Journal of Psycholinguistic Research, Vol. 4, No. 1, 1975
Introduction
The investigation of cognitive planning processes underlying the generation of spontaneous speech has recently focused attention on what might be called the macrostructure of hesitation. Previous studies on the significance of hesitation have dealt with the microstructure of speech, namely the location and duration of individual pauses. These studies revealed relationships between particular pauses and the process of lexical selection.
When the overall proportions of pausing are considered, rather than each individual pause, several factors are associated with variations in this proportion. These reflect aspects of the cognitive process concerned with content rather than structure (Goldman-Eisler, 1961, 1968; Lay and Paivio, 1969; Reynolds and Paivio, 1968).
Henderson et al. (1966) asked whether content planning occurs sporadically throughout an utterance or at identifiable periods during it. In samples of spontaneous speech, they found that hesitant periods, predominantly silence, alternated with fluent periods, predominantly phonation. This cyclic pattern of speech frequently showed a direct relationship between the amount of pausing in a hesitant phase and the amount of speech in the following fluent phase. They termed this pattern “rhythmic speech.”
From the mathematical dependency between hesitant and fluent phases, they inferred a psychological dependency. They proposed that the amount of speech produced in a fluent phase required planning time provided by the preceding pause. Hesitant phases exhibited not only more silence but also more hesitation phenomena of other kinds, including filled pauses such as “ah’s” and “um’s,” and interruptions at grammatically inappropriate locations.
As Goldman-Eisler (1968, p. 83) observed:
“These results show that at the time of uttering fluent speech the speaker’s pausing is under control and relatively well integrated into the syntactic structure, serving the function of communication rather than being symptomatic of internal processes.”
Conversely, pauses in hesitant phases are less controlled and appear to reflect more active planning.
Method
Collection of Speech Samples
The speech samples were gathered as part of a broader research programme. The corpus comprised nearly three and a half hours of spontaneous speech from eight male native British English speakers aged between 20 and 30.
Participants discussed social and political propositions in a conversational setting with the experimenter. Their speech was recorded and later transcribed. Visual analogues were prepared to mark periods of silence and phonation so that temporal patterns could be examined.
Identification of Rhythmic Segments
Speech and pause durations were plotted sequentially. Alternating periods emerged in which long pauses paired with short utterances were followed by short pauses paired with longer utterances. These alternating slopes defined temporal cycles.
Segments containing three or more complete cycles were analysed further. Ten were mathematically rhythmic in the sense described by Henderson et al.
Cycle lengths ranged from approximately 11 to 39 seconds, with a mean of 18 seconds.
Identifying Semantic Units
To identify meaningful divisions in speech, independent judges segmented transcripts into what were termed “ideas.” This everyday term was deliberately chosen to avoid structural assumptions and to allow flexible interpretation.
Participants were asked simply to divide the text wherever they felt one idea ended and another began.
Results
Quantitative Findings
Judges showed considerable variation in the number of idea divisions they made. However, when only locations chosen by more than half the judges were considered, a consistent pattern emerged.
These criterial idea boundaries were found to correspond significantly with the beginnings of temporal cycles. In other words, where speakers paused and then resumed speech often aligned with meaningful conceptual boundaries.
Ideas and Syntax
Most divisions selected by participants coincided with clause boundaries, and many aligned with sentence boundaries. Although not every sentence represented a complete idea, idea boundaries were rarely located mid-clause.
Thus, clause boundaries appear to be a necessary but not sufficient condition for idea segmentation. Larger semantic units often consisted of several clauses or sentences grouped together.
Discussion
The temporal cycles observed in spontaneous speech are clearly associated with semantic and syntactic structure. This finding challenges the view that these cycles are random and supports the hypothesis that they reflect cognitive planning processes.
Speakers appear to plan ahead in terms of familiar linguistic units such as clauses and sentences. At the same time, they are able to group several of these into larger, semantically unified structures. These larger structures likely represent higher-level planning units.
Smaller units, such as phonemic clauses, may be highly automated and require little planning time. Hesitations therefore occur primarily where higher-level processes are active, such as lexical choice or conceptual organisation.
The variability among participants’ segmentations suggests that semantic structure is hierarchical and flexible rather than fixed. Different segmentations may all be plausible depending on the task or perspective.
Another possible explanation is that cycles reflect underlying biological rhythms that planning processes exploit. However, the close alignment between cycles and linguistic structure suggests that planning plays a primary role.
Conclusion
Temporal cycles in spontaneous speech correspond closely with meaningful linguistic and semantic units. This supports the view that hesitation is not random but reflects the cognitive demands of planning and organising speech.
Speech production, therefore, involves cycles of preparation and execution, with pauses serving as integral components of higher-level cognitive processing.
References
Boomer, D. S. (1965). Hesitation and grammatical encoding. Language and Speech, 8, 148–158.
Boomer, D. S., and Laver, J. D. M. (1968). Slips of the tongue. British Journal of Disorders of Communication, 3, 2–11.
Butterworth, B. L. (1972). Semantic analyses of the phasing of fluency in spontaneous speech. PhD dissertation, University College London.
Goldman-Eisler, F. (1958–1972). Various works on hesitation and spontaneous speech.
Henderson, A., Goldman-Eisler, F., and Skarbek, A. (1966). Sequential temporal patterns in spontaneous speech. Language and Speech, 9, 207–216.
Maclay, H., and Osgood, C. E. (1959). Hesitation phenomena in spontaneous English speech. Word, 15, 19–44.
Miller, G. A., Galanter, E., and Pribram, K. H. (1960). Plans and the Structure of Behaviour.
Wundt, W. (1912). Die Sprache.