A Single‐Stage Approach to Learning Phonological Categories: Insights From Inuktitut

Abstract

To acquire one’s native phonological system, language-specific phonological categories and relationships must be extracted from the input. The acquisition of the categories and relationships have each in their own right been the focus of intense research. However, it is remarkable that research on the acquisition of categories and the relations between them have proceeded, for the most part, independent of one another. We argue that this has led to the implicit view that phonological acquisition is a ‘two-stage’ process: phonetic categories are first acquired, and then subsequently mapped onto abstract phoneme categories. We present simulations that suggest two problems with this view: first, the learner might mistake the phoneme-level categories for phonetic-level categories and thus be unable to learn the relationships between phonetic-level categories; on the other hand, the learner might construct inaccurate phonetic-level representations that prevent it from finding regular relations among them. We suggest an alternative conception of the phonological acquisition problem that sidesteps this apparent inevitability, and acquires phonemic categories in a single stage. Using acoustic data from Inuktitut, we show that this model reliably converges on a set of phoneme-level categories and phonetic-level relations among subcategories, without making use of a lexicon.

Weak semantic context helps phonetic learning in a model of infant language acquisition

“…In theory, semantic information could offer a valuable cue for phoneme induction 1 by helping infants distinguish between minimal pairs, as linguists do (Trubetzkoy, 1939). However, due to a widespread assumption that infants do not know the meanings of many words at the age when they are learning phonetic categories (see Swingley, 2009 for a review), most recent models of early phonetic category acquisition have explored the phonetic learning problem in the absence of semantic information (de Boer and Kuhl, 2003;Dillon et al, 2013;Feldman et al, 2013a;McMurray et al, 2009;Vallabha et al, 2007).…”

Statistical Learning, Inductive Bias, and Bayesian Inference in Language Acquisition

“…Although these results are preliminary and still need to be extended to more realistic lexicons, they provide intriguing evidence that simultaneously learning linguistic generalizations at multiple levels (phonetic categories and word forms) can actually make the learning problem easier than learning in sequence. Dillon, Dunbar, & Idsardi (2011) have also recently studied the acquisition of phonemes and phonological rules from acoustic data, using a Bayesian model. Like Feldman et al (2009b, submitted), they recognize that word forms are comprised of phonetic categories.…”

“…While this is a reasonable model of acquisition, it nonetheless implies a learning sequence where learning phonetic categories must happen before learning phonemes and phonological rules. Dillon et al (2011) explore whether relaxing this assumption could lead to better learning.…”

“…In particular, Dillon et al (2011) investigate the vowel system of Inuktitut, which has three phonemes with two allophones each (for a total of six phonetic categories). This kind of vowel system is not uncommon in the world's languages (e.g., it is shared by Quechua and many dialects of Arabic), and so represents a realistic learning problem.…”

“…Their findings are similar to Feldman et al's more general finding: Learning multiple levels of representation simultaneously can be a better strategy than trying to learn them in sequence. In particular, Dillon et al (2011) found that the learner who only identifies phonetic categories will converge on phonetic categories that make it much harder to formulate the correct phonological rule (and so define the correct phonemes). This problem occurs because the learner disregards the linguistic context when identifying its categories, and uses only the acoustic information.…”