Miller's limitation of the processing capacity from the production side with speech : structural differences in spontaneous speech compared to read-out speech

Abstract

In the past century, psychologists, cognitive scientists, and psycholinguists have dealt with the question of how we perceive language and the world around us and have provided insights into the limited processing capacity of our working memory system. One famous approach is Miller (1956), who concluded that we can perceive and process 7 plus or minus 2 chunks at once, where a chunk corresponds to a monosyllabic word or another item such as an object on a sheet of paper. Miller's findings were much discussed in the literature. However, no study investigates the language production side. Speech is a natural form of human communication and it is also natural to pause in order to take a breath or sort one's thoughts. Assuming that pauses are an indication of the processes of working memory, what speech structure results from it? Can a pattern of 7 plus or minus 2 chunks in a group be identified in speech as well? Do spontaneous and read-out speech differ in this respect? In the present work, I present a pause segmentation structure of spontaneous speech, compared to read-out speech, that is quite similar to the results by Miller: around 7 words are produced at a time. However, different minimal pause lengths result in different numbers. With a minimal pause length of 100 ms around 5 words belong to one group and a minimum of 500 ms results in 10 words. No major differences between spontaneom; and read speech were found, although one could expect spontaneous speech to be more segmented by pauses.