Speech predictability can hinder communication in difficult listening conditions

Area of Science:

• Auditory Perception
• Psychoacoustics
• Speech Processing

Background:

• Difficult listening conditions, such as noisy environments, typically involve challenges to speech intelligibility.
• It is hypothesized that speech intelligibility may improve over time due to noise adaptation and speech predictability.
• These factors are expected to facilitate the recognition of upcoming words in sentences.

Purpose of the Study:

• To investigate how speech intelligibility changes across word positions within sentences in challenging auditory environments.
• To determine the influence of noise adaptation and speech predictability on word recognition over time.
• To examine the underlying mechanisms, including prediction accuracy, contributing to speech intelligibility fluctuations.

Main Methods:

• 100 normal-hearing listeners were presented with sentences in quiet and noisy conditions.
• Word recognition was measured as a function of word position within the sentence.
• Masker sounds were used at individualized levels to achieve a 50% sentence recognition probability.

Main Results:

• Contrary to expectations, word recognition was highest for the initial word and progressively declined with later word positions.
• This decline in recognition was not attributed to differences in word audibility, word type, age, or working memory capacity.
• A probabilistic model indicated that inaccurate predictions, rather than accurate ones, led to the observed decline, overriding noise adaptation benefits.

Conclusions:

• Speech predictability can hinder rather than help word recognition as a sentence progresses.
• Inaccurate predictions generated by misunderstandings can outweigh the benefits of accurate predictions.
• The study concludes that prediction inaccuracies significantly impact speech intelligibility in challenging listening scenarios.