Speakers and listeners: some problems of generalising from a common speech pathology research design

Experimental studies of speech and language behaviour often have difficulty measuring the phenomena that they seek to investigate. A common solution has been to collect data in the form of listeners' judgements of the speech of members of a clinical group of interest. If the speakers are taken to be a random sample of the clinical group and the listeners a random sample of those who they may communicate with, it is tempting to believe that results from such studies may be generalised to both other speakers and listeners, thus allowing a general statement about the communicative abilities of the clinical group to be made. Two related problems with this form of design are discussed. Both lead to the conclusion that greater caution is required in the interpretation of studies in which listeners' judgements of speakers are analysed. Potential solutions to the problem are offered.