Contingent categorization in speech perception.

Area of Science:

• Psycholinguistics
• Auditory Perception
• Speech Science

Background:

• Speech signals exhibit significant variability due to factors like talker identity and phonetic context.
• This acoustic variability has spurred various theories on speech recognition mechanisms.
• Contingent categorization, a theory suggesting speech perception relies on computed expectations relative to acoustic cues, has gained traction through computational models.

Purpose of the Study:

• To empirically investigate the role of contingent encoding in human speech perception.
• To determine if listeners utilize contextual information beyond bottom-up acoustic cues.

Main Methods:

• Participants categorized fricatives in synthesized CV syllables.
• Syllables were created by splicing fricatives from one syllable with vowels from another.
• Experimental manipulation involved matching and mismatching the vowel and/or talker context between spliced components.

Main Results:

• Listener accuracy in fricative identification was reduced when contextual cues mismatched the acoustic information.
• Response times were significantly slower for fricatives presented in mismatching contexts.
• These findings indicate a reliance on contextual information during speech perception.

Conclusions:

• Empirical evidence supports the theory of contingent categorization in speech perception.
• Listeners integrate contextual expectations with acoustic information for robust speech recognition.
• Speech perception is not solely based on bottom-up acoustic processing but involves top-down influences.