Core phonology: Evidence from grammatical universals

The human capacity for language is one of the most contentious topics in cognitive science. While some researchers attribute language to domain-general mechanisms, others postulate a specialized language system. When it comes to the phonological component, however, even proponents of domain-specificity concede that specialization is unlikely (Fitch et al., 2005). Phonological competence, in this view, is the product of experience, auditory perception, and motor control. And indeed, phonological systems are intimately grounded in phonetics. But while the domain-general perspective can account for this fact, it offers no explanation for several key features of language. It fails to explain why all languages—signed and spoken—have a phonological system, why phonological systems emerge spontaneously, in the absence of a model (Sandler et al., in press), and why the cultural invention of reading and writing invariably recapitulates phonological principles. Such observations, however, are readily explained by the view of phonology as a core knowledge system (Berent, in press). Core knowledge systems are specialized, early knowledge frameworks and scaffolds for subsequent learning (Carey & Spelke, 1996). Crucially, core systems manifest a unique, universal design—it is the argument from design that presents the strongest test for specialization. Applying this test to phonology, one would expect that, if phonology is a specialized core-knowledge system, then all grammars should exhibit universal principles that are largely invariant across individuals. Like other species-specific communication systems (e.g., birdsong; Fehér et al., 2009), universal phonological principles could be triggered by experience, but they are unattainable by learning (e.g. by statistical learning, induction, etc., Samuels, 2004). Optimality Theory (Prince & Smolensky, 1993/2004) offers specific, amply testable hypotheses concerning those principles. In this account, all grammars include a common set of markedness constraints that express (dis)preferences for certain linguistic structures (e.g., structure A≻B). These constraints form part of the grammars of all speakers, irrespective of whether the structures under consideration (e.g., A, B) are present in one’s language or absent. In this talk, I summarize the results of an experimental research program that tests of this hypothesis. I proceed in two steps. First, I demonstrate that individual speakers’ preferences mirror the language typology. I next gauge the source of these preferences by pitting universal-grammatical explanations against various extra-grammatical alternatives. My specific case study concerns the sonority-related restrictions on onset structure (e.g., bl in block). Across languages, onsets such as bl are preferred (e.g., they are more frequent) relative to onsets like bdif, which, in turn, are preferred to lbif. Figure 1 illustrates this regularity in a representative sample of 90 diverse languages (data from Greenberg, 1978). These systematic distributional regularities are attributed to sonority (s)—a scalar phonological property that correlates with intensity: liquids (e.g., l) are more sonorous (s=3) than nasals (n,m; s=2), which in turn, are more sonorous than stops (e.g., b,p; s=1). Accordingly, onsets such as bl manifest a large cline in sonority (∆s=3), bn has a smaller cline (∆s=3), bd manifests a sonority plateau, whereas lb exhibits a sonority fall. (∆s=-3). The typological preference (e.g., blif≻bnif≻bdif≻lbif) is thus captured by grammatical constraints that favor onsets of large sonority clines—the larger the cline, the less marked the onset (Clements, 1990). Of interest is whether those constraints are universally active in the grammars of individual speakers. To address this question, I examine whether people favor onsets of large sonority clines over smaller clines even when such onsets are unattested in their language. Participants’ preferences, in turn, are gauged from their capacity to accurately identify these onsets. Past research has shown that speakers systematically misidentify ill-formed onsets that are unattested in their language (e.g., tlatәla; Pitt, 1998). Grammatical ill-formedness presents one explanation for this phenomenon (Anttila, 1997). In this account, the markedness of ill-formed onsets prevents their faithful encoding by the grammar, and consequently, such onsets must be systematically recoded as better-formed structures (e.g., as tela). To the extent small sonority clines are illformed, then onsets with small sonority clines (e.g., lb) should be recoded, hence, more likely to be misidentified than better-formed onsets (e.g., bd).