Neural commitment to the "wrong" sounds: can native language phonological expertise help explain the sensitive period for language acquisition?

Amy Finn, University of California, Berkeley
Carla L. Hudson Kam, University of California, Berkeley
Marc Ettlinger, Northwestern University
Mark D'Esposito, University of California, Berkeley

Throughout development, input is crucial for pruning many sensory circuits. Indeed, circuits expert in auditory and perhaps also language processing are pruned based on exposure to native language sounds early in life. Here we ask if such experience provides the basis for the neural representation of language, and further, if this early neural commitment to the phonetic and phonological properties of the first language might contribute to the sensitive period for language acquisition more broadly. In particular, we examine if a commitment to one's native language sounds carries a cost for adults attempting to learn a new language with a phonology that is highly distinct from that of their first language.

We explore behavioral and neural consequences of this hypothesis by teaching one of two miniature artificial languages (MALs) to native-English-speaking adults over the course of 4 days, and testing and scanning them afterward. Crucially, the phonological inventory of one MAL overlaps extensively with English (OMAL), while the other is maximally different from English (DMAL). Otherwise, the two MALs are structurally the same and different from English: each has 30 nouns, 4 verbs , SOV word order, regular noun class suffixes, and subject-verb agreement. Post-exposure, participants were tested on their knowledge of the MAL grammars: word-order, noun class suffix pairings, and verb agreement. They were then scanned while assessing the grammaticality of MAL sentences. Since both MALs are structurally different from English, while only one is phonologically different, we can probe the effect of phonological differences on grammar learning and on the corresponding neural recruitment.

Twenty participants (10/condition) participated. Participants successfully learned all components of the languages regardless of condition. However, DMAL learners took more time to learn the vocabulary and while they were not worse overall on measures of noun-class suffixes and verb-agreement, they were worse when tested on novel, as opposed to familiar, items in these tests. DMAL learners were therefore less likely to generalize their knowledge than OMAL learners.

If early neural commitment to one's native language phonology accounts for these learning differences, it should follow that neural recruitment of language areas be substantially different for the two MALs. In particular, we might expect that recruitment for DMAL be less localized to classic language areas. The fMRI results showed that neural recruitment for DMAL was significantly more diffuse than recruitment for OMAL. Individuals processing DMAL sentences recruit more auditory, phonological, and frontal areas. Auditory areas include classic primary auditory cortex and other regions along the Superior Temporal Gyrus. Phonological areas recruited include the Supramarginal Gyrus and WernickeÕs areas. DMAL learners also recruited more Dorsal Lateral Prefrontal Cortex (DLPFC), the primary substrate for working memory. Together, behavioral and neural data suggest that learning of L2 grammar is affected by the phonological and auditory scaffolding that support it. Because the DMAL phonology is different from English, learners are less able to take advantage of their existing phonological knowledge and its corresponding neural circuitry.