Perceptual learning of co-articulation in speech

Area of Science:

• Psychology
• Linguistics
• Speech Perception

Background:

• Speech perception involves understanding sounds in context.
• Coarticulation, where sounds influence each other, poses a challenge to speech perception.
• Previous research has not fully explored how listeners learn to accommodate coarticulation.

Purpose of the Study:

• To investigate perceptual learning of coarticulation in speech.
• To determine if training can improve the ability to process coarticulated speech sounds.
• To examine the conditions under which perceptual learning of coarticulation generalizes.

Main Methods:

• Four experiments were conducted using behavioral tasks.
• Experiment 1 established a co-articulatory mismatch effect.
• Experiments 2 and 3 involved training on rhyme judgment tasks followed by phoneme monitoring, manipulating training and test stimuli.
• Experiment 4 further explored generalization with varying acoustic cues.

Main Results:

• A co-articulatory mismatch effect was observed without training.
• Training eliminated the mismatch effect, though word training led to slower reaction times.
• Perceptual learning generalized to new speech tokens only when the coarticulatory information was acoustically prominent.
• The effect of training depended on whether identical or novel mismatching tokens were used.

Conclusions:

• Listeners can learn to accommodate the coarticulated nature of speech.
• Perceptual learning of coarticulation is possible through targeted training.
• Generalization of this learning depends on the acoustic salience of coarticulatory cues and stimulus variability.