The Prosodic Organization of Conversational Structures

It is a basic premise of conversation analysis that interactions are both structurally organized and contextually oriented (Heritage 1984), and a primary goal of the analysis is elucidating these structures and determining how they are interrelated. A working model of conversational structure must minimally account for a mechanical structure composed of turns or exchanges; the propositional content of the conversation, analyzed here as consisting of information units; and shifts in speaker-hearer roles. The definition of each of the components—turns, information units, and even speaker—is far from problematic. One way to focus these definitions is to look at unit boundaries, where there are changes in turns, topics and speakers.

The focus of the present study is an examination of the correlation between such shifts and changes in the prosodic structure of a conversation. Intonation has been studied from two essentially different views: the acoustic approach measures intonation in terms of changes in fundamental frequency (F0), while the perceptual approach relies on auditory perception, and intonation can be defined in terms of pitch. Changes in F0 can be tracked through conversation with reference to a declination unit (DU), a term which refers to the downdrift phenomenon of F0 over the course of speech (Cohen and ’t Hart 1967; see Ladd (1993) for discussion and refinement). By contrast, the perceptual approach sees changes in the intonational contour with respect to the intonation unit (IU) (Chafe 1987). Instrumental measurements for this study are done on the Computerized Speech Lab by Kay Elemetrics; the analysis is based on a corpus of spontaneous Russian conversations recorded in St. Peterburg.

A preliminary analysis suggests the following results: (1) There is a high correlation between an information unit and an IU, and a correlation between an information unit and syntactic completion; (2) The correlation between DUs and IUs is not absolute: in Russian conversation, a DU often includes more than one IU, but there is a strong tendency for the boundary of a DU to match that of an IU (a finding which matches that of English; see Schuetze-Coburn et al 1991); (3) There is a high coincidence between the boundaries of turns, DUs, and IUs, sometimes correlated with a pause. Significantly, the pause alone does not signal turn completion without corresponding prosodic cues, and even relatively long pauses (such as 2.21 seconds in the following example) may be treated as turn-internal:

(1)	ja	pytalas’	uexat’	včera	(2.21 sec)
	I	tried	to.leave	yesterday
Hz	208				217-294
(2)	dumaju	ja	ego	esli 	pojmaju
	I.think	I	him	if	I.catch
Hz	227-269	312			345
(3)	ja	ego	razorvu	voobšče	na 	časti
	I	him	rip		in.general	to	pieces
Hz	270						170	175

This example is divided into one IU per line; each IU corresponds roughly to one syntactic clause (two clauses/IUs with the parenthetical verb dumaju ‘I think’ in line 2), with one DU corresponding to the three lines. Each line also corresponds to a single information unit with these organized into a larger topical unit which stretches over more of the conversation (a story about this speaker’s cat). Here, the ongoing discourse-level topic and high intonation before the pause signal that this is not a turn boundary.


Chafe, W. 1987. “Cognitive Constraints on Information Flow.” In R. S. Tomlin, ed., Coherence and Grounding in Discourse, 21–51. Amsterdam: John Benjamins.

Cohen, A. and J. ’t Hart. 1967. “On the Anatomy of Intonation.” Lingua 19: 177–192.

Heritage, J. 1984. Garfinkel and Ethnomethodology. Oxford: Basil Blackwell.

Ladd, D. R. 1993. “On the Theoretical Status of “the Baseline” in Modelling Intonation.” Language and Speech 36(4): 435–459.

Schuetze-Coburn, S., M. Shapley and E.G. Weber. 1991. “Units of Intonation in Discourse: A Comparison of Acoustic and Auditory Analyses.” Language and Speech 34(3): 207–234.