Over 60 years in a cocktail party: brain research on selective listening by Kimmo Alho

In 1953, Colin Cherry published results from an experiment where participants were instructed to attend to and repeat (‘shadow’) a speech message delivered to one ear while ignoring a concurrent message delivered to the other ear. He observed that the task was quite easy and that the participants did not process much the to-be-ignored message. This ‘cocktail-party phenomenon’, that is, the ability to selectively attend to particular sounds in noisy situations, has been studied in numerous subsequent studies which led to theories proposing that during selective listening, the attended speech is selected for further (e.g., semantic) processing, and the processing of unattended speech is attenuated, at an early auditory processing stage on the basis of the pitch and/or location of speaker’s voice. Some 30 years later it was shown in electroencephalography (EEG) experiments using tones rather than speech, that indeed during selective listening based on the location and/or pitch of attended sounds, processing of these sounds is enhanced in the auditory cortex. Subsequent magnetoencephalography (MEG), positron emission tomography (PET) and functional magnetic resonance imaging (fMRI) studies have supported this finding for both non-speech sounds and speech. Moreover, PET and fMRI studies on selective listening to tones have shown coinciding activity in dorsolateral prefrontal and parietal cortical areas that are presumably involved in top-down control of attentional selection occurring in the auditory cortex. However, in PET and fMRI studies on selective listening to continuous speech, no such prefrontal and parietal activations have been observed, except during orienting of attention to a particular location or pitch. This suggests that after initial top-down tuning of auditory cortex to favor a particular speech stream, the auditory cortex is able to carry out attentional selection without the support of the fronto-parietal attention network.