A computational model to investigate assumptions in the headturn preference procedure

Area of Science:

• Cognitive Science
• Developmental Psychology
• Computational Linguistics

Background:

• The Headturn Preference Procedure (HPP) is widely used to study infant speech perception.
• Interpreting HPP results relies on several underlying assumptions about infant processing and experimental design.
• A computational approach can rigorously test these assumptions.

Purpose of the Study:

• To computationally model infant speech processing using the HPP.
• To investigate assumptions regarding processing, recognition, segmentation, and inter-infant variability.
• To examine the impact of experimental factors like voice stimuli and headturn criteria.

Main Methods:

• Development of a cognitively plausible computational model processing real speech.
• Simulation of internal speech processing, recognition, and headturn generation.
• Analysis of model outputs under varying assumptions and experimental parameters.

Main Results:

• Model simulations supported assumptions that behavioral differences arise from distinct processing and involve recognition.
• Explicit word segmentation was not required to replicate observed infant behaviors.
• Infant attention span, experimenter decision criteria, and voice stimuli significantly influenced simulated HPP outcomes.

Conclusions:

• Computational modeling provides insights into the assumptions of HPP studies.
• The necessity of explicit word segmentation in early speech processing remains an open question.
• Recommendations are provided for refining HPP methodology to account for identified influencing factors.