Normal Non-Fluency in Adult Males: An Intra- and Inter-Speaker Study

AbstractOur study is a detailed exploration of non-fluency in adults who do not stutter and a discussion of possible clinical implications of our findings. There have been explorations of aspects of non-fluency as part of the study of the effect of cognitive load upon speech planning (e.g. Goldman-Eisler, 1968) or the role of fluency disruptions in conversation management (e.g. Eklund, 2004). There have been few descriptive studies exploring the type and frequency of all non-fluencies in an environment in which aspects of both cognitive load and the conversation structure are controlled. The University of Cambridge Dynamic Variability in Speech (DyViS) corpus (Nolan et al., 2009) contains long samples of speech elicited in two consistent contexts. One hundred 18-25 year old males with the same accent of British English participated in the study. None of the speakers had a speech or hearing disorder. They completed a mock police interview and a more informal telephone conversation which covered the same topics. The corpus aims to provide a set of baseline measures for research into aspects of speech of interest to researchers and practitioners working in the field of forensic speech analysis. In forensic speech casework the presence of non-fluencies is sometimes discussed, but quantitative analysis of its range, frequency and consistency among speakers has not previously been available. In order to address this gap the DyViS corpus was used to investigate individual variation in the two speaking contexts. The study examined the speech of 20 of the speakers in both contexts. Disfluencies were defined as phenomena which interrupt the flow of speech and were counted as the number of occurrences per 100 syllables. In addition to filled and silent pauses all speakers used repetitions, prolongations and self-interruptions in varying amounts. Statistical analysis demonstrated that the profile of types of non-fluency used by speakers tends to be distinctive. Fluency disruptions therefore contribute to the overall characteristics of a person's speech. These characteristics were relatively consistent within a speaker and across the contexts examined. The range of phenomena found will be demonstrated and comparisons with phenomena found in stuttered speech will be explored.From a clinical perspective the issues of interest include such matters as:•non-fluency as a characteristic of all speech;•the similarity of types of non-fluency in non-stuttered and stuttered speech;•the contribution of non-fluencies to the characterisation of individual speakers;•the similarity of some of the phenomena to those found in stuttered speech and therelevance of this when assessing the frequency of stutters;•providing a means of helping speakers who stutter to understand the normal andnatural role of non-fluency in speech