Prosodic cues to word boundaries in a segmentation task assessed using reverse correlation.

Area of Science:

• Psycholinguistics
• Speech Perception
• Auditory Neuroscience

Background:

• Listeners utilize acoustic cues to segment continuous speech into words.
• Traditional methods involve direct manipulation of hypothesized segmentation features.
• A novel approach using reverse correlation minimizes assumptions about relevant acoustic features.

Purpose of the Study:

• To investigate acoustic segmentation cues in speech perception using a data-driven method.
• To identify which prosodic features contribute to word boundary perception.
• To evaluate the efficacy of reverse correlation in speech segmentation research.

Main Methods:

• Employed reverse correlation to analyze perception of word boundaries.
• Used phonemically identical French sentence pairs with manipulated prosody.
• Introduced random fundamental frequency (f0) trajectories and segment durations as stimuli.

Main Results:

• The study identified significant perceptual roles for specific acoustic features.
• A rising fundamental frequency (f0) contour at the beginning of content words was a key cue.
• Increased vowel duration at the beginning of content words also played a prominent role.

Conclusions:

• Reverse correlation provides a valuable, assumption-light method for studying speech perception.
• Prosodic features, specifically f0 rise and vowel duration, are crucial for identifying word boundaries in French.
• These findings advance our understanding of how the brain processes spoken language segmentation.