Elsevier

NeuroImage

Volume 36, Issue 3, 1 July 2007, Pages 912-923
NeuroImage

Age of acquisition modulates neural activity for both regular and irregular syntactic functions

https://doi.org/10.1016/j.neuroimage.2007.02.055Get rights and content

Abstract

Studies have found that neural activity is greater for irregular grammatical items than regular items. Findings with monolingual Spanish speakers have revealed a similar effect when making gender decisions for visually presented nouns. The current study extended previous studies by looking at the role of regularity in modulating differences in groups that differ in the age of acquisition of a language. Early and late learners of Spanish matched on measures of language proficiency were asked to make gender decisions to regular (-o for masculine and -a for feminine) and irregular items (which can end in e, l, n, r, s, t and z). Results revealed increased activity in left BA 44 for irregular compared to regular items in separate comparisons for both early and late learners. In addition, within-group comparisons revealed that neural activity for irregulars extended into left BA 47 for late learners and into left BA 6 for early learners. Direct comparisons between groups revealed increased activity in left BA 44/45 for irregular items indicating the need for more extensive syntactic processing in late learners. The results revealed that processing of irregular grammatical gender leads to increased activity in left BA 44 and adjacent areas in the left IFG regardless of when a language is learned. Furthermore, these findings suggest differential recruitment of brain areas associated with grammatical processing in late learners. The results are discussed with regard to a model which considers L2 learning as emerging from the competitive interplay between two languages.

Introduction

Is second language (L2) processing fundamentally similar or different from first language (L1) processing? Work investigating language processing in bilingual aphasics (Fabbro, 1999, Fabbro et al., 2000, Junque et al., 1995, Paradis, 1987, Paradis, 1995a, Paradis, 1995b) and healthy normals using fMRI (Abutalebi et al., 2001, Chee et al., 2001, Dehaene et al., 1997, Hernandez et al., 2001, Hernandez et al., 2000, Illes et al., 1999, Perani et al., 1998, Rodriguez-Fornells et al., 2002, Wartenburger et al., 2003) has reported evidence that both languages are most likely processed by similar neural circuits. Despite the fact that brain activity is mediated by mostly overlapping neural circuits, recent work in the neuroimaging literature suggests that both age of second language acquisition (AoA) and language proficiency modulate the magnitude of such activity (Chee et al., 2001, Perani et al., 1998, Wartenburger et al., 2003). Furthermore, results from a recent fMRI study suggest a dissociation between syntax and semantics (Wartenburger et al., 2003). In this study, comparisons were made between groups that differed on 2nd language (L2) age of acquisition (AoA) but were matched on proficiency. In a second comparison, groups that differed on proficiency but were matched on L2 AoA were compared. Results revealed that syntax was more strongly modulated by 2nd language (L2) age of acquisition (AoA) whereas semantic processing was more strongly modulated by language proficiency. Finally, studies have found that L2 learners show increased neural activity for irregular items relative to regular items (Tatsuno and Sakai, 2005). The current study sought to extend the findings of previous studies by testing early and late Spanish learners matched on language proficiency using a grammatical gender decision task. Both regular and irregular gender marked items were used. These results should help to elucidate whether differences in AoA modulate neural activity differentially for regular and irregular gender marked items.

As noted above, earlier studies have not found a clear relationship between second language AoA and recovery of function in bilingual aphasics or in the location or amount of neural activity. However, two studies suggest that age of acquisition plays a more critical role than proficiency in determining the neural activity associated with grammatical processing (Wartenburger et al., 2003, Weber-Fox and Neville, 1996). In a seminal study, Weber-Fox and Neville (1996) presented a group of early and late L2 learners sentences that contained semantic or syntactic anomalies while observing neural activity with Event Related Potentials (ERPs). Results across both error types revealed differences in the neural signatures of these effects when L2 was learned later in life. Differences were observed when comparing the data obtained for L2 learners with that which had been collected in other studies with native speakers. For syntactic anomalies, second language speakers’ data were found to diverge from that seen in native speakers even when the former began L2 acquisition between 1 and 3 years of life. For semantic processing, however, differences between monolinguals and L2 speakers were observed only in individuals who learned L2 after the age of 11. That is, the neural signatures associated with syntactic anomalies are influenced to a greater extent by the age of L2 acquisition than those associated with semantic anomalies.

More recently, Wartenburger et al. (2003) asked participants to detect syntactic anomalies (violations of case, gender or number) or semantic anomalies in a set of visually presented sentences. Like Weber-Fox and Neville (1996), the investigators were interested in understanding the influence of language proficiency and AoA on the neural signatures associated with semantic and syntactic processing. In order to investigate these differences, Wartenbuger et al. included one early high-proficiency group, one late high-proficiency group and one late low-proficiency group. Differences in neural activity due to AoA were detected via comparisons between early and late high-proficiency subjects. Differences in neural activity due to proficiency were estimated by comparing late high and low proficiency subjects. The results from their study were in line with those observed by Weber-Fox and Neville (1996). For semantic judgments, there were differences between high and low proficiency late bilinguals, but no differences between early and late high-proficiency subjects. For syntactic anomalies, there were significant differences between the two high-proficiency groups, which differed on AoA, and very subtle differences between the two late AoA groups that differed on proficiency.

Taken together, the two studies by Weber-Fox and Neville and Wartenburger et al. suggest that second language AoA may have an important influence on the nature of neural activity during grammatical processing. However, some of these effects could be due to methodological limitations of the previous studies. First, many of the conclusions attributed to AoA by Weber-Fox and Neville (1996) are based on comparisons between the data collected in their study on L2 learners with other studies in monolinguals. This leaves open the possibility that differences attributed to maturational changes may be due in part to differences in methodologies, procedures and data analytic methods across studies. Follow-up studies should compare native and nonnative speakers directly. Second, both studies do not address the issue of whether these effects appear for all syntactic functions. In this respect, a recent study by Rossi et al. (in press) has found that differences between native speakers and high-proficiency L2 learners may not appear for all types of syntactic violations. Subsequent studies should consider carefully under what conditions L2 AoA effects appear. In this respect, Wartenburger et al. (2003) did not consider differences in regularity of a syntactic function as a factor that may influence AoA effects.

It is precisely the issue of regularity and how it is modulated in bilingual speakers that was addressed by two seminal studies (Sakai et al., 2004, Tatsuno and Sakai, 2005). In the first study, Sakai et al. (2004) tested pairs of twins that were trained on the past tense conjugations of verbs in English over a 2 month period. Participants were then asked to perform a verb matching task or a past tense identification task. The results revealed increased activity in left dorsal Inferior Frontal Gyrus (IFG) as accuracy improved. This appeared in both a comparison between post- and pre-training periods as well as in the individual performance of each subject. In a second study, Tatsuno and Sakai (2005) used the same task to test a group of Japanese native speakers who had learned English as a second language. Comparable to Sakai et al. (2004), the results revealed increased neural activity in the superior and anterior portion of left BA 45 with decreased proficiency in English. The results revealed increased activity in the lower portion of left IFG (BA 45/47) during processing of irregular (but not regular) past tense items compared to processing of similar items in a verb matching task. These two studies indicate the importance of inferior portions of left IFG in the processing of past tense verbs in late learners of a second language. Furthermore, these results suggest that a more superior and anterior portion of left BA 45 may be more sensitive to proficiency in a language as measured by improved accuracy in the task.

The cognitive and neural mechanism underlying the computation of inflectional morphology is a point of debate. In the cognitive literature, researchers are divided on whether regular and irregular inflectional morphology is handled by separate systems or by a single system. A prominent variant of the separate system theory is the “Words and Rules” theory, developed by Pinker (Pinker, 1991, Pinker and Ullman, 2002). In this theory, the regular past tense is generated by a combinatorial grammatical system which applies a rule without reference to the phonology or semantics of the stem. Irregular past tense forms, on the other hand, are stored in the lexicon and retrieved via an associative memory mechanism. More recently, Ullman and colleagues (Ullman, 2001a, Ullman, 2004) have extended this account by proposing that regular and irregular grammatical forms are handled by distinct neural systems. Because it is a subsystem of declarative memory, lexical memory should be modulated by the medial temporal and temporo-parietal regions which are considered to be responsible for the consolidation and long-term retention of declarative memories. Consequently, regular inflection, which is more dependent on grammatical processing and more reliant on the procedural system, should involve the basal ganglia, Broca’s area and neighboring anterior regions. Results from a series of studies have lent support to this view (Patterson et al., 2001, Tyler et al., 2002, Ullman et al., 1997). Ullman and colleagues have termed this framework the Declarative Procedural (DP) model due to the involvement of different memory systems during language processing.

Dissociations between regular and irregular processing can also be explained using an interactive/activation account (Rumelhart and McClelland, 1986). Although the interactive/activation (IA) model has a single system for all past tense forms, retrieval of regular and irregular forms depends on differential contributions of phonological and semantic components. In this view, irregular inflection may be more vulnerable to disruption from a semantic deficit whereas regular past tense forms place a greater demand on the phonological system (Bird et al., 2003, Burzio, 2002, Lambon Ralph et al., 2005). Unlike the irregular past tense, however, the transformation into the regular past tense in English should be more vulnerable after phonological impairments because of the addition of the ending “ed” after regular forms. This theory has received support from connectionist simulations (Joanisse and Seidenberg, 1999), work with language impaired populations (Bird et al., 2003, Joanisse, 2004), as well as newer work with neuroimaging methodologies (Desai et al., 2006, Joanisse, 2004).

A number of studies have found increased activity for irregular relative to regular past tense items in English (Desai et al., 2006, Jaeger et al., 1996, Sahin et al., 2006). Desai et al. (2006) found that processing of the irregular relative to the regular past tense involves increased activity in left IFG in BA 45/47 (Desai et al., 2006). This region has been associated with semantic processing (Bookheimer, 2002, Dapretto et al., 1999, Devlin et al., 2003, Hagoort et al., 2006, Hagoort et al., 2004, Poldrack et al., 1999). Hence, these results are supportive of the IA claim that generation of the irregular past tense involves greater semantic processing than that of the regular past tense.

Desai et al. also found increased activity in another locus of the IFG near the anterior insula. This increased activity is in the right homologue of the area traditionally associated with motor planning and/or articulatory processing (Bates et al., 2003b, Dronkers, 1996, Wise et al., 1999). One potential interpretation of this finding is that right insula and IFG activity involve recruitment of right hemisphere homologues during more difficult processing. Sahin et al. (2006), however, have found that processing of the irregular past tense involves increased activity in the medial portion of the supplementary motor area and the anterior cingulate gyrus, suggesting that processing of irregular items may involve more motor and attentional control. Taken together these results suggest that processing of the irregular past tense may also involve additional articulatory processing, motor planning and execution as well as attentional control.

An important question that arises with regard to the DP model is whether fMRI studies as currently designed are a valid test of this model. To date most studies have employed an overt past tense generation task in which participants are asked to take a present tense form and generate the past tense. This task clearly asks participants to make a conscious explicit judgment in generating the past tense. Hence, it is unclear if the past tense generation task can adequately test the DP model.

Finally, one key aspect of the DP and the IA models of past tense hinges on whether they are able to generalize beyond the English past tense to irregular/regular distinctions in other domains or languages. Of particular interest to the current study is the adaptation of either of these models for the processing of grammatical gender. We will return to models of gender processing after introducing the concept of grammatical gender.

Grammatical gender is a pervasive phenomenon in many of the world’s languages. However, this gender is not based on biological gender. Across languages, translation equivalents of the same lexical item may have completely different grammatical gender. Furthermore, gender marked nouns have to agree with determiners and adjectives. Hence, gender plays a role at the syntactic level.

Like the English past tense, there are two basic models for grammatical gender. Some models posit that grammatical gender involves a separate node or abstract representation (Caramazza and Miozzo, 1997, Levelt et al., 1999). However, others suggest that grammatical gender involves a combination of semantic and phonological information (Corbett, 1991, Zubin and Koepcke, 1981). For example, Corbett (1991) suggests that grammatical gender in most languages utilizes a semantic core by employing a range of morphonological information to build up separate gender classes. More recently, Mirkovic et al. (2005) simulated the acquisition of gender using connectionist models in which gender was represented explicitly or as a combination of semantic and phonological information. Both models were able to learn gender accurately and within the same number of learning cycles. In addition, both models were able to generalize accurately to a set of items which were outside the training set. These results suggest that a network which represents gender explicitly or via a combination of phonological and semantic rules can show equivalent performance. This conceptualization is compatible with previous neuroimaging work in our laboratory which has found that processing of irregular items may rely on areas that are both more closely associated with syntactic processing and which have been less directly linked to these processes (Hernandez et al., 2004). More specifically these results suggest that grammatical gender may be represented as both a syntactic rule and as a combination of phonological and semantic information.

Like other gender marked languages, Spanish has both phonologically regular items (-o for masculine, -a for feminine) and irregular items (with items forming smaller neighborhoods of items ending in e, l, n, r, s, t and z). Phonologically regular endings mark both biological gender (son, hijo, daughter, hija) as well as non-biological gender (moon, luna, river, rio). However, irregular endings are not associated with any particular biological gender. Previous studies have found that grammatical gender of nouns is learned easily by children (Devescovi et al., 1998, MacWhinney et al., 1989, Pizzuto and Caselli, 1992), can be detected by adults (Bates et al., 1995, Brooks et al., 1993, Cole and Segui, 1994, Radeau et al., 1989) and can be primed by gender marked adjectives (Akhutina et al., 1999, Bates et al., 1996, Bates et al., 2001, Cole and Segui, 1994, Deutsch et al., 1999, Friederici and Jacobsen, 1999, Gurjanov et al., 1985, Hagoort and Brown, 1999, Jescheniak and Schriefers, 1999). Furthermore, gender decisions are slower for irregularly marked nouns than for regularly marked nouns (Bates et al., 1995). Hence, retrieval of irregular gender is more effortful and may require more complex lexical processing (Bates et al., 1995). The effect of regularity on grammatical gender provides an interesting parallel to the effects that have been observed for past tense processing in English.

To date, only two published studies have considered the neural bases of grammatical gender processing. In a seminal study, Miceli et al. (2002) asked Spanish monolinguals to make gender (with regular and irregular items combined), semantic or phonological judgments about a series of visually presented words. Results revealed that processing of grammatical gender relative to semantic processing leads to increased activity in mid left BA 44/45. In an extension of this work, Hernandez et al. (2004) asked a group of Spanish speakers to monitor the gender of irregular and regular gender marked items. Like Miceli et al. (2002), results revealed increased activity in left BA 44/45. Increased activity was also observed in left superior BA 44 and left BA 45/47 when comparing gender decisions for irregular items to those of regular items (Hernandez et al., 2004). Hence, processing of both regular and irregular gender marked items relative to a semantic task leads to increased activity in left BA 44/45 (Miceli et al., 2002). Furthermore, comparisons between irregular items and regularly marked items reveal increased activity in additional areas both superior and inferior to the activity observed in left BA 44/45, an area which has been posited to be important for syntactic processing.

The presence of neural activity in left hemisphere including superior BA 44, BA 44/45, BA 45/47 and the insula sheds light on the differences in the manner in which processing of regular and irregular grammatical gender occurs. Left superior BA 44 has been observed across a wide range of studies which involve articulation and is thought to be involved in phonological encoding and retrieval (Wise et al., 1999, Zatorre et al., 1996). This area also shows increased activity when comparing generation of a determiner to picture naming (Heim et al., 2002) suggesting that the retrieval of a gender marked noun’s determiner may involve more complex phonological encoding and retrieval. The region of activity in left BA 44/45 is within a region that is known to be involved in a number of language processes including grammatical processing (Bookheimer, 2002, Dapretto et al., 1999, Friederici, 2002, Friederici et al., 2006, Miceli et al., 2002) as well as semantic processing (Vigliocco et al., 2006). Finally, a number of studies have linked neural activity in the left anterior insula with articulation. Using a lesion overlap method in two separate patient studies, Bates et al. (2003b) and Dronkers (1996) found speech difficulties to be associated with brain lesions in the left anterior insula. Furthermore, Ackermann and Riecker (2004) found increased activity in the left anterior insula during overt pronunciation of memorized syllables. This suggests that the left anterior insula can be directly linked to the coordination of muscle movements needed for articulation. In summary, fMRI results suggest that when processing irregularly marked gender items in Spanish, native speakers engage areas in left IFG associated with language related processes including phonological retrieval, syntactic processing and articulation.

Of particular interest to the current study is the increase in neural activity observed in left BA 45/47 for irregular items in Hernandez et al.’s study. This particular area is proximal to that observed for processing of the irregular past tense items in monolingual English speakers (Desai et al., 2006) and late learners of L2 (Sakai et al., 2004). The results from Hernandez et al. (2004) suggest that increased activity in left BA 45/47 for irregular items generalizes beyond the irregular past tense to other languages and processing of other types of grammatical functions. As noted earlier, this area has been linked to semantic processing. As such, it suggests that processing of irregular items may involve deeper semantic processing. It remains to be seen if this activity in this area differentiates between native and non-native speakers of a language.

The current study employed a unique approach in order to elucidate the difference between first and second language acquisition. Unlike previous studies which compared early and late learners of an L2 (Wartenburger et al., 2003, Weber-Fox and Neville, 1996), the present study compared two groups of Spanish speakers that differ in AoA but were matched in proficiency. Early Spanish–English bilinguals were chosen such that they learned Spanish first and were exposed to English by the age of 5. Late English–Spanish bilinguals were chosen such that they learned Spanish after the age of 12. Both groups were matched for proficiency on measures of word reading, vocabulary and basic grammar. To our knowledge, it is the first published study that has attempted to directly compare native and non-native speakers who are being scanned with fMRI while processing regular and irregular gender marked items.

In line with the literature reviewed so far, we expect to find the following: (1) speakers of Spanish as their first or second language will differ in neural activity during syntactic processing even when the former group learns L2 very early in life; (2) regular and irregular marked items differ in the amount of grammatical processing and may also differ in the co-activation of semantic and phonological information; (3) processing of irregular grammatical functions will lead to increased activity in left BA 45/47, which is thought to be due to the increased need for semantic retrieval when processing these items.

The main aim of the current study is to distinguish processing of an early learned L1 from a late learned L2. There is evidence that late learners have particular difficulty learning irregularly marked items (Birdsong and Flege, 2001, Flege et al., 1999). Hence, there should be larger cross-group differences for irregular items than for regular items. As noted above, processing of irregular items leads to increased activity in left BA 45/47 across at least two different grammatically irregular forms in two different languages. It will be of particular interest to observe whether activity in left BA 45/47 differentiates early and late learners. Furthermore, activity in additional areas recruited by non-native speakers involves other portions of left IFG devoted to language related processes. The presence of differences between these groups should elucidate how AoA influences the processing of regular and irregular morphology.

Section snippets

Early Spanish learners

This group was composed of twelve subjects (7 female and 5 male) with a mean age of 22.3 (SD = 1.35, range 20 to 25) who learned Spanish as a native language but were dominant in English, their second language. None had any past medical history or had used medication. All were right handed as assessed by our internal handedness questionnaire and reported no left-handed members in their immediate family.

Late Spanish learners

This group was composed of twelve subjects (7 female and 5 male) with a mean age of 24.5 (SD = 

Discussion

Results from the current experiment extend findings from previous studies by revealing the role that regularity and AoA play in modulating neural activity during grammatical processing. To our knowledge, this is the first study which has shown that regularity can modulate differences in early and late learners matched in language proficiency.

Acknowledgments

This material is based upon work supported by the National Science Foundation Grants “fMRI studies of language and task switching” (BCS0092043) and “Neural correlates of grammatical gender processing” (INT0202686) as well as a grant from the National Institute of Child Health and Development (1 R03HD050313-01).

References (88)

  • S. Heim et al.

    Broca’s area in the human brain is involved in the selection of grammatical gender for language production: evidence from event-related functional magnetic resonance imaging

    Neurosci. Lett.

    (2002)
  • A.E. Hernandez et al.

    In search of the language switch: an fMRI study of picture naming in Spanish–English bilinguals

    Brain Lang.

    (2000)
  • A.E. Hernandez et al.

    Language switching and language representation in Spanish–English bilinguals: an fMRI study

    NeuroImage

    (2001)
  • A. Hernandez et al.

    The emergence of competing modules in bilingualism

    Trends Cogn. Sci.

    (2005)
  • J. Illes et al.

    Convergent cortical representation of semantic processing in bilinguals

    Brain Lang.

    (1999)
  • M.A. Lambon Ralph et al.

    What underlies the neuropsychological pattern of irregular > regular past-tense verb production?

    Brain and Language

    (2005)
  • B. MacWhinney et al.

    Language learning: cues or rules?

    J. Mem. Lang.

    (1989)
  • J.L. McClelland et al.

    Rules or connections in past-tense inflections: what does the evidence rule out?

    Trends Cogn. Sci.

    (2002)
  • G. Meschyan et al.

    Impact of language proficiency and orthographic transparency on bilingual word reading: an fMRI investigation

    Neuroimage

    (2006)
  • K. Patterson et al.

    Deficits in irregular past-tense verb morphology associated with degraded semantic knowledge

    Neuropsychologia

    (2001)
  • S. Pinker et al.

    The past and future of the past tense

    Trends Cogn. Sci.

    (2002)
  • R.A. Poldrack et al.

    Functional specialization for semantic and phonological processing in the left inferior prefrontal cortex

    NeuroImage

    (1999)
  • C.J. Price et al.

    The myth of the visual word form area

    NeuroImage

    (2003)
  • N.T. Sahin et al.

    Abstract grammatical processing of nouns and verbs in Broca’s area: evidence from fMRI

    Cortex

    (2006)
  • M.T. Ullman

    Contributions of memory circuits to language: the declarative/procedural model

    Cognition

    (2004)
  • I. Wartenburger et al.

    Early setting of grammatical processing in the bilingual brain

    Neuron

    (2003)
  • R.J. Wise et al.

    Brain regions involved in articulation

    Lancet

    (1999)
  • J. Abutalebi et al.

    The bilingual brain as revealed by functional neuroimaging

    Bilingualism: Lang. Cogn.

    (2001)
  • T. Akhutina et al.

    Processing of grammatical gender in a three-gender system: experimental evidence from Russian

    J. Psycholinguist. Res.

    (1999)
  • R.d.D. Balaguer et al.

    Regular and irregular morphology and its relationship with agrammatism: evidence from two Spanish–Catalan bilinguals

    Brain Lang.

    (2004)
  • E. Bates et al.

    Gender and lexical access in Italian

    Percept. Psychophys.

    (1995)
  • E. Bates et al.

    Gender priming in Italian

    Percept. Psychophys.

    (1996)
  • E. Bates et al.

    Timed picture naming in seven languages

    Psychon. Bull. Rev.

    (2003)
  • E. Bates et al.

    Voxel-based lesion-symptom mapping

    Nat. Neurosci.

    (2003)
  • D. Birdsong et al.

    Regular–irregular dissociations in the acquisition of English as a second language

  • S. Bookheimer

    Functional MRI of language: new approaches to understanding the cortical organization of semantic processing

    Annu. Rev. Neurosci.

    (2002)
  • P. Cole et al.

    Grammatical incongruency and vocabulary types

    Mem. Cogn.

    (1994)
  • G.G. Corbett

    Gender

    (1991)
  • S. Dehaene et al.

    Anatomical variability in the cortical representation of first and second language

    NeuroReport

    (1997)
  • R. Desai et al.

    FMRI of past tense processing: the effects of phonological complexity and task difficulty

    J. Cogn. Neurosci.

    (2006)
  • A. Deutsch et al.

    Semantic influence on processing gender agreement: evidence from Hebrew

    J. Psycholinguist. Res.

    (1999)
  • A. Devescovi et al.

    The development of sentence comprehension in Italian and Serbo-Croatian: local versus distributed cues

  • J.T. Devlin et al.

    Semantic processing in the left inferior prefrontal cortex: a combined functional magnetic resonance imaging and transcranial magnetic stimulation study

    J. Cogn. Neurosci.

    (2003)
  • N.F. Dronkers

    A new brain region for coordinating speech articulation

    Nature

    (1996)
  • Cited by (0)

    View full text