In her Keynote Article, Paradis reviews evidence from bilingual language development to assess the claims of two opposing theoretical views of language disorders. Specifically, she examines the evidence for similarities in language profiles of typically developing (TD) sequential bilingual (second language [L2]) children and monolingual children with specific language impairment (SLI) with respect to Rice's extended optional infinitive (EOI) account. A limited processing capacity (LPC) account of SLI, Leonard's surface hypothesis, is evaluated within the context of comparisons among bilingual children with SLI, monolingual children with SLI, and TD bilingual children. Paradis concludes that the evidence from bilingual children poses challenges for both accounts of SLI.
The observation that in early stages of L2 acquisition, TD children may present with morphological error patterns that characterize monolingual children with SLI is an incredibly important one. It contributes to the ever-growing body of literature that suggests overlap in language processing patterns between these two populations. Various studies have reported considerable similarities in language profiles of typical L2 learners and monolingual children with SLI (Crago & Paradis, Reference Crago, Paradis, Levy and Schaeffer2003; Håkansson & Nettelbladt, Reference Håkansson, Nettelbladt, Johnson and Gilbert1996; Paradis & Crago, Reference Paradis and Crago2000; Windsor & Kohnert, Reference Windsor and Kohnert2004). Such an overlap poses a serious clinical problem of identifying language impairment in children who are L2 speakers. Thus, the clinical contributions of this work cannot be overstated. In addition to informing the clinical practice with L2 learners, Paradis uses the observation that L2 speakers and monolingual children with SLI share certain morphological error patterns as the basis for considering two competing theoretical accounts of SLI.
A main thesis of the Paradis article is that data from bilingual children can inform theoretical debates concerning the nature of SLI. This is certainly an intriguing notion that is worthy of additional investigation. In this Commentary we will examine two of the underlying assumptions and related logical arguments within the Paradis Keynote. One underlying assumption of this work is that the EOI account of SLI applies cross-linguistically. The other underlying assumption of this work is that the LPC account of SLI would predict a “double deficit” in bilingual children with SLI compared to monolingual children with SLI.
The motivation behind examining the utility of the EOI account for explaining commonalities between children with SLI and L2 learners is the observation that the EOI profile maintains cross-linguistically. Verb morphology is especially problematic for children with SLI acquiring English and a number of other languages. Errors in marking tense and agreement, as well as omissions of auxiliary verbs in obligatory contexts have been noted in children acquiring English, Swedish, and German. As a result, prior research by Paradis and colleagues has focused on characterizing cross-linguistic SLI data in terms of Rice's EOI hypothesis (e.g., Paradis & Crago, Reference Paradis and Crago2001), extending this account to bilingual development (Paradis, Reference Paradis2007) and to comparisons of L2 and SLI language profiles (Paradis & Crago, Reference Paradis, Crago, Prévost and Paradis2004). However, by no means is cross-linguistic work unequivocal. In Spanish, Italian, Hebrew, and possibly French, there is evidence that the errors in the use of noun and adjective morphology, such as articles and clitics, may be a better index of SLI than tense morphology (e.g., Bedore & Leonard, Reference Bedore and Leonard2001; Bortolini, Caselli, & Leonard, Reference Bortolini, Caselli and Leonard1997; Bottari, Cipriani, Chilosi, & Pfanner, Reference Bottari, Cipriani, Chilosi and Pfanner1998; Dromi, Leonard, Adam, & Zadunaisky-Ehrlich, Reference Dromi, Leonard, Adam and Zadunaisky-Ehrlich1999; Jakubowicz, Nash, Rigaut, & Gerard, Reference Jakubowicz, Nash, Rigaut and Gerard1998; Leonard & Bortolini, Reference Leonard and Bortolini1998). Moreover, in some languages, children with SLI do not have inordinate difficulties with acquiring tense or agreement inflectional morphemes, and perform on a par not only with their language-matched TD peers but also with age-matched TD children. For instance, Leonard, Sabbadini, Leonard, and Volterra (Reference Leonard, Sabbadini, Leonard and Volterra1987) found that Italian-speaking children with SLI 4 to 6 years of age showed extensive use of third-person singular. This was in contrast to their English-speaking SLI peers, who demonstrated significant difficulties with third-person singular at the same age. Similarly, Thordardottir and Namazi (Reference Thordardottir and Namazi2007) showed that in French-speaking children with SLI, verb finiteness errors occurred rarely, and with the same frequency as in age-matched and language-matched peers. Further, Dromi, Leonard, and Shteiman (Reference Dromi, Leonard and Shteiman1993) compared Hebrew-speaking children with SLI to English-speaking children with SLI from another study (Leonard, Bortolini, Caselli, McGregor, & Sabbadini, Reference Leonard, Bortolini, Caselli, McGregor and Sabbadini1992) and showed that Hebrew-speaking children with SLI had significantly less difficulty than English-speaking children with SLI on verb inflection morphology.
Together, cross-linguistic studies indicate that the specific SLI profile depends on properties of the particular language being acquired. In languages like Italian and Hebrew, where inflectional verb morphology is a frequent and necessary cue to meaning, children with SLI can acquire tense and agreement markers on a time line similar to that of TD children. In languages like English and French, where word order is a more reliable cue than inflectional verb morphology, children with SLI can demonstrate significant difficulties with acquiring tense and agreement markers compared to TD children. This is precisely the type of evidence that motivated LPC accounts of SLI to take into account the strength of cues in a language to explain the cross-linguistic variability in morphological error patterns that characterize children with SLI. Cross-linguistic differences in morphological acquisition are accommodated by theories of language development positing that the availability (or cue validity) of a grammatical marker drives the acquisition patterns of that marker in TD children (e.g., Aksu-Koc & Slobin, Reference Aksu-Koc, Slobin and Slobin1985; Nicoladis, Palmer, & Marentette, Reference Nicoladis, Palmer and Marentette2007; Plunkett & Marchman, Reference Plunkett and Marchman1991) and in children with SLI (see Lukács, Leonard, Kas, & Pléh, Reference Lukács, Leonard, Kas and Pléh2009; Norbury, Bishop, & Briscoe, Reference Norbury, Bishop and Briscoe2001). Although cross-linguistic differences in SLI profiles do not invalidate the EOI account of SLI, cross-linguistic work is far from unambiguous in demonstrating universal difficulties with tense and agreement morphology in children with SLI.
Although Paradis states that research findings pose difficulties for both theories of SLI, the weight of the evidence presented favors the EOI account. It is important to note that different contexts (i.e., comparisons of different populations) are employed to evaluate the EOI and LPC theories. A finding that morphological error patterns of sequential bilingual children acquiring English or French are similar to the morphological characteristics of monolingual children with SLI acquiring English or French is used to suggest the viability of the EOI account. A finding of comparable effect sizes for comparison of bilingual children with SLI to their TD bilingual peers and of monolingual children with SLI to their TD monolingual peers is viewed as lack of support for the LPC account. We would like to suggest that these conclusions warrant reexamination.
First, just because similar patterns are observed in L2 learners and children with SLI, this does not mean that the morphological errors in their productions have the same explanatory mechanism. The same error pattern can be elicited via very different means. For example, Hayiou-Thomas, Bishop, and Plunkett (Reference Hayiou-Thomas, Bishop and Plunkett2004) demonstrated that the verb-use error pattern that characterizes speech production in English-speaking children with SLI (omission of past-tense -ed and third-person singular -s) can be induced in TD children by increasing the cognitive load of the task either by simulating limitations in working memory or processing speed. We contend that observation of comparable error patterns in L2 learners and children with SLI says very little about the mechanisms underlying the linguistic behavior in the two populations. Children with SLI show more errors than TD monolingual children because of an underlying difficulty acquiring language (whatever that may be, a representational morphological deficit or limited processing capacity). Children acquiring an L2 show more errors in L2 production than monolingual children because of their lack of experience with L2. Although it is quite possible that L2 acquisition is a qualitatively different process than L1 acquisition, certainly no one would argue that L2 speakers have an underlying difficulty learning language (because L1 acquisition was successful). A similar argument was made by Kohnert (Reference Kohnert, Altarriba and Heredia2008), who pointed out that children with SLI are presumed to have inefficient language learning systems, but L2 children have had less exposure or fewer opportunities to learn the L2. Paradis (2009) also acknowledges this, stating that these different groups of learners have similar profiles even though they have “dissimilar input experiences and internal learning mechanisms.” Given that underlying mechanisms must be different in the L2 and SLI populations, we question what superficial similarities in error patterns reveal except that the topology of a given language leads to similar patterns of what is difficult to learn. We argue that even if there were no differences at all observed between L2 and monolingual SLI, this would neither support nor refute the EOI theory or the LPC account. Simply put, similar profiles do not necessarily imply similar underlying mechanisms.
Second, the examination of cumulative effects in the bilingual SLI data in no way invalidates the LPC account of SLI. Paradis argues that there is no “double-deficit” in bilingual children with SLI, something that you should expect if they had processing capacity limitations. The data used by Paradis to test the double-deficit hypothesis are cross-sectional, and the severity of language impairment across the groups is not known to be equivalent: this is critical for assessment of “effect size” comparisons. Setting aside the fact that these are not longitudinal data, bilingual children with SLI are outperformed by both their TD bilingual peers and by monolingual peers with SLI, actually supporting LPC accounts. Other researchers have also contrasted the LPC and the EOI accounts, but have done so within a single set of participants. Restrepo and Kruth (Reference Restrepo and Kruth2000) conducted detailed case studies comparing two sequential Spanish–English bilingual children: one with SLI and one without. Based on their results, they contended that data from bilingual children with SLI did not support the notion of a universal grammatical marker of SLI but were compatible with a limited processing capacity account. Similarly, Sanz-Torrent, Serrat, Andreu, and Serra (Reference Sanz-Torrent, Serrat, Andreu and Serra2008) examined developmental language patterns in Catalan–Spanish bilingual preschool children, and concluded that the patterns of correct and incorrect verb form usage were not consistent with the EOI account but instead supported the surface hypothesis.
In the end, we would argue that descriptive data from bilingual children with SLI is equivocal at best, with respect to supporting or refuting either the EOI or the LPC account, largely because cross-linguistic transfer, language proficiency and exposure, age of acquisition associated with the L2 are all likely to play a role in the morphological error patterns observed for bilingual children. This variability within the bilingual population may mask the “mechanism” that underlies the error pattern, and any work that attempts to identify such a mechanism must account for all the other variables first. This difficulty is exacerbated by the findings of facilitation effects associated with bilingualism, which may (or may not) apply to bilingual children with SLI. Paradis herself brings up the possibility that improved executive function observed in bilinguals may have a mitigating effect on purported processing limitations of bilingual children with SLI. The lack of differences in the effect sizes for the bilingual TD versus SLI comparison and for the monolingual TD versus SLI comparison is therefore not interpretable at this point, and explicit experimental manipulations are required before a meaningful juxtaposition of the LPC and EOI accounts can be made in this population.
In conclusion, the work identifying overlaps and divergences in error patterns exhibited by children with SLI and L2 learners is extremely important from a clinical perspective. The comparisons of descriptive profiles across bilingual and monolingual children with and without SLI certainly provide a starting point for exploring the theoretical implications of morphological difficulties associated with L2 acquisition and SLI. Further work is needed to experimentally examine theoretical claims regarding the underlying nature of SLI with respect to bilingual data. One particularly promising area of investigation is Paradis' suggestion regarding the possible compensatory, facilitative effects of bilingualism on language processing in children with SLI; this issue has significant theoretical as well as clinical implications.
You have
Access
In her Keynote Article, Paradis reviews evidence from bilingual language development to assess the claims of two opposing theoretical views of language disorders. Specifically, she examines the evidence for similarities in language profiles of typically developing (TD) sequential bilingual (second language [L2]) children and monolingual children with specific language impairment (SLI) with respect to Rice's extended optional infinitive (EOI) account. A limited processing capacity (LPC) account of SLI, Leonard's surface hypothesis, is evaluated within the context of comparisons among bilingual children with SLI, monolingual children with SLI, and TD bilingual children. Paradis concludes that the evidence from bilingual children poses challenges for both accounts of SLI.
The observation that in early stages of L2 acquisition, TD children may present with morphological error patterns that characterize monolingual children with SLI is an incredibly important one. It contributes to the ever-growing body of literature that suggests overlap in language processing patterns between these two populations. Various studies have reported considerable similarities in language profiles of typical L2 learners and monolingual children with SLI (Crago & Paradis, Reference Crago, Paradis, Levy and Schaeffer2003; Håkansson & Nettelbladt, Reference Håkansson, Nettelbladt, Johnson and Gilbert1996; Paradis & Crago, Reference Paradis and Crago2000; Windsor & Kohnert, Reference Windsor and Kohnert2004). Such an overlap poses a serious clinical problem of identifying language impairment in children who are L2 speakers. Thus, the clinical contributions of this work cannot be overstated. In addition to informing the clinical practice with L2 learners, Paradis uses the observation that L2 speakers and monolingual children with SLI share certain morphological error patterns as the basis for considering two competing theoretical accounts of SLI.
A main thesis of the Paradis article is that data from bilingual children can inform theoretical debates concerning the nature of SLI. This is certainly an intriguing notion that is worthy of additional investigation. In this Commentary we will examine two of the underlying assumptions and related logical arguments within the Paradis Keynote. One underlying assumption of this work is that the EOI account of SLI applies cross-linguistically. The other underlying assumption of this work is that the LPC account of SLI would predict a “double deficit” in bilingual children with SLI compared to monolingual children with SLI.
The motivation behind examining the utility of the EOI account for explaining commonalities between children with SLI and L2 learners is the observation that the EOI profile maintains cross-linguistically. Verb morphology is especially problematic for children with SLI acquiring English and a number of other languages. Errors in marking tense and agreement, as well as omissions of auxiliary verbs in obligatory contexts have been noted in children acquiring English, Swedish, and German. As a result, prior research by Paradis and colleagues has focused on characterizing cross-linguistic SLI data in terms of Rice's EOI hypothesis (e.g., Paradis & Crago, Reference Paradis and Crago2001), extending this account to bilingual development (Paradis, Reference Paradis2007) and to comparisons of L2 and SLI language profiles (Paradis & Crago, Reference Paradis, Crago, Prévost and Paradis2004). However, by no means is cross-linguistic work unequivocal. In Spanish, Italian, Hebrew, and possibly French, there is evidence that the errors in the use of noun and adjective morphology, such as articles and clitics, may be a better index of SLI than tense morphology (e.g., Bedore & Leonard, Reference Bedore and Leonard2001; Bortolini, Caselli, & Leonard, Reference Bortolini, Caselli and Leonard1997; Bottari, Cipriani, Chilosi, & Pfanner, Reference Bottari, Cipriani, Chilosi and Pfanner1998; Dromi, Leonard, Adam, & Zadunaisky-Ehrlich, Reference Dromi, Leonard, Adam and Zadunaisky-Ehrlich1999; Jakubowicz, Nash, Rigaut, & Gerard, Reference Jakubowicz, Nash, Rigaut and Gerard1998; Leonard & Bortolini, Reference Leonard and Bortolini1998). Moreover, in some languages, children with SLI do not have inordinate difficulties with acquiring tense or agreement inflectional morphemes, and perform on a par not only with their language-matched TD peers but also with age-matched TD children. For instance, Leonard, Sabbadini, Leonard, and Volterra (Reference Leonard, Sabbadini, Leonard and Volterra1987) found that Italian-speaking children with SLI 4 to 6 years of age showed extensive use of third-person singular. This was in contrast to their English-speaking SLI peers, who demonstrated significant difficulties with third-person singular at the same age. Similarly, Thordardottir and Namazi (Reference Thordardottir and Namazi2007) showed that in French-speaking children with SLI, verb finiteness errors occurred rarely, and with the same frequency as in age-matched and language-matched peers. Further, Dromi, Leonard, and Shteiman (Reference Dromi, Leonard and Shteiman1993) compared Hebrew-speaking children with SLI to English-speaking children with SLI from another study (Leonard, Bortolini, Caselli, McGregor, & Sabbadini, Reference Leonard, Bortolini, Caselli, McGregor and Sabbadini1992) and showed that Hebrew-speaking children with SLI had significantly less difficulty than English-speaking children with SLI on verb inflection morphology.
Together, cross-linguistic studies indicate that the specific SLI profile depends on properties of the particular language being acquired. In languages like Italian and Hebrew, where inflectional verb morphology is a frequent and necessary cue to meaning, children with SLI can acquire tense and agreement markers on a time line similar to that of TD children. In languages like English and French, where word order is a more reliable cue than inflectional verb morphology, children with SLI can demonstrate significant difficulties with acquiring tense and agreement markers compared to TD children. This is precisely the type of evidence that motivated LPC accounts of SLI to take into account the strength of cues in a language to explain the cross-linguistic variability in morphological error patterns that characterize children with SLI. Cross-linguistic differences in morphological acquisition are accommodated by theories of language development positing that the availability (or cue validity) of a grammatical marker drives the acquisition patterns of that marker in TD children (e.g., Aksu-Koc & Slobin, Reference Aksu-Koc, Slobin and Slobin1985; Nicoladis, Palmer, & Marentette, Reference Nicoladis, Palmer and Marentette2007; Plunkett & Marchman, Reference Plunkett and Marchman1991) and in children with SLI (see Lukács, Leonard, Kas, & Pléh, Reference Lukács, Leonard, Kas and Pléh2009; Norbury, Bishop, & Briscoe, Reference Norbury, Bishop and Briscoe2001). Although cross-linguistic differences in SLI profiles do not invalidate the EOI account of SLI, cross-linguistic work is far from unambiguous in demonstrating universal difficulties with tense and agreement morphology in children with SLI.
Although Paradis states that research findings pose difficulties for both theories of SLI, the weight of the evidence presented favors the EOI account. It is important to note that different contexts (i.e., comparisons of different populations) are employed to evaluate the EOI and LPC theories. A finding that morphological error patterns of sequential bilingual children acquiring English or French are similar to the morphological characteristics of monolingual children with SLI acquiring English or French is used to suggest the viability of the EOI account. A finding of comparable effect sizes for comparison of bilingual children with SLI to their TD bilingual peers and of monolingual children with SLI to their TD monolingual peers is viewed as lack of support for the LPC account. We would like to suggest that these conclusions warrant reexamination.
First, just because similar patterns are observed in L2 learners and children with SLI, this does not mean that the morphological errors in their productions have the same explanatory mechanism. The same error pattern can be elicited via very different means. For example, Hayiou-Thomas, Bishop, and Plunkett (Reference Hayiou-Thomas, Bishop and Plunkett2004) demonstrated that the verb-use error pattern that characterizes speech production in English-speaking children with SLI (omission of past-tense -ed and third-person singular -s) can be induced in TD children by increasing the cognitive load of the task either by simulating limitations in working memory or processing speed. We contend that observation of comparable error patterns in L2 learners and children with SLI says very little about the mechanisms underlying the linguistic behavior in the two populations. Children with SLI show more errors than TD monolingual children because of an underlying difficulty acquiring language (whatever that may be, a representational morphological deficit or limited processing capacity). Children acquiring an L2 show more errors in L2 production than monolingual children because of their lack of experience with L2. Although it is quite possible that L2 acquisition is a qualitatively different process than L1 acquisition, certainly no one would argue that L2 speakers have an underlying difficulty learning language (because L1 acquisition was successful). A similar argument was made by Kohnert (Reference Kohnert, Altarriba and Heredia2008), who pointed out that children with SLI are presumed to have inefficient language learning systems, but L2 children have had less exposure or fewer opportunities to learn the L2. Paradis (2009) also acknowledges this, stating that these different groups of learners have similar profiles even though they have “dissimilar input experiences and internal learning mechanisms.” Given that underlying mechanisms must be different in the L2 and SLI populations, we question what superficial similarities in error patterns reveal except that the topology of a given language leads to similar patterns of what is difficult to learn. We argue that even if there were no differences at all observed between L2 and monolingual SLI, this would neither support nor refute the EOI theory or the LPC account. Simply put, similar profiles do not necessarily imply similar underlying mechanisms.
Second, the examination of cumulative effects in the bilingual SLI data in no way invalidates the LPC account of SLI. Paradis argues that there is no “double-deficit” in bilingual children with SLI, something that you should expect if they had processing capacity limitations. The data used by Paradis to test the double-deficit hypothesis are cross-sectional, and the severity of language impairment across the groups is not known to be equivalent: this is critical for assessment of “effect size” comparisons. Setting aside the fact that these are not longitudinal data, bilingual children with SLI are outperformed by both their TD bilingual peers and by monolingual peers with SLI, actually supporting LPC accounts. Other researchers have also contrasted the LPC and the EOI accounts, but have done so within a single set of participants. Restrepo and Kruth (Reference Restrepo and Kruth2000) conducted detailed case studies comparing two sequential Spanish–English bilingual children: one with SLI and one without. Based on their results, they contended that data from bilingual children with SLI did not support the notion of a universal grammatical marker of SLI but were compatible with a limited processing capacity account. Similarly, Sanz-Torrent, Serrat, Andreu, and Serra (Reference Sanz-Torrent, Serrat, Andreu and Serra2008) examined developmental language patterns in Catalan–Spanish bilingual preschool children, and concluded that the patterns of correct and incorrect verb form usage were not consistent with the EOI account but instead supported the surface hypothesis.
In the end, we would argue that descriptive data from bilingual children with SLI is equivocal at best, with respect to supporting or refuting either the EOI or the LPC account, largely because cross-linguistic transfer, language proficiency and exposure, age of acquisition associated with the L2 are all likely to play a role in the morphological error patterns observed for bilingual children. This variability within the bilingual population may mask the “mechanism” that underlies the error pattern, and any work that attempts to identify such a mechanism must account for all the other variables first. This difficulty is exacerbated by the findings of facilitation effects associated with bilingualism, which may (or may not) apply to bilingual children with SLI. Paradis herself brings up the possibility that improved executive function observed in bilinguals may have a mitigating effect on purported processing limitations of bilingual children with SLI. The lack of differences in the effect sizes for the bilingual TD versus SLI comparison and for the monolingual TD versus SLI comparison is therefore not interpretable at this point, and explicit experimental manipulations are required before a meaningful juxtaposition of the LPC and EOI accounts can be made in this population.
In conclusion, the work identifying overlaps and divergences in error patterns exhibited by children with SLI and L2 learners is extremely important from a clinical perspective. The comparisons of descriptive profiles across bilingual and monolingual children with and without SLI certainly provide a starting point for exploring the theoretical implications of morphological difficulties associated with L2 acquisition and SLI. Further work is needed to experimentally examine theoretical claims regarding the underlying nature of SLI with respect to bilingual data. One particularly promising area of investigation is Paradis' suggestion regarding the possible compensatory, facilitative effects of bilingualism on language processing in children with SLI; this issue has significant theoretical as well as clinical implications.