Temporal co-ordination: the lynch-pin of language production

Abstract

Pauses in spontaneous speaking constitute a rich source of data for several disciplines. They have been used to enhance automatic segmentation of speech, classification of patients with communication disorders, the psycholinguistic models of speaking, and the analysis of psychological disorders. However, although pause analyses have been with us for more than 40 years, interpretation has been compromised by several complex issues. The first of these is the fact that pause distributions are skewed, making the arithmetic mean a poor measure of central tendency. The second problem is that there appear to be two types of pause rather than one, an issue that has usually been ‘solved’ by discarding pauses that fall short of some arbitrarily defined criterion. The third problem stems from the fact that application of arbitrary criteria to all individuals must, inevitably, produce errors of both inclusion and exclusion, thereby compromising the measurement of pause duration. Recent developments in the analysis of pause distributions have solved each of these problems. We have demonstrated that the overall pause distribution comprises two log-normal component distributions, the component distributions can be separated by application of signal detection theory, and that the proportion of misclassifications minimised and estimated (Kirsner, Dunn, Hird, Parkin & Clark (2002). In this paper we will describe applications of the new paradigm to recall, amnesia, second language learning and acquired and degenerative communication disorders. Our findings are consistent with the following propositions: (1) short pause duration distributions are sensitive to acquired neurogenic communication disorders, and therefore occupy a critical niche in language production, (2) while the processes underlying the short and long pause duration distributions probably recruit information from qualitatively different sources, they meet complementary requirements in language production, (3) the short and long pause duration distributions together provide rich information about both acquired and degenerative communication disorders, and point toward a new approach to classification, and, (4) the short and long pause distributions reflect the central cohesive element in language production. In summary, the new paradigm poses a significant challenge for models of speaking. Models of language production do not include provision for the presence of two pause duration distributions in spontaneous speaking, or for their statistical properties. Furthermore, because each distribution draws on information from a variety of sources involving distinct ‘domains’, their presence constitutes a challenge to modular approaches to language production. And, finally, because the new paradigm involves spontaneous, contextualised speech, it poses a significant challenge to models that rely on data from de-contextualized tasks, tasks that have not only failed to achieve their original objectives involving the identification and measurement of ‘pure’ processes, but which in addition restrict consideration by design to frozen moments in the operation of what is, patently, a dynamic system. [Authors' abstract]