The Theory

• Objects of speech perception are the intended gestures of the speaker, represented as invariant motor control structures for linguistically significant movements.
• Speech production and perception are innately linked in a specialized "module" responsible for both perception and production of phonetic structures. The perceptual side of this module converts from acoustic signal to gesture automatically.

Why speech perception requires some theory

Relationship between gesture and signal is not straightforward.

Gestures for successive phonological units overlap in time ("coarticulation").

Sound is determined by multiple overlapping gestures: resonances are determined by overall shape of the vocal tract (x-ray data from Ohman, 1967).

idi	ada	udu

Relation of gesture to signal is

• complex (due to overlap)
• but systematic (determined by physics).

Due to coarticulation, the same gesture will be represented in different contexts by different sounds.

For example, in CV syllables, information about constrictor ("place" information) is carried by the transition of the second formant. For a tongue-tip (coronal) stop, this will be rising in some vowel context, but falling in others.

This accounts for lack of invariance of formant transitions.

However...

In isolation, these varying transitions sound like completely different sounds (chirps).

How is it that in a speech context, they yield a percept of the same consonant?