Perception of anticipatory coarticulation effects

Area of Science:

• Speech Perception
• Phonetics
• Psychoacoustics

Background:

• Articulatory and acoustic studies indicate anticipatory coarticulation spans multiple speech segments.
• Previous perceptual studies suggested limited use of coarticulatory information to adjacent segments only.

Purpose of the Study:

• To investigate whether speech perception utilizes coarticulatory information beyond immediately adjacent segments.
• To determine the extent to which listeners are sensitive to anticipatory coarticulation effects.

Main Methods:

• Ten talkers produced sentences with varying consonant (C), vowel 1 (V1), and vowel 2 (V2) contexts.
• Cross-splicing techniques exchanged final syllables to create mismatches in coarticulatory information.
• Listener recognition time (RT) for the final vowel (V2) was measured in crossed and intact sentences.
• Linear Predictive Coding (LPC) analysis examined acoustic variations (F2 frequency) in pre-target segments.
• Experiments were replicated using synthetic speech with manipulated pre-target F2 variations.

Main Results:

• Recognition time for V2 was slower in crossed sentences, indicating perceptual interference.
• The degree of interference was influenced by the preceding consonant (C) and vowel (V1) context.
• Interference increased with the amount and proximity of inappropriate pre-cross acoustic information.
• Pre-target F2 variations, dependent on C, V1, and V2, predicted the amount of RT interference.
• Results with synthetic speech corroborated findings from real speech experiments.

Conclusions:

• Speech perception is not limited to processing information from immediately adjacent speech segments.
• Listeners are sensitive to acoustic effects of coarticulation across multiple segments in an utterance.
• The perceptual significance of coarticulation extends beyond local phonetic contexts.