The Influence of Different Prosodic Cues on Word Segmentation

Area of Science:

• Psycholinguistics
• Speech Perception
• Language Acquisition

Background:

• Segmenting continuous speech into words is crucial for spoken language learning.
• Listeners utilize both statistical transitional probabilities and prosodic cues to identify word boundaries.
• The relative importance and interaction of these cues, especially various prosodic features, are not fully understood.

Purpose of the Study:

• To investigate the relative importance of different prosodic cues in speech segmentation.
• To compare the effects of various prosodic cues against statistical cues in artificial language learning.
• To determine how prosodic cues interact with and potentially override statistical cues.

Main Methods:

• Artificial language learning experiments with native German speakers.
• Participants learned meaningless trisyllabic words from a continuous speech stream.
• Compared a baseline of statistical cues only with conditions manipulating word-final syllable prosody (pause, lengthening, shortening, pitch changes).
• Evaluated segmentation performance across three tasks of varying difficulty.

Main Results:

• Pauses and final syllable lengthening significantly facilitated speech segmentation, with pauses being most effective.
• Final syllable shortening impaired segmentation, demonstrating that prosodic cues can override statistical cues when in conflict.
• Pitch variations (lower or higher) had minimal impact on speech segmentation.
• Segmentation performance varied across tasks, indicating robustness and generality of the findings.

Conclusions:

• Prosodic cues, particularly pauses and duration, play a significant role in speech segmentation.
• Prosodic cues can override statistical information, highlighting their importance in boundary detection.
• Syllable duration appears more critical than pitch for speech segmentation in this context.
• Findings contribute to understanding universal boundary cues versus language-specific stress patterns in speech processing.