3). 503–535.
Kohnert, Kathryn J., Elizabeth Bates & Arturo E. Hernandez. 1999. Balancing bilinguals: Lexical-semantic produc-
tion and cognitive processing in children learning Spanish and English. Journal of Speech, Language, and Hearing
Research 42(6). 1400–1413.
McCarthy, Corinne. 2008. Morphological variability in the comprehension of agreement: An argument for representation
over computation. Second Language Research 24(4). 459–486.
Montrul, Silvina. 2004. The acquisition of Spanish: Morphosyntactic development in monolingual and bilingual
L1 acquisition and adult L2 acquisition. Amsterdam: John Benjamins.
Montrul, Silvina. 2008. Incomplete acquisition in bilingualism: Re-examining the age factor. Amsterdam: John
Benjamins.
Montrul, Silvina, Justin Davidson, Israel de la Fuente & Rebecca Foote. 2014. Early language experience facilitates the
processing of gender agreement in Spanish heritage speakers. Bilingualism: Language and Cognition 17(1). 118–138.
Montrul, Silvina, Rebecca Foote & Silvia Perpiñán. 2008. Gender agreement in adult second language learners and
Spanish heritage speakers: The effects of age and context of acquisition. Language Learning 58(3). 503–553.
AGE OF ACQUISITION AND SENSITIVITY TO GENDER 385

Montrul, Silvina & Kim Potowski. 2007. Command of gender agreement in school-age Spanish-English bilingual
children. International Journal of Bilingualism 11(3). 301–328.
Paradis, Michel. 2004. A neurolinguistic theory of bilingualism. Amsterdam: John Benjamins.
Polinsky, Maria. 2008. Gender under incomplete acquisition: Heritage speakers’ knowledge of noun categorization. The
Heritage Language Journal 6(1). 40–71.
Sabourin, Laura, Laurie A. Stowe & Ger J. de Haan. 2006. Transfer effects in learning a second language grammatical
system. Second Language Research 22(1). 1–29.
Sagarra, Nuria & Julia Herschensohn. 2010. The role of proficiency and working memory in gender and number
agreement processing in L1 and L2 Spanish. Lingua 120(8). 2022–2039.
Sagarra, Nuria & Julia Herschensohn. 2011. Proficiency and animacy effects on L2 gender agreement processes during
comprehension. Language Learning 61(1). 80–116.
Schriefers, Herbert & Jörg D. Jescheniak. 1999. Representation and processing of grammatical gender in language
production: A review. Journal of Psycholinguistic Research 28(6). 575–600.
Sebastián Gallés, Núria, Fernando Cuetos, Manuel Carreiras & M. Antònia Martí. 2000. Léxico informatizado del
español. Barcelona, Spain: University of Barcelona.
Downloaded by [Radcliffe Infirmary] at 03:44 15 July 2016

Tanenhaus, Michael K. & Margery M. Lucas. 1987. Context effects in lexical processing. Cognition 25(1/2). 213–234.
Ullman, Michael T. 2001. The neural basis of lexicon and grammar in first and second language: The
declarative/procedural model. Bilingualism: Language and Cognition 4(2). 105–122.
White, Lydia, Elena Valenzuela, Martyna Kozlowska-Macgregor & Yan-Kit Ingrid Leung. 2004. Gender agreement in
nonnative Spanish: Evidence against failed features. Applied Psycholinguistics 25(1). 105–133.

Submitted 1 July 2013


Final version accepted 19 December 2013

APPENDIX

TABLE A1
Full List of Experimental Stimuli—Word Repetition Task

Masculine Feminine

Transparent Opaque Transparent Opaque

banco ‘bank’ valle ‘valley’ cuna ‘cradle’ col ‘cabbage’


barco ‘ship’ bigote ‘mustache’ blusa ‘blouse’ base ‘base’
campo ‘field bosque ‘forest’ boca ‘mouth’ calle ‘street’
cuadro ‘picture’ color ‘color’ bolsa ‘bag’ carne ‘meat’
cuarto ‘room’ diente ‘tooth’ caja ‘box’ clase ‘class’
cuello ‘neck’ país ‘country’ cama ‘bed’ piel ‘skin’
cuerpo ‘body’ pan ‘bread’ cara ‘face’ torre ‘tower’
queso ‘cheese’ papel ‘paper’ carta ‘letter’ voz ‘voice’
beso ‘kiss’ parque ‘park’ casa ‘house’ tarde ‘afternoon’
pelo ‘hair’ pie ‘foot’ granja ‘farm’ tesis ‘thesis’
piso ‘floor’ diamante ‘diamond’ playa ‘beach’ cárcel ‘jail’
plato ‘plate’ traje ‘suit’ taza ‘cup’ catedral ‘cathedral’
camino ‘road’ deporte ‘sport’ tela ‘cloth’ pared ‘wall’
vaso ‘glass’ paisaje ‘landscape’ bebida ‘drink’ cumbre ‘summit’
canto ‘song’ baile ‘dance’ botella ‘bottle’ tribu ‘tribe’
vino ‘wine’ vegetal ‘vegetable’ camisa ‘shirt’ vocal ‘vowel’