Part 1
Neurolinguistic Perspective
Chapter 1
Identifying the Neural Substrates of Second Language Acquisition: What is the Contribution from Functional and Structural MRI?
MAURITS VAN DEN NOORT, PEGGY BOSCH, TARIK HADZIBEGANOVIC, KATRIEN MONDT, MARCO HAVERKORT and KENNETH HUGDAHL
The Critical Period Hypothesis
In the language acquisition field, the proposition that there is a critical or sensitive period for language acquisition is an influential one. Penfield and Roberts (1959) and Lenneberg (1967) were among the first to hypothesize that there was a critical period for language acquisition. The idea of a critical period was originally inspired by research in biology referring to the phenomenon that certain skills, such as a particular type of birdsong, might only be acquired if adequate external stimulation is available during a certain temporal frame in the ontogenetic evolution of the organism. If this temporal frame is missed, acquisition of the specific skill becomes impossible or will at least end up as incomplete (Indefrey & Gullberg, 2006).
Since the 1960s, much ink has been spilt trying to explain the role of age in second and foreign language acquisition (Birdsong, 2006). Scientists often talk about 'the critical period for language acquisition' as if there were only one critical period for all features of language. It is, however, widely agreed that there are different critical phases for different linguistic subsystems, for example phonology and syntax (see e.g. Long, 1990; Seliger, 1978), and there is no unique sensitive period for all parts of language (Van Boxtel, 2005). In addition, the notion of different critical periods is consistent with current neurobiological views on critical periods in other contexts (Knudsen, 2004).
The negative correlation between age of learning onset and eventual asymptotic performance is best described by an 'earlier is better' rule. Much recent experimental data have accumulated supporting this generalization, largely offering a maturational account of age effects and indicating a particular sensitive period that limits both first language (L1) acquisition and second language (L2) acquisition (Bates et al., 1992; Birdsong & Molis, 2001; Huttenlocher, 1990; Long, 1990).
If a critical period puts constraints on L2 acquisition, one would suppose that the observed performance should correlate negatively with the age at which L2 learning starts. Moreover, according to this hypothesis, the effect should already be visible in individuals who started L2 learning before the end of maturation. In addition, it should be impossible for all or at least most late learners to perform in the range of native controls. However, if constraints put on attainment are largely maturational in nature, then they should pertain to L2 acquisition in general. Following these lines of reasoning (Birdsong & Molis, 2001), Johnson and Newport (1989) proposed a maturational model of L2 attainment. Although the reported findings and corresponding interpretations of Johnson and Newport have been widely acknowledged, another line of evidence is directed against this model.
For instance, in several experimental studies, postmaturational age effects were reported (Bialystok & Hakuta, 1994, 1999; Birdsong, 1992; Flege, 1999). Moreover, many studies have found native-likeness results among late L2 learners (Birdsong, 2006), who tended to perform similarly to native speakers on a variety of language tasks (Bongaerts, 1999; Cranshaw, 1997; Van Boxtel, 2005). Taken together, a rather controversial idea about the existence of a critical period for language acquisition is still an unsolved and highly debated issue (Birdsong, 1999; Hyltenstam & Abrahamsson, 2001; Marinova-Todd et al., 2000; Scovel, 2000; Singleton, 2001; Van Boxtel, 2005; Wartenburger et al., 2003). As a consequence, we feel it is natural to ask what could functional magnetic resonance imaging (fMRI) and related non-invasive structural neuroimaging techniques add to this vivid discussion?
Neuroimaging Research
Neuroimaging techniques seem to be very useful for investigating the issue of eventual neurophysiological modifications in the brain before the end of maturation that can account for the general tendency to attain a higher level of proficiency in early versus late learners. For instance, if late learners could be shown to process their L2 in different brain areas than early learners, this would further support the Critical Period Hypothesis (CPH) for L2 acquisition (e.g. see Van Boxtel, 2005).
The question about how second and/or third languages are represented and maintained in the brain is not new and has been the focus of much previous neuroimaging and related research. One important proposal is that, in bilinguals and multilinguals, different languages are represented and processed in distinct brain regions, resulting thus in multiple and language-specific neural network systems. Some indirect support for this hypothesis comes from studies with bilingual and /or multilingual patients diagnosed with aphasia (Paradis, 1989). In some cases, it was found that only one of the acquired languages was affected (Albert & Obler, 1978; Paradis, 1995). Moreover, fairly often, it has been reported that different languages can recover to different degrees or even that there is an antagonistic pattern of recovery between L1 and L2 (Paradis, 1977). In addition, in other cases, neurosurgery was shown to have a selective impairment of only one language in bilinguals (Gomez-Tortosa et al., 1995). Finally, the hypothesis that different languages are represented and processed in distinct brain regions is supported via findings uncovered in electrical stimulation studies. These findings demonstrated that in multilingual subjects, different languages may be disrupted selectively (Black & Ronner, 1987; Ojemann & Whitaker, 1978; Roux & Trémoulet, 2002).
In several event-related potential (ERP) studies, which have a high temporal resolution, it has been demonstrated that the temporal aspects of high-proficient L2 use are by and large comparable to those of L1 use, importantly, even when L2 acquisition was initiated at the age of 12 or after that (Hahne, 2001; Hahne & Friederici, 2001; Ojima et al., 2005; Proverbio et al., 2002; Stowe & Sabourin, 2005). The findings of the ERP studies demonstrated that as L2 proficiency increases, the ERP components in the L2 become more similar to L1 (Birdsong, 2006).
More recently, positron emission tomography (PET) and fMRI have enabled a more direct study of the neural representation of language in bilinguals and multilinguals since both PET and fMRI have a better spatial resolution than ERP. However, the overall results that have been reported in neuroimaging studies so far are not quite consistent. For instance, in some investigations (Dehaene et al., 1997; Kim et al., 1997; Mondt, 2007; Perani et al., 1996; Van den Noort et al., 2006c; Yetkin et al., 1996) at least partially separate representations for different languages were detected. However, in other investigations (e.g. Hasegawa et al., 2002; Hernandez et al., 2001; Illes et al., 1999; Klein et al., 1995) no evidence was found, indicating that languages have been organized in distinct brain regions. But again, only in a few studies (Chee et al., 1999; Klein et al., 1999; Pu et al., 2001; Vingerhoets et al., 2003) common neural representations were reported (Van den Noort et al., 2006a).
The problem with PET and fMRI approach to L2 acquisition and the associated age factor is that the current results are, as they stand, far from consistent, which can partially be explained by two major problems: (1) different research groups tend to use different kinds of experimental paradigms, making comparisons among bilingual and multilingual studies very difficult if not impossible and (2) the experimenters do not use the same selection criteria for the inclusion of their subjects. As a consequence, most of the emerging differences reported in PET and fMRI studies can be accounted for by differences of the related proficiency level, L2 onset and L2 exposure of the bilingual and multilingual subjects (Indefrey, 2006).
Nevertheless, reliable differences among the hemodynamic activation patterns in L1 and L2 language processing have been reported, but only for subgroups of bilingual speakers and largely in the direction of more pronounced activation patterns during L2 processing (Abutalebi et al., 2001; Indefrey, 2006; Stowe & Sabourin, 2005). In particular, the relative effect of the three important factors (onset, proficiency and exposure) seems to vary between the employed experimental paradigms and between the language-processing components involved. In word-level production, all three factors seem to play an important role, whereas for word-level semantic processing in comprehension tasks, this does not seem to be the case. Here, L2 onset and exposure do not seem to represent major factors. On the other hand, L2 onset seems to be the most relevant factor for activation differences associated with syntactic processing in sentence comprehension tasks. With respect to this, however, it is necessary to note here that even in late L2 learners, pronounced L2 syntactic processing activation patterns only seem to become apparent when participants are asked to make explicit metalinguistic judgments (for a more detailed discussion, see Indefrey, 2006).
Additionally, it is important to note here that most neuroimaging investigations do not or only rarely focus on specific linguistic subsystems, like syntax, and as a consequence, the interpretation of the results in general is a rather difficult endeavor. For instance, we still have a rather incomplete picture about the exact function and interactions of brain areas involved in language production and perception. Moreover, it still remains unclear how exactly the brain works with respect to L1 and we know even less about the underlying neurological processes involved in L2 acquisition. Finally, it cannot be totally excluded that certain approaches may simply not be appropriate enough to reveal subtle differences (Van Boxtel, 2005).
Structural Brain Changes
Inferior parietal region
Recently, structural magnetic resonance imaging (MRI) has been used in L2 research as well (Van den Noort et al., 2005). Could structural MRI provide more direct evidence on whether there are neuroanatomical changes in the brain prior to the end of maturation that can further explain why early L2 learners attain a higher level of proficiency than late L2 learners?
In a recent investigation by Mechelli et al. (2004), voxel-based morphometry (Ashburner & Friston, 2000; Good et al., 2002) was employed to analyse structural plasticity in healthy right-handed English and Italian bilinguals and to assess the differences in the grey- and white-matter density between bilingual and monolingual subjects. In total, 83 subjects entered their study: 25 were monolingual who had little or no previous exposure to L2 and 25 were 'early' bilinguals, who had learned a second European language before the age of 5 and who had practised it regularly since. A total of 33 subjects were 'late' bilinguals who had learned a second European language between the ages of 10 and 15 and practised it regularly for at least 5 years. Moreover, all participants were native English speakers and factors like age and level of education were controlled for in this study.
Voxel-based morphometry showed that grey-matter density in one specific area in the brain, namely the inferior parietal cortex, was greater in bilinguals than monolinguals. This effect was significant in the left hemisphere and a trend was also observable in the right hemisphere. Although the observed increase of grey-matter density in the inferior parietal cortex was common to both early and late bilingual subjects, the effect was stronger in the early bilinguals in both hemispheres. However, no other significant effects were found in either grey or white matter.
In addition, Mechelli et al. (2004) were interested in the question whether there was a relation between brain structure and L2 proficiency and age of acquisition. In order to test this, 22 additional participants entered their study. All participants were native Italian speakers, who had learned English as an L2 when they were between 2 and 34 years old. L2 reading, writing, production and speech comprehension abilities were addressed using a battery of standardized neuropsychological tests. The results indicated that the overall proficiency, as indexed by principal component analysis (PCA), correlated negatively with age of acquisition. Quite remarkably, voxel-based morphometry further revealed that L2 proficiency correlated with grey-matter density in exactly the same brain area, namely the left inferior parietal region. Furthermore, grey-matter density in this region correlated negatively with the age of L2 acquisition. Finally, no other significant effects were reported in either grey or white matter (Mechelli et al., 2004).
Taken together, Mechelli et al. (2004) detected an increase in the density of grey matter in the left inferior parietal cortex of bilinguals relative to monolinguals, which is more pronounced in early rather than late bilinguals, and have additionally demonstrated that the density in this region increases with L2 proficiency but decreases as the age of acquisition increases. The reported findings could originate from a genetic predisposition to increased density, or from a structural experience-induced modification (Golestani et al., 2002). It is most likely that early bilinguals acquire an L2 via social experience, rather than as a result of a genetic predisposition. According to the authors, the findings indicate that the structure of the human brain is altered by the experience of acquiring an L2. Previous L2 studies with fMRI had also shown activation in the inferior parietal region, for example during verbal-fluency tasks (Poline et al., 1996; Warburton e...