Modeling the Acquisition of Phonological Interactions: Biases and Generalization

Brandon Prickett,Gaja Jarosz

doi:10.3765/amp.v9i0.4901

Abstract

In this paper we computationally implement four different theories for representing opaque and transparent phonological interactions: Harmonic Serialism, Stratal OT, Two-Level Constraints, and Indexed Constraints. We then show that these theories make unique predictions on two tasks: (1) a learning-bias task, based on previous experimental work with humans and (2) a novel generalization task that no human data exists for. Our results in (1) show that serial models predict that transparent languages should be easier to acquire, while parallel models do not. Furthermore, the results for (2) show that all four of the theories we test make unique predictions for how humans should generalize to novel phonological interaction types.

Highlights

Phonological processes in a language have the potential to interact with one another in numerous ways (Kiparsky 1968, 1971)
It was first extended to serial HS by Jarosz (2016), and for the current project, we extended it to a two-level Stratal OT framework
We considered a bias to be present in any model for which the accuracy across language conditions for the relevant word types exhibited the relative preferences observed in Prickett’s study (2019)

Summary

Introduction

Phonological processes in a language have the potential to interact with one another in numerous ways (Kiparsky 1968, 1971). Counterbleeding and counterfeeding interactions are typically called opaque, because in the surface forms of the language, certain processes (in this case, palatalization) seem to either over- or under-apply in places where they shouldn’t (Kiparsky 1971, McCarthy 1999, Baković 2011). Little work has directly compared the differences in predictions made by various theories of opacity regarding learning and generalization This is the focus of the present paper. The paper proceeds as follows: §2 summarizes existing work on both human and machine learning of phonological interactions, §3 presents the analyses for opaque interactions that each of the theories of interest use, §4 describes the novel computational modeling experiments we ran, §5 presents the results of those simulations, and §6 interprets them and discusses their implications

Background

Representing phonological interactions

Simulations

Results

Discussion