Objective measurement of fluency in natural language production: A dynamic systems approach
Language research is dominated by the concept of modularity. The basic assumptions involve neural localization of function, and adoption of tasks that tap into specialized functions, involving words or phonemes for example. The tasks that emerge to support this research are generally de-contextualized. Recent work in neuroscience has identified large-scale self organizing neural networks. It is our contention that the advanced neuro-imaging procedures demand an equivalent refinement in the language sampling domain. The collection of natural speaking samples, and an objective approach to fluency, are critical to the understanding of language production. This paper describes a measurement system designed to quantify fluency in natural spoken language. The system classifies environmental and breathing noise, and estimates means and standard deviations for the three lognormal distributions associated with spontaneous speaking: short pauses, long pauses and speech segment duration. The analysis of natural samples produced by three diverse aphasic speakers demonstrates the sensitivity of the fluency measure as well as the profile of independent or correlated changes across the parameters. The system yields objective and sensitive measures of communicative efficiency for individuals across a variety of speaking contexts.