Talker information influences spectral contrast effects in speech categorization

Area of Science:

• Auditory Perception
• Speech Processing
• Psychoacoustics

Background:

• Spectral contrast effects magnify spectral differences, aiding speech categorization.
• Previous research debated whether talker variability impacts these effects.

Purpose of the Study:

• Investigate how talker variability and spectral properties affect spectral contrast effects in vowel categorization.
• Clarify conflicting findings on the role of talker information in speech perception.

Main Methods:

• Listeners categorized vowels after hearing sentences under three talker conditions: one talker/one sentence, one talker/many sentences, and many talkers/many sentences.
• Spectral peaks in sentences were amplified (+20 dB or +5 dB) to manipulate spectral contrast.

Main Results:

• With large spectral peaks (+20 dB), talker information did not alter spectral contrast effects.
• With modest spectral peaks (+5 dB), spectral contrast effects were smaller when listeners heard many talkers compared to a single talker.
• Constant recalibration to new talkers diminished sensitivity to modest spectral cues, reducing contrast effects.

Conclusions:

• Talker information does not influence spectral contrast effects when spectral cues are robust.
• Talker variability can diminish spectral contrast effects when spectral cues are weak, due to listener recalibration.
• Findings reconcile conflicting reports on talker influence in speech categorization.