You are viewing the SNL 2017 Archive Website. For the latest information, see the Current Website.

Poster C68, Thursday, November 9, 10:00 – 11:15 am, Harborview and Loch Raven Ballrooms

The language network of polyglots

Olessia Jouravlev1,2, Zachary Mineroff1, Evelina Fedorenko1,3,4;1Massachusetts Institute of Technology, 2Carleton University, 3Harvard Medical School, 4Massachusetts General Hospital

The neurocognitive mechanisms of impaired language processing have received a lot of attention in the last few decades. Much less is known about individuals with a special talent for language, like those able to master multiple languages (Erard, 2012). We report the first fMRI investigation of seventeen polyglots (M(languages)=11.6; range(languages)=5-55). We asked two questions: (1) Does the language network of polyglots differ from that of non-polyglots?; and (2) How are multiple languages represented in polyglots’ brains? Study 1: The language network was identified in each individual with a language localizer that contrasted responses to sentences and nonwords (Fedorenko et al., 2010). The polyglots were compared to carefully pairwise-matched (including on IQ) non-polyglots, and a larger population of non-polyglots (n=217) all of whom had performed the same localizer. The polyglots showed both less extensive activation (p<0.001) and a smaller Sentences>Nonwords effect (p<0.001). No group differences were observed in two control brain networks (the Multiple Demand network and the Default Mode Network), arguing against ubiquitous differences in information processing between polyglots and non-polyglots. Why might polyglots use smaller patches of cortex to process language? One possibility is that they process language more efficiently from birth, even as they acquire their first language. Another possibility is that language processing becomes more efficient as a result of acquiring multiple languages. Without establishing a genetic basis for polyglotism combined with longitudinal investigations of individuals as they acquire new languages, both possibilities remain viable. Study 2: We examined the polyglots’ neural responses to passages in their native language (L1), three non-native languages of high to moderate proficiency (L2-L4), two cognates of languages familiar to the participant (L5-L6), and two completely unfamiliar languages (L7-L8). Each language included a control scrambled-speech condition matched to the critical conditions acoustically but with no discernable linguistic content (Overath et al., 2015). The Intact>Scrambled contrasts for the different languages activated highly overlapping areas within the language network. The Intact>Scrambled effect was reliable in all languages (ps<0.03), but its size generally scaled with proficiency, decreasing from L2 to L8, except for the response to L1, which was relatively low. Statistical comparisons revealed that a) the response to L1 was marginally lower than to familiar non-native languages (p=0.05); b) the response to both L1 and familiar non-native languages was stronger than to unfamiliar (non-cognate) languages (ps<0.004); c) the response to familiar non-native languages was stronger than to unfamiliar cognate languages (p<0.001); and d) the response to unfamiliar cognate languages was stronger than to unfamiliar non-cognate languages (p=0.05). Thus the ability to extract high-level linguistic information from the speech signal appears to lead to stronger responses in the language regions. Further, the response scaled with proficiency: the Intact>Scrambled effect decreased from L2 to L3 to L4, and could be predicted from self-rated proficiency ratings (p=0.038), in line with the idea that as proficiency increases one is able to extract progressively more meaning from the signal. However, one’s native language constituted an exception: the response was lower than to familiar non-native languages, perhaps reflecting greater efficiency.

Topic Area: Multilingualism

Back to Poster Schedule