# A role for features in speech perception

## TL;DR

This paper argues that abstract phonological features influence how second language learners perceive speech, based on two studies comparing French and English speakers.

## Contribution

The paper provides empirical evidence supporting a role for phonological features in L2 speech perception, specifically through the influence of L1 phonological status.

## Key findings

- French listeners perceive novel nasal vowels better than English listeners due to differences in the phonological status of [nasal] in their L1.
- The absence of the [SG] feature in French hinders perception of English /h/ and affects VOT values and aspiration patterns in L2 learners.
- Phonological features are shown to shape behavior across multiple segment groups, supporting a unified phonological account.

## Abstract

Archibald’s article makes a strong case for abstract symbolic representations in the phonological grammars of second language learners/users (L2ers). The evidence he brings to bear on this comes principally from the prosodic domain. However, the case for abstractness is hardest to defend in the segmental domain, specifically when it comes to motivating a role for features. The goal of this commentary is to show that L2 speech perception is mediated, in part, by features and thereby provide support for Archibald’s claim. Two speech perception studies are discussed. The first study shows that the status of the feature [nasal] in vowels in the first language (L1) grammar, as contrastive (French) or allophonic (English), impacts naive perception of novel nasal vowels. French listeners successfully perceive the novel vowels; English listeners’ success is hindered by the phonological status of [nasal] in the L1 grammar. The results are proposed to support a role for abstract phonological representations: for features; for the conditions under which they must be shared across segments; and for a theory of licensing that can capture the licensing potential of different prosodic positions. The second study shows that the absence of the feature [SG] from the L1 grammar of French negatively impacts the ability to perceive and build an appropriate representation for English /h/. The lack of [SG] is proposed to account for three types of behaviour displayed by L2ers: their failure to perceive [h] as distinct from Ø; their VOT values for ‘voiced’ and ‘voiceless’ stops which fall between those of L1 French and those of target L2 English; and their overapplication of aspiration to stops in sC clusters. It is shown that these patterns cohere under a phonological account, as features have a classificatory function and are thereby expected to shape phonological behaviour across multiple groups of segments.