The cocktail party phenomenon: A review of research on speech intelligibility in multiple-talker conditions

It is in most cases relatively easy to understand one talker even if other persons are talking at the same time. Over the last decades, this so-called cocktail party phenomenon has been the subject of a considerable number of studies. In the present paper, an overview is given of part of this research, specifically those studies dealing with the intelligibility of speech presented against a background of competing speech. In the first section of the review, the properties of speech are considered that are relevant when it acts as an interfering sound: the long-term average frequency spectrum, and the modulation spectrum. In the middle sections, speech intelligibility data are reviewed for monaural and binaural listening and with one or multiple interfering voices (or speech-like sounds). It appears from these data that speech intelligibility depends in a complex manner on the properties of the interfering signal(s), the number of signals, the spatial configuration of the sources, and the acoustic environment. This dependency can be predicted in part by existing models. For the prediction of effects of voice similarity and interferer modulations, however, no suitable models are available. The fifth section of the review is devoted to research on effects of hearing impairment and the use of hearing aids. These studies demonstrate that speech intelligibility in conditions involving interfering speech is significantly poorer for the average hearing-impaired listener than for the normal hearing. Hearing aids can only be effective when they yield both selective amplification and an increase of the signal-to-noise ratio at the ears (e.g. by using directional microphones). The review ends by pointing out areas of interest for future research.