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, synthesising 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, the fact that the ecology of the learning environment is unsupervised does not rule out more powerful error driven learning mechanisms. 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 organising the auditory space, but true categorisation 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.

KEYWORDS:

• Speech perception
• speech categorization
• infancy
• category learning
• development
• language development
• unsupervised learning
• statistical learning

Acknowledgements

The author would like Samantha Chiu for key insights while developing the ideas presented here, Keith Apfelbaum and Ethan Kutlu for comments on an earlier draft; Michael Ramscar for helpful discussions about discriminative and associative learning; and Jessie Nixon for patience and insight during the editorial process.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1 Few extant theories wholly embrace the canonical view—many people espouse a kind of hybrid, and some of the research challenging it was conducted by the people who developed the canonical view. Nonetheless this is the view that appears in most textbooks (Gerken, 2009; Keil, 2014; Traxler, 2012) and in the kinds of review papers used in graduate seminars (Werker, 2018). It continues to offer foundational assumptions that still shape much of the field, such as the idea that speech category learning occurs early in development, that speech is perceived categorically, and that unsupervised learning is the only viable mechanism. Indeed I have found no reviews or texts even mention older development, or any alternative to categorical perception, much less grapple with the implications for infant development.

2 This was such a pervasive belief in infant speech category learning that when I interviewed for graduate school, I told a famous connectionist that the problem of learning speech categories was so clearly unsupervised that backpropagation was clearly worthless and not to be pursued further. Needless to say, I did not get into that programme, and this footnote serves as a mea culpa. Please forgive my naivete, Dave.

3 Note that in some literatures, the term associative learning has been applied to both Hebbian style and discriminative learning; here I’m using it in the more narrow framing, for lack of a better umbrella term.

Additional information

Funding

This work was supported by the National Institute on Deafness and Other Communication Disorders [grant number DC-00242, DC-008089].