Multiplicity and Equivalence of Cues
Each gestural contrast exhibits acoustic patterns that differ along a number of acoustic dimensions.
These dimensions can be artificially manipulated to evoke the percept of one or another phonetic category. As such these dimensions have been dubbed "cues."
What do we know (experimentally) about these cues:
- Every potential cue is an actual one.
- No cues are necessary to the percept of a given category:
Cues may engage in "trading relations" with one another.
Sine wave replicas of speech lack any of the cues of normal speech.
Thus, there is no way to define categories in auditory terms, which would be explained if the categories were not intrinsically auditory, but rather the acoustic signal was being used to serve as as source of information about gestures.
On this view, gestures define the categories. And there are necessary articulatory properties associated with these categories (for example "d" always involves raising of the tip of the tongue toward the back of the upper teeth).
Segmentation of signal and percept
Percepts are segmented into discrete units, (corresponding to gestures or sets of them), but the acoustic signal is not segmented in the same way.
The acoustic signal does appear to be divided into discrete chunks, but these do not correspond the the perceptual units in a one-to-one fashion.
Could such segmentation result from auditory processes or interactions?
Context-dependence of acoustic information
Same phonetic category is cued by different acoustic patterns in different contexts. This is very general phenomenon.
Conversely, the same acoustic pattern has very different phonetic consequences in different contexts.
e.g., "say" vs. "stay"
formant transitions result in percept of stop after silence, but are part of fricative percept otherwise.
Duplex Perception
Same signal can perceived as both a speech event and a non-speeech sound at the same time.
The percepts are quite different, and they change in different ways as a function of changes in context.
Reasonable conclusion that there are two modules that produce simultaneous representations of distal objects.
Audio-Visual Integration
Auditory and visual stimuli are seamlessly integrated into an amodal phonetic percept,without any awareness as to where the information came from. More on audio-visual integration
Phonetic percepts cannot be auditory, therefore.
Even pre-linguistic infants (4-month olds) appreciate the relation between audio and visual signals.
When is a signal perceived as speech?
No necessary conditions in the surface properties of the sound for predicting whether a signal will be perceived as speech.
Even pure tones varying in frequency can be perceived as speech if they follow the dynamic patterns of the resonances of the vocal tract.
Acoustic patterns seem to be perceived as speech just when the phonetic module can can interpret the pattern as resulting from a gestural structure.
Phonetic and auditory responses to the cues
If there are distinct modules for speech and general auditory processing, then the sensitivities of the systems should not (except by accident) be the same.
Experiments show that boundaries of phonetic categories are not fixed; they move in accordance with the articulatory adjustments caused by context, dialect, rate of speech.
While the auditory system may exhibit discontinuities, many ad hoc assumptions would be required to explain how these discontinuities shift in exactly those circumstances in which the relevant articulatory adjustments take place.
Non-speech sounds may exhibit categorical perception, but the boundaries turn out to not be in exactly the same positions when the stimuli are perceived as auditory events as when they cue the perception of phonetic categories.
In general, response to very same stimuli are different when they are perceived as speech vs. non-speech events:
- use of sine wave analogs, where instructions to subjects will determine whether they are perceived as speech or not. Patterns of discrimination are quite different. Example
- duplex perception--discrimination functions are quite different when subject is responding to the speech or non-speech side of the percept.