RELATING GENDER ERRORS TO MORPHOSYNTAX AND
LEXICON IN ADVANCED FRENCH INTERLANGUAGE
Jean-Marc Dewaele
Daniel Véronique
Abstract:
An correlational analysis between accuracy levels of
gender agreement -based on an analysis of 519 gender
errors out of 9378 modifiers in the advanced French
interlanguage of 27 Flemish L1 speakers- and
morphosyntactic and lexical variables, revealed a clear
negative relationship between the number of gender errors
and fluency variables. No relation was found between
gender errors and other types of agreement errors. This
suggests that more advanced learners, whose interlanguage
speech production process is more automatised or
proceduralised, do commit fewer gender errors but that
good mastery of gender agreement does not imply an
equally good mastery of other types of agreement.
1
Introduction*
Perfect command of morphology seems to elude even the
most advanced second language learners (Bartning 1997).
One particular morphological problem that besets learners
of French in particular is mastering gender agreement.
Numerous studies have shown that even very advanced
learners continue to make gender errors (Surridge &
Lessard 1984, Carroll 1989, Surridge 1993, Dewaele 1994,
Bartning 1999a & b, Dewaele & Véronique 1999).
The question we will address in the present paper is
whether these advanced learners display similar mastery
of other morphological and lexical rules and in what
measure production in French interlanguage (IL) is
determined by the processing of gender.
The resulting picture should allow us to develop a
better understanding of the link between different
subsystems in the IL and could provide an insight in the
processes of gender retrieval and agreement during IL
speech production.
2
Gender in French
French distinguishes two grammatical genders: masculine
and feminine (Grevisse 1980, Surridge 1996). Gender is
partly an inherent diacritic feature of French nouns, the
value of which has to be acquired individually for every
lexical entry stored in the mental lexicon (gender
attribution). Koehn (1994) found that gender of
morphologically simple nouns is partly predictable on the
basis of their phonological characteristics. Her corpus
analysis of the Petit Larousse revealed that more than
97% of the nouns with the endings: are
masculine.
Gender is also ‘a derivative property of specifiers such
as determiners and adjectives’ (Carroll 1989:545). The
gender feature to the noun can thus trigger gender
agreement among modifying expressions under precise
syntactic conditions (1989:546). The conceptual gender
difference in generally associated with the grammatical
gender difference.
Gender opposition is inaudible in two thirds of the
adjectivesi. This proportion decreases to one half in
3
written language (Riegel et al. 1994:359). The form-
function relationship for gender in adjective is
relatively simple according to Schane (1968) and Blanche
Benveniste (1990). Masculine forms are created by
suppressing the final ‘e’ in feminine forms (see table
1).
INSERT TABLE 1
Gender opposition in determiners only exists in the
singular as can be seen in the following tables:
INSERT TABLE 1
INSERT TABLE 2
Previous work
There are three other studies that use Pienemann’s
Processability Theory (1998) as a theoretical framework
for analysis of gender agreement, namely Bartning (1999a
and 1999b), using the InterFra corpus, and Dewaele &
Véronique (1999) who set out to investigate the factors
that affect accuracy rates for gender agreement in
advanced French IL and present a number of
psycholinguistic hypotheses concerning the ontogenesis of
4
gender errors. As Pienemann’s (1998) predicts that
“structures involving no exchange of grammatical
information between constituent can be processed before
structures that do require such information exchanges”
(1998:76), it seemed reasonable to assume that if this
prediction is correct, one would still be able to detect
traces in more advanced ILs. Structures are that acquired
early could be expected to attain a higher degree of
automaticity and accuracy in later stages. Accuracy rates
for gender agreement were however found to be higher
within phrase constituents (involving determiners and
attributive adjectives) than across constituents
(predicative adjectives), but a t-test revealed that this
difference is not significant (t = -1.4, df = 26, ns).
This means that constituent boundaries do not
significantly affect accuracy rates in gender agreement
in the data. Similarly, Bartning (1999a) in her
longitudinal analysis of advanced French IL produced by 4
Swedish L1 speakers, found that accuracy of gender
agreement for 189 attributive adjectives was not
consistently higher than that for 205 predicative
5
adjectives. In a follow-on paper, Bartning (1999b)
extended her analysis of gender agreement to determiners
in the IL of 6 advanced speakers and also included
material from 9 pre-advanced learners. The earlier
finding was confirmed for advanced learners but not for
pre-advanced learners. The latter were found to obtain
higher adjectival gender agreement accuracy rates in
attributive adjectives than in predicative adjectivesii,
which would be in line with Pienemann’s predictions.
Pienemann (1998) assumes that synchronic variation in
accuracy rates results from different use of grammatical
rules across different tasks (1998:297) and specific
lexical requirements produced by the individual
communicative tasks (1998:308). Synchronic variation has,
according to Pienemann, more to do with concepts like
“data density” and “percentage of rule application”.
Dewaele & Véronique (1999) found a large amount of intra-
individual and interindividual variation. The non-
application of the agreement rule which stops the
diacritic information from arriving at its destination is
only one of several possible scenarios to account for
6
gender errors. The correct diacritic information could be
attached to the lemma, but the phrasal or S-procedures
may fail to deliver the information systematically to all
modifiers, in which case the monitor can intercept
certain errors and produce the correct form later.
Another possibility is that no diacritic information
concerning gender is attached to the lemma, resulting in
free variation or transfer of the gender feature from
other activated lemmata in the speaker’s L1 or other ILs.
A final possibility is that the lemma contains the wrong
diacritic information resulting in systematic gender
errors for certain nouns. This could the consequence of
overgeneralisation of certain phonological gender rules
(Koehn 1994).
An ANOVA revealed that interindividual variation
depended less on the amount of formal instruction the
learner had enjoyed in the target language and more on
the frequency with which s/he declared to use the TL
outside the classroom (Dewaele & Véronique 1999).
7
Rationale
Linguistic variables in interlanguage studies are
usually analysed in isolation (Bartning 1997). We feel
that by looking at the relation between different aspects
of the learner’s speech production process and by
relating morphology, lexicon and syntax we might gain a
better understanding of the interlanguage system as a
whole. Our decision to examine the relationship between
gender agreement and other types of agreement errors,
iFootnotes
* We wish to thank Inge Bartning and Jürgen Meisel for
their excellent suggestions.
? The proportions are very similar in our interlanguage
corpus produced in the informal situation (cf. section on
methodology) where 65% of adjective types were invariable
(representing 40% of adjective tokens). The proportion of
invariable adjective types dropped to 47% in the corpus
produced in the formal situation (representing 39% of
adjective tokens).
ii This result should be interpreted with caution as it is
based on a very small sample.
8
fluency, complexity, lexical richness, and style choice
in our corpus of advanced French IL is based on the fact
that these variables encompass the working of the whole
model as described by Levelt (1989) and Pienemann (1998).
The nature of these variables is very different but they
were shown to be highly interrelated (Dewaele 1999). We
will, more specifically, address the following five
research questions:
1) Do determiners or adjectives best reflect the
learner’s knowledge of gender for a particular noun ?
Dewaele & Véronique (1999) found that gender agreement is
more often correct in determiners than in adjectives.
This finding would be reinforced if we were to find
higher correlations between gender accuracy rates for
determiners (rather than adjectives) and that of the
various linguistic variables.
2) Are gender agreement errors in modifiers caused by
the problems at the level of diacritical information
exchange ? If that were the case, one would expect
similar problems for number and person agreement. An
absence of correlation between these different types of
9
agreement errors would suggest that agreement errors in
advanced ILs are no longer caused by problems at the
level of diacritical information exchange.
3) Are gender agreement errors linked to processing
problems ? Uncertainty about the gender of a noun might
force the speaker to slow his production down and
allocate extra resources to solve the problem (Towell et
al. 1996). This increased monitoring could be reflected
in lower speech rates, an increase in filled pauses and
in shorter utterances.
4) Is the mastery of gender agreement related to the
size of the learner’s actively used vocabulary ? More
advanced learners could be expected to have a larger
vocabulary but could that lead to higher accuracy in
gender agreement ? These advanced learners could have
developed a larger number of rules of thumb (Carroll
1989) or phonological gender rules (Koehn 1994) allowing
them to make less gender errors.
5) Is the decrease in the number of gender errors linked
to the learner’s capacity to produce more implicit speech
(Dewaele 1993a) ? One could expect a significant
10
correlation between these two indicators of IL
development.
Methodology
Participants
Twenty-seven university students, 8 female and 19 male,
aged between 18 and 21, participated in the experiment.
They had taken French at a high school level (3 to 5 hours
a week) for 6 to 8 years. Their interlanguage could be
described as “advanced” (Bartning 1997). The subjects were
administered a sociobiographical questionnaire.
Linguistic material
Conversations were recorded with the 27 subjects in a
stressful and a neutral situation. The stressful situation
consisted of an oral exam of about ten minutes that aimed
at evaluating the learners’ proficiency in the target-
language. Topics were politics, economics and the
subjects’ performance for other exams. In all 5 hours of
speech (17,613 words) were recorded. The neutral situation
involved conversations between the same researcher and
subjects in a relaxed atmosphere. There was no time-
11
restriction. Topics covered studies, hobbies, politics. In
all 15 hours of speech (35,021 words) were recorded. The
recordings were transcribed by the researcher into
orthographical French. These transcriptions were then
coded at the word level according to their grammatical
nature and possible lexical or morphological errors. We
singled out 519 gender errors out of 9378 modifiers.
We reported earlier that the identification of gender
errors is not always easy in French as some adjectives
share the same form for the masculine and the feminine.
In these cases we did not count the form as a gender
error even thought is probably was. This means that the
total number of gender errors for adjectives in our
corpus is probably higher. The following utterance
illustrates the problem:
Richard Informal 1132. Acheter un télé, une, un auto
très cher(e), mais lentement je veux consumer les dix
millionsiii.
iii Every utterance is preceded by the (fictional) name of
the speaker, the type of interview ( Informal or Formal)
and the number of the utterance in the corpus.
12
‘To buy a television, a (fem.), a (masc.) very expensive
(masc. or fem.) car , but I want to spend the ten
millions very slowly.’
The speaker clearly hesitates about the gender of
‘auto’, first using the correct feminine indefinite
article, followed by a pause and then using the incorrect
masculine indefinite, the noun, an adverb and then an
adjective where the gender difference is inaudible.
Analysis
We will now examine the accuracy rates (percentage of
target-like usage) for gender agreement and their
relations to the other morphosyntactic and lexical
variables. Dewaele & Véronique (1999) assumed that the
most advanced learners had a greater automatisation of
the linguistic knowledge which allowed them to spend less
time on the retrieval of gender information. It is
therefore highly probable that accuracy rates for gender
agreement will correlate with temporal variables as these
reflect the smooth running of the learner’s speech
production process.
13
Proportion of other types of agreement errors
Six types of morphological and 4 types of lexical errors
were calculated in Dewaele (1993a, 1994) as well as the
variation in morpholexical accuracy rates for all word
classes as well as the variation in proportions of
particular error types across word classes. The present
analysis concentrates on the 3 types of morphological
errors involving agreement: gender agreement, number
agreement and person agreement in verbs. We checked to see
whether the proportion of gender agreement errors is
linked to the two other types of agreement errors. This
turned out not to be the case: the correlation with the
proportion of number agreement errors was entirely non-
significant in the informal situation (r = - .13, ns.) and
the formal situation (r = - .02, ns); the correlation with
the proportion of person agreement errors in finite verbs
was negative without reaching statistical significance in
the informal situation (r = - .36, p < .07) and was close
to zero in the formal situation (r = .01, ns).
INSERT TABLE 4
14
Speech rates
The speech rate gives us a global idea of the efficiency
of the speech production process (Levelt 1989). Higher
speech rates suggest a higher degree of automaticity in
the speech production, lower speech rates on the other
hand can be an indication of more controlled processing,
requiring speakers to pay attention to the processing
while it is happening (Towell et al. 1996, Dewaele 1998a).
These controlled processes are tightly capacity-limited.
A Pearson correlation revealed a strong positive
correlation between accuracy rates for gender agreement of
determiners and speech rate, in the informal situation (r
= .74, p < .0001) as well as in the formal situation (r
= .47, p < .015). A slightly weaker but still significant
correlation was found between speech rate and accuracy
rates for gender agreement of adjectives in the formal
situation (r = .46, p < .02) but the correlation was non-
significant in the informal situation (r = .24, ns). These
results suggest that fluent speakers make fewer gender
errors.
15
Hesitation phenomena
Hesitations, often followed by editing expressions such
as ‘er’, are generally interpreted as evidence of
cognitive activity. The ‘er’ may serve to signal to the
addressee that there is trouble and that the source of the
trouble is still actual (Levelt 1989:484). These editing
expressions are especially common in L2 production after
errors or before lexical gaps (Dewaele 1996b). They can,
however, also be a sign that the incremental processing is
breaking down (Levelt 1989).
The proportion of filled pauses correlated negatively
with accuracy rates for gender agreement of determiners in
the informal situation (r = - .57, p < .002) and in the
formal situation (r = - .55, p < .003). A similar
correlation appeared between the proportion of filled
pauses and gender agreement of adjectives in the formal
situation (r = - .48, p < .02), but the difference was
non-significant in the informal situation (r = -.10, ns).
This suggests that a higher proportion of filled pauses is
linked to lower accuracy rates for gender agreement.
16
Length of utterance
Mean length of utterance can provide an interesting
insight in the learners’ capacity to build complex
structures in their interlanguage and is often considered
as another measure of fluency (Dewaele 1995b). It also
reflects the development of the IL (Towell et al.
1996:112). We used a formula based on the mean length of
the 3 longest utterances in a speech extract (MLU3).
A non-significant negative Pearson correlation emerges
between the MLU3 scores and accuracy rates for gender
agreement in determiners in the informal situation (r =
-.36, p < .07) which becomes positive and significant in
the formal situation (r = .46, p < .02). Accuracy rates
for gender agreement in adjectives did not correlate with
MLU3 scores in the informal situation (r = -.06, ns) nor
in the formal situation (r = .15, ns). This suggests that
speakers who produce their longest utterances in the
informal situation make more gender errors in determiners
while those who produce their longest utterances in the
formal situation make fewer gender errors in determiners.
This could be interpreted as a sign that the learners
17
producing their longest utterance in the informal
situation are not the same as the ones producing their
longest utterances in the formal situation. A similar
conclusion was reached in an earlier study where it
appeared that more introvert speakers produce their
longest utterances in the informal situation but not in
the formal situation (Dewaele 1999).
Lexical richness
Dewaele (1993b) found significant negative correlations
between lexical richness scores and measures of fluency in
the formal situation. When more cognitive resources are
diverted to lexical searching, the speech production slows
down. Is there a link however between the size of
learners’ active vocabulary and their knowledge of
morphological information of the lemmata ?
A Pearson correlation suggests that this is partly true
but that is varies according to the situation. Lexical
richness scores correlated positively with accuracy rates
for gender agreement in determiners in the informal
situation (r = .37, p < .06) but not in the formal
situation (r = -.19, ns). Accuracy rates for gender
18
agreement in adjectives did not correlate with lexical
richness scores in the informal situation (r = .06, ns)
nor in the formal situation (r = .11, ns)
The choice of speech-style
In Dewaele (1995a) it was explained that the perception
that speakers have of the formality of the situation leads
to make different pragmatic choices. The choice of speech
style will depend on the need of the speaker to be
unambiguously understood and is decided in the
conceptualiser (Levelt 1989). This decision is reflected
in the proportion of ‘deictical’ word classes in the
speech extracts and was measured as follows: for each of
two situations (informal and formal), a separate factor
analysis was performed on the proportion at token-level of
nouns, determiners, prepositions, verbs, pronouns, adverbs
and conjunctions in the French interlanguage of the
participants. Each time, two main orthogonal factors
appeared. The first dimension, which explains over 50% of
the variance, was called ‘implicitness/explicitness’. The
nouns, modifiers and prepositions obtained strong negative
loadings on this factor, as opposed to the pronouns,
19
adverbs, and verbs which obtained high positive loadings.
The nouns, modifiers and prepositions are thus situated
near the explicit end of this dimension, in contrast to
the pronouns, adverbs, and verbs on the implicit end on
the continuum. This grouping of word classes does not
reflect the traditional dichotomy of ‘grammatical’ versus
‘lexical’ words (Dewaele 1995a, 1996a). A speaker who
wants to avoid ambiguity and misinterpretation of his/her
words, she/he relies as little as possible on the spatio-
temporal context they share with the interlocutor(s). This
is achieved by explicit and precise description of the
elements of the context needed to disambiguate the
expression, hence the decrease of deictical words. As
these deictical words are short and of high-frequency,
they can be retrieved and articulated more quickly. Any
decrease of deictical words will hence inevitably hamper
fluency.
A Pearson correlation between individual factor scores of
our subjects on the ‘implicitness/explicitness’ dimension
and accuracy rates for gender agreement in determiners
revealed positive relations in the informal situation (r =
20
.48, p < .02) and non-significant negative relations in
the formal situation (r = -.28, ns). The speakers who
opted for more implicit speech styles in the informal
situation thus tended to make fewer gender errors but that
relation disappeared in the formal situation. Accuracy
rates for gender agreement in adjectives did not correlate
significantly with implicitness scores in the informal
situation (r = .28, ns) nor in the formal situation (r =
-.29, ns)
Tables 5 and 6 present an overview of the results of the
correlation analyses between the different variables and
the gender accuracy rates for the determiners and the
adjectives.
INSERT TABLE 5
INSERT TABLE 6
Discussion
It is striking to observe how much weaker the
correlations are between the linguistic variables and
gender accuracy rates for adjectives compared to those
for determiners. This finding further supports the
finding that gender accuracy rates for adjectives and
21
determiners are very different in advanced ILsiv, the
former being significantly lower (Dewaele & Véronique
1999, Bartning 1999b). It thus seems that determiners
present a better reflection of the advanced learner’s
knowledge of the diacritic information concerning gender
for a particular noun than adjectives. The latter are
lower frequency words of higher morphophonological
complexity bound to present lexeme selection problems for
the learner. In other words, a gender error in an
adjective is more likely to be only an “apparent” gender
error. The learner, possessing the correct diacritic
information for gender and knowing the agreement rules,
activates the matching lemma, but might be unable to
produce the lexeme with the specification (+ feminine).
Meisel (personal communication) suggests that learners
associate a particular determiner with any new noun they
acquire. Hence the increased likelihood of the correct
determiner being produced with any noun, contrary to
adjectives whose link with nouns is much weaker.
The lack of significant correlations between gender
errors and other types of agreement errors suggests that
22
the errors in advanced IL do not occur because of a
deficient exchange of diacritical information. Gender
errors seemed rather linked to processing problems.
Indeed, lower gender accuracy was correlated to speech
rates, filled pauses and utterance length. Speakers with
a poorer knowledge of the gender of nouns might monitor
this problem more intensely and more often, hence
creating an adverse effect on their fluency. This
reinforces the view expressed in Dewaele & Véronique
(1999) that gender errors can have many different causes
and do not necessarily reflect agreement problems.
The link between mastery of gender agreement and lexical
richness was found to be situation-dependent and/or
discourse domain-dependent. It is positive in the
informal situation but disappears in the formal
situation. One possible explanation is that speakers may
have speedy access to the gender for words belonging to
discourse domains they are more familiar with, but that
this process is slower and more prone to errors once the
conversation covers less familiar discourse domains asiv Bartning (1999b) found that pre-advanced leaners had
lower accuracy rates for determiners than for adjectives.
23
was the case in the formal situation. These advanced
learners could indeed have developed a larger number of
rules of thumb (Carroll 1989) or phonological gender
rules (Koehn 1994) allowing them to make less gender
errors, but the stress of the formal situation might
impair their application.
The final research question was concerned with the link
between gender errors and style-choice. The relation
turned out to be situation-dependent too. Speakers who
make fewer gender errors choose more implicit speech-
styles in the informal situation, which can be
interpreted as a sign of confidence and of a more
developed IL. The link disappears in the formal
situation. Possible reasons for this are less familiar
discourse domains and the stress of the exam situation
which strongly affects the performance of more
introverted speakers (Dewaele 1999).
Conclusion
We analysed the link between accuracy rates for gender
agreement and morpho-lexical variables, indicators of
fluency, lexical richness and utterance length in order
24
to gain a better view of the link between different
subsystems in the IL. The analyses allow us to draw the
following portrait of the learners who make fewer gender
errors: they are fluent, confident enough in the informal
situation to use implicit styles. This suggests that for
these speakers lemma selection and retrieval of diacritic
information concerning gender does not mobilise too great
a share of available cognitive resources, although
increased formality of the situation might alter the
picture.
It was suggested that gender errors are not necessarily
the result of problems with agreement rules, neither do
they automatically imply that the diacritic information
for gender in the noun is absent or wrong. Determiners
were found to give a better reflection of the learner’s
knowledge of the gender of a noun than adjectives.
References
BARTNING, I. 1997. L’apprenant dit avancé et son
acquisition d’une langue étrangère. Tour d’horizon et
esquisse d’une caractérisation de la variété avancée. AILE
9, 9-5.
25
BARTNING, I. 1999a. Stratégies d’acquisition dans l’emploi
genre et de l’accord adjectival en français. in: Mélanges
de syntaxe et de sémantique. Hommages à Andrée Borillo. Toulouse.
BARTNING, I. 1999b. Gender agreement in L2 French: Pre-
advanced versus advanced learner. Manuscript.
BLANCHE-BENVENISTE, C. 199. Le français parlé. Paris: Editions
du CNRS.
CARROLL, S. 1989. ‘Second Language Acquisition and the
Computational Paradigm.’ Language Learning 39/4: 535-594.
DEWAELE, J.-M. 1993a. Variation in the morphosyntactic
and lexical systems of French-based interlanguages. In
B. Ketteman & W. Wieden (Eds.), Current Issues in European
Second Language Acquisition Research (pp. 130-140). Tübingen:
Narr.
DEWAELE, J.-M. 1993b. Extraversion et richesse lexicale
dans deux styles d'interlangue française. I.T.L. Review of
Applied Linguistics 100, 87-105.
DEWAELE, J.-M. 1994. Variation synchronique des taux
d'exactitude. Analyse de fréquence des erreurs
morpholexicales dans trois styles d'interlangue
française. IRAL, 33, 4, 275-30.
26
DEWAELE, J.-M. 1995a. Style-shifting in oral
interlanguage: Quantification and definition. In L.
Eubank, L. Selinker & M. Sharwood Smith (eds.), The
Current State of Interlanguage (pp. 231-238). Amsterdam-
Philadelphia: John Benjamins.
DEWAELE, J.-M. 1995b. Variation dans la longueur
moyenne d’énoncés dans l’interlangue française. In L.
Beheydt (Ed.), Linguistique appliquée dans les années 9. (Special
issue), ABLA Papers 16, 43-58.
DEWAELE, J.-M. 1996a. Variation dans la composition
lexicale de styles oraux. IRAL 34, 4, 261-282.
DEWAELE, J.-M. 1996b. Les phénomènes d'hésitation dans
l'interlangue française: analyse de la variation
interstylistique et interindividuelle. Rassegna Italiana
da Linguistica Applicata, 28, 1, 87-103.
DEWAELE, J.-M. 1998a. Speech rate variation in 2 oral
styles of advanced French interlanguage. In V. Regan
(ed.), Contemporary Approaches to Second Language Acquisition in
social context: crosslinguistic perspectives. (pp. 113-123).
Dublin: University College Academic Press.
27
DEWAELE, J.M. 1998b. Lexical inventions: French
interlanguage as L2 versus L3, Applied Linguistics 19, 4,
471-49.
DEWAELE, J.M. 1999. L’effet de l’extraversion sur la
production du discours de bilingues. AILE. Numéro spécial,
vol. 1, 111-126.
DEWAELE, J.M. & VERONIQUE, D. 1999. Gender agreement in
advanced French interlanguage. Unpublished ms.
GREVISSE, M. 198. Le Bon Usage. Grammaire française. Gembloux:
Duculot.
KOEHN, C. 1994. The acquisition of gender and number
morphology within NP. In: J. M. Meisel (ed.) Bilingual First
Language Acquisition. French and German grammatical development.
Amsterdam-Philadelhia: John Benjamins, 29-52.
LEVELT, W.J.M. 1989. Speaking. From Intention to Articulation.
Cambridge: The MIT Press.
PIENEMANN, M. 1998. Language processing and second language
development : processability theory. Amsterdam-Philadelpia: John
Benjamins.
28
RIEGEL, M.; PELLAT, J.-C. & R. RIOUL 1994. Grammaire
méthodique du français. Paris: Presses universitaires de
France.
SCHANE, S. 1968. French phonology and morphology. Cambridge:
MIT Press.
SURRIDGE, M. & G. LESSARD 1984. Pour une prise de
conscience du genre grammatical. La Revue canadienne des
langues vivantes 41/1: 43-52.
SURRIDGE, M. 1993. Gender Assignment in French: The
Hierarchy of Rules and the Chronology of Acquisition.
IRAL, 31/2, 77-95.
SURRIDGE, M. 1996. Le ou La? The Gender of French Nouns.
Clevedon, England: Multilingual Matters.
TOWELL, R.; HAWKINS, R. & BAZERGUI, N. 1996. The
development of fluency in advanced learners of French.
Applied Linguistics 17, 1, 84-119.
All correspondence concerning this article should be
addressed to:
Jean-Marc Dewaele, School for Languages, Literature and
Culture, Birkbeck College, University of London, 43
29
Gordon Square, London WC1H 0PD. e-mail:
Daniel Véronique, UFR Didactique du Français langue
étrangère, Paris III, Sorbonne Nouvelle.
30
Table 1: Form/function relationship for French adjectives
singular plural
Masculine - e - e(s)
Feminine + e + e(s)
Table 2: Form/function relationship for French articles
indefinite definite partitive
singular plural singularplural singular plural
Masculine un des le les du des
Feminine une des la les de la des
31
Table 3: Form/function relationship for other French
determiners
singular plural
masculine à l’/au aux
ce ces
son ses
feminine à la aux
cette ces
sa ses
Table 4: Correlations between the proportion of gender
agreement errors and other types of agreement errors :
Variable Informal
Situation
Formal
Situation
number agreement -.13 -.02
person agreement -.36 .01
* p < .05
32
Table 5: Correlations between accuracy in gender
agreement for determiners and different linguistic
variables in the informal and formal situations
Variable Informal
Situation
Formal
Situation
Speech rates .74*** .47*
Proportion of filled
pauses
-.57** -.55**
Lexical richness .37 -.19
MLU3 -.35 .46*
Implicit speech-
style
.48* -.28
* p < .05
** p < .01
*** p < .001
33
Table 6: Correlations between accuracy in gender
agreement for adjectives and different linguistic
variables in the informal and formal situations
Variable Informal
Situation
Formal
Situation
Speech rates .24 .46*
Proportion of filled
pauses
-.10 -.48*
Lexical richness .06 .11
MLU3 -.06 .15
Implicit speech-
style
.28 -.29
* p < .05
** p < .01
34
Top Related