Can cognitive performance predict individual differences in speech recognition at a cocktail party?

Although cognition is widely assumed to contribute to speech recognition at a cocktail party, the empirical evidence supporting this assumption is surprisingly limited. One reason for this lack of evidence is that cognitive contributions are often tangled with confounding factors such as age and hearing loss. Another reason is that while speech recognition is measured in various backgrounds, cognitive performance is typically measured only in a quiet background. To address these two issues, we measured speech recognition and cognitive performance in both a steady-state noise background and a two-talker babble background in 50 young adults with normal hearing (30 females). To minimize auditory masking, we presented the auditory stimuli for selective attention and digit span tasks to one ear and the background to the other ear. We found that the individual variance in speech recognition was three times greater in the babble background than in the noise background. Compared to the quiet background, noise had no effect but babble degraded performance in selective attention and digit span tasks. Correlation and regression analyses showed that cognitive performance did not significantly contribute to speech-in-noise recognition (r² = 0-11 %), but it explained 26-68 % of the variance in speech-in-babble recognition, with the greatest variance explained when cognitive performance was measured in the babble background. The present study demonstrates that under controlled conditions, cognitive performance can predict individual differences in speech recognition at a simulated cocktail party.