Abstract
Aims and objectives/purpose/research questions:
We explored the advantage of bilingualism and the effect of linguistic proximity on the acquisition of determiner-noun agreement, an aspect of inflectional morphology, in a French immersion setting. To our knowledge, third language acquisition literature has yet to provide evidence of a bilingual advantage in learning a particular feature of morphology in an additional language among young children.
Design/methodology/approach:
We compared accuracy in determiner-noun agreement on a narrative task in three groups of grade 1 and 2 students (n = 15 in each group): a group of third language French learners whose first language marked gender, a group of third language French learners whose first language did not mark gender, and a group of English learners learning French as a second language.
Data and analysis:
If the determiner and the noun agreed in gender, gender on the determiner was considered correct. A repeated measures factorial ANCOVA was carried out to compare the performance among the three language groups on the proportion of correctly marked masculine and feminine nouns.
Findings/conclusions:
All of the children achieved high levels of accuracy on masculine nouns with no difference among the three language groups on the proportion of correctly marked masculine nouns. There was a significant difference in the proportion of correctly marked feminine nouns in favour of the group whose first language marks gender compared to the other two groups.
Originality:
The current study supports transfer in the domain of morphology among primary-school-aged children (who are first language speakers of diverse minority languages) in the early stages of third-language French acquisition.
Significance/implications:
This finding provides preliminary evidence of a bilingual advantage in the domain of morphology among young, emergent trilinguals whose first language and third language share gender marking as a linguistic feature, supporting the linguistic proximity model.
Keywords
Previous studies attest to a bilingual advantage over monolinguals in acquiring general proficiency in an additional language. The advantage has been demonstrated within the context of additive bilingual education programmes through which students acquire language and literacy skills in their first two languages (first language: L1; second language: L2) and instruction in a third language (L3) (usually foreign) (e.g. Brohy, 2001; Cenoz & Valencia, 1994; Lasagabaster, 2001). The advantage afforded to L3 learners has been most pronounced where the L3 learner’s linguistic repertoire is comprised of languages with overlapping properties, suggesting that language proximity facilitates cross-linguistic influence, the transfer of linguistic competencies acquired in one language to the learner’s other languages (e.g. Schepens, van der Slik, & van Hout, 2016).
Indeed, the Linguistic Proximity Model (LPM) (Mykhaylyk, Mitrofanova, Rodina, & Westergaard, 2015; Westergaard, Mitrofanova, Mykhaylyk, & Rodina, 2017) theorizes that cross-linguistic influence in L3 acquisition results when a linguistic feature ‘reveals’ (Westergaard et al., 2017, p. 670) abstract structural similarity with a linguistic feature of the L1 and/or L2. Third language acquisition is characterized as incremental property-by-property learning in which both facilitative and non-facilitative influence are discernible. According to the LPM, all previously acquired languages are active during the learning process, regardless of order of acquisition or general typological grouping. 1
Thus, the LPM is particularly useful as a framework to account for the bilingual advantage in acquiring specific features within a domain of language. In their review of the L3 literature, Hirosh and Degani (2018) cite studies comparing monolingual and bilingual performance on measures targeting specific linguistic features, including features of grammar. The authors conclude that in the domain of grammar, evidence is suggestive of an advantage. However, they point to the dearth of studies examining the potential advantage of bilinguals relative to monolinguals in grammar learning, calling for ‘much needed’ (p. 907) systematic research in this domain. Moreover, the studies reviewed focus on the acquisition of syntax in the late elementary grades to adulthood (e.g. Abu-Rabia & Sanitsky, 2010; Klein, 1995; Stafford, Sanz, & Bowden, 2010). To our knowledge, the L3 literature has yet to provide evidence of a bilingual advantage in learning a particular feature of morphology in an additional language among young children. The current study aims to address this gap in the literature.
We situate our study within the framework of the LPM as we explore the potential advantage of bilingualism and the effect of linguistic proximity on the acquisition of French gender agreement in an elementary French immersion (FI) setting. In light of studies in the L2 literature related to gender marking in monolinguals and bilinguals (to be reviewed), we aim to determine if L3 learners of French whose L1 marks gender are more successful in gender marking in French than children whose L1 does not mark gender. Specifically, we compare accuracy in determiner-noun agreement in French on a narrative task in three groups of grade 1 and 2 students: a group of L3 French learners whose L1 marks gender, a group of L3 French learners whose L1 does not mark gender, and a group of English learners of L2 French.
Before doing so, however, we begin with a brief overview of the socio-linguistic and educational context within which the current study took place in order to highlight the unique features of FI student populations. In Canada, French-English bilingualism is a national policy and a valued societal goal. FI programmes, which offer integrated language and content instruction to non-francophone children, are seen by Canadian parents as a means to ensure their children achieve that goal (Genesee, 2004). However, French is rarely heard or spoken in English-dominant parts of the country, including the large urban centre in which our study took place. At school entry in kindergarten or grade 1, FI classrooms are comprised of majority English speakers for whom French is the L2 and minority language speakers for whom English is the L2 and French the L3. English language arts are generally introduced in grade 4. Furthermore, FI students tend to come from families of middle to high socio-economic status (SES) (Allen, 2004).
The current study investigates a potential bilingual advantage in learning a particular aspect of French morphology (i.e. determiner-noun agreement) among primary-school-aged children who are L1 speakers of diverse minority languages (of which half mark gender) relative to one another and to their monolingual majority language peers. Previous studies exploring the bilingual advantage in L3 learners who are speakers of a minority L1 have typically involved comparisons of outcomes on broad measures of school-based foreign language learning between minority and majority language speakers educated primarily in the majority language (Abu-Rabia & Sanitsky, 2010; Edele, Kempert, & Schotte, 2018; Maluch & Kempert, 2017; Maluch, Neumann, & Kempert, 2016; Van Gelderen et al., 2003). Overall, the studies yielded mixed results. Likewise, studies conducted within the context of Canadian FI programs in which outcomes of minority and majority language learners were compared have produced conflicting results with respect to a bilingual advantage (Bérubé & Marinova-Todd, 2012; Mady, 2017; Swain, Lapkin, Rowan, & Hart, 1990).
Importantly, the discrepancies in research findings suggest additional background and individual factors that condition L3 language learning in minority language speakers. Background factors include socio-economic status (e.g. Cenoz, 2013), parental education levels (De Angelis, 2015) and quality of the home literacy environment (e.g. Maluch, Kempert, Neumann & Stanat, 2015). Individual factors include age of L2 acquisition (Maluch & Kempert, 2017), patterns of L1/L2 use (e.g. Maluch & Kempert, 2017; Maluch et al., 2016), L1/L2 proficiency (e.g. Dewaele, 2010; Edele et al., 2018), and acquisition of literacy (e.g. Rauch, Naumann, & Jude, 2011; Swain et al., 1990). Minority language speakers develop very different linguistic profiles as a function of variations in the socio-linguistic environment and their impact can be discerned on L3 learning. In contrast to previous studies, we provide evidence of equivalency with respect to a number of these factors across the two minority language groups and, where appropriate, the majority language group. Specifically, we present data establishing equivalent levels of parental education, French literacy skills, patterns of L1 use, age of L2 acquisition, and the quality of the home literacy environment in an effort to isolate the effect of linguistic proximity in facilitating L3 learning in the domain of morphology.
Grammatical gender
Grammatical gender is an aspect of inflectional morphology found in many languages. In these languages, every noun belongs to a gender system. Grammatical gender systems vary cross-linguistically on a number of dimensions (Comrie, 1999). There are four main gender systems: masculine-feminine (e.g. French), masculine-feminine-neuter (e.g. Russian), common-neuter (e.g. Dutch) and animate-inanimate (e.g. Basque) (Corbett, 1991). Agreement is a syntactic property that refers to the gender and number marking between adjectives, determiners, numerals, past participles and pronouns as a function of the noun with which they co-occur (Grüter, Lew-Williams, & Fernald, 2012).
In French, the noun of the nominal phrase will determine the gender of its determiners, quantifiers and adjectives. For example, leMASC petit garçon means ‘the little boy’, whereas, laFEM petite fille means ‘the little girl’. In this example, the definite article (le/la) and the adjective (petit/petite) change their form according to the gender of the noun. Likewise, indefinite articles differ in form in the masculine and feminine (e.g. unMASC chat ‘a cat’; uneFEM souris ‘a mouse’). Whereas definite and indefinite articles in French differ in form for the masculine and feminine in the singular, they do not differ in the plural (lesPL/desPL ‘the/some’) (Dewaele, 2015).
Gender in French is also marked on past participles; for the most part, however, the masculine and feminine form are phonologically indistinct (il/elle est allé MASC /alléeFEM ‘he/she went’), as they are in the case of a large number of adjectives (joli MASC/jolieFEM ‘pretty’) and most nouns (ami MASC/amieFEM ‘friend’). Conversely, French gender is unambiguously marked on determiners (Kupisch, Müller, & Cantone, 2002), with the exception of singular nouns beginning with a vowel or h muet ‘silent h’ which are subject to élision, making the definite articles le and la phonologically indistinct as a result of the hiatus imposed by the ‘silent h’ (l’animalMASC ‘the animal’, l’abeilleFEM ‘the bee’, l’homme MASC ‘the man’, l’herbeFEM ‘the grass’) (Kupisch, Akpinar, & Stöhr, 2013). Because of its relative lack of ambiguity, we chose to focus our study on singular noun-determiner agreement within the context of an oral narrative task.
Grammatical gender in French categorizes nouns according to semantic principles and formal principles to form rules of assignment (Comrie, 1999; Montrul & Potowski, 2007). As a result, the gender attribution system is opaque; grammatical gender corresponds to natural gender in only about 10% of nouns. Approximately, 58.4% of French nouns are masculine (Corbett, 1991; Séguin, 1969). Noun endings may provide a useful strategy for gender system assignment (Lyster, 2006); however, these cues are unreliable because the French gender system has many exceptions. For example, nouns ending with -age are usually masculine (leMASC fromage, ‘the cheese’) but there are some exceptions, such as uneFEM image, ‘a picture’, laFEM plage, ‘the beach’ and laFEM page, ‘the page’. Similarly, nouns ending with -ion are usually feminine (laFEM station, ‘the station’), yet unMASC avion, ‘a plane’, unMASC lion, ‘a lion’ and unMASC camion, ‘a truck’ are masculine. Despite the system’s opacity, native French-speaking children make relatively few errors in determiner-noun agreement in spontaneous speech even at a very young age (Kupisch et al., 2002). Conversely, grammatical gender has been described as ‘notoriously difficult’ (Ayoun, 2007, p. 137) to acquire for learners of French as an additional language. However, if additional language learning is viewed within the framework of the LPM (Westergaard et al., 2017), the question arises as to whether grammatical gender marking is facilitated in French in learners whose linguistic repertoire includes a language that features gender marking as a structural property. The model would predict that this should be the case.
Indeed, previous studies investigating grammatical gender marking in an L2 support the prediction (Ellis, Conradie, & Huddlestone, 2012; Sabourin, Stowe, & de Haan, 2006; Schwartz et al., 2015). For example, Sabourin et al. (2006) compared the ability to mark gender agreement among adult L2 Dutch speakers of L1 German, French, Italian, Spanish or English. As expected, the German and Romance speakers performed significantly above chance, whereas the English group performed at chance, suggesting cross-linguistic influence of gender-marking ability from L1 to L2. Ellis et al. (2012) investigated L2 acquisition of grammatical gender in German by adult L1 speakers of Afrikaans, English and Italian using picture naming and sentence completion tasks. Neither Afrikaans nor English express grammatical gender whereas Italian does. The results revealed an advantage in favour of the L1 Italian speakers. The authors suggest their results provide evidence for transfer of abstract grammatical properties (i.e. deep transfer) since incongruities in the gender systems of Italian and German preclude transfer of knowledge of the grammatical gender of specific nouns (i.e. surface transfer) across the two languages (cf. Sabourin et al., 2006). More recently, Schwartz et al. (2015) examined the influence of L2 grammar on the acquisition of noun-adjective gender agreement in L1 Russian speakers by comparing four- to five-year-old bilingual children who spoke English, Finnish, German or Hebrew as their L2 and age-matched monolingual children. Children whose L2 marked grammatical gender outperformed the children whose L2 did not mark grammatical gender, suggesting transfer from L2 to L1.
The current study
Building on previous studies investigating a bilingual advantage in L3 learning and on studies reported in the L2 literature examining cross-linguistic influence of gender marking, we compared accuracy in determiner-noun agreement in French among three groups of French immersion students in grades 1 and 2. The first two groups were comprised of children learning French as their L3 through formal schooling who spoke English as their L2. The children in one of the two groups spoke a L1 that marked determiner-noun agreement; the children in the second group spoke a L1 that did not mark determiner-noun agreement. 2 The second group of L3 learners was included to test the possibility that bilingualism per se, or the cognitive and linguistic factors that mediate it, may facilitate L3 learning regardless of linguistic proximity. The final group was comprised of L2 learners, all of whom spoke English as their L1. English does not assign nouns to gender categories or mark determiner-noun agreement.
In view of previous findings in the L2 literature (Ellis et al., 2012; Sabourin et al., 2006; Schwartz et al., 2015), and consistent with the LPM, we predicted that accuracy in determiner-noun agreement would be dependent on L1 status. That is, children who had been exposed to a language that marked gender at home were expected to correctly mark determiner-noun agreement more often than children who had been exposed to a language that did not mark gender.
Method
Participants
Participants of the present study were selected from a larger study (n = 228) that examined children’s language and literacy development in English and French (see Au-Yeung et al., 2015). As part of the inclusion criteria, children selected for this study were non-native speakers of French. The children began French instruction in the fall of grade 1; the majority had attended a full-day, publicly funded, English-language kindergarten program the previous year. Furthermore, children who were identified as at-risk readers based on standardized screening measures and thus participated in an intervention study (Wise, D’Angelo, & Chen, 2016) were excluded from the study. Using a demographic questionnaire, we identified 15 children who spoke a home language that marked grammatical gender and 15 children who spoke a home language that did not mark gender. These participants were matched on French receptive vocabulary with children who only spoke English at home. The matching resulted in a total of 45 children (Mage = 84.99 months, SD = 4.93, 28 male, age range = 76.3 – 98.6 months, n = 33 in grade 1, n = 12 in grade 2): 15 children whose home language marked gender (Mage = 85.5 months, SD = 5.40, 8 males, 13 born in Canada), 15 children whose home language did not mark gender (Mage = 85.2 months, SD = 4.58, 12 males, 13 born in Canada) and 15 children who spoke English at home (Mage = 84.32 months, SD = 5.03, 8 males, all born in Canada). The L3 children were a linguistically diverse group, comprised of speakers of Russian (n = 12), Romanian, Serbian, Spanish (ns =1), all of which mark gender; Chinese (n = 8), Korean (n = 4), Farsi (n = 2) and Hungarian (n = 1), none of which mark gender. Herein, the L3 group whose L1 marked gender was labelled L3MG, the L3 group whose L1 did not mark gender was labelled L3NMG, and the L2 group was labelled L2NMG.
According to parent report, approximately 47% of the children in the L3MG group always spoke their L1 at home, and 53% spoke their L1 frequently. In the L3NMG group, 60% of children always spoke their L1 at home, and 40% reported using their L1 frequently. All children in the L3 groups were reported to speak French rarely or never at home. 3 In terms of the L2NMG group, 93% spoke English always at home and 7% used English frequently. Ninety-two percent of the children in the L2NMG group spoke French rarely or never at home. 4 A one-way analysis of variance found no significant difference in age (F (2, 41) = .22, p = .801), parental education (F (2, 44) = .38, p = .687) and the number of children’s books in the home (F (2, 44) = 1.87, p = .166) among the three language groups.
Measures
Covariates
In order to establish that all three language groups were equivalent with respect to factors that contribute to additional language learning, we first present data on several covariates using standardized and experimental measures validated in previous research.
Parental demographic questionnaire
Parents completed a demographic questionnaire designed by the researchers targeting the following: parent education (i.e. highest level of education received on a 6-point scale from (1) some high school to (6) post-graduate degree); age at which their child first spoke English (in months); the frequency with which the child communicates in his or her first language (5-point scale from (0) never to (4) always); and the quality of the children’s home literacy environment as indicated by the number of books in the home (5-point scale from (0) less than 10 to (4) more than 100).
Nonverbal reasoning
General intelligence contributes to additional language learning (Cenoz, 2013). Consequently, the Matrix Analogies Test (Naglieri, 1985) was used to assess nonverbal reasoning. This measure included 4 subtests of 16 items each: pattern completion, reasoning by analogy, serial reasoning and spatial visualization. For each item within a subtest, children were presented with an incomplete visuospatial matrix and asked to select one of six options that completed it. Subtests were discontinued when children made four consecutive errors. Cronbach’s alpha reliability was .95.
French receptive vocabulary
Form A of the Échelle de Vocabulaire en Images Peabody (Dunn, Theriault-Whalen, & Dunn., 1993), normed on a sample of Canadian children who were native speakers of French, was used to assess students’ French receptive vocabulary. A target word was spoken and children identified one of four pictures that best illustrated it. Three practice items preceded testing. Testing was discontinued when the child made six errors on eight consecutive items. Cronbach’s alpha reliability was .96.
French word reading accuracy
An experimental measure designed by Au-Yeung et al. (2015) was used to assess word reading accuracy in French. There were 120 items arranged in 15 sets of 8 words each. The children were asked to read the words. Testing was discontinued when the children misread five or more words within a set of eight words. Raw scores represent the number of words correctly read. Cronbach’s alpha reliability was .98.
Gender-marking measures
Children’s skill at marking noun-determiner agreement was assessed using a story generation task in French. Narrative samples are considered to be ecologically valid measures of discourse-level language proficiency in children from linguistically diverse backgrounds (Bedore, Peña, Gillam, & Tsung-Han, 2010; Cleave, Girolametto, Chen, & Johnson, 2010).
Narrative generation task
The child was presented with one of four wordless picture books by Mercer Mayer: A Boy, a Dog and a Frog (1967), Frog, Where Are You? (1969), Frog on His Own (1973) or One Frog Too Many (1975). Four books were used because of the longitudinal nature of the broader research study. Assignment to books was random. Following Paris and Paris (2003), shortened versions of the books were created, all comprised of 16 pictures. The procedure for testing was as follows: the experimenter began by showing the child a copy of the wordless picture book, pointing out the absence of text. The child was asked to look carefully at the pictures in the book to figure out the story they were telling. The child was told to keep the story a secret from the experimenter until all pictures had been seen and he or she had ‘figured out what was happening’. When the child was ready to tell the story, the experimenter took back the book, opening it to page 1. The child was asked to tell a story in French that the experimenter would understand without looking at the pictures. The experimenter remained silent except to encourage the child to remain on task or to provide supportive feedback without influencing the content of the story. The child’s story was audio recorded for later transcription.
Trained research assistants who were competent speakers of French transcribed the narratives which were then segmented into utterances consisting of a single main clause and any dependent clauses. Following Craig, Connor, and Washington (2003), successive main clauses linked by simple coordinate conjunctions (et, ‘and’; mais, ‘but’) were segmented into separate utterances if the second clause included a subject. When the second clause elliptically omitted the subject, the two clauses were considered a single utterance. Twenty percent of the transcriptions were randomly selected and re-transcribed by a research assistant for reliability purposes. Interrater reliability was calculated using the following formula: (number of agreements / (number of agreements + disagreements)) × 100 (Sackett, 1978) and yielded 97% for words (n = 1407) and 85% for utterance boundaries (n = 172).
Following transcription, the narratives were coded by the first author. All nouns that were preceded by a determiner were coded, including repetitions since children were inconsistent in their use of gender markings on the same noun. If the determiner and the noun agreed in gender, one point was awarded. If the determiner and the noun disagreed in gender, no point was awarded. We calculated the proportions of correctly marked masculine and feminine nouns for each group of children by dividing the number of correctly marked nouns in each category by the total number of nouns in that category. Only singular definite, indefinite and possessive determiners were included in the coding scheme (definite articles subject to élision were excluded). For reliability purposes, 20% of the transcripts were randomly selected and recoded by a research assistant. Sackett’s (1978) formula was applied again and yielded 96% for masculine and 97% for feminine noun-determiner agreement.
Procedure
Parents of the participating children completed a consent form and a questionnaire to collect basic demographic information and background on children’s home language environment. Testing was done on an individual basis for all students during regular school hours by trained research assistants. Testing ran 45–60 minutes over 1–2 sessions. Nonverbal reasoning skills were measured in English in the fall. The children were assessed on all other measures the following spring. Instructions for the French tasks were given in both French and English.
Results
Descriptive statistics
Table 1 summarizes the descriptive statistics for all of the measures. There were no extreme outliers (z < ± 3.29; Tabachnick & Fidell, 2007) and very few missing values (less than 3%). Both raw and standard scores are reported for all the standardized measures; raw scores were used for the statistical analyses. The standard scores for the Matrix Analogies Test and the French receptive vocabulary task are based on a distribution with a mean of 100 and a standard deviation of 15. With respect to nonverbal reasoning, the standard scores were one standard deviation above average relative to age-level peers. The standard scores for French receptive vocabulary indicated that the children in our sample were one standard deviation below average relative to francophone children.
Descriptive statistics.
Note. L3MG = a group of third language French learners whose first language marked gender; L3NMG = a group of third language French learners whose first language did not mark gender; L2NMG = a group of English learners learning French as a second language.
We conducted a preliminary descriptive investigation of the data, in order to establish that all three language groups were equivalent with respect to several factors that contribute to additional language learning. A one-way analysis of variance with language group as the between-groups factor found no significant difference in nonverbal reasoning (F (2, 44) = .39, p = .678), French receptive vocabulary (F (2, 44) = .03, p = .970) and French word reading accuracy (F (2, 44) = 2.44, p = .100). With respect to the age at which the child first spoke English, there was a significant difference between the L2NG and the two L3 groups (F (2, 42) = 8.27, p = .001). Notably, the L2NG group first spoke English significantly earlier compared to the L3MG and the L3NMG groups, respectively. No difference was found between the L3MG and L3NMG. An independent samples t-test revealed significantly higher values among grade 2 students in terms of age (M = 91.33, SD = 3.05; t (40) = −7.76, p < .001), French receptive vocabulary (M = 73.58, SD = 22.66; t (43) = −5.11, p = .001) and French word reading (M = 75.75, SD = 20.83; t (43) = −2.75, p = .009), when compared to grade 1 children. In terms of nonverbal reasoning, there was a significant difference in favour of grade 1 children (M = 25.82, SD = 13.00; t (43) = 2.13, p = .039). As a result, we included these variables as covariates in the analysis.
Determiner-noun agreement
A 3 Language Group (L3MG, L3NMG, L2NMG) x 2 Stimulus Type (masculine, feminine) repeated measures factorial ANCOVA was carried out to compare the performance among the three language groups on the proportion of correctly marked masculine and feminine nouns. 5 Stimulus type was a within-subjects factor, language group was a between-subjects factor, and age, nonverbal reasoning, French receptive vocabulary and French word reading were the covariates. A non-significant Box’s M test (p =.509) indicated equality of variance-covariance matrices of the proportion of masculine and feminine nouns correctly marked across the three language groups. Mauchly’s test revealed that the assumption of sphericity was met. The main effect of stimulus type failed to reach significance, F (1, 35) = 2.10, p = .156. By contrast, there was a strong effect of language group, F (1, 35) = 4.75, p = .015, partial η² = .213.
As shown in Figure 1, the interaction between language group and stimulus type was significant, F (2, 36) = 10.73, p < .001, partial η² = .380. Bonferroni post-hoc pairwise comparisons revealed that there were no significant differences among the three language groups on the proportion of correctly marked masculine nouns. 6 However, as shown in Figure 1, there was a significant difference in favour of the L3MG group compared to the L2NMG group (p < .001) and the L3NMG group (p = .007) on the proportion of correctly marked feminine nouns. There was no significant difference between the L3NMG group compared to the L2NMG group (p = 1.00). There were no significant interactions between stimulus type and the covariates: age F (1, 35) = .02, p = .903, nonverbal reasoning F (1, 35) = .68, p = .414, French receptive vocabulary F (1, 35) = .29, p = .592, and French word reading accuracy, F (1, 35) = 1.46, p = .235. 7
Mean proportion scores by condition and language group.
Discussion
In this study, we explored the advantage of bilingualism and the effect of linguistic proximity on the acquisition of gender agreement, an aspect of inflectional morphology, in a French immersion (FI) setting. To do so, we compared accuracy in determiner-noun agreement in French on a narrative task in three groups of grade 1 and 2 students: a group of L3 French learners whose L1 marked gender, a group of L3 French learners whose L1 did not mark gender, and a group of English learners of L2 French. Importantly, we established that the two L3 groups were equivalent with respect to a number of factors that contribute to additional language learning, particularly among minority language speakers. This allowed us to isolate the effect of linguistic proximity in determining outcomes in the two L3 groups. Our results indicated that children’s ability to correctly mark determiner-noun agreement on feminine nouns was dependent on L1 status, whereas their ability to correctly mark determiner-noun agreement on masculine nouns was independent of L1 status. Overall, our results provide some support for the LPM (Westergaard et al., 2017) which theorizes that cross-linguistic influence in L3 acquisition results when a linguistic feature of the L1 and/or L2 is structurally similar to a linguistic feature of the L3.
First, our results revealed that L3 French learners whose L1 marked gender were more accurate in marking determiner-noun agreement on feminine nouns than either the L3 French learners whose L1 did not mark gender or the French L2 learners. Our results extend those reported in the L2 literature indicating cross-linguistic influence from the L1 to the L2 (e.g. Schwartz et al., 2015) to the context of third language acquisition. In particular, the current study provides evidence of cross-linguistic influence of gender marking, a feature of inflectional morphology, from the L1 to the L3. Moreover, it provides evidence of transfer in the domain of morphology among primary-school-aged children (6–7 years of age) in the early stages of L3 French acquisition. Consistent with the LPM (Westergaard et al., 2017), we attribute this finding to the advantage bilingual children experience in learning an additional language when that language shares structural features with the L1 and/or L2.
Conversely, our results revealed that performance in marking masculine nouns was independent of L1 status. No group differences were found as all children, regardless of language background, achieved approximately 90% accuracy on masculine nouns. The level of accuracy achieved in marking masculine determiner-noun agreement stands in stark contrast to that achieved in marking feminine determiner-noun agreement. Indeed, accuracy on feminine nouns across the three groups stood at 48%, 29% and 19% for the L3 gender marking group, the L3 non-gender marking group and the L2 group, respectively. Thus, all children, regardless of L1 background, demonstrated a clear tendency to employ a masculine-as-default strategy in which they over-generalized the masculine form of determiners. Logically, overuse of the masculine form implied underuse of the feminine form since the French language is comprised of only two gender systems. Notably, Boloh and Ibernon (2010) observed the use of the masculine-as-default strategy in young native speakers of French at age four and seven albeit under strictly controlled experimental conditions. Our results suggest that children who learn French as an additional language, be it their L2 or L3, follow a similar initial pattern of development and that the masculine-as-default strategy is evident in the naturalistic context of their storytelling.
Since approximately 58.4% of French nouns are masculine (Corbett, 1991; Séguin, 1969), children have less exposure to the feminine form of determiners. Thus, the feminine may require more time to acquire than the masculine form during French L1 development (Boloh & Ibernon, 2010) and during the acquisition of French as an additional language. The difficulty in accurately marking feminine determiner-noun agreement in additional language learners was clearly evident in the narrative context, although significantly less so among L3 learners whose L1 marked gender than among their peers – both L2 and L3 – whose L1 did not mark gender. Harley (1993) suggests that the difficulty FI students encounter in acquiring grammatical gender is related, in part, to its low communicative load. That is, speakers can understand the gist of a communication without attending to articles or adjective endings. Indeed, the French language counts only about 50 nouns that are categorized as both masculine and feminine, and whose meaning differs as a function of categorization (e.g. unMASC tour ‘a turn’; uneFEM tour ‘a tower’) (Presson, MacWhinney, & Tokowicz, 2014). With their clear focus on the development of communicative skill (Genesee, 2004), FI classrooms may not prioritize explicit instruction in marking gender, particularly in the early years.
In an effort to isolate the effect of linguistic proximity on outcomes in determiner-noun agreement, the groups in our study were compared on a number of variables demonstrated in previous research to impact language learning, particularly in minority language children. With respect to L3 group comparisons, no differences were found in terms of the age of acquisition of L2 English and both groups were found to begin speaking English significantly later than the L1 English group. Further, equal numbers of L3 parents reported that the L1 was always or frequently spoken at home. In addition, comparisons across all three groups determined comparable pre-school language experiences (i.e. English kindergarten). Taken together, these data establish both L3 groups as late simultaneous/early sequential bilinguals, with comparable levels of exposure to and use of their L1 and L2.
To further isolate the effect of linguistic proximity, the three groups were compared with respect to socio-cultural variables that bear on language learning (parental education, the quality of the home literacy environments as indicated by the number of books in the home) (Maluch & Kempert, 2017). Again, our groups were found to be equivalent. Cognitive and literacy variables were also considered; neither discriminated among the three groups. In addition, French was reported by all but one parent to be spoken rarely if ever at home, limiting exposure to French among children to the equivalent amounts and types of exposure afforded to them in school. Having accounted for these relevant factors, the finding of a bilingual advantage in French feminine gender marking among the L3 children whose L1 marked gender suggests the important role linguistic proximity may play in facilitating cross-linguistic influence within the domain of morphology even among young children in the early stages of L3 acquisition. Cross-linguistic influence, in turn, is theorized to support the bilingual advantage that underlies the superior outcomes of the L3 gender marking group in marking feminine determiner-noun agreement in French. This interpretation of the results is bolstered by the failure to find a significant difference between the L2 French learners and the L3 non-gender marking group on gender marking accuracy as it suggests that bilingualism in itself – or the cognitive advantages it is theorized to engender – do not necessarily lead to an advantage in learning specific aspects of L3 morphology.
In the present study, 80% of the children whose L1 marked gender spoke Russian, a language with a three-tiered masculine-feminine-neuter gender system. In comparison, French has a two-tiered gender system that consists of masculine and feminine nouns. The two systems are, therefore, incongruent. Accordingly, we concur with Ellis and colleagues (2012) who suggested that their results provided evidence of deep transfer of abstract grammatical properties since incongruities in the gender systems of the languages under study precluded surface transfer of knowledge of the grammatical gender of specific nouns across the two languages. Again, this perspective is consistent with the LPM. However, to provide a comprehensive picture of cross-language transfer of grammatical gender marking, future research should test the notion that gender acquisition may be impacted not only by the presence of gender as an abstract category, but by the similarity of categorization, as suggested by Schwartz and colleagues (2015).
Our study had limitations. First, because it was concurrent in nature, longitudinal follow-up is necessary to examine the development of gender acquisition over time. Second, because of our small sample size, our findings should be considered preliminary. Future studies involving larger numbers of participants per group are needed to test their robustness. Third, our study was conducted within the context of a Canadian French immersion program. The particularities of the program and its participants were discussed in the introduction. They relate, among other things, to the status of French as an additional language and to the socio-economic status of the majority of families in FI. Future research is needed to investigate the bilingual advantage in L3 morphology acquisition among minority language children in other social and educational settings to address the generalizability of our findings. Finally, future studies should explore whether acquisition of grammatical gender in an additional language is affected by the structural similarity of gender marking systems between previously acquired languages and the additional language.
Despite the limitations, our study is novel in a number of ways. It is one of a very few studies to examine the bilingual advantage in the domain of morphology by comparing groups of L3 learners to one another and to their L2 learner peers. In doing so, it extends findings previously limited to the L2 literature. In addition, the study involved primary-school-aged children in their first and second years in French immersion. Few studies have examined the bilingual advantage in such young children in the initial stages of additional language learning. Furthermore, it provides evidence of equivalency among groups on a number of variables that have been shown to influence third language learning, particularly among minority language learners, thereby isolating as much as possible the effect of linguistic proximity in cross-linguistic influence, a factor that undergirds the bilingual advantage.
Footnotes
Acknowledgements
This research would not have been possible without the support of the teachers, parents and children at our partner schools, in addition to all of the research assistants of the Multilingualism and Literacy Lab.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was funded by the Social Sciences and Humanities Research Council (Grant Number 494811) to the third author. Opinions reflect those of the authors and do not necessarily reflect those of the funding agency.
