The Weekly Newsletter of MIT Linguistics

Phonology Circle - May 5 - Franz Cozier, Adam Albright

This week’s Phonology Circle will feature the first round of WCCFL practice talks, with presentations by Franz Cozier and Adam Albright.

Time: Monday 5 May, 5pm
Location: 32-D831

Franz Cozier
Encoding perceived contrast between CC-clusters and simplified counterparts in coda CC simplification

This paper examines grammatical constraints on word-final consonant cluster inventories (VC1C2#). Crosslinguistically, languages such as Trinidad dialectal English (TE), African American English, Cameroon English, Quebec French, and Catalan show striking consistency in the set of clusters that are illicit word-finally as shown in (1) (cf. Côté 2004, Green 1992, Bobda 1994, Mascaro 1976). Languages that ban these clusters do not release their final stops (cf. Archambault & Dumochel 1993). This makes it seem likely that simplification is related to the perceptibility of C2 in the absence of release. The central claims of this paper are (1) that C2 deletion is triggered when the distinctiveness of VC1C2 and VC1, as a function of phonetic cues, falls below a particular threshold and (2) that speakers encode this perceptually based difference between simplified and preserved clusters in their grammars. Experimental results will show how the synchronic grammar of TE reflects the historical simplification process. A second experiment will confirm that perceptibility is not just something that causes loss of C2 over time but that the grammar attributes simplification to perceptual difficulty raised by unreleased C2’s

Adam Albright
Chaotic evolution in an unbiased learner

A premise of channel-base) explanations of typology is that isolated misproductions or miscategorizations may cause the signal to deviate from the speaker’s original intent in a way that may be misinterpreted as a phonetically natural change. For example, /np/ may be perceived as [mp] due to articulatory overlap and the difficulty of distinguishing coarticulated [np] from [mp]. Over time, deviations are assumed to create patterns corresponding to cross-linguistically common processes, which may then be learned and reinforced even by unbiased learners. Numerous studies have investigated whether human infants or adults behave like unbiased learners, while less attention has been paid to a prior question: are series of misperceptions actually sufficient to create the patterns observed typologically?

In this talk, I report a series of simulations designed to address this question. An unbiased inductive learner was used to investigate what patterns might arise in languages partway through a phonetically motivated change. I consider languages with a typologically dispreferred contrast such as [np] vs. [mp], with [np] words occasionally reanalyzed as [mp]. I explored the properties of hypothetical languages at a stage with a 3:1 preference for [mp] by generating 1,000 artificial lexicons, each containing 50 words with nasal+[p] clusters. Lexical items were randomly constructed to obey basic syllable constraints, with a skew towards shorter (di- or trisyllabic) words. In all of these languages, there is a 75% tendency for labials before [p] (nasal place assimilation). The question of interest is whether there are even stronger statistical patterns, due to coincidences elsewhere in the word. To test this, I submitted all 1,000 languages to an inductive model of phonological constraint discovery, which compares words that share a particular property (such as [n] or [m]) to determine the best predictors in the surrounding phonological context (such as a following [p]). It emerged that in 678/1000 languages, the algorithm found specific contexts that were more reliable predictors of nasal place. If taken seriously and extended productively to derived contexts, this constraint could lead to highly unnatural alternations. Thus, it appears that rather than leading to neutralization, phonetically natural changes may be derailed, creating unnatural statistical correlations that may be picked up and extended by an unbiased learner. I consider several biases which would prevent the learner from being lead astray by such patterns, and provide a more accurate account of the attested typology.