Ongoing and future research

Letting go of the 'imitation' assumption

The idea that children learn pronunciation by imitation is deeply embedded in speech science. It's partly true, of course. The pronunciation of most individual words is copied from how the child hears others pronounce them. However, this need only concern the identity and serial order of the speech sounds that make up these words. (The 'ho' 'ri' 'pi' 'lation' of horripilation, for example.) The systemic aspects of their pronunciation need not be copied in this process at all: they will have been learnt previously. That there was any form of imitation during those prior processes of learning is only an assumption, lacking any evidence to support it. Many fundamental problems in the field are resolved by understanding that it is probably wrong, and appreciating what the alternative mechanism is.

Similarly, some aspects of prosody are imitated. But since English-speaking children adopt stress-accent as a prominence mechanism for syllables, one does not need to assume that the speech timing phenomena that characterise the language are learnt by imitation. It is much more likely that they arise as a result of the constraints imposed by aerodynamics and embodiment.

Here are just a few examples of what can now be explained:

    • Is speech an acoustic code reproduced gesturally, or a gestural code made audible?
    • What explains the properties of the tense and lax vowel classes of English? What is the underlying link between tense vowels and diphthongs?
    • Why is the /iː/ vowel in seat shorter than (the 'same') vowel in seed? Why does the /æ/ vowel shorten in triplets like ram, ramp, ramped as the number of consonants in the coda increases?
    • How can one speaker shadow another with such short latency?
    • Why don't children's patterns of speech development fit what we would expect from a 'copying' mechanism? (There are unexpected regressions and plateaus in so-called speech timing phenomena.)
    • Is there a ‘rhythm’ to speech production? Is English ‘stress-timed’?
    • Why is pronunciation teaching so ineffective?

This is a very diverse list and it can be extended. But it is not surprising that so much that was mysterious can now be explained. The two alternative mechanisms to imitation that I describe are fundamental to speech production and the effects of us having learned to pronounce in these ways are pervasive.

Rethinking the role of the power supply (speech breathing) and a new understanding of how speech is represented in the brain together have implications for many aspects of speech research.

Where my ideas came from

I didn't come up with the ideas for my research by myself. Caleb Gattegno claimed that children don't learn to talk by imitation and I came across his work as a result of learning Japanese by the Silent Way, the approach to language learning that he developed. The name reflects the silence of the teacher, not, of course, silence on part of the students. It was a shock, at first, to be asked to experiment with making sounds for myself rather than being given models to mimic from the teacher or a tape. But very quickly the teacher's silence became very liberating, and my pronunciation ended up being really quite good and certainly far better than that of friends who had learnt by the usual 'listen and repeat' approach.

This experience was one of the spurs to me trying to find out if Gattegno was right about how pronunciation is learnt (both by adults and by young children).

Current projects

If it was easy to prove how children learn to pronounce then it would have been done.

But of the various proposals in Part 1 of the thesis, some would be reasonably straightforward to test; for example, that respiratory system activity underlies the phenomenon of perceptual-centres (P-centres), as described in s.6.4.1.

In Part 2 of the thesis, I propose that the bootstrap to learning to pronounce (i.e. learning to pronounce speech sounds, as opposed to learning to pronounce words) comes from mirroring interactions between mother and child. Ian Howard and I tested this proposal through computer modelling of an infant and infant/caregiver interactions, and our results supported the proposal. Our papers are available on the downloads page.