Bilingualism: Language and Cognition 9 (1), 2006, 1532

2006 Cambridge University Press doi:10.1017/S136672890500235X

15

Cross-linguistic transfer in adjectivenoun strings by preschool bilingual children*

E L E NA N I C O L A D I S
University of Alberta

One hypothesis holds that bilingual childrens cross-linguistic transfer occurs in spontaneous production when there is structural overlap between the two languages and ambiguity in at least one language (D opke, 1998; Hulk and M uller, 2000). This study tested whether overlap/ambiguity of adjectivenoun strings in English and French predicted transfer. In English, there is only one order (adjectivenoun) while in French both adjectivenoun and nounadjective order are allowed, with the latter as the default. Unidirectional transfer from English to French was predicted. 35 FrenchEnglish preschool bilingual children (and 35 age-matched English monolinguals and 10 French monolinguals) were asked to name pictures by using an adjectivenoun string. In addition to the reversing adjectivenoun strings in French as predicted by the overlap/ambiguity hypothesis, the bilingual children reversed more adjectivenoun strings in English than monolinguals. It is proposed that cross-linguistic transfer might better be understood as an epiphenomenon of speech production.

Introduction Children who grow up hearing two languages can clearly differentiate these two languages early in development (see Nicoladis and Genesee, 1997, for a review of the literature). Nevertheless, it is clear that bilingual children differ from monolingual children in several ways. First, in some instances or with some domains of language, bilingual children can be delayed relative to monolingual children (e.g. Bosch and Sebasti an-Gall es, 2003). Delay may be particularly likely for aspects of language for which developmental timing is sensitive to frequency of use or frequency in the input (e.g. M uller and Hulk, 2001). Delay is probably due to the fact that bilingual children use and hear less of either language than monolingual children (Pearson, Fern andez, Lewedag and Oller, 1997). In some cases, however, the knowledge a bilingual child has in one language can facilitate the acquisition of similar knowledge in the other language. For example, GawlitzekMaiwald and Tracy (1996) showed that a GermanEnglish

* This study was funded through an operating grant from the Natural Sciences and Engineering Research Council of Canada. Many thanks to the parents who allowed their children to participate and the daycares and preschools who welcomed us while we ran this study. Christie Ladouceur, Amy Ryniak and Nathalie Savoie helped with nding children and with the testing. Monique Charbonnier, Andrea Krott, Nathalie Savoie, Marie-Jos ee Verret and Robert Wilcocks all provided their intuitions about the grammaticality of adjectivenoun strings in various languages. Gary Dell and Christina Gagne discussed with me exactly how a speech production model could account for the present data. Anamaria Popescu read and criticized an earlier version of this paper.

bilingual child used her knowledge of some aspects of German to produce these structures in English (see also Pearson, Fern andez and Oller, 1993, for similar arguments about vocabulary). A nal difference from monolingual children is that bilingual children sometimes use constructions that are either ungrammatical or sound odd in the target language. For example, a Spanish English bilingual child might refer to his fathers car as the car of my father (perhaps modeled on the Spanish possessive string el carro de mi papa). This construction is grammatical in English although it is more natural to say my fathers car. In this paper, I use the terms crosslinguistic transfer (or transfer) to refer to the structural inuence of one language on another (see Meisel, 1983), regardless of whether the inuence facilitates language acquisition or results in a grammatical construction. This formulation of cross-linguistic transfer crucially implies that bilingual children use language differently from monolingual children (D opke, 1998; Hulk and M uller, 2000) and that difference cannot be accounted for by a difference in the rate of acquisition (such as delay or acceleration; see also Paradis and Genesee, 1996; cf. M uller and Hulk, 2001). Cross-linguistic transfer has been observed in bilingual children in phonology (Holm and Dodd, 1999; Paradis, 2001; Barlow, 2002; Brulard and Carr, 2003), in complex word formation (Nicoladis, 1999; 2002a; 2003) and in syntax (Hulk and van der Linden, 1996; Hulk, 1997; D opke, 1998; M uller, 1998; Yip and Matthews, 2000; Paradis and Navarro, 2003). All instances of crosslinguistic transfer that have been reported in preschool children have been in the PRODUCTION of language

Address for correspondence University of Alberta, Department of Psychology, P-217 Biological Sciences Building, Edmonton AB, T6G 2E9, Canada E-mail: elenan@ualberta.ca

16

Elena Nicoladis M uller and Hulk, 2001). The data in the present study cannot differentiate between the level at which overlap and ambiguity might be important, so I refer to both versions as the structural overlap/ambiguity hypothesis. In support of this hypothesis, D opke (1998), for example, found that EnglishGerman bilingual children overgeneralized -VO word order in German. German allows both -VO and -OV word orders, depending on whether the construction in question is a main or dependent clause, while English allows -VO. Working within the Competition Model (Bates and MacWhinney, 1989), D opke argued that her young subjects were prone to overgeneralize -VO word order in their German because the -VO order was reinforced on the surface of both the German and the English input they heard. Working within a UG framework, Hulk and M uller (2000) and M uller and Hulk (2001) have similarly argued that there has to be an overlap of the underlying structures of the two languages, with ambiguity in at least one language, particularly at the pragmatic/syntactic interface, for cross-linguistic transfer to occur. Structural overlap/ambiguity can explain many cases of cross-linguistic transfer in child language acquisition, like placement of object clitics in French (Hulk, 1997), object drop in a DutchFrench bilingual child and a GermanItalian bilingual child (Hulk and M uller, 2000), verbobject compounds in FrenchEnglish bilingual children (Nicoladis, 2003) and word order in subordinate clauses in German (D opke, 1998; M uller, 1998). Furthermore, if this explanation is correct, bilingual children should have no difculty in acquisition when there is no overlap or ambiguity in their two languages. Indeed, transfer is not seen in bilingual children in some cases when there is no overlap or ambiguity, as in strong/clitic pronoun use, verb niteness and placement of negative markers relative to the verb in French and English (Paradis and Genesee, 1996) and use of root innitives in a FrenchDutch bilingual child and an ItalianGerman bilingual child (Hulk and M uller, 2000). Nevertheless, there are a few reports of cross-linguistic transfer in the absence of structural overlap/ambiguity. For example, nounnoun compound words are exclusively left-headed in French and exclusively right-headed in English and yet FrenchEnglish bilingual children make more reversals in English nounnoun compounds than English children (Nicoladis, 2002a). Similarly, relative clauses are exclusively pre-nominal in Cantonese and exclusively post-nominal in English, yet a CantoneseEnglish bilingual child made reversals in both languages (Yip and Matthews, 2000). These reports of cross-linguistic transfer in the absence of structural overlap/ambiguity raise some doubts that it is the perfect explanation. It is nevertheless possible that it is a very good explanation and its applicability still needs to be assessed. The purpose of the present study is to test if structural overlap/ambiguity explains cross-linguistic transfer in

rather than comprehension (see Nicoladis, 2002a, 2003; cf. Meisel, 1983). For example, Nicoladis (2002a) showed that FrenchEnglish bilinguals made reversals when producing novel nominal compounds (like chairs owers to refer to chairs with owers on them). However, the bilinguals were equivalent to monolinguals in their comprehension of novel nominal compounds that relied on order alone (for example, picking out a clown balloon from an array of four pictures, including a clown holding a balloon and a balloon in the shape of a clown). Based on results such as these, Nicoladis (2003) argued that transfer was a phenomenon of speech production. In spite of the above observations, in devising an explanation of cross-linguistic transfer, researchers have thus far generally focused on a representational level that does not distinguish between production and comprehension, and exclusively at the level of syntax. Here, I briey touch on three explanations of crosslinguistic transfer: 1) idiosyncratic input, 2) language dominance and 3) structural overlap/ambiguity. Each of these explanations is considered in turn. Idiosyncratic input could account for some cases of cross-linguistic transfer. Paradis and Navarro (2003) reported that a SpanishEnglish bilingual child produced more overt subjects in Spanish than two monolingual Spanish-speaking children of the same age. This could be seen as inuence from English, a language that requires overt subjects. However, both parents of the bilingual child also produced more overt subjects than did the mothers of the monolingual children. Bilingual children are often more procient in one of their languages than another (Schlyter, 1993; Pearson et al., 1997). A bilingual childs more procient language is sometimes referred to as his/her dominant language (Genesee, Nicoladis and Paradis, 1995). Children might be more likely to incorporate structures from their dominant into their weaker language, than vice versa (D opke, 1998; Yip and Matthews, 2000; Bernardini, 2003). For example, Yip and Matthews (2000) found evidence of transfer from Cantonese to English in a Cantonese English bilingual child during a period when he was dominant in Cantonese; transfer was evident in three areas: whin-situ interrogatives, null objects, and relative clauses. It should be noted that in some cases of cross-linguistic transfer, dominance or degree of prociency in the target language cannot account for the pattern of results (M uller, 1998; M uller and Hulk, 2001; Nicoladis, 2002a). A third possible explanation of transfer is when one language has two possibilities (ambiguity) and shares one of those possibilities with the other language (overlap; D opke, 1998; Hulk and M uller, 2000; M uller and Hulk, 2001). The implicated structures could exhibit their overlap and ambiguity either at the level of the surface structure (D opke, 1998) or in the underlying abstract categories of the languages (Hulk and M uller, 2000;

Transfer in AN strings FrenchEnglish bilingual childrens adjectivenoun strings. Before turning to specic predictions for this study, I briey describe English and French adjective noun strings and review the literature on the acquisition of these strings. Order of adjectivenoun strings in English and French and acquisition Describing the order of adjectivenoun strings depends on whether the focus is on the surface realization of order or on the underlying structure. In this section, I describe rst the surface realization of adjectivenoun strings in English and French and then summarize arguments for the underlying structure. With a few exceptions, simple English adjectives appear pre-nominally (e.g. big dog, green tree). The exceptions include strings with some quantiers (e.g. something big, anyone new; cf. a big something) and a few adjectives, usually borrowed from French (e.g. knights errant, boyfriend extraordinaire). Heavy or modied adjectives can appear post-nominally (e.g. the pony faster than the zebra or the road less traveled by). In French, there is a small number of adjectives, most with high token frequency, that are usually placed pre-nominally (e.g. grand big/tall, petit little, beau beautiful, nouveau new, vieux old; as in grand arbre lit. tall tree). However, most adjectives typically appear post-nominally (e.g. chien noir lit. dog black, livre ouvert lit. book open). It should be noted that almost any French adjectives can be used grammatically in both pre-nominal and post-nominal positions. The change from the typical position results in a change in the nuanced meaning of the adjective and occurs only rarely when a speaker wishes to emphasize the correct nuance of his/her meaning. When an adjective that is typically used in a post-nominal position is used in a pre-nominal position, the result is often a change to a metaphoric or abstract interpretation (e.g. in la noire cr eature du diable the black creature of the devil the word noire takes on the meaning of evil in addition to its literal meaning of black). When an adjective that is typically used in a pre-nominal position is used in the post-nominal position, the speaker usually wants to focus on a concrete meaning of the adjective (e.g. in la personne grande the person big, the word grande might mean only tall and would not have the nuance of either adult or important, as the pre-nominal order could). There is no research on when French-speaking children master the association between the change from the typical position with a change in meaning. For the purpose of this study, I have assumed that this knowledge is acquired after the preschool years. As for the underlying structure of adjectivenoun strings, Kayne (1994) argued that adjectives are always

17

generated in pre-nominal position in all languages. If this were true, then the surface position of simple English adjectives corresponds to the underlying structure. In French, however, adjectives that appear post-nominally must undergo noun-raising (see also Granfeldt, 2000; Bernstein, 2001, for further arguments that French adjectives are generated pre-nominally). Note that this situation would apply to most French adjectives, since most adjectives typically appear post-nominally in French. To the extent that the underlying structure impacts on acquisition, French-speaking children might make more reversals from the typical surface order with adjectives that typically appear post-nominally (so they might say le noir chien the black dog) than they do with adjectives that typically appear pre-nominally (e.g. le chien grand the dog big). There is little research on childrens acquisition of the order of adjectivenoun strings. Most acquisitional research has been on childrens understanding of the semantics of adjectives and/or how the syntactic frame determines childrens understanding of the semantic category of an adjective (e.g. Hall, 1994; Hall and Moore, 1997; Klibanoff and Waxman, 2000). However, the extant evidence on childrens spontaneous speech has shown that children are usually accurate in their ordering of adjectives and nouns. For example, Brown (1973) observed that English-speaking children used adjectives and nouns in the correct order from the age of two years on, except when the copula was thought to be missing in a sentence. Bernardini (2003) summarizes evidence suggesting that children who speak French might use pre-nominal adjectives earlier in development than postnominal-adjectives. This nding might be expected, given the high token frequency of pre-nominal adjectives (compare Marcus, Pinker, Ullman, Hollander, Rose and Xu, 1992, who show that English-speaking children use irregular past tense verbs earlier than regular past tense verbs, perhaps because of the high token frequency of irregular verbs in English). There are some hints in the literature that bilingual childrens acquisition of two overlapping/ambiguous adjectivenoun orders might result in some errors. In the following examples, all the children were learning a Germanic language (Swedish, German or English) in which the typical position of adjectives is pre-nominal and a Romance language (Italian or French) in which the postnominal position is considered typical but the pre-nominal position is allowed. Schlyter (2001) mentioned that one FrenchSwedish bilingual child occasionally produced adjectivenoun reversals in Swedish spontaneous speech (i.e. toward the typical nounadjective French order), although she did not report the frequency of the reversals. Volterra and Taeschner (1978) reported that an ItalianGerman bilingual made a few errors in ordering her German adjectives, toward the default post-nominal

18

Elena Nicoladis The present study As noted above, the purpose of the present study is to test the hypothesis that structural overlap/ambiguity leads to cross-linguistic transfer. While D opke (1998) and M uller and Hulk (2001) developed their ideas within different theoretical frameworks, both versions generate the same predictions for the present study. I briey explain how. If the surface level is crucial, then English adjectives consistently appear pre-nominally while in French, most adjectives appear post-nominally but some adjectives appear with high frequency in pre-nominal position. Both languages provide children with evidence for pre-nominal adjective usage but, in French, there will be competition with the post-nominal position. If overlap/ambiguity predicts transfer, then FrenchEnglish bilingual children should produce more reversals in French than monolingual children (with post-nominal adjectives; e.g. un blanc chien a white dog where the typical order is un chien blanc a dog white). The same prediction results if the level of underlying structures is crucial. If all languages generate adjectives pre-nominally (Kayne, 1994), then French-speaking children produce any post-nominal adjectives through N-raising. French English bilingual children should raise fewer nouns, resulting in more frequent reversals with post-nominal adjectives (e.g. un blanc chien a white dog), in French than monolinguals. Both versions also predict that there should be no difference in the rate of reversals of pre-nominal adjectives in French by bilingual and monolingual children. Finally, bilingual children should not reverse English adjectives any more than English monolingual children. Both versions of the overlap/ambiguity hypothesis suggest that the difference between bilinguals and monolinguals will be in terms of the rate of use (D opke, 1998; M uller and Hulk, 2001). Because these hypotheses are cast in terms of comparisons with monolingual children, it is important to compare bilinguals and monolinguals. For English monolinguals, it seems unlikely that they would produce many reversals, given the consistent evidence that they receive about prenominal adjective placement (Bates and MacWhinney, 1989) and/or the correspondence of the surface order with the underlying order (Kayne, 1994). French monolinguals might make more reversals with postnominal adjectives than with pre-nominal adjectives if adjectives are underlyingly pre-nominal and they have not fully acquired N-raising (Kayne, 1994). Alternatively if French monolinguals are sensitive to the type frequency of post-nominal adjectives (see Bybee, 1995), they might make more reversals with pre-nominal adjectives than with post-nominal adjectives (i.e. overregularize). The children were also given a vocabulary test in order to estimate their language development relative

Italian order.1 Bernardini (2003) pointed out that a Swedish-dominant SwedishItalian bilingual made a few errors in Italian, notably using adjectives that could appear in either order in the incorrect order as dened by the pragmatic context. In these three studies, the children only produced a few examples of adjectivenoun strings. In another study in which the children produced a larger number of adjectivenoun strings, FrenchEnglish bilingual children ordered adjectives correctly over 90% of the time in spontaneous speech in both languages from at least the age of two and a half and older (Nicoladis, 1999, 2002b). For FrenchEnglish bilingual children, the correct ordering holds true even for codemixed utterances; that is, they follow adult norms in ordering adjectives relative to nouns from the age of two years on (Paradis, Nicoladis and Genesee, 2000). On the basis of these studies, it would seem that the acquisition of either one or two adjectivenoun orders might lead to a few errors, but generally is not difcult for children. The conclusion that children acquire the order of adjectivenoun strings relatively easily rests on data collected from spontaneous speech (see discussion in M uller and Hulk, 2001). There is some evidence that data from spontaneous speech may overestimate childrens abilities. For example, when Nicoladis (2002b) collected data from eight FrenchEnglish bilingual children on the ordering of novel nounnoun compounds (like ower socks), both from spontaneous speech and an elicitation task, she found that, in spontaneous speech, an average of 10% of English compounds were misordered while in the elicitation task, an average of 51% of compounds were misordered. In spontaneous speech, children can choose what to say and what not to say so they may avoid difcult strings. This is not to say that experimental tasks measure childrens abilities more accurately. In fact, such tasks may underestimate childrens abilities because they rely on what children do on a small number of items in short period of time. To determine childrens abilities, then, it is necessary to have evidence from both spontaneous speech and a variety of experimental tasks. As we already know that childrens production of adjectivenoun ordering is often correct in spontaneous speech, we now need to know what they do in an experimental task. In this study an elicitation task was used to assess childrens ordering of adjectives and nouns.

Volterra and Taeschner (1978) also reported several apparent incorrect orderings of Italian adjectives as well, one of which Bernardini (2003) cites as ungrammatical. My consultation with a native Italian speaker suggests that, like in French, both orders are potentially grammatical in Italian. Because Volterra and Taeschner (1978) did not give the context of the utterances, it might be hasty to conclude that the childs Italian adjectives were misordered.

Transfer in AN strings Table 1. Background characteristics of French monolinguals, age-matched bilinguals and vocabulary-matched bilinguals
French monolinguals Number Age range in years;months Average age in months (SD) Average French PPVT score

19

Age-matched bilinguals 10 3;35;1 47.7 (5.8) 51.6 (19.2)

Vocabulary-matched bilinguals 10 4;04;10 54.6 (3.0) 66.4 (13.7)

10 3;35;0 46.9 (6.1) 67.7 (11.9)

PPVT = Peabody Picture Vocabulary Test: Raw score.

to the other children in the study. Vocabulary size can be an indicator of prociency in each language (Genesee et al., 1995). By correlating vocabulary scores with reversals of adjectivenoun strings, we can explore whether prociency within a language might be related to childrens reversals. Also, by assigning the children a dominant language depending on their higher vocabulary score, we can look for possible effects of dominance on reversals in both languages.

Methods Participants Children were included in this study only if they were judged to be typically developing by their parents and/or teachers. Thirty-ve FrenchEnglish bilingual children participated in this study: 10 boys and 25 girls. All of the children had been exposed regularly to both languages from at least the time they were 12 months of age. The range in ages was from 2;11 to 5;3 (M = 51.5 months, SD = 6.1). Forty-four English monolingual children were tested as a control group and we include data here from 35 children matched on age (the matching was done within four months of age; range from 3;3 to 5;7) to the bilingual children (M = 54.2 months, SD = 8.9). There were 23 boys and 12 girls in the monolingual English sample. There are few French monolingual children in the area in which this study was carried out (Edmonton, Alberta, Canada). Nevertheless, 10 additional children were identied by parents as French monolinguals, 5 boys and 5 girls, ranging in age from 3;3 to 5;0 (M = 46.9 months, SD = 6.1). When comparing the French monolinguals to the bilinguals, three comparisons are made: 1) with all of the bilinguals, 2) with an agematched sample of bilinguals (N = 10), and 3) with a vocabulary matched sample of bilinguals (N = 10). The background characteristics are summarized in Table 1. The age-matched sample was matched within three months of age. The vocabulary-matched sample was matched within 10 points on the raw score of the French version of the Peabody Picture Vocabulary Test (described

below). None of the same children were included in both the age-matched and the vocabulary-matched sample. By comparing the bilingual childrens performance to both an age-matched and a vocabulary-matched sample, we can test if the observed results are due to delay in language acquisition. For example, if an age-matched sample produces more reversals than monolinguals but a vocabulary-matched sample produces the same rate of reversals, it is possible that the difference between monolinguals and bilinguals is due to a difference in the rate of acquisition of one language (cf. M uller & Hulk, 2001).

Materials All children were administered the Peabody Picture Vocabulary Test III (PPVT; Dunn and Dunn, 1997). The PPVT is a standardized test of comprehension vocabulary and assesses childrens knowledge of a variety of different lexical categories, including nouns, verbs and adjectives. Form IIIA was given in English. A translated version of version IIIB the PPVT was given in French. The French version has not been standardized but the raw score correlates positively with age in monolingual Frenchspeaking children (Nicoladis, under review). To elicit adjectivenoun strings, the children were shown 20 pictures per language. Each picture was composed of a target element among elements that differed on the dimension of interest. For example, for the target element a sad dog, the picture consisted of a crying dog among a number of dogs jumping for joy (see Figure 1). The reason for including the differing dimension is that adjectives are often used to differentiate one object from another (Waxman and Klibanoff, 2000). For the French version, 8 adjectives that are typically placed pre-nominally and 12 that are typically placed post-nominally were chosen as the targets (summarized in Table 2; the target strings are in Appendix A). For the English version, translations of the French adjectives were used (with one exception where the target English version had the word purple for the French vert green), but with different nouns.

20

Elena Nicoladis

Figure 1. Picture for target Sad dog.

The result of this manipulation means that about half the target adjectives in English would be pre-nominal if they were in French and half would be post-nominal (summarized in Table 2; the target strings are in Appendix A). As will be seen, this fact turned out to be interesting for the analysis. The pictures were randomized once for each language and put in a binder. All children saw the pictures in the same order. Procedure Children were tested in a quiet space in their daycares or preschools by a native or uent speaker of the relevant language. Children were rst given the PPVT and then the adjectivenoun elicitation task.

For the elicitation task, the experimenter rst familiarized the children with each picture by showing each picture and saying Look at this! (in French Regarde c a!). Then the experimenter pointed out the objects in the picture that differed from the target string and named them. For example, for the target sad dog, the experimenter pointed out all of the jumping dogs and said These dogs are all happy. By familiarizing the children with the pictures, we hoped to bias children to notice how the target element differed from all of the others so they would produce the target adjective. Note that the experimenters phrasing did not give any hints as to the typical order of adjectives and nouns during the familiarization phase. After all 20 pictures had been seen once, the experimenter explained that she wanted the child to name a

Transfer in AN strings Table 2. Target adjectives in French and English

Pre-nominal-in-French English version Big Small Short Beautiful Long Old Fat Mean French version Grande Petits Courte Belle Long Vieille Gros M echantes Post-nominal-in-French English version Square Striped Full Broken Open Red Wet Closed Purple Sick Sad Happy French version Carr e Ray e Plein Bris e Ouverte Rouge Mouill e Ferm ee Vert Malade Triste Heureuse

21

particular thing in each picture. For each picture, she asked Whats that? (in French Cest quoi c a?), pointing to the target element (such as the sad dog for the picture in Figure 1). The childrens name for the target element of each picture was written down on the spot. If the childrens answer did not include both an adjective and noun, the experimenter probed again by asking Whats that one? (in French Cest quoi celui-ci?), pointing to the target object. Even if the second answer did not include both an adjective and noun, the experimenter went on to the next picture. When all the pictures were named, the experimenter thanked the child for his/her participation.

Coding The children did not always produce the target strings (see Appendix A). Many of the childrens responses were a single noun or a single adjective (49% of the English items presented to the bilingual children, 53% of the French items to the bilingual children, 19% of the English items to the monolinguals and 30% of the French items to monolinguals). The analyses are only of the childrens adjectivenoun strings. The bilingual children produced signicantly fewer adjectivenoun strings in English (M = 7.1, SD = 5.9) than the English monolingual children (M = 14.8, SD = 6.0). This difference may have been due at least partially to the English vocabulary differences between the two groups. For the bilingual children the number of adjectivenoun strings produced was positively but not signicantly correlated with their PPVT score, r (33) = .277, p < .10. For the monolingual children there was a signicant and

positive correlation between the number of adjective noun strings produced and PPVT score, r (33) = .500, p < .01. The correlation between vocabulary scores and the number of adjectivenoun strings could be due to the fact that as childrens vocabulary increased, they could include more adjectives in multiple-word strings. In French, the monolingual children produced an average of 10.0 (SD = 5.5, N = 10) adjectivenoun strings and the bilingual children an average of 5.9 (SD = 5.3, N = 35). The bilingual childrens French PPVT scores were positively and signicantly correlated with the number of adjectivenoun strings, r (33) = .577, p < .01. There was no correlation between the monolingual childrens French PPVT scores and the number of adjectivenoun strings produced, r (8) = .119, ns. The children did not always produce the expected adjective (or the noun). For this reason, a coding system was devised for the adjectives the children actually produced (see Appendix B). Every different adjective and noun combination was counted as one adjectivenoun string, where the English word one was counted as a noun. For example, a child who said both big giraffe and big caterpillar was credited with two adjective noun strings. It was necessary to count this way because many children produced the same adjective in different positions (e.g. one child said both the mouse mad and a mad sh). If a child used the same words to describe several pictures (e.g. a big one), the child was credited with only one adjectivenoun string. Note that the variation in childrens responses to different pictures made an item analysis impossible. The children did not produce any modied adjectives in English. All modied adjectives in French (e.g. un cochon tout content a pig all happy or des cochons pas maigres some pigs not thin) were excluded from the analysis because adults often use modied pre-nominal adjectives post-nominally. This study was not designed to test childrens knowledge of this rule. The dependent variable is the number of reversals out of each childs total number of adjectivenoun strings produced, multiplied by 100 to create a percentage of adjectivenoun strings reversed. For English adjective noun strings, any post-nominal use of an adjective was considered a reversal (e.g. a book open). One child used an adjective in the middle of a word (i.e. a garbage yucky can). That item was excluded from the analysis. For French adjectivenoun strings, a single default position was assigned for each French adjective used by the children, based on how frequently that adjective occurs pre-nominally or post-nominally in child-directed French, as decided in consultation with native and uent French-speaking adults (see Appendix B). This forced categorization was done in spite of the fact that both pre-nominal and post-nominal positions can be grammatical for many French adjectives (e.g. one child

22

Elena Nicoladis
70 60 % reversals 50 40 30 20 10 0 Monolingual (French) Bilingual (French: all) Bilingual (French: age-match) Bilingual (French: PPVT-match)

Figure 2. Average rate of reversals of all French adjectives.

said des sapins gros some pine-trees big/fat when the contrasting trees in the picture were short; if the child meant to emphasize the concrete interpretation of bigness, this utterance could be considered grammatical). A reversal was coded when an adjective was used in the order that deviated from our assigned default position (e.g. une personne grand a person big). Note that, like in the last example, children sometimes made gender (or number) agreement errors that were disregarded in the analyses for this study. The result of the forced categorization of adjectives as either pre- or post-nominal is that some of the childrens French productions were counted as reversals, even though they might have been grammatical. This coding scheme may underestimate childrens ability to switch adjective order to emphasize a nuance of the meaning but allows for a straightforward categorization of childrens adjectivenoun ordering that does not rely on monolingual performance.2 Not all children produced adjectivenoun strings or both post-nominal and pre-nominal strings. These children are sometimes included in the analyses, if appropriate. Dominance groups To check for the effects of language dominance on reversals, I divided the children into three dominance groups on the basis of their vocabulary scores: French Dominant, Balanced and English Dominant. A child was classied as dominant in one language or the other if

the difference between the two vocabulary scores was 9 points or higher.3 The cut-off of 9 points was chosen somewhat arbitrarily and resulted in approximately equal sized groups. According to this classication, there were nine French Dominant children, 15 Balanced and 11 English Dominant. Results French Figure 2 summarizes the rate of reversals for all French adjectives. As can be seen, the bilingual children reversed more adjectives relative to nouns than the monolinguals. This difference did not reach signicance when all the bilingual children were compared to the monolinguals, t (38) = 1.68, p = .10, or if only the age-matched bilingual children were compared, t (18) = 1.91, p = .07. When only the vocabulary-matched bilingual children were compared with the monolingual children, the rate of reversals was signicantly higher for the bilingual children, t (18) = 2.42, p = .03. Figure 3 summarizes the rate of reversals for French adjectives by their default position. Here, the same pattern of results was observed in all three comparisons: all children made more reversals with pre-nominal adjectives (like une personne grand lit. a person big) than with post-nominal adjectives (like un ray e dinosaure lit. a striped dinosaur) and the bilingual children made more of both kinds of reversals than monolingual

Two observations are in order here. First, this categorization was made under the assumption that preschool children are sensitive to the token frequency of adjectivenoun ordering in their input and are not sensitive to the change in meaning associated with the change in adjective position. If this assumption is found to be incorrect, then a more sensitive coding scheme would be necessary. Second, recall that the hypotheses were phrased in terms of a comparison between bilingual and monolingual performance, so I could not rely on what the monolingual children did to categorize the French adjectives as pre- or post-nominal.

Note that this is a very crude way of estimating language dominance for at least two reasons. First, the vocabulary test only ESTIMATES a childs prociency within a language. Second, the French version of the vocabulary test we used is a translation of the English version. While this translated version correlates with age in monolingual children (Nicoladis, 2003), we have no guarantee that the scores on the French version are on the same scale as the scores on the English version. As the results will show no signicant effects by Dominance, it should be kept in mind that this could be due, at least in part, to an insensitive measure of Dominance.

Transfer in AN strings
100 90 80 70 60 50 40 30 20 10 0 Monolingual (French) Bilingual (French: all) Bilingual (French: agematch) Bilingual (French: PPVTmatch)

23

% reversals

Prenominal Postnominal

Figure 3. Average rate of reversals of French adjectives by default position.

Table 3. Correlations between French adjective reversals and age and vocabulary
% All adjectives reversed Bilingual (French) Age French PPVT English PPVT Age French PPVT .075 .207 .360* .136 .228 % Pre-nominal reversed (e.g. beau) .122 .235 .039 .114 .055 % Post-nominal reversed (e.g. ouvert) .180 .176 .303 .405 .011

Monolingual (French)

p < .05.

children. This pattern was observed when the bilingual children were compared to the monolingual children on a 2 2 [Group Adjective Position] ANOVA, with Adjective Position as a repeated variable. The rst analysis compared all the bilingual children with the monolingual children. There was a main effect for Adjective Position, F (1, 26) = 5.25, p = .03, and a main effect for Group, F (1, 26) = 6.73, p = .02. There was no interaction effect (F < 1). This same pattern of results was observed when the age-matched bilingual group was compared to the monolinguals. There was a signicant main effect of Adjective Position, F (1, 14) = 6.40, p = .02, and a signicant main effect of Group, F (1, 14) = 10.19, p < .01. There was no interaction effect, F (1, 14) = 1.67, p = .22. Finally, a similar pattern of results was observed for the vocabulary-matched bilinguals and the monolinguals. The main effect of Adjective Position did not reach signicance, F (1, 18) = 3.86, p = .07, but there was a signicant main effect of Group, F (1, 18) = 6.24, p = .02. There was no interaction effect (F < 1). The nal analysis on the childrens use of French adjectives investigated correlations between their rate of adjective reversals and age and vocabulary scores. These correlations are summarized in Table 3. The only correlations that reached signicance were between English vocabulary scores and all French adjective reversals. These correlations mean that the higher the

childrens English vocabulary score, the more likely they were to reverse French adjectives. English The rate of reversals of all English adjectives is considered rst. The bilingual children reversed an average of 12.4% (SD = 22.4%, N = 27) of their English adjectives and the monolingual children an average of 4.3% (SD = 16.2%, N = 34). This difference is signicant on a one-way ANOVA, F (1, 59) = 5.75, p = .02. Next, the rate of reversals in English adjectives depending on the usual placement of their translation equivalents in French (e.g. big was considered a prenominal adjective because grand usually goes before nouns in French) is considered. These results are summarized in Figure 4. To compare the two groups, a 2 2 [Group Positionin-French] ANOVA was performed, with Position-inFrench as a repeated measure. The results did not reach signicance for the main effect for Position-in-French, F (1, 54) = 3.22, p = .08, or the main effect for Group, F (1, 54) = 3.28, p = .08. The interaction did not reach signicance, F (1, 54) = 2.26, p = .14. The correlations between age and vocabulary and rate of reversals are next considered (see Table 4). The only signicant correlations were between English vocabulary

24

Elena Nicoladis
60 50 % reversals 40 30 20 10 0 Monolingual (English) Bilingual (English) Prenominal in French Postnominal in French

Figure 4. Average rate of reversals of English adjectives by default position in French.

Table 4. Correlations between English adjective reversals and age and vocabulary
% All adjectives reversed Bilingual (English) Age English PPVT French PPVT Age English PPVT .068 .452** .032 .116 .135 % Pre-nominal-in-French reversed .026 .192 .045 .134 .143 % Post-nominal-in-French reversed .007 .341* .152 .122 .136

Monolingual (English)

p < .01, p < .05.

Table 5. Bilinguals average (SD) percentage of reversals in French and English, by Dominance Group
French Dominant French English 26.0 (22.8) 19.3 (21.2) English Dominant 36.9 (44.3) 3.8 (6.9)

Balanced 27.2 (34.3) 14.8 (29.8)

in French than the Balanced or the French Dominant children and the French Dominant children made more reversals in English than the Balanced and the English Dominant children. These differences were not signicant for either English, F (2, 27) = 1.20, p = .32, or for French, F (2, 27) = 0.26, p = .77. Discussion Before turning to the results of this study in terms of crosslinguistic transfer, it should be noted that the bilingual children were always more often correct in their adjective placement than they were incorrect. This nding supports the conclusion that bilingual children differentiate their languages at a syntactic level in the preschool years (see de Houwer, 2002). The effects of knowing another language are clearly secondary to knowing the syntactic strings in each language. If cross-linguistic transfer were due to structural overlap/ambiguity (D opke, 1998; Hulk and M uller, 2000), FrenchEnglish bilingual children should produce more reversals of post-nominal French adjectives than monolinguals. For example, ray e striped typically l appears post-nominally in French (e.g. un e ephant ray e lit. an elephant striped) so bilingual children might say l un ray e e ephant (lit. a striped elephant) more often than monolingual children. This study has shown that

scores and the rate of all adjective reversals as well as the post-nominal-in-French adjectives. As would be expected, these correlations are negative, meaning that the greater the English vocabulary, the less likely bilingual children were to produce reversed adjectivenoun strings.

Bilinguals English and French and Dominance Effects There was no correlation between the rate of reversals in English and French, r (32) = .075, ns. The lack of correlation held true for both pre-nominal adjectives, r (32) = .215, ns, and post-nominal adjectives, r (32) = .105, ns. Table 5 summarizes the average rates of reversals by language and Dominance Group. As can be seen in this table, the English Dominant children made more reversals

Transfer in AN strings this was indeed the case. The bilingual children reversed more post-nominal adjectives in French (M = 20%) than monolinguals (M = 1%). Also, in line with the predictions from the structural overlap/ambiguity hypothesis, the bilingual children were more likely to make adjective reversals in French, a language with two adjective noun orders, (M = 27%) than in English, with only one (M = 12%). This could have been due to the fact that there are two possible orders for French adjectivenoun strings and only one in English. Note that the bilinguals high rate of reversals in French could not be attributed to dominance or delay in acquisition (cf. M uller and Hulk, 2001), at least as dominance4 and delay were operationalized here. If the high rate of reversals had been due to delay, the vocabulary-matched group should have produced a similar rate of reversals to the monolinguals. In fact, the vocabulary-matched group produced more reversals than the monolinguals. There were, however, two ways in which the bilingual children in this study differed from monolingual children that were not predicted by the structural ambiguity of the two languages: 1) the bilingual children made more reversals of pre-nominal French adjectives than monolingual children (e.g. une personne grand lit. a person big) and 2) the bilingual children made more reversals of English adjectives than monolingual children, particularly those adjectives that, if they were in French, would be placed post-nominally (e.g. a monkey purple). Each of these ndings is discussed in turn. The bilingual children made more reversals of prenominal French adjectives (e.g. une personne grand lit. a person big) than monolingual children. It should be noted, however, that this type of reversal was also more frequent than post-nominal adjective reversals among the monolingual children.5 It is quite possible that the coding scheme used for this study (Appendix B) was too harsh (see footnote 2). A more sensitive coding scheme that would test childrens sensitivity to the effects of order
4

25

There are several reasons why this study cannot absolutely rule out language dominance as playing a role in cross-linguistic transfer. First, it is possible that the effect of dominance is linear rather than categorical. This would explain why there were correlations between the bilinguals English vocabulary and the rate of reversals in both English and French. Second, it is possible that vocabulary scores, the measures used to estimate dominance here, are not strongly related to childrens meaningful production of syntactic constructions. Third, there was a fairly small number of children in each dominance group, so there may not have been enough power in the sample to show the effect of dominance. It is possible that the effects of dominance could be seen in a study with more children and with a more valid measure of dominance. This nding could pose a challenge to Kaynes (1994) argument that adjectives are underlyingly pre-nominal. Naturally, French-speaking children could acquire N-raising early in acquisition (see Bernardini, 2003, for just such an argument about Italian-speaking children). It is not clear why bilingual children would raise nouns more frequently than monolingual children.

changes on meaning could clearly result in different results. Nevertheless, it is interesting that bilingual children would choose the post-nominal option (i.e. opposite to English) for pre-nominal adjectives more often than the monolingual children.6 It is possible that at least some of the reversals of the pre-nominal adjectives can be conceptualized as overregularizations. That is, even monolingual children pass through a stage when they place a lot of pre-nominal adjectives in the default (i.e. high type frequency) post-nominal position. Overregularizations toward a default are commonly reported in child language acquisition (e.g. Bybee, 1995). The curious nding here is that the bilingual children made this kind of overregularization more often than monolingual children. This fact cannot be directly attributed to the bilinguals knowledge of English because to make these overregularizations, the bilingual children were using the opposite order to English. Even when vocabulary was controlled for (i.e. in the vocabularymatched group), the bilingual children still made more pre-nominal adjective reversals than monolinguals. For that reason, the greater rate of overregularization is unlikely to be due to frequency differences in exposure to French compared to monolingual children (cf. Patterson and Pearson, 2004). A more likely interpretation of the high rate of overregularizations is that this is a property of learning two languages with two different default orders. In other words, the bilingual children may have tried to make cleaner rules about the two languages than monolinguals in order to simplify the task of knowing two languages. That is, they may have tried to make a simple pre-nominal adjective placement rule about English and a simple post-nominal adjective placement rule about French. If this interpretation is correct, it is, to my knowledge, the rst report of such a strategy in the literature (although see Gawlitzek-Maiwald and Tracy, 1996, for arguments of strategic acquisition by a bilingual child). If this reasoning is correct, then this effect is not so much due to cross-linguistic transfer (i.e. the effect of English on French) but rather an effect of bilingualism more generally (i.e. the effect of knowing two languages with different rules). The second result that contradicted the predictions from the structural overlap/ambiguity hypothesis was that the bilingual children tended to make more reversals of English adjectives than monolingual children, particularly those adjectives that, if they were in French, would be placed post-nominally. That is, the bilingual children were more likely to make reversals with adjectives like closed and green that, in French, typically appear post-nominally
6

The bilingual children did not use the post-nominal position more overall compared to the monolingual children in French. The average rate of post-nominal adjective placement across all adjectives was 45% for the bilingual children and 46% for the monolingual children.

26

Elena Nicoladis transfer is not limited to cases of structural overlap (Yip and Matthews, 2000; Nicoladis, 2002a). However, as this study has shown, transfer may be more likely when there is structural overlap. Fourth, this study has shown that the typical placement of a particular word in one language can affect the placement in another language. It was argued here that that must be due to a single concept underlying words from the two languages. If this interpretation is correct, then it is also possible that pragmatic and/or conceptual variables might limit where cross-linguistic transfer can occur (e.g. there might be no transfer where there is structural overlap/ambiguity in two languages but the two structures have different meanings or are used in different pragmatic contexts; cf. Hulk and M uller, 2000). Fifth, as noted in the introduction, cross-linguistic transfer can occur at a phonological level, as well as a syntactic level (e.g. Paradis, 2001). An explanation of cross-linguistic transfer that could account for phonological transfer as well as syntactic transfer would clearly be preferable to one that could only account for the latter. I propose to consider cross-linguistic transfer as a manifestation of speech production errors, in other words an epiphenomenon of speech production. In the next section, I present the outlines for a possible model of crosslinguistic transfer in terms of a speech production model, based loosely on other models (Dell, Chang and Grifn, 1999; Dell, Reed, Adams and Meyer, 2000; Ferreira and Dell, 2000; Costa, 2004). It is not necessarily my intention here to provide the correct model, but rather to demonstrate that it is theoretically possible to account for cross-linguistic transfer in bilingual children using a model of speech production. I will show how such a model could explain the present results. Cross-linguistic transfer as an epiphenomenon of speech production The rst phase of speech production according to many models is the message (or the concepts) that the speaker wishes to convey. During this phase, I assume that a speakers knowledge about the world constrains the kinds of concepts that are considered in combination (see Medin and Shoben, 1988; Murphy, 2002). There is no consensus as to whether the choice of a bilinguals languages affects the speakers message choice (Costa, 2004). At the second phase (sometimes referred to as the lemma stage; Ferreira and Dell, 2000), a speaker chooses the specic words and the syntactic framework in which the words will appear. At this phase, the language of the words and the syntactic framework would have to be chosen. There is some evidence that there is interaction between the specic words and the syntactic framework (Ferreira and Dell, 2000), suggesting that the lexical and the conceptual levels are co-determined. If there are multiple

than with adjectives like big and beautiful that, in French, typically appear pre-nominally. This result would suggest that childrens knowledge of the typical placement of the same-meaning adjective affects their placement in the other language. This effect cannot be at the syntactic level (i.e. all the English adjectives used here are prenominal) and is unlikely to be at the lexical level (i.e. why would simultaneous bilingual children have a direct association between vert and green? cf. Kroll and Dussias, 2004, for why sequential bilinguals might). It seems likely that there is a connection at the conceptual level (e.g. the meaning of green). If this interpretation is correct, then conceptual knowledge associated with a single word can affect syntactic realization in childrens language production (cf. Hulk and M uller, 2000). Before turning to a more general discussion about what we know about cross-linguistic transfer, I would like to point out that the English monolinguals did make some reversals in the production of adjectivenoun strings (M = 4%). This result seems surprising, given Englishspeaking childrens extensive evidence that adjectives are placed pre-nominally (see also Kayne, 1994). The children were all around three years of age or older so their reversals are unlikely to be due to copula omission (cf. Brown, 1973). It is possible that the fact that English adjectives can occur post-nominally (e.g. the dog bigger than a German Shepherd) affects childrens ideas of simple adjective placement. It is also possible that childrens acquisition of lexical categories like noun and adjective is incomplete and/or imperfectly mapped onto words in use. For example, children might not distinguish between Thats a cow dancing and Thats a dancing cow. Also, nouns are frequently used as modiers in compounds in English (e.g. police car). Monolingual English-speaking children make some reversals in novel nounnoun compounds (e.g. chair owers to refer to a chair with owers on it; Nicoladis, 2002a). Some of the childrens reversals in this study could be due, then, to a failure to distinguish perfectly between nouns as modiers and adjectives (cf. Nicoladis, 2002b). In any case, the monolinguals performance in this study reinforces the point that an identication of cross-linguistic transfer requires comparison with monolinguals (M uller and Hulk, 2001). This study and others (e.g. D opke, 1998; Hulk and M uller, 2000) have shown a number of variables affecting cross-linguistic transfer in preschool children. First, crosslinguistic transfer cannot explain the majority of what bilingual children do with their two languages the majority of what bilingual children generally do is correct in both of their languages. Second, cross-linguistic transfer has been reported in speech production rather than comprehension (Nicoladis, 2003). Third, structural overlap is predictive of morphosyntactic transfer (D opke, 1998; Hulk and M uller, 2000), although cross-linguistic

Transfer in AN strings
CONCEPTS

27

GREEN

APPLE

green

apple

vert

pomme

SYNTAX English: A-N French: N-A French: A-N

PHONOLOGY

Figure 5. Schematic diagram of model of concept and lemma-level effects in English (post-nominal in French) results.

possible syntactic frameworks that could convey roughly the same message, there is competition between those frameworks (sometimes resulting in errors). The third and nal phase of speech production is the choice of the phonological frame. The phonological frame species which consonants are legal onsets and codas (Dell et al., 1999). The phonological form can be thought of as somewhat separate from the lemma level, although there can be interactions between the two (Dell et al., 1999). The results of the present study are mostly concerned with the lemma level of speech production. At any point when a speaker has to make a choice, there is likely to be competition, with some activation of the other options that were not chosen by the speaker (Dell et al., 1999). At the lemma level, a bilingual must choose words from the appropriate language as well as from the appropriate syntactic frame. For the present study, there would be competition between the syntactic frames for English adjectivenoun strings as well as both French options. Figure 5 shows what the activation might be like for a child who wished to produce green apple (the translation for green, vert, is a post-nominal adjective). The thicker lines indicate greater activation. This gure shows that when activating green and apple in English, both vert and pomme apple are also activated, although to a lesser

extent. For this reason the syntactic frame for a postnominal adjective in French also gets some activation. The result is that children are most likely to produce the correct order, but, when they make errors, are more likely to produce reversals for English adjectives that would be post-nominal in French. Figure 6 shows a schematic diagram of the activation to produce une pomme verte lit. an apple green. For bilingual children, the choice of the concepts GREEN and APPLE would activate both English and French lexicalizations of these concepts, although the activation for French would be greater because that is the target language. The French adjective depicted in Figure 6 is postnominal. If it had been a French pre-nominal adjective that had been chosen, both French syntactic frames would be activated because the post-nominal adjectives have the higher type frequency frame and frequency is thought to lead to greater activation (Dell et al., 1999). This would then explain why French-speaking children all make more reversals with pre-nominal adjectives than with postnominal adjectives. Note that this kind of model does not explain why the bilingual children in this study made more reversals with French pre-nominal adjectives than monolingual children.

28

Elena Nicoladis
CONCEPTS

GREEN

APPLE

vert

pomme

green

apple

SYNTAX French: N-A English: A-N French: A-N

PHONOLOGY

Figure 6. Schematic diagram of model of concept and lemma-level effects in post-nominaladjective French results.

As noted earlier, this nding could be due to a strategy of learning rules in two languages and so would have to do with the instantiations of the syntactic frames in the model. While the schemas outlined here may not yet be correct in their detail. Nevertheless, this kind of speech production model could explain the fact that when there are two syntactic rules in one language, cross-linguistic transfer is more likely to be observed in preschool children than when there is only one (e.g. D opke, 1998; Hulk and M uller, 2000; the French results in this study). That is because there is competition between the two rules at the lemma level. The fact that cross-linguistic transfer can be observed even when there is only one option in each language (e.g. Yip and Matthews, 2000; Nicoladis, 2002a; the English results in this study) suggests that there is competition between two languages syntactic rules at the lemma level. Furthermore, a more detailed speech production model could potentially account for phonological transfer by the competition of similar sounds

both within and between languages (cf. Dell et al., 1999, 2000). A number of predictions could be made about crosslinguistic transfer with this account. For example, all competing properties assigned at the lemma level should be subject to cross-linguistic transfer (e.g. grammatical gender, syntactic framework). Another prediction can be made from the nding that recent practice can affect speech production errors (Dell et al., 2000). For that reason, cross-linguistic transfer should be affected by recent language choice (see Grosjean, 2001; see also Volterra and Taeschner, 1978). In this study, for example, the bilingual children were all tested in French daycares and preschools. It is possible that the degree of crosslinguistic transfer in English observed in this study was partially due to their recent exposure to French. This interpretation could be tested by systematically varying the exposure to a particular language before testing for cross-linguistic transfer. Another prediction is that the same pattern of cross-linguistic transfer observed in

Transfer in AN strings children should also be observed in highly procient bilingual adults, if the task demands are increased for adults (see, for example, Bybee and Slobin, 1982, for a monolingual example). If these effects by age are to be observed, they will probably be observed beyond the age range of three to ve years, as there was no correlation between age and the rate of reversals within that age range in the present study. One nal prediction that can be made here is that phonological transfer should be more likely in words that 1) have similar meanings and/or 2) sound similar in the two languages.

29

In closing, it is interesting to consider the possibility that cross-linguistic transfer could be an epiphenomenon of speech production. If this explanation is correct, then cross-linguistic transfer might be manifested in comprehension as well, but perhaps in different ways than in production (cf. Nicoladis, 2003; see, for example, Jared and Kroll, 2001; see also Dell et al., 1999). A model of cross-linguistic transfer in terms of speech production would clearly need to be spelled out in more detail. Nevertheless, there are testable predictions that can be made even with the outline of a model proposed here.

Appendix A. Target strings

French version Pre-nominal Une grande abeille Des petits sapins Une courte queue Une belle madame Un long crayon Une vieille femme Des gros cochons Des m echantes araign ees Post-nominal Un bouton carr e Un dinosaure ray e Un verre plein Une ampoule bris ee Une bouche ouverte Un cheval rouge Un chien mouill e Une poubelle ferm ee Un e ephant vert l Un chat malade Une banane heureuse Un chien triste English version Pre-nominal-in-French A big ower A small giraffe A short caterpillar A beautiful buttery A happy dog A long snake An old man A fat sh A mean mouse A sad penguin Post-nominal-in-French A square globe A striped elephant A full garbage A broken pencil An open book A red cat A wet boy Closed scissors A purple monkey A sick dragon

Appendix B. Adjectives used by children and codes

French adjectives Beau Bleu Bon Bris e Carr e Cass e Content Enorme F ach e Ferm e F eroce Gentil Grand Gros Jaune Code Pre-nominal Post-nominal Pre-nominal Post-nominal Post-nominal Post-nominal Post-nominal Post-nominal Post-nominal Post-nominal Post-nominal Pre-nominal Pre-nominal Pre-nominal Post-nominal English adjectives Angry Bad Beautiful Big/Bigger Blue Broken Brown Closed Circle Crying Dry Fat Flat Full/Fulled Giant Code Post-nominal-in-French Pre-nominal-in-French Pre-nominal-in-French Pre-nominal-in-French Post-nominal-in-French Post-nominal-in-French Post-nominal-in-French Post-nominal-in-French Post-nominal-in-French Post-nominal-in-French Post-nominal-in-French Pre-nominal-in-French Post-nominal-in-French Post-nominal-in-French Pre-nominal-in-French

30

Elena Nicoladis

Appendix B. Continued
French adjectives Jeune Long Maigre Malade M echant Mouill e Noir Ouvert Petit Plein Ray e Rempli Rose Rouge Sec Triste Vert Vide Vieux Vivant Code Pre-nominal Pre-nominal Post-nominal Post-nominal Pre-nominal Post-nominal Post-nominal Post-nominal Pre-nominal Post-nominal Post-nominal Post-nominal Post-nominal Post-nominal Post-nominal Post-nominal Post-nominal Post-nominal Pre-nominal Post-nominal English adjectives Good Green Happy Healthy Huge Humungous Laughing Little Long Lousy Mad Mean Nice Old/Older Open/Opened Orange Pink Pretty Purple Raining Red Round Sad Sick Silly Shut Short Striped/Stripey Tall Ugly Wet White Young Code Pre-nominal-in-French Post-nominal-in-French Pre-nominal-in-French Post-nominal-in-French Post-nominal-in-French Post-nominal-in-French Post-nominal-in-French Pre-nominal-in-French Pre-nominal-in-French Post-nominal-in-French Post-nominal-in-French Pre-nominal-in-French Pre-nominal-in-French Pre-nominal-in-French Post-nominal-in-French Post-nominal-in-French Post-nominal-in-French Pre-nominal-in-French Post-nominal-in-French Post-nominal-in-French Post-nominal-in-French Post-nominal-in-French Post-nominal-in-French Post-nominal-in-French Post-nominal-in-French Post-nominal-in-French Pre-nominal-in-French Post-nominal-in-French Pre-nominal-in-French Pre-nominal-in-French Post-nominal-in-French Post-nominal-in-French Pre-nominal-in-French

References
Barlow, J. A. (2002). Error patterns and transfer in Spanish English bilingual phonological development. In B. Skarabela, et al. (eds.), Boston University Conference on Language Development 26 Proceedings, pp. 6071. Somerville, MA: Cascadilla Press. Bates, E. & MacWhinney, B. (1989). Functionalism and the Competition Model. In B. MacWhinney & E. Bates (eds.), The crosslinguistic study of sentence processing, pp. 157 193. Cambridge: Cambridge University Press. Bernardini, P. (2003). Child and adult acquisition of word order in the Italian DP. In N. M uller (ed.), (In)vulnerable

domains in multilingualism, pp. 4181. Amsterdam: John Benjamins. Bernstein, J. B. (2001). Focusing the right way in Romance determiner phrases. Probus, 13, 129. Bosch, L. & Sebasti an-Gall es, N. (2003). Simultaneous bilingualism and the perception of a language-specic vowel contrast in the rst year of life. Language and Speech, 46, 217243. Brown, R. (1973). A rst language: The early stages. Cambridge, MA: Harvard University Press. Brulard, I. & Carr, P. (2003). FrenchEnglish bilingual acquisition of phonology: One production system or two? International Journal of Bilingualism, 7, 177202.

Transfer in AN strings
Bybee, J. (1995). Regular morphology and the lexicon. Language and Cognitive Processes, 10, 425455. Bybee, J. L. & Slobin, D. I. (1982). Rules and schemas in the development and use of the English past tense. Language, 58, 265289. Costa, A. (2004). Speech production in bilinguals. In T. K. Bhatia & W. C. Ritchie (eds.), The handbook of bilingualism, pp. 201223. Malden, MA: Blackwell Publishing. de Houwer, A. (2002). How different are monolingual and bilingual acquisition? Ilha do Desterro, 43, 127148. Dell, G. S., Chang, F., & Grifn, Z. M. (1999). Connectionist models of language production: Lexical access and grammatical encoding. Cognitive Science, 23, 517 542. Dell, G. S., Reed, K. D., Adams, D. R. & Meyer, A. S. (2000). Speech errors, phonotactic constraints, and implicit learning: A study of the role of experience in language production. Journal of Experimental Psychology: Learning, Memory, and Cognition, 26, (13551367). D opke, S. (1998). Competing language structures: The acquisition of verb placement by bilingual GermanEnglish children. Journal of Child Language, 25, 555584. Dunn, L. M. & Dunn, L. M. (1997). Peabody Picture Vocabulary Test, third edition. Circle Pines, MN: American Guidance Service. Ferreira, V . S. & Dell, G. S. (2000). Effect of ambiguity and lexical availability on syntactic and lexical production. Cognitive Psychology, 40, 296340. Gawlitzek-Maiwald, I. & Tracy, R. (1996). Bilingual bootstrapping. Linguistics, 34, 901926. Genesee, F., Nicoladis, E. & Paradis, J. (1995). Language differentiation in early bilingual development. Journal of Child Language, 22, 611631. Granfeldt, J. (2000). The acquisition of the determiner phrase in bilingual and second language French. Bilingualism: Language and Cognition, 3, 263280. Grosjean, F. (2001). The bilinguals language modes. In J. L. Nicol (ed.), One mind, two languages: Bilingual language processing, pp. 122. Oxford: Blackwell. Hall, D. G. (1994). Semantic constraints on word learning: Proper names and adjectives. Child Development, 65, 12991317. Hall, D. G. & Moore, C. E. (1997). Red bluebirds and black greenies: Preschoolers understanding of the semantics of adjectives and count nouns. Journal of Experimental Child Psychology, 67, 236267. Holm, A. & Dodd, B. (1999). A longitudinal study of the phonological development of two CantoneseEnglish bilingual children. Applied Psycholinguistics, 20, 349376. Hulk, A. (1997). The acquisition of French object pronouns by a Dutch/French bilingual child. In A. Sorace, C. Heycock & R. Shillcock (eds.), Language acquisition: Knowledge, representation and processing. Proceedings of the GALA 1997 Conference on Language Acquisition, pp. 521526. Edinburgh: University of Edinburgh. Hulk, A. & M uller, N. (2000). Bilingual rst language acquisition at the interface between syntax and pragmatics. Bilingualism: Language and Cognition, 3, 227244. Hulk, A. C. J. & van der Linden, E. (1996). Language mixing in a FrenchDutch bilingual child. In E. Kellerman, et al. (eds.),

31

Eurosla 6: A selection of papers, pp. 89103. Amsterdam: John Benjamins. Jared, D. & Kroll, J. F. (2001). Do bilinguals activate phonological representations in one or both of their languages when naming words? Journal of Memory and Language, 44, 231. Kayne, R. S. (1994). The antisymmetry of syntax. Cambridge, MA: The MIT Press. Klibanoff, R. S. & Waxman, S. R. (2000). Basic level categories support the acquisition of novel adjectives: Evidence from preschool-aged children. Child Development, 71, 649 659. Kroll, J. F. & Dussias, P. E. (2004). The comprehension of words and sentences in two languages. In T. K. Bhatia, & W. C. Ritchie, (eds.), The handbook of bilingualism, pp. 169 200. Malden, MA: Blackwell Publishing. Marcus, G. F., Pinker, S., Ullman, M. T., Hollander, M., Rose, T. J. & Xu, F. (1992). Overregularization in language acquisition. Monographs of the Society for Research in Child Development, 57 (4, Serial No. 228). Medin, D. L. & Shoben, E. J. (1988). Context and structure in conceptual combination. Cognitive Psychology, 20, 158 190. Meisel, J. (1983). Transfer as a second language strategy. Language and Communication, 3, 1146. M uller, N. (1998). Transfer in bilingual rst language acquisition. Bilingualism: Language and Cognition, 1, 151171. M uller, N. & Hulk, A. (2001). Crosslinguistic inuence in bilingual language acquisition: Italian and French as recipient languages. Bilingualism: Language and Cognition, 4, 121. Murphy, G. L. (2002). The big book of concepts. Cambridge, MA: MIT Press. Nicoladis, E. (1999). Where is my brush-teeth? Acquisition of compound nouns in a bilingual child. Bilingualism: Language and Cognition, 2, 245256. Nicoladis, E. (2002a). Whats the difference between toilet paper and paper toilet? FrenchEnglish bilingual childrens crosslinguistic transfer in compound nouns. Journal of Child Language, 29, 843863. Nicoladis, E. (2002b). The cues that children use in acquiring adjectival phrases and compound nouns: Evidence from bilingual children. Brain and Language, 81, 635648. Nicoladis, E. (2003). Cross-linguistic transfer in deverbal compounds of preschool bilingual children. Bilingualism: Language and Cognition, 6, 1731. Nicoladis, E. (under review). Acquisition of deverbal compounds by French-speaking preschoolers. Nicoladis, E. & Genesee, F. (1997). Language development in preschool bilingual children. Journal of SpeechLanguage Pathology and Audiology, 21, 258270. Paradis, J. (2001). Do bilingual two-year olds have separate phonological systems? International Journal of Bilingualism, 5, 1938. Paradis, J. & Genesee, F. (1996). Syntactic acquisition in bilingual children: Autonomous or interdependent? Studies in Second Language Acquisition, 18, 125. Paradis, J. & Navarro, S. (2003). Subject realization and crosslinguistic interference in the bilingual acquisition of

32

Elena Nicoladis
Progression and regression in language: Sociocultural, neuropsychological, and linguistic perspectives, pp. 289 308. Cambridge: Cambridge University Press. Schlyter, S. (2001). Pragmatic rules, C-domain, and language dominance. Bilingualism: Language and Cognition, 4, 40 42. Volterra, V. & Taeschner, T. (1978). The acquisition and development of language by bilingual children. Journal of Child Language, 5, 311326. Waxman, S. R. & Klibanoff, R. S. (2000). The role of comparison in the extension of novel adjectives. Developmental Psychology, 36, 571581. Yip, V. & Matthews, S. (2000). Syntactic transfer in a Cantonese English bilingual child. Bilingualism: Language and Cognition, 3, 193208. Received June 13, 2005 Revision received September 2, 2004 Accepted September 17, 2005

Spanish and English: What is the role of the input? Journal of Child Language, 30, 371393. Paradis, J., Nicoladis, E. & Genesee, F. (2000). Early emergence of structural constraints on code-mixing: Evidence from FrenchEnglish bilingual children. Bilingualism: Language and Cognition, 3, 245261. Patterson, J. L. & Pearson, B. Z. (2004). Bilingual lexical development: Inuences, contexts, and processes. In B. A. Goldstein (ed.), Bilingual language development and disorders in SpanishEnglish speakers, pp. 77104. Baltimore: Paul H. Brookes. Pearson, B. Z., Fern andez, S., Lewedag, V . & Oller, D. K. (1997). Input factors in lexical learning of bilingual infants (ages 10 to 30 months). Applied Psycholinguistics, 18, 4158. Pearson, B. Z., Fern andez, M. C. & Oller, D. K. (1993). Lexical development in bilingual infants and toddlers: Comparison to monolingual norms. Language Learning, 43, 93120. Schlyter, S. (1993). The weaker language in bilingual Swedish French children. In K. Hyltenstam & . Viberg, (eds.),