Chapter 3: Input in multilingual language acquisition
3.1 The role of input in bilingual language acquisition
3.1.1 Input quantity
3.1.1.2 Input quantity and grammatical acquisition
According to Blom (2010:422), the relevance of input has been established for lexical acquisition, but much less so for grammatical acquisition. If quantity of input does indeed influence grammatical acquisition, one would expect to find, as with lexical acquisition, evidence of delayed development in young bilinguals, as such learners are naturally exposed to less input in each of their languages than are monolingual learners of those languages (Blom, 2010:423). This prediction does not seem to be upheld by overviews in the literature (cf., for example, Genesee, 2001; Meisel, 2001; Nicoladis & Genesee, 1997; Paradis & Genesee, 1996), although it should be noted that the research focus has been almost exclusively on the dominant language in terms of input (Blom, 2010:423; Meisel, 2007b:496; Paradis, Crago, Genesee, & Rice, 2003:3).
Studies that do indicate a relationship between input quantity and grammatical development in bilingual children and that also investigate the weaker language in terms of input include, in chronological order, Schlyter (1993); Schlyter and Håkansson (1994); Gathercole (2002a, 2002b, 2002c); Meisel, 2007b; Paradis, Nicoladis and Crago (2007); and Blom (2010). Hoff, Welsh, Place and Ribot (in press), using English-Spanish participants; Meisel (2007b), using
52 German-French participants; as well as Schlyter (1993) and Schlyter and Håkansson (1994), using Swedish-French participants, all report a lower mean length of utterance in young bilinguals’ weaker language than in their stronger language, which suggests delayed development in the weaker language when compared to the stronger language. Paradis et al. (2007) used slightly older bilinguals (mean age = 4;9) speaking English and French to investigate the use of the regular and irregular past tense in both the stronger and weaker language in terms of input. The amount of input each child received in each of the two languages was calculated on grounds of their length of exposure since birth and their current exposure in the home and daycare/preschool environment. On grounds of the input amounts, participants were divided into two groups: an English-dominant group of 14 and a French- dominant group of 11. Results showed the English-dominant group to outperform the French- dominant group on both regular and irregular English verbs, and the French-dominant group to outperform the English-dominant group on both regular and irregular French verbs, indicating that input dominance does affect the acquisition of verb morphology.
Gathercole (2002a, 2002b, 2002c) used a grammaticality judgement task to test Spanish- English bilingual second- and fifth-graders’ knowledge of the mass/count distinction in English, grammatical gender in Spanish and that-trace structures in both Spanish and English. On all three measures, bilingual participants could not meet monolingual norms. Additionally, input quantity (estimated on grounds of home language, SES and medium of instruction (MoI) at school) was found to explain differences between the different bilinguals’ scores.
According to Blom (2010:424), the studies by Paradis et al. (2007) and Gathercole (2002a, 2002b, 2002c) only confirm the relevance of input to grammatical acquisition in older children, who might already be losing their innate ability to acquire grammar and therefore be relying more heavily on distributional properties of the input than younger children do. For this reason, she investigated grammatical development in four two- to three-year-old Turkish- Dutch bilingual children with differing input situations, as well as two Turkish and two Dutch age-matched monolingual controls. Spontaneous speech data were collected from the participants at three month intervals over the course of a year and a half, when they were aged between 2;0 and 3;6. During the data collection sessions, participants wore a jacket
53 fitted with a wireless transmitter and a microphone in order to capture all linguistic interaction. It was found that the three-year-old bilinguals’ mean length of utterance, proportion of finite utterances and finite verb type profile was comparable to that of two- year-old monolinguals, indicating protracted grammatical development in the bilinguals (Blom, 2010:439). However, she points out that this difference between bilinguals and monolinguals was found only in the case of the bilinguals’ weaker language in terms of input, suggesting that “children’s ability to acquire grammar is robust: only clearly reduced input will result in a protracted development” (Blom, 2010:439).
An important contribution of the above study is its investigation of a group of bilinguals with sociolinguistic backgrounds that differ from the “one-parent, one-language, middle-class” set-up investigated in many other bilingualism studies: here, participants formed part of a large ethnic minority (i.e. Turkish labour immigrants to the Netherlands) and resided in bilingual communities that exposed them to “code-switching, convergence and ethnic varieties” (Blom, 2010:440). Blom (2010:441) suspects that aspects of this social setting may explain some of her findings, but concedes that such an investigation fell outside of the scope of her study. For this reason, she calls for studies that address not only the effect of input quantity but also input quality and social setting on the bilingual acquisition process (Blom, 2010:441).
Two important limitations of studies investigating the effect of input quantity on early bilingual grammatical development identified by Unsworth (2013a:86) are that they typically focus (i) on the child’s current exposure situation, without consideration of the amount of exposure that the child has accumulated over time, and (ii) mostly on successive L2 learners as opposed to simultaneous bilinguals. Addressing this dearth in the literature, Unsworth (2013a) set out to investigate the effect of input quantity, measured in terms of both current and what she terms “cumulative” exposure, on Dutch grammatical gender acquisition by 136 simultaneous bilingual English-Dutch children (ranging in age from three to 17 years). For control purposes, she also collected data from 26 monolingual Dutch four- to six-year-olds and employed data from the most comparable study available, i.e. that by Blom, Polišenskà and Weerman (2008), for comparisons with her three- and seven-year-olds.
54 Research (e.g. van der Velde, 2003) has shown that, in contrast to the same process in other languages, the acquisition of Dutch grammatical gender is a lengthy process, with the production of errors in this area by monolingual Dutch speakers continuing until at least age six (Unsworth, 2013a:87). Although monolingual children do eventually acquire the target system, it is not yet clear from studies using bilingual participants whether bilinguals are able to advance beyond the stage at which, like young monolinguals, they produce errors caused by overgeneralisation (Unsworth, 2013a:87). On grounds of the results of previous studies on the acquisition of Dutch grammatical gender, Blom et al. (2008:323) suggest that bilinguals might fail to acquire the relevant rules due to the fact that their necessarily reduced amount of exposure to Dutch, compared to monolinguals, prevents them from reaching the required “critical mass” of grammatical evidence needed to infer these rules before the end of what might be a critical or sensitive period, ending around six or seven years of age (cf. Meisel, 2007a; Meisel, 2009; and, for Welsh grammatical gender, Gathercole & Thomas, 2005). An alternative explanation of the fact that bilingual/L2 children struggle with Dutch gender marking relates to linguistic competence versus linguistic performance: as posited by the Missing Surface Inflection Hypothesis (MSIH) (Haznedar & Schwartz, 1997; Prévost & White, 2000), bilinguals might have acquired the relevant rules, but have problems in applying them during production.
The bilingual participants in Unsworth’s (2013a) study were all exposed to both English and Dutch in the home, mostly in an OPOL situation, and were divided into groups based on age. The following vocabulary tests were used as a measure of general proficiency: for English, depending on the child’s specific variety, either the Peabody Picture Vocabulary Test IV (Dunn & Dunn, 2007) or British Picture Vocabulary Scale (Dunn, Dunn, Whetton, & Burley, 1997); for Dutch, the PPVT-III-NL (Dunn, Dunn, & Schlichting, 2005). To test gender- marking on definite determiners, two elicited production tasks ̶ one picture description and one story-telling task ̶ and one grammaticality judgement task were used; to test gender- marking on adjectives in indefinite determiner phrases, only the picture description task was used (Unsworth, 2013a:91). Lastly, a very detailed parental questionnaire was used to collect information on the bilingual participants’ current exposure to Dutch and their exposure to this language over time, i.e. cumulative length of exposure, which was calculated on grounds of
55 the estimated “proportion of each one-year period which included exposure to Dutch” (Unsworth, 2013a:91). The latter measure was introduced as a variable to be taken into account when estimating input quantity, because it recognises the inherent variation in this regard in bilingual settings and so enables a more accurate comparison of the language development of monolingual and bilingual participants (Unsworth, 2013a:86,95). As the questions asked in Unsworth’s (2013a) questionnaire incidentally coincided quite closely with those in the questionnaire originally designed for the purposes of the current study and were later used as guideline for some alterations thereto, the design of Unsworth’s questionnaire is discussed in detail in the chapter on methodology, Chapter 5.
What is important to note here is that, whereas length of exposure in the traditional sense would be equal to chronological age in the case of simultaneous bilinguals who were exposed to two languages from birth, cumulative length of exposure in this study proved to be, on average, only about half of this value (Unsworth, 2013a:91). This meant that, using cumulative length of exposure as the basis for comparison, the bilingual participants were better compared to monolinguals of an age equal to the amount of years of cumulative exposure that the bilinguals have had (in this case, monolinguals who were about half the age of the bilinguals) rather than to age-matched monolinguals (Unsworth, 2013a:95). An age- matched monolingual-bilingual comparison showed bilinguals to fare significantly worse than monolinguals with Dutch gender-marking (Unsworth, 2013a:95). However, when matched on cumulative length of exposure, the differences between monolinguals and bilinguals disappeared in the case of both production and grammaticality judgement tasks, with bilinguals scoring as high as or even higher than monolingual counterparts matched in
this manner (Unsworth, 2013a:103).14 As regards the effect of the two different measures of
input quantity investigated in this study, i.e. current amount of exposure and cumulative length of exposure, both were found to be significant predictors of accuracy with the marking of grammatical gender on Dutch determiners, but not on adjectives (Unsworth, 2013a:106).
14 Note, however, that the older bilinguals would have had higher levels of cognitive maturity and metalinguistic knowledge at their disposal than would the younger monolinguals, which may skew a comparison between these groups. Cf. Unsworth and Blom (2010) for methodological suggestions as to how to control for these types of differences.
56 As far as the relationship between input quantity and proficiency in the case of bilingual language acquisition is concerned, there is (at least to my knowledge) no available literature on investigations into this phenomenon in the context of the acquisition of the passive, the grammatical construction of interest in the present study.