The acquisition of speech categories: beyond perceptual narrowing, beyond unsupervised learning and beyond infancy

Abstract

An early achievement in language is carving a variable acoustic space into categories. The canonical story is that infants accomplish this by the second year, when only unsupervised learning is plausible. I challenge this view, synthesizing five lines of developmental, phonetic and computational work. First, unsupervised learning may be insufficient given the statistics of speech (including infant-directed). Second, evidence that infants "have" speech categories rests on tenuous methodological assumptions. Third, when the ecology of the learning environment is distinguished from the mechanisms of learning, more powerful error-driven mechanisms may contribute. Fourth, several implicit supervisory signals are available to older infants. Finally, development is protracted through adolescence, enabling richer avenues for development.Infancy may be a time of organizing the auditory space, but true categorization only arises via complex developmental cascades later in life. This has implications for critical periods, second language acquisition, and our basic framing of speech perception.

Relationships between native and non-native speech perception.

“…This is consistent with the more modern view that speech perception follows a more protracted developmental pattern, in which perceptual representations continue to be refined (this includes becoming more graded) at least throughout adolescence (Hazan & Barrett, 2000; Idemaru & Holt, 2013; McMurray, 2022; McMurray et al, 2018; Nittrouer, 2004; Zevin, 2012). Thus, graded perception of speech seems to be the norm (McMurray, 2022; McMurray et al, 2018; Toscano et al, 2010), though individuals differ in the extent of this gradiency (Fuhrmeister & Myers, 2021; Kapnoula et al, 2021, 2017; Kong & Edwards, 2016). However, these differences may not have much of an impact on how accurately listeners can perceive non-native speech sounds.…”

“…This is consistent with the more modern view that speech perception follows a more protracted developmental pattern, in which perceptual representations continue to be refined (this includes becoming more graded) at least throughout adolescence (Hazan & Barrett, 2000;Idemaru & Holt, 2013;McMurray, 2022;McMurray et al, 2018;Nittrouer, 2004;Zevin, 2012). Thus, graded NATIVE AND NON-NATIVE SPEECH PERCEPTION 9 perception of speech seems to be the norm (McMurray, 2022;McMurray et al, 2018;Toscano et al, 2010), though individuals differ in the extent of this gradiency (Fuhrmeister & Myers, 2021;Kapnoula et al, 2021Kapnoula et al, , 2017Kong & Edwards, 2016). However, these differences may not have much of an impact on how accurately listeners can perceive non-native speech sounds.…”

Infant Phonetic Learning as Perceptual Space Learning: A Crosslinguistic Evaluation of Computational Models

“…Traditionally, these accounts have assumed that infants learn phonetic categories, and the emergence of such language‐specific categories affects infants' ability to perceive differences between some phones in other languages. Recently, such accounts have been challenged (Feldman et al., 2021; McMurray, 2023), and an alternative view was proposed (Feldman et al., 2021): infants start by learning perceptual spaces, while category learning comes later in life. Under this view, infants' discrimination ability changes due to the transformation of the acoustic similarity space.…”

Moving away from deficiency models: Gradiency in bilingual speech categorization

“…First, even in monolingual children, speech perception skills develop slowly. This is unlike the most common views of speech categorization, which argues that speech categorization stabilizes in infancy ( Werker and Curtin, 2005 ) (but see McMurray, 2022 ); critically in the context of multilingualism, it offers a gentle challenge to the notion that plasticity tapers off at later ages due to some kind of critical period – in fact, speech perception is developing quite slowly, implicating plasticity that may be available throughout the lifespan. Second, a gradient representation, rather than a discrete or categorical one, seems to be something desirable that people are attempting to develop.…”

“…The question is why? Classic developmental work argued that speech categories are formed during the first year of life and that the emergence of these categories and their structure was associated with a sensitive period ( Werker and Tees, 1984 ) (though see McMurray, 2022 ). If we assume CP as a model of speech perception, this can then explain adult learners: many new L2 distinctions comprise within-category distinctions in the native language (e.g., the English l/r distinction which lies within the Japanese category).…”