Identifying Links Between Latent Memory and Speech Recognition Factors.

Area of Science:

• Cognitive psychology
• Auditory neuroscience
• Speech processing

Background:

• Speech recognition accuracy is influenced by memory abilities.
• Traditional correlation methods may oversimplify the relationship between memory and speech recognition.
• Latent factor modeling offers a more nuanced approach to understanding cognitive structures.

Purpose of the Study:

• To investigate the underlying structure of memory and its association with speech recognition.
• To compare latent factor models with traditional correlational approaches.
• To determine how different memory tasks contribute to speech recognition in noise.

Main Methods:

• Fitting latent factor models to trial-level data from memory and speech recognition tasks.
• Analyzing vocoded sentence recognition in a two-talker masker across varying signal-to-noise ratios.
• Comparing models with different numbers of latent factors using Expected Log pointwise Predictive Density.

Main Results:

• Speech recognition accuracy was best explained by a single latent factor.
• Memory performance was best explained by two latent factors: a general factor and a digit-span-specific factor.
• The general memory factor strongly correlated with speech recognition, while the digit-span factor did not.

Conclusions:

• Latent factor analysis provides a robust method for exploring cognitive structures.
• A common latent memory factor underlies individual differences in speech recognition.
• Interpretations of memory task performance in relation to speech recognition require careful consideration of task-specific factors.