lournal o/ Personality and Social Psychology 1972, Vol. 23, No. 2, 283-292 SOME SIGNALS AND RULES FOR TAKING SPEAKING TURNS IN CONVERSATIONS1

STARKEY DUNCAN, JR." University of Chicago

The turn-taking mechanism, whereby participants manage the smooth and appropriate exchange of speaking turns in face-to-face interaction, was studied. Three basic signals for this mechanism are described: (o) turn-yielding signals by the speaker, (6) attempt-suppressing signals by the speaker, and (c) back- channel signals by the auditor. These signals are used and responded to in a relatively structured manner, describable in terms of a set of rules. Behaviors in every communication modality examined—content, syntax, intonation, para- language, and body motion—were active as elements of the turn-taking signals.

E. Goffman (personal communication, has commented that this phenomenon "is August 7, 1969) has asked rhetorically how nearly the most obvious aspect of conversa- people manage to walk down the street with- tion [p. 568J." Jaffe and Feldstein (1970) out continually bumping into each other. also refer to the saliency of turn taking and Part of the answer is that in our culture the importance of avoiding interruptions. there are rules for walking down the street They cite Sullivan (1947), who observed and for managing situations in which indi- careful turn taking in conversations between viduals find themselves on collision courses. chronic mental hospital patients, and Miller Goffman (1963) has suggested some rules (1963), who suggests that turn taking is a for these situations. language universal. Kendon (1967) deals in Just as it is desirable to avoid bumping detail with the role of gaze direction in turn into people on the street, it is desirable to taking. Schegloff (1968) proposed the "basic avoid in conversations an inordinate amount rule for conversations: one party at a time of simultaneous talking. Beyond consider- [p. 1076, italics in original]," and discussed ations of etiquette, it is difficult to maintain some implications of this rule. Leighton, adequate mutual comprehensibility when Stollak, and Ferguson (1971) found more participants in a conversation are talking at interrupting and simultaneous talking in the the same time. interaction of families waiting for psycho- The fact that participants in a conversa- therapy than in the interaction of "normal" tion tend to take turns in speaking and listen- families. ing has been frequently observed and dis- The question may be asked, again rhetori- cussed by other investigators. Yngve (1970) cally, how participants in a conversation can 1 This study was supported in part by Grants avoid continually bumping into each other in MH-16,210 and MH-l7,756 from the National In- a verbal sense. The thesis of this paper is stitute of Mental Health, and by Grant GS-3033 that there is a regular communication mecha- from the Division of Social Sciences of the National nism in our culture for managing the taking Science Foundation. The author is grateful to Erving Goffman, Adam Kendon, and Allen Dittmann, who of speaking turns in face-to-face interaction made valuable constructive criticisms of an earlier (Goffman, 1963). Through this mechanism, draft of this paper. The author wishes to express participants in an interaction can effect the his appreciation to Susan Beekman, Mark Cary, smooth and appropriate exchange of speaking Diane Martin, Ray O'Cain, Tom Shanks, and An- turns. (The term "turn taking" has been drew Szasz, who contributed to the transcriptions and data analysis. I am endebted to Dick Jenney, independently suggested by Yngve, 1970, and Wayne Anderson, and the client, who generously by Goffman, personal communication, June 5, consented to serve as participants in this study. 1970). 2 Requests for reprints should be sent to Starkey The proposed turn-taking mechanism is Duncan, Jr., Department of Psychology, University of Chicago, 5848 South University Avenue, Chicago, mediated through signals composed of clear- Illinois 60637. cut behavioral cues, considered to be per- 283 284 STARKEY DUNCAN, JR.

ceived as discrete. The turn-taking signals address, etc., to more emotionally laden discussion of are used and responded to according to rules. the client's reasons for applying for therapy. At the same time, there is a strong intrinsic motivation for Signals, cues, and rules are described in de- the interview, namely, an application for therapy, tail below. thereby avoiding the more artificial experimental Turn taking is considered to be one of a situation in which unacquainted subjects are brought number of communication mechanisms, such together and asked to discuss anything which might as those discussed by Scheflen (1968), oper- be of mutual interest. The client was in her early twenties, working as ating in face-to-face interaction. These mecha- a secretary, and had not completed college. The nisms serve the function of integrating the therapist-interviewer was a 40-year-old-male, an performances of the participants in a variety experienced therapist, who had been doing prelimi- of ways, for example, regulating the pace at nary interviews for many years. which the communication proceeds, and moni- The second interview was between the therapist who participated in the first interview, and a second toring deviations from appropriate conduct. male therapist, also 40 years old. The two therapists Goffman (19SS) commented on these inte- were good friends and had known each other for grating mechanisms in general and on turn about 10 years. Their interaction was relaxed and taking in particular: lively. The topic in this case was another client whom the first therapist had seen in a preliminary In any society, whenever the physical possibility interview, and whom the second therapist had at that of spoken interaction arises, it seems that a system time seen in therapy for two interviews. of practices, conventions, and procedural rules comes The preliminary interview is designated at Inter- into play which functions as a means of guiding and view 1, and the second, peer interaction, is desig- organizing the flow of messages [p. 2261. nated as Interview 2. The client is designated as Participant A; the preliminary interviewer, B; and The notion that a set of rules operates to the second therapist, C. Thus, the participants in integrate the turn-taking behavior of partici- Interview 1 were A and B, and the participants in pants in a conversation is supported by Jaffe Interview 2 were B and C. and Feldstein (1970), who also studied tem- Videotaping poral patterns of speech and silence in dyadic To videotape the interactions, the camera was conversations. Although they limited their placed so that both participants in each interaction data to the information provided by a "pair were fully visible from head to foot on the tape of voice-actuated relays which treat any sound at all times. No zoom techniques or other special above threshold as equivalent [p. 113]," focusing effects were used. A single camera was set their findings suggested to them up in full view of the participants. The camera and tape were left running prior to the participants' entry further interactional rules that govern the matching into the room and were not touched again until of speech rates of the participants, the prohibition after the interview. of interruption, and the requirement for properly Despite the fact that a wide-angle lens was used, timed signals that acknowledge understanding and the camera was necessarily at such distance from confirm the continued attention of the listener [p. 6]. the participants that more subtle details of facial expressions were not discriminable on the videotape. SOURCE OF DATA Less subtle expressions, such as broad smiles and grimaces, were readily discernable. In contrast, very Interviews small movements of the hands and fingers, for The results to be reported were based on meticu- example, were clearly evident on the tapes, so that lous transcriptions of speech and body motion be- fine discriminations of these movements could be haviors during the first 19 minutes of two dyadic made and were on the transcription. A high-quality interviews, as recorded on videotape. monophonic, audiotrack was obtained on the video- The first interview was a preliminary interview tape. held at the Counseling and Psychotherapy Research Center at the University of Chicago. This prelimi- Transcription nary interview is part of the routine intake procedure For this study, the principal requirements for the at the Counseling and Psychotherapy Research transcription were those of maximum behavioral Center, and the client was a regular applicant for therapy. A preliminary interview was chosen for breadth and of continuity (no breaks or interrup- intensive transcription of communication behaviors tions). Maximum breadth is desirable in analysis because within a rather compressed period of time because it is not yet known which behavioral cues a wide variety of types of interaction may be en- are the primary mediators of any given communica- countered, from simple information giving, such as tion function. Continuity of transcription permits SOME SIGNALS AND RULES IN CONVERSATIONS 285 the complete analysis of sequences of events: the tic behaviors was actually encountered in the two basic concern of this study. dyads and included in the transcriptions. In terms of size, two 19-minute transcriptions of interaction are simultaneously very small and very Body Motion large, depending upon one's perspective. From the In contrast to paralanguage, there was no available point of view of the wealth of communication en- transcription system for body motion which could gaged in each day by an individual, the transcrip- be readily applied to our videotapes. This situation tions are quite brief. On the other hand, these tran- led to a transcribing method based on the behaviors scriptions are believed to be unique in their breadth actually found in each interview. The transcription and duration. As described below, there was a de- system for the first interview was created by first tailed transcription of English segmental and supra- making an inventory of the movements used by the segmental phonemes, paralanguage, and body motion two participants and then assigning either arbitrary of both participants in the two interviews. The time or descriptive labels to these movements. This sys- involved in making the two transcriptions was tem was then applied to the second interview, after great, involving the better part of 2 academic years. expanding it to include new movements observed Phonemes in the second interview. While there is no pretense that the resulting Segmental phonemes. Transcription of segmental transcription system is able to encompass all move- phonemes, which describe the way syllables are ments occurring in this culture, every attempt was pronounced within the framework of the English made to include all movements observed in the dyads sound system, followed the scheme developed by under study. The transcription was in this sense Trager and Smith (19S7). The segmental phonemes comprehensive. Included were (a) head gestures and were the least important components of the study. movements (nodding, turning, pointing, shaking, etc.) Suprasegmental phonemes. The suprasegmental and direction of head orientation; (b) shoulder phonemes are commonly referred to as intonation. movements (e.g., shrugs); (c) facial expressions, They include the phenomena of stress, pitch, and such as could be clearly seen; (d) hand gestures juncture. and movements of all sorts (each hand transcribed The Trager-Smith scheme for transcribing supra- independently) ; (e) foot movements (each foot segmental phonemes was used, with minor modifica- independently); (/) leg movements; (g) postures tions identical to those described in previous studies and posture shifts; and (h) use of artifacts, such as by the present author (Duncan, Rosenberg, & pipe, kleenex, papers, and clipboard. Finkelstein, 1969; Duncan & Rosenthal, 1968). Three terminal junctures—rising, falling, and Coordination of Body Motion and Speech sustained—were transcribed in accordance with the Transcriptions Trager-Smith system. These junctures are composed of contours of pitch, intensity, and duration occurring Speech syllables were used to locate all tran- on the final syllable of phonemic clauses. scribed events. Thus, the movements of both partici- pants in an interview were located with respect to The point of departure for all subsequent analysis in this study was the phonemic clause (Trager & the syllables emitted by the participant who hap- Smith, 1957). A phonemic clause is a phonological pened to be speaking at the time, or to the pause between two syllables. unit, defined by Trager and Smith as containing one and only one primary stress and one terminal juncture. Transcribing primary stresses and terminal THE TURN-TAKING MECHANISM junctures automatically identifies the phonemic The variables for the turn-taking mecha- clauses in a corpus. nism were formulated as signals by which Paralanguage each participant indicated his state with re- Paralanguage refers to the wide variety of vocal gard to the speaking turn. Given the display behaviors that occur in speech but that are not part or absence of a given turn-taking signal by- of the sound system of language, as traditionally one participant, rules delimit the appropriate conceived. Comprehensive catalogs of paralinguistic responses by the other participant. behaviors have been compiled by Trager (1958), Crystal and Quirk (1964), and Crystal (1969). Any The rules and signals, considered together, one speaker will probably use only a small fraction establish empirical expectations with respect of the total behaviors available. The following list, to turn-taking activities at any given moment which uses Trager's (1958) terminology, includes only those behaviors that play a part in the turn- in a conversation, assuming that the partici- taking signals: (a) intensity (overloud-oversoft); pants in the conversations under analysis are (6) pitch height (overhigh-overlow); and (c) ex- rule abiding for the most part. Data relevant tent (drawl-clipping of individual syllables). The terms in parentheses define the anchor point for each to evaluating the turn-taking mechanisms are behavioral continuum. A wide variety of paralinguis- presented in the Results section. 286 STAEKEY DUNCAN, JR.

Definitions Under proper operation of the turn-taking Definitions of "speaker," "auditor," and mechanism, if the auditor acts to take his "simultaneous turns" are required to begin. A turn in response to a yielding signal by the speaker is a participant in a conversation speaker, the speaker will immediately yield who claims the speaking turn at any given his turn. moment. An auditor (Kendon, 1967) is a A state of simultaneous turns can be cre- participant who does not claim the speaking ated in two ways: (a) if the auditor attempts turn at any given moment. to take his speaking turn in the absence of a In general, with the help of the turn-taking turn-yielding signal by the speaker or (b) if mechanism, the turn need not be disputed in the speaker displays a yielding signal, and the course of the conversation. However, the auditor acts to take his turn, and the when both participants claim the speaking original speaker then continues to claim his turn at the same time, then the state of speaking turn. Neither of these sequences of simultaneous turns obtains in the conversa- events will occur in an interaction when the tion and the turn-taking mechanism may be participants are adhering to the turn-taking said to have broken down, or perhaps to have mechanism. been discarded, for the duration of that state. The auditor is not obliged to take his The term "simultaneous turns" is used here speaking turn in response to a regular turn- instead of the more usual term "simultaneous yielding signal by the speaker. The auditor talking" because there will be certain circum- may alternatively communicate in the back stances, described below, in which talking by channel (Yngve, 1970) or remain silent. the auditor, even simultaneously with the (Portions of the back channel are briefly speaker, does not imply a claim for the speak- described below.) ing turn. Therefore, the two terms should be Signal differentiated. When simultaneous turns occurs, the audi- A turn-yielding signal for a speaker is de- tor attempts to take his turn, and the original scribed as the display of at least one of a set speaker continues with his turn. The turn- of six discrete behavioral cues. These cues taking mechanism is not designated to explain may be displayed either singly or together. how the state of simultaneous turns is re- When displayed together, they may occur solved, that is, which one of the two speakers either simultaneously or in tight sequences. will continue and which one will fall silent. The six turn-yielding cues are listed below. Meltzer, Morris, and Hayes (1971) reported 1. Intonation: the use of any pitch level- some interesting findings, based on techniques terminal junction combination other than 2 21 of "social psychophysics [p. 401]," on one at the end of a phonemic clause. Following mode of resolution. the Trager-Smith (1957) notation, the 2 re- Just as both participants may become fers to an intermediate pitch level, neither speakers, claiming simultaneous turns, both high (3) nor low (1). The single bar junc- participants may also become auditors, not ture "|" at the end of the clause refers to a claiming a turn. In this case, the obvious re- sustention of the pitch at the level previously sult is silence for the duration of that mutual indicated. Thus, 2 2 refers to a phonemic state. The phenomenon—apparent avoidance clause ending on an intermediate pitch level, of the turn by both participants—was not which is sustained, neither rising nor falling, observed in the two interviews analyzed for at the juncture between clauses. this study and thus will not be discussed in A frequent pattern of phonemic clauses in this paper. American English is a series of 2 2| clauses, which terminates in a final clause with a rising Turn Yielding or falling pitch level-juncture combination. Rule Any of these rising or falling combinations The auditor may take his speaking turn qualifies as a regular turn-yielding cue. A when the speaker gives a turn-yielding signal. phonemic clause containing one of these SOME SIGNALS AND RULES IN CONVERSATIONS 287 rising or falling combinations is referred to text. These expressions do not add substantive as a "terminal clause." information to the speech content that they 2. Paralanguage: Drawl: drawl on the final follow. Instances in which the auditor pro- syllable or on the stressed syllable of a ceeded to take his speaking turn during the terminal clause. completion of a sociocentric sequence are not 3. Body motion: the termination of any considered as a state of simultaneous turns hand gesticulation (Kendon, 1967) used in the conversation. Rather, such an act is during a speaking turn or the relaxation of considered to be an instance of permissible a tensed hand position (e.g., a fist) during simultaneous talking. a turn. 5. Paralanguage: Pitch/loudness: a drop To account for the gesticulations observed in paralinguistic pitch and/or loudness in in the two interviews, it is sufficient to define conjunction with one of the sociocentric se- gesticulations as those hand movements gen- quences described above. When used, these erally away from the body, which commonly expressions typically followed a terminal accompany, and which appear to bear a direct clause but did not often share the same para- relationship to, speech. language. Specifically excluded from the definition of 6. Syntax: the completion of a gram- gesticulation are self-adaptors and object matical clause, involving a subject-predicate adaptors (Ekman & Friesen, 1969). Self- combination. adaptors involving the hands are movements in which the hand comes in contact with one's Attempt-Suppressing Signal own body, often with the appearance of Rule grooming. Examples would be rubbing the chin, scratching the cheek, smoothing the An attempt-suppressing signal displayed by hair, brushing off the pants leg, picking lint the speaker maintains the turn for him, re- (real or imaginary) from the socks, etc. Such gardless of the number of yielding cues con- self-adaptors are very frequent for many indi- currently being displayed. Auditors almost never attempted to take their turn when this viduals both while they are speakers and signal was being displayed. while they are auditors. For present purposes, the definition of object adaptors is construed Signal more narrowly than in Ekman and Friesen's definition. Movements considered to be object The attempt-suppressing signal consists of adaptors in the first interview were those one or both of the speaker's hands being movements having to do with Participant B's engaged in gesticulation. Self- and object maintaining his pipe, and some movements in adaptors do not operate as attempt-suppress- which the paper on his clipboard was wrinkled. ing signals. Dropping of the gesticulating Highly similar behaviors, termed "self- hand or hands into a rest position, as on the manipulatory gestures," were also studied by arm of a chair, constitutes a turn-yielding cue Rosenfeld (1966). (see 3 described above). It should be noted 4. Sociocentric sequences: the appearance that much speech is not accompanied by of one of several stereotyped expressions, gesticulation, and therefore neither the at- typically following a substantive statement. tempt-suppressing signal nor its coordinate Examples are "but uh," "or something," or yielding cue would be applicable for that "you know." The three participants varied in speech. the particular expressions used and in the fre- Back-Channel Communication quency of their use. They were most common for the client. They were generally preceded Rule by other yielding cues, and were often accom- The term "back channel" was introduced panied by a marked paralinguistic "trailing in the context of turn taking by Yngve off" effect. The term sociocentric sequences (1970) to cover such messages as "mm-hmm" was coined by Bernstein (1962), who com- and head nods from the auditor. In this mented on these expressions in another con- sense, the term is roughly equivalent to 288 STARKEY DUNCAN, JR.

Kendon's (1967) "accompaniment behavior" behaviors in syntax, intonation, paralanguage, and Dittmann and Llewellyn's (1968) "listener and body motion. response." Specifically, boundaries of the units were For the purposes of this study, it is suf- defined as being (a) at the ends of phonemic ficient to point out that a back-channel com- clauses (b) which additionally were marked munication does not constitute a turn or a by the display of one or more of the floor- claim for a turn. To the contrary, it appears yielding cues described above, and/or by the that, when a speaker is displaying a turn- display of one or more of the following cues. yielding signal, the back channel is often used 1. Unfilled pause: an appreciable unfilled by the auditor to avoid taking his speaking pause following the phonemic clause. turn. In this sense, taking a turn and com- 2. Head direction: turning of the speaker's municating in the back channel may be con- head toward the auditor. This cue is identical sidered to be contrasting tacks an auditor, to the speaker's part of the gaze-direction faced with a yielding signal, may take. pattern discovered by Kendon (1967). It is important to note that, because back- 3. Paralanguage: a drop in paralinguistic channel signals do not constitute a turn or a pitch and/or loudness in conjunction with a claim for a turn, their display by the auditor phonemic clause, either across the entire simultaneously with the speaker's speech is clause or across its final syllable or syllables. not considered to be a state of simultaneous 4. Body motion (for Participant A only): turns in the dyad. a relaxation of the foot from a marked dorsal flexion. (Throughout the interview the client's Signal legs were stretched out in front of her and It is suspected that back-channel communi- were crossed at the ankle.) From time to time cation comprises a large and complex set of one or both feet would be flexed dorsally, signals, which at present may not be well such that they assumed a nearly perpendicu- understood. These signals may participate in lar angle to the floor. Their falling, as a result a variety of communication functions, in- of relaxing the flexion, was the cue. cluding the regulation of speaking turns. It is Each of these behaviors appeared to play proposed that "back channel" be expanded to a part in the speaker's segmenting of his refer to this broader class of signals, as it is communication and in the timing and place- progressively documented. ment of the auditor's turn-taking and back- In addition to the back-channel signals such channel responses. as "mm-hmm," "yeah," and head nods men- tioned above, the following back-channel sig- RESULTS nals were observed in our interviews: (a) sen- The rules for turn yielding and attempt tence completions, in which the auditor com- suppressing lead to the expectations that (a) pleted a sentence that the speaker had begun the occurrence of simultaneous turns will be (independently reported by Yngve, 1970); associated primarily with the auditor's turn- (b) brief requests for clarification; and (c) taking attempts when zero yielding cues (i.e., restatement in a few words of an immediately the absence of a yielding signal) are being preceding thought expressed by the speaker. displayed by the speaker, and (b) display of the attempt-suppressing signal by the speaker Unit of Analysis will sharply reduce the auditor's turn-taking In order to subdivide the interviews, a unit attempts in response to yielding cues. In of analysis was chosen which in size lay be- addition, auditor attempts to take the floor tween the phonemic clause and the speaking appeared to vary as a function of number of turn. Every cue display and every instance of yielding cues conjointly displayed. Data smooth exchange of turns or simultaneous analyses are presented relevant to these three turns occurring in the interviews was located empirical issues. with respect to these units. As with the yield- Table 1, on which the analyses were based, ing signal, the unit was defined in terms of presents the data, summed over the two inter- the display of at least one of a number of views, for auditor turn-taking attempts and SOME SIGNALS AND RULES IN CONVERSATIONS 289

TABLE 1 AUDITOR TURN-TAKING ATTEMPTS AND RESULTING SIMULTANEOUS TURNS AS A FUNCTION OF NUMISER OF TURN-YIELDING CUES DISPLAYED AND THE DISPLAY OF THE ATTEMPT-SUPPRESSING SIGNAL

Simultaneous turns resulting Speaker yielding cue display Auditor turn-taking attempt from auditor attempt

(A) (B) (C) (D) (E) N conjointly Frequency of PS, pa displayed display N SD* If SDb

No attempt-suppressing signal displayed

0 52 5 .10 .04 5 1.00 .00 1 123 12 .10 .03 2 .17 .11 2 146 25 .17 .03 2 .08 .05 3 89 29 .33 .05 2 .07 .05 4 47 15 .32 .07 0 .00 .00 5 9 4 .44 .17 0 .00 .00 6 2 1 .50 .35 0 .00 .00 E 468 91 11

Attempt-suppressing signal displayed

0 56 7 .13 .04 7 1.00 .00 1 109 0 .00 .00 2 138 0 .00 .00 3 105 2 .02 .01 1 .50 .35 4 6 0 .00 .00 5 3 0 .00 .00 E 417 9 8

" column B/column A. b Standard error of the proportioin -~ -\w/PQ 0 column D/column B.

resulting simultaneous turns, in terms of (a) When no attempt-suppressing signal was the number of turn-yielding cues conjointly displayed, the correlation between number of displayed and (b) the display or absence of yielding cues displayed and percentage of the attempt-suppressing signal. Percentages auditor turn-taking attempts was .96 (dj = of auditor attempts in response to a given 4). This correlation accounts for 92% of the number of cues were calculated as the num- variance. The analysis of variance for the ber of attempts divided by the number of regression is presented in Table 3. The F displays of those cues. Percentages of simul- value for the analysis was SO.S3. (When df taneous turns were calculated by dividing the — 1/4, an F value of 21.2 is significant number of simultaneous turns by the number at the .01 level of significance.) The data of attempts. points and the regression line are plotted in The relationship between the display of the left half of Figure 1. It should be noted zero yielding cues by the speaker and the that the data point for the display of six cues was not included in the regression be- occurrence of simultaneous turns resulting cause there were only two such cases, thereby from a turn-taking attempt by the auditor giving a relatively unreliable estimate for may be tested by applying chi-square to that point. It happens, however, that the Table 2, a 2X2 contingency table derived data point falls precisely on the regression from Table 1 (x2 = 52.31, corrected for con- line, as may be seen in Figure 1. tinuity, dj - 1. When dj = 1, a chi-square of The results on auditor attempts were 10.8 has an associated probability of .001). sharply different when the speaker was dis- 290 STARKEY DUNCAN, JR.

TABLE 2 TABLE 3 SMOOTH EXCHANGES OF TURNS AND SIMULTANEOUS PERCENTAGE or AUDITOR TURN-TAKING ATTEMPTS TURNS RESULTING FROM AUDITOR'S TURN-TAKING IN RESPONSE TO NUMBER or YIELDING CUES ATTEMPTS IN RESPONSE TO THE SPEAKER'S DISPLAYED: ANALYSIS or VARIANCE DISPLAY OF ZERO YIELDING CUES AND FOR THE REGRESSION OF >1 YIELDING CUES Source of variation df .S.S MS F No. yielding cues Smooth exchange Simultaneous displayed of turns turns Attributable to regression 1 .09367 .09367 50.53 Deviation from regression 4 .00741 .00185 0 0 12 Total 5 .10109 >1 81 7

Note.—N = 100. cues were conjointly displayed, the prob- playing an attempt-suppressing signal along ability of a turn-taking attempt by the audi- with his yielding cues. Data on auditor at- tor increased in a strictly linear fashion. tempts in the presence of a suppressing sig- On the other hand, the display of an nal are presented in the lower half of Table 1 attempt-suppressing signal essentially elimi- and in the right half of Figure 1. With the nated the auditor's tendency to take his turn, exception of the display of zero yielding regardless of the number of yielding cues con- cues, the auditor attempt curve was virtually comitantly displayed. Because the auditor's flat at 0%, with no increase of turn-taking attempts were so thoroughly suppressed by attempts as the number of yielding cues the signal, it was not possible to infer from increases. the data the likelihood of simultaneous turns resulting from turn-taking responses to the DISCUSSION suppressing signal. The results reflect strong regularities in It should be noted that the display of any interview behaviors with respect to turn number of yielding cues by the speaker, in taking. It should be borne in mind that the the absence of a suppressing signal, did not data presented here were generated through automatically result in an attempt by the the coordinated action of two individuals. auditor. At best, the probability of an audi- As an integrating mechanism, turn taking ap- tor attempt appears to be about .50. Thus, pears capable of being remarkably successful the auditor retains considerable discretion in dyadic conversations. over his responses. The chance of simultaneous turns was Either the speaker or the auditor may dis- sharply decreased when the auditor attempted regard the turn-taking mechanism, so that a to take his turn after the display of a yield- state of simultaneous turns is produced. In the ing signal by the speaker. As more yielding present data the occurrence of auditor turn-

so No Attempt- Suppressing Signal Attempt- Suppressing Displayed Signal Displayed

(89) • ^ < 30 S, <" S-J §^2° .(146) (52 i (56) £' 10"/»(123) v(109) ( 012345601 23 Number of Yielding Cues Conjointly Displayed by the Speaker FIG. 1. Auditor's turn-taking attempts in response to the display of yielding cues and attempt-suppressing signals (ws are shown in parentheses). SOME SIGNALS AND RULES IN CONVERSATIONS 291 taking attempts when zero yielding cues were The findings on the turn-yielding signal being displayed may be straightforwardly in- provide an example of the usefulness of be- terpreted as an interruption of the speaker. haviorally comprehensive research. The cues On those six occasions on which simultaneous comprising this signal were found in every turns were associated with the display of communication modality examined: content, yielding cues by the speaker, the blame may syntax, intonation, paralanguage, and body be laid to him for not properly yielding when motion. he had so signaled. The behavioral breadth in the yielding sig- This study was based on two behaviorally nal provides it with the desirable property of inclusive transcriptions of dyadic face-to-face flexibility. No single communication modality interaction. The emphasis on inclusiveness is required in order to display a signal. The required joint consideration of the linguistic, yielding signal may also be said to possess paralinguistic, and body motion components the property of generality, in that the cues of face-to-face interaction, as opposed to for the signal are formulated in terms of focusing exclusively on any one or two of general properties of behaviors, rather than these modalities. As mentioned above, this specific acts. For example, it is not a specific behavioral inclusiveness is desirable at this intonation pattern that served as a cue, but stage of our understanding of face-to-face simply any deviation from the 2 2 pattern; interaction because it is not known a priori not a specific gesticulation, but cessation of which behaviors in the stream of communica- the gesture, or relaxation of a tensed hand tion are the important cues for any given position; not a specific paralinguistic pat- communication function. tern, but a drop from the preceding pattern A primary obstacle to research of this type in pitch and/or loudness, and so on. is the laboriousness of making fine-grained Further research is underway on various transcriptions of multiple interaction be- aspects of the turn-taking mechanism using haviors. Despite the difficulty of the task, our transcriptions. The distribution and func- these detailed comprehensive transcriptions tions of the impressively large and complex are valuable for their potential contribution class of back-channel signals are being in- to the discovery and documentation of vari- vestigated. The notion of floor-requesting ous communication functions. It is important signals by the auditor, suggested by Yngve at this stage of research to be able to specify, (1970), is being explored, including the pos- quite accurately, what happens where in inter- sibility that there may in some sense be an actions. Considering the wealth of transcribed ongoing negotiation for the floor by speaker data on languages throughout the world, the and auditor. raw transcribed data available on face-to-face By positing three types of signals and as interaction in its broader aspect are deplor- many rules, a turn-taking mechanism can be ably scant. described which accounts for extensive por- Once important signals are identified for tions of the turn-taking behavior in the two any given communication function, further interviews under examination. The overall research on that function can proceed at a strength of the results underscores the po- much more rapid pace. In Scheflen's (1966) tential of further research on the rule- words, the signals become "recognizable at a governed aspects of behavior. glance and recordable with a stroke [p. 277, The rules for turn taking were designed to original italics]." Accordingly, work is pres- be applicable across a wide range of individual ently underway to transcribe the turn-taking styles and communication contexts. There is signals (and a few other potentially signifi- room for appreciable variation in their use. cant behaviors) for brief sections of addi- For example, variation may be found in the tional interviews. In this manner it is pos- use of (a) the attempt-suppressing signal, sible to capitalize on existing knowledge of (b) back-channel communications in lieu of turn taking, both to extend our understanding turn-taking attempts, (c) the number of si- of it and to validate further our original multaneous turns created either by the findings. speaker or by the auditor, and (d) the num- 292 STAEKEY DUNCAN, JR. her of turn-yielding signals composed of large JAFFE, J., & FELDSTEIN, S. Rhythms of dialogue. numbers of cues. If such variations actually New York: Academic Press, 1970. KENDON, A. Some functions of gaze-direction in exist, they may be related to other variables social interaction. Ada Psychologica, 1967, 26, of interest. These variations in the use of 22-63. basic structural elements of conversations may LEIOHTON, L. A., STOLLAK, G. E., & FERGUSON, L. R. be the source of many of the subtle effects Patterns of communication in normal and clinic which, while difficult to specify explicitly, families. Journal of Consulting and Clinical Psy- chology, 1971, 36, 252-256. often have telling consequences on impres- MELTZER, L., MORRIS, W. N., & HAYES, D. P. In- sion formation and on the developmental terruption outcomes and vocal amplitude: Ex- course of interactions. plorations in social psychophysics. Journal of Personality and Social Psychology, 1971, 18, 392- REFERENCES 402. MILLER, G. A. Review of J. H. Greenberg (Ed.), BERNSTEIN, B. Social class, linguistic codes, and Universals of language. Contemporary Psychology, grammatical elements. Language and Speech, 1962, 1963, 8, 417-418. S, 221-240. ROSENFELD, H. M. Instrumental affiliative functions CRYSTAL, D. Prosodic systems and intonation in of facial and gestural expressions. Journal of English. Cambridge: Cambridge University Press, Personality and Social Psychology, 1966, 4, 65-72. 1969. SCHEFLEN, A. E. Natural history method in psy- CRYSTAL, D., & QUIRK, R. Systems of prosodic and chotherapy: Communicational research. In L. A. paralinguistic features in English. The Hague: Gottschalk & A. H. Auerbach (Eds.), Methods of Mouton, 1964. research in psychotherapy. New York: Appleton- DITTMANN, A. T., & LLEWELLYN, L. G. Relation- Century-Crofts, 1966. ship between vocalizations and head nods as SCHEFLEN, A. E. Human communication: Behavioral listener responses. Journal of Personality and Social programs and their integration in interaction. Psychology, 1968, 9, 79-84. Behavioral Science, 1968, 13, 44-55. DUNCAN, S. D., Jr., ROSENBERG, M. J., & FINKEL- SCHEGLOI-F, E. A. Sequencing in conversational open- STEIN, J. The paralanguage of experimenter bias. ings. American Anthropologist, 1968, 70, 1075- Sociometry, 1969, 32, 207-219 1095. DUNCAN, S. D., Jr., & ROSENTHAL, R. Vocal em- SULLIVAN, H. S. Conceptions of modern psychiatry. phasis in experimenters' instruction reading as New York: Norton, 1947. unintended determinant of subjects' responses. TRACER, G. L. Paralanguage: A first approximation. Language and Speech, 1968, 11, 20-26. Studies in Linguistics, 1958, 13, 1-12. EKMAN, P., & FRIESEN, W. V. The repertoire of non- TRACER, G. L., & SMITH, H. L., JR. An outline of verbal behavior: Categories, origins, usage, and of English structure. Washington, D.C.: American coding. Semiotica, 1969, 1, 49-98. Council of Learned Societies, 1957. GOFFMAN, E. On face work: An analysis of ritual YNGVE, V. H. On getting a word in edgewise. Papers elements in social interaction. Psychiatry, 1955, from the sixth regional meeting of the Chicago 18,213-231. Linguistic Society. Chicago: Chicago Linguistic GOFFMAN, E. Behavior in public places: Notes on the Society, 1970. social organization of gatherings. New York: Free Press of Glencoe, 1963. (Received for early publication October 5, 1971)