Rational speech comprehension: Interaction between predictability, acoustic signal, and noise

Area of Science:

• Cognitive Science
• Auditory Perception
• Speech Processing

Background:

• Speech comprehension integrates multiple information sources.
• Existing models often overlook sentence-level context and noise effects.
• Prior research has limitations in controlling speech sounds and noise types.

Purpose of the Study:

• To investigate how listeners combine contextual and acoustic information during speech comprehension in noise.
• To examine the influence of top-down predictability, noise type, and acoustic signal characteristics.
• To address conflicting results regarding noise types and their impact on speech intelligibility.

Main Methods:

• An online word recognition experiment was conducted.
• Manipulated top-down predictability, background noise type, and acoustic signal properties.
• Created conditions varying in speech sound masking based on established principles.

Main Results:

• Listeners increased reliance on sentence context when acoustic signals were degraded.
• This effect persisted even with minor intelligibility differences caused by noise-speech interactions.
• Listeners probabilistically combined contextual predictions with acoustic information.

Conclusions:

• Speech comprehension involves a probabilistic trade-off between contextual and acoustic information.
• The balance between top-down and bottom-up processing is influenced by specific noise-speech sound combinations.
• Listeners adapt their information weighting based on signal clarity and noise characteristics.