Convergence of Mean Vocal Intensity in Dyadic ......CONVERGENCE OF MEAN VOCAL INTENSITY 793 social...

15
Journal oi Personality and Social 1'sycholoxy 1975, Vol. 32, No. 5, 790 804 Convergence of Mean Vocal Intensity in Dyadic Communication as a Function of Social Desirability Michael Natale Ohio University Convergence of mean vocal intensity was demonstrated under two conditions simulating the natural dialogue. In the first experiment, the vocal intensity of the interviewer as heard by the subjects over two speakers placed at a 10-foot (3-m) distance, was restricted to three different intensity levels: 80-83, 86-89, and 92-95 db. SPL. The order of the presentation was counterbalanced across subjects. The second experiment used unstructured conversation, with each pair of subjects conversing for three 1-hour sessions. Prior to the first session, the subjects filled out the Marlowe-Crowne Social Desirability Scale. The results indicated that the lowering or raising of the interviewer's vocal level produced a corresponding change in the vocal level of the subject. Also, the degree to which a subject will match another subject's vocal level was predicted by the social desirability of the individual. These and earlier results are discussed in terms of conceptual frameworks. Speech behavior contains both verbal and vocal components, the latter being denned as the noncontent aspects of speech (Stark- weather, 19S6). With reference to vocal characteristics of speech, an area of research that has received increased attention over the last 20 years is the "synchrony" or "convergence" between conversationalists on various vocal parameters: speech latency durations (the reaction time before respond- ing to the other speakers; Matarazzo & Wiens, 1967), length of silence (Jaffe, 1967), duration of utterances (Matarazzo, Weitman, Saslow, & Wiens, 1963), vocal intensity (Black, 1949a, 1949b; Welkowitz, Feldstein, Finkelstein, & Ayelsworth, 1972), and various speech rates (Webb, 1967). The primary purpose of this study was to investigate the relevance of vocal intensity as an interpersonal cue in verbal communica- tion. More specifically, it was expected that speaking partners would converge in their average vocal intensity. The reliability of vocal intensity as a speech parameter for an individual has recently been demonstrated in a study by Welkowitz et al. (1972). Black Requests for reprints should be sent to Michael Natale, Number 2 Wolfe Street Apartments, Athens, Ohio 43 701. (1949a, 1949b), in a series of experiments, established that "the listener responds with greater intensity as he hears more intense signals." The subjects in Black's study re- peated statements or answered questions. It should be noted that in Black's study, the subjects listened through headphones and therefore were prevented from receiving normal sidetone feedback (hearing oneself). Moreover, the length of the experimenter's statements was a mere five syllables. The convergence of vocal intensity that Black demonstrated cannot be generalized to normal dialogues but can only be related to subjects' synchronization with specifically structured statements. The noninteractive structure of the conversation used by Black precludes the interpretation of the results as representative of a normal dialogue, which has no distinct response cues. Concerning the autoregulation of vocal intensity, of great importance are the long- known Lombard and Fletcher functions. These refer to the automatic raising of vocal intensity by a speaker with increased noise level and the raising of one's vocal activity with a decrease in the level at which one hears oneself (sidetone). These regulations by a speaker of vocal intensity for noise and sidetone compensation has long been con- 790

Transcript of Convergence of Mean Vocal Intensity in Dyadic ......CONVERGENCE OF MEAN VOCAL INTENSITY 793 social...

Page 1: Convergence of Mean Vocal Intensity in Dyadic ......CONVERGENCE OF MEAN VOCAL INTENSITY 793 social desirability, as measured by the Marlowe-Crowne scale (Crowne & Marlowe, 1960). In

Journal oi Personality and Social 1'sycholoxy1975, Vol. 32, No. 5, 790 804

Convergence of Mean Vocal Intensityin Dyadic Communication as a Function

of Social Desirability

Michael NataleOhio University

Convergence of mean vocal intensity was demonstrated under two conditionssimulating the natural dialogue. In the first experiment, the vocal intensityof the interviewer as heard by the subjects over two speakers placed at a10-foot (3-m) distance, was restricted to three different intensity levels: 80-83,86-89, and 92-95 db. SPL. The order of the presentation was counterbalancedacross subjects. The second experiment used unstructured conversation, witheach pair of subjects conversing for three 1-hour sessions. Prior to the firstsession, the subjects filled out the Marlowe-Crowne Social Desirability Scale.The results indicated that the lowering or raising of the interviewer's vocallevel produced a corresponding change in the vocal level of the subject.Also, the degree to which a subject will match another subject's vocal levelwas predicted by the social desirability of the individual. These and earlierresults are discussed in terms of conceptual frameworks.

Speech behavior contains both verbal andvocal components, the latter being dennedas the noncontent aspects of speech (Stark-weather, 19S6). With reference to vocalcharacteristics of speech, an area of researchthat has received increased attention overthe last 20 years is the "synchrony" or"convergence" between conversationalists onvarious vocal parameters: speech latencydurations (the reaction time before respond-ing to the other speakers; Matarazzo &Wiens, 1967), length of silence (Jaffe, 1967),duration of utterances (Matarazzo, Weitman,Saslow, & Wiens, 1963), vocal intensity(Black, 1949a, 1949b; Welkowitz, Feldstein,Finkelstein, & Ayelsworth, 1972), and variousspeech rates (Webb, 1967).

The primary purpose of this study was toinvestigate the relevance of vocal intensityas an interpersonal cue in verbal communica-tion. More specifically, it was expected thatspeaking partners would converge in theiraverage vocal intensity. The reliability ofvocal intensity as a speech parameter foran individual has recently been demonstratedin a study by Welkowitz et al. (1972). Black

Requests for reprints should be sent to MichaelNatale, Number 2 Wolfe Street Apartments, Athens,Ohio 43 701.

(1949a, 1949b), in a series of experiments,established that "the listener responds withgreater intensity as he hears more intensesignals." The subjects in Black's study re-peated statements or answered questions. Itshould be noted that in Black's study, thesubjects listened through headphones andtherefore were prevented from receivingnormal sidetone feedback (hearing oneself).Moreover, the length of the experimenter'sstatements was a mere five syllables. Theconvergence of vocal intensity that Blackdemonstrated cannot be generalized to normaldialogues but can only be related to subjects'synchronization with specifically structuredstatements. The noninteractive structure ofthe conversation used by Black precludes theinterpretation of the results as representativeof a normal dialogue, which has no distinctresponse cues.

Concerning the autoregulation of vocalintensity, of great importance are the long-known Lombard and Fletcher functions.These refer to the automatic raising of vocalintensity by a speaker with increased noiselevel and the raising of one's vocal activitywith a decrease in the level at which onehears oneself (sidetone). These regulationsby a speaker of vocal intensity for noise andsidetone compensation has long been con-

790

Page 2: Convergence of Mean Vocal Intensity in Dyadic ......CONVERGENCE OF MEAN VOCAL INTENSITY 793 social desirability, as measured by the Marlowe-Crowne scale (Crowne & Marlowe, 1960). In

CONVERGENCE OF MEAN VOCAL INTENSITY 791

sidered the function of a private loop (Azzi,1956; Becker, 1932; Charlip & Burk, 1967;Egan, 1967; Kerrison, 1918). But recentlyLane and Tranel (1971), in summarizing re-search evidence, have demonstrated that therole of the Lombard and Fletcher functionsis an attempt by the speaker to increase hiscommunicativeness and intelligibility and notthe result of an auditory feelback loop. Theproof of this hypothesis is detailed, but ithas been borne out in several experiments(Irwin & Mills, 196S; Lane, 1962; Lane,Catania, & Stevens, 1961; Black, Note 1).

Even more interestingly, Lane and Tranel(1971) illustrated that in regulation of vocalintensity for communicative and social pur-poses, not only does a speaker use privatefeedback loops, but he often uses a publicloop as well. More specifically, the speakeroften observes the consequences of his vocalbehavior on the vocal behavior of the audi-ence and adjusts his own vocal behavioraccordingly. In other words, since the speakeruses vocal intensity level as a mechanism forpromoting intelligibility and communicability,an excellent guideline for choosing the properlevel of vocal activity is to use, and hencematch, the vocal intensity level of the personwith whom he is conversing. Evidence forthis matching of vocal intensity is presentedby Black (1949a, 1949b), but as previouslynoted, the methodology of these experimentsdoes not allow one to interpret the resultsas proof of a vocal intensity convergencephenomenon that takes place in normaldyadic communication.

Recently Baircl and Tice (1969) havedemonstrated that when reading a list ofwords, individuals will evince imitative mod-eling of the vocal intensity level of the listof words previously recited to them; this wasin the absence of all overt social rewards.Again, this experiment does not simulate thenormal dialogue in many respects, but it doesprovide evidence that convergence of vocalintensity in spontaneous dialogues might beexpected to occur.

Equally germane to the phenomenon ofconvergence of vocal intensity in dyadic com-munication is the finding that absolute dif-ferences of vocal intensity in dyads are af-fected by length of interpersonal contact and

experimental manipulation of set (Welkowitzet al., 1972). The results of the Welkowitzet al. (1972) study cannot be consideredconclusive because the data were solely cor-relational in nature and obtained from a smallsample of eight dyads. However, researchseems to indicate that vocal intensity is astable vocal characteristic (Welkowitz et al.,1972) and that it is susceptible to interper-sonal influences (Baird & Tice, 3969; Black,1949a, 1949b; Welkowitz et al., 1972).

Of increasing interest is the examinationof social variables that interact with non-content speech convergence. Research in thisarea is scant. Matarazzo's (196S) studiesindicate that perceived dominance is a strongfactor contributing to convergence of meanduration of utterances; specifically, patientsconverged to a greater extent when the psy-chotherapist assumed a more directive role intherapy. The relationship of social variablesto convergence of noncontent speech behav-iors in dyads was more clearly demonstratedin an experiment by Welkowitz and Feldstein(1969), which found that perception of simi-larity significantly increased the convergenceof the average length of pauses and switchingpauses in dyads.

The possibility of personality variables in-teracting with noncontent speech convergencehas received a trickle of evidence. Matarazzo(1962) found that schizophrenics did notconverge in their duration of utterances inthe same manner as normals and neurotics.Even more interesting is the fact that psycho-logical differentiation (field dependence/inde-pendence) was found to be inversely relatedto the convergence of length of pauses innormals (Marcus, Welkowitz, Feldstein, &Jaffe, Note 2).

Other investigations have indicated (Bieri,Bradburn, & Galinsky, 1958; Crutchfield,Woodworth, & Albrecht, Note 3) that psycho-logical differentiation is inversely related tothe amount of interpersonal and social aware-ness and responsiveness that a person pos-sesses. Consequently, it seems that a person'sability to adapt socially or to comply isrelated to the degree that the person willmatch other people's speech behaviors. Thisinterpretation would be consonant withMatarazzo's (1962) study of schizophrenics

Page 3: Convergence of Mean Vocal Intensity in Dyadic ......CONVERGENCE OF MEAN VOCAL INTENSITY 793 social desirability, as measured by the Marlowe-Crowne scale (Crowne & Marlowe, 1960). In

792 MICHAEL NATALE

and the Marcus, Welkowitz, Feldstein, &Jaffe (Note 2) study on psychological dif-ferentiation as related to speech convergence.

Accordingly, the author feels that a mea-sure of a person's conformity would betteraccount for the convergence of noncontentspeech behaviors (note that Marcus et al.,Note 2, accounted for only 10% of the vari-ance), with the conformity being denned asthe giving up of a personal norm (Asch,1952), in this case, the individual's meanvocal intensity. It is logical to expect thatconformity constitutes a behavioral categorylikely to be engaged in by a person seekingto gratify a need for approval (Marlowe &Crowne, 1961). Hence, it is expected that so-cial desirability as measured by the Marlowe-Crowne Social Desirability Scale (Crowne &Marlowe, 1960) can significantly predict theconvergence of vocal intensity in a dyad.Research on verbal conditioning (Crowne &Strickland, 1962; Marlowe, 1962) indicatesthat the need, for approval does affect therealm of a person's speech behavior in thatverbal conditioning is differentially affectedby the social desirability of the subject in-volved. Even more direct evidence for therelationship between the amount of conver-gence of noncontent speech behavior, in thiscase vocal intensity, and the magnitude ofan individual's social desirability lies in thefact that Salman (Note 4) found that fielddependence was positively correlated withsocial desirability (Marlowe-Crowne scale).In light of the previously demonstrated rela-tionship between psychological differentiationand matching of vocal parameters of speech(Marcus ct al., Note 2), it seems that theconvergence of vocal intensity of speakingpartners would occur as a function of theindividual's social desirability, as measuredby the Marlowe-Crowne scale. A relationshipbetween social desirability of the individualand the magnitude of vocal intensity conver-gence in a dyad is completely consonant withthe previously described evidence of Laneand Tranel (1971), which indicates that anindividual regulates his vocal intensity forsocial purposes, that is, to increase his com-municativeness and intelligibility, and that hewill probably use the vocal level of the personhe is speaking with as a guideline in choosing

the effective level (amplitude) of his ownspeech.

In any research examining the occurrenceof approaching similarities of a particularbehavior in pairs of sets of individuals, itcould be asked, "What is the relationshipbetween the converging activity and time?"Previous research (Ray & Webb, 1966;Welkowitz & Feldstein, 1969) has indicatedthat speech convergence of mean durationof utterance increases over time. Also, asmentioned earlier, Welkowitz et al. (1972)found correlational evidence suggesting thatthe length of interpersonal contact may affectconvergence of vocal intensity in pairs ofspeakers. Therefore, it is also expected thatconvergence of mean vocal intensity in adyad will occur as a function of time; thatis, the differences between the mean vocalintensity of the two conversationists willdecrease over occasions.

Concerning the Marlowe-Crowne SocialDesirability Scale, it has been determined(Crowne & Marlowe, I960)) that the inter-nal consistency of the final form of the scaleis .88 (Kuder-Richardson Formula 20) andthat the need for approval is independent ofthe sex of the individual.

Hypotheses

On the basis of Baird and Tice's (1969)and Black's (1949a, 1949b) findings, a con-vergence of mean vocal intensity in a dyadwould be predicted. However, methodologicalapproaches used in these experiments obscureinterpretation of the results as indicatingthe existence of a natural vocal phenome-non in dyadic communication. Research byWelkowitz et al. (1972) obtained correla-tional evidence that suggests the followinghypotheses:

1. Hypothesis 1: Mean vocal intensitylevels change for an individual while con-versing with another person if that personchanges his speech behavior (loudness).

The findings of Lane and Tranel (1971),Marcus et al. (Note 2), Matarazzo (1962),and Salman (Note 4) indicate a secondhypothesis:

2. Hypothesis 2: Speech convergence ofmean vocal intensity in a dyad can be sig-nificantly accounted for by an individual's

Page 4: Convergence of Mean Vocal Intensity in Dyadic ......CONVERGENCE OF MEAN VOCAL INTENSITY 793 social desirability, as measured by the Marlowe-Crowne scale (Crowne & Marlowe, 1960). In

CONVERGENCE OF MEAN VOCAL INTENSITY 793

social desirability, as measured by theMarlowe-Crowne scale (Crowne & Marlowe,1960). In other words, the propensity of anindividual to act in a "sociable manner" isdirectly related to the degree to which hewill match another person's speech behavior,particularly, vocal intensity.

Previous research on psychological differen-tiation (Welkowitz & Feldstein, 1969;Marcus et al., Note 2) has indicated that theinteraction variable A X B significantly pre-dicts the convergence of noncontent speechbehavior (pauses and switching pauses). Inview of this empirical evidence, the followinghypothesis is offered:

3. Hypothesis 3: It is expected that theinteraction (multiplicative) of the partici-pants' social desirability would significantlyaccount for the convergence of mean vocalintensity. This interaction variable is termedthe social desirability of the dyad.

Research by Ray and Webb (1966),Welkowitz and Feldstein (1969), and Welko-witz et al. (1972) suggests a final hypothesis:

4. Hypothesis 4: Convergence of meanvocal amplitude occurs over time; that is, theabsolute differences of the conversationists'mean vocal intensity decreases over occasions.

Hypothesis 1 is explored in the first experi-ment, while the remaining hypotheses areexamined in the second experiment.

EXPERIMENT 1

The first experiment focused on the match-ing of a subject's mean vocal intensity withthat of the interviewer's, whose vocal inten-sity was experimentally controlled. A modi-fied standardized interview was used to ex-amine the hypothesis that the vocal intensityof the subject would change correspondinglywith that of the interviewer.

MethodSubjects

Interviewees were 21 male students of Ohio Uni-versity, enrolled in an introductory psychologycourse, who volunteered and received course creditfor their participation.

ApparatusThe conversation between the subjects and the

interviewer (experimenter) took place in soundproofrecording studios. The conversationalists were in sepa-

rate booths, not in view of each other. The speakingpartners heard each other's voices through speakersplaced at a 10-foot (3-m) distance. The use ofthis arrangement eliminated facial expression, headnods, and other postural cues.

Both participants' speech was recorded by a Sony350 tape recorder. The subject spoke into a uni-directional lavalier microphone set at 4 inches (.1 m)from his lips. The intensity of the experimenter'sspeech behavior was controlled by use of a Columbia"limiting" amplifier. This device absolutely fixedthe upper limit of the speaker's vocal intensity, withthe lower limit being established by a compressingamplifier that is contingent on the individual's powerand frequency. The lower limit of the experimenter'svocal intensity as amplified by the "limiter" wasfound to be no more than 3 db. below the fixedupper limit. Hence, the vocal intensity of the experi-menter, as heard by the subject, could be controlledwithin a specific range of 4 db. Another character-istic of the limiting amplifier is that it also auto-matically condenses the pitch range of the speakerto a characteristic level; therefore, inflection of theinterviewer's speech was held constant across allsubjects. It should be noted that the fidelity of thehuman voice is preserved by the "limiter," despitethe truncated range of pitch.

The mean vocal intensity of each subject wasmeasured by using a General Radio 1521-A graphiclevel recorder. The servo pen of the graphic levelrecorder operated at a writing speed of SO.8 cm persecond with a paper speed of .64 cm per second(.25 inches per second). This recorder displays ongraph paper the peak intensity on a 0-8-db. scale.The subject's mean vocal intensity was defined bythe average of the recorded peak amplitudes on thegraph at approximately 10-sec intervals. The peaklevel closest to the 10-scc mark was used as thedatum.

It is also possible to monitor the taped dialoguesas they arc fed into the graphic level recorder.Nonspecch behavior such as yawning, whistling, etc.,were eliminated from the data domain, approxi-mately 1%, thus rendering the reading of vocalintensity as a distinct paralinguistic behavior.

Procedure

All subjects were told, "This is merely an experi-ment on interpersonal communication; the only taskis to freely talk with each other." The subjectswere also told that their conversational partner wasin another room and that they were separated be-cause part of the purpose of the experiment wasto evaluate how people communicate under suchconditions. It should be noted that the subject didnot know that his partner was the interviewer.

The subjects were also given an additional set ofinstructions to ensure proper tape recording and toguarantee that a normal dialogue (no reading ofpoetry or prose) would take place. The general topicof the conversations was held constant across allsubjects by the interviewer, who initiated a con-versation on a topic related to the world situation.

Page 5: Convergence of Mean Vocal Intensity in Dyadic ......CONVERGENCE OF MEAN VOCAL INTENSITY 793 social desirability, as measured by the Marlowe-Crowne scale (Crowne & Marlowe, 1960). In

794 MICHAEL NATALE

The content of the interviewer's statementsadhered closely to Matarazzo's (1962) rules lor thestandardized interview:

1. The interviewer introduces the standardizedportion of the interview by a 5-sec utterance(following his signal to the observer).

2. All interviewing must be "nondirectivc." Nodirect questioning, no probing, and no depth inter-viewing. The experimenter can reflect, ask for clari-fication, ask for more information, or introduce anew general topic area. In general, the interviewer'scomments should be nonchallenging and open-endedand related to cither the subject's past commentsor some new general topic.

3. All of the interviewer's action must be verbalonly or verbal and gestural at the same time; thatis, the interviewer cannot use head nods and othergestures alone.

4. All of the interviewer's utterances must beS sec in duration.

5. Each period of the interview consists of 20utterances.

As stated above, the duration of the interviewer'sspeech sample was controlled across all subjects.Such exact control by the interviewer of the lengthand amount of his own utterances, with the aid ofa stopwatch and counter, was deemed feasible, sinceMatarazzo (1962) demonstrated that such finecontrol can easily be obtained with practice.

Even more important to the methodology of theexperiment was that the dialogues were unobtru-sively divided into three periods. By the previouslydescribed use of the "limiter," the vocal intensityof the interviewer across all subjects was specificfor each of the three periods: 80-83, 85-89, and92-95 db. SPL, these being the three experimentaltreatment conditions. When the prccstablishcd length(20 utterances by the interviewer) of a period wasmet, the interviewer gave a visual signal to theobserver, who in a separate booth out of view fromthe subject, manipulated the "limiter" and, hence,established the vocal intensity level of the inter-viewer as heard by the subject. The order of thethree controlled vocal intensity periods of the inter-viewer was counterbalanced across subjects in aLatin square design. The interviewer was not cog-nizant of the order of the presentations of the vocalintensity levels, thus creating a "blind" interviewerwho did not have knowledge of his vocal amplitudelevel as heard by the subject.

Therefore, the methodology of the present experi-ment eliminated visual-gestural cues, along with con-trolling verbal content (directive vs. nondirective,general topic of conversation). Also controlled werethe interviewer's inflection, number, and duration ofthe statements.

Prior to the experiment, practice sessions wereconducted to determine the specific vocal intensityrange of each period in order to ensure that thedifferent vocal intensity ranges would not be tooabrupt, in change or discomforting in any mannerto the subject. After the, experiment was completed,

a brief interview was conducted in which the sub-jects were asked about their understanding of thepurpose of the experiment.

Results

Confirmation of the validity of the experi-mental procedure was first examined. Post-experimental questioning revealed that noneof the subjects were aware that their speakingpartner was a specific "interviewer" ratherthan a fellow subject. The specified controlby the interviewer of his mean duration ofutterance was also demonstrated, X ~ 4.87sec, SD — .46 sec.

In the first experiment, it was predictedthat the manipulation of the interviewer'svocal intensity would produce a correspond-ing change in the mean vocal intensity ofthe subject. In Table 1, the means and stand-ard deviations of the treatment groups arepresented. Simple observation indicates thatthe average vocal intensity of the subjectsrose in a manner corresponding to theascending vocal intensity of the interviewer.

An analysis of variance (simple Latinsquare) was performed to check for the hy-pothesized change of the subject's mean vocalintensity. Order and group effects were foundto be nonsignificant. It may be noted thatthe order effect approached significance, F(2, 36) =3.13, p < .06, which indicates thateven in a short time, interval convergence ofvocal intensity does occur as a function oftime. This order effect was not predicted inthis experiment because the author suspectedthat the short length of interpersonal con-tact might preclude its occurrence, as indi-cated in Table 1. The main effect of vocalintensity was highly significant, F ( 2 , 36) —31.03, p < .001, which indicates that con-vergence of vocal intensity in dyadic com-munication is a powerful phenomenon. Winer(1962) pointed out that violation of the as-sumptions of equality and symmetry of co-variance matrices in a repeated-measures de-sign can likely produce an inflated F. Thiscan be offset by using a conservative F testfirst proposed by Greenhouse and Geiser(19S9). When such an analysis is performed,the main effect of vocal intensity is stillhighly significant, F(l, 18) = 31.03, p<.001.

Page 6: Convergence of Mean Vocal Intensity in Dyadic ......CONVERGENCE OF MEAN VOCAL INTENSITY 793 social desirability, as measured by the Marlowe-Crowne scale (Crowne & Marlowe, 1960). In

CONVERGENCE OP MEAN VOCAL INTENSITY 795

TABLE 1

MEANS AND STANDARD DEVIATIONS OF MEAN VOCALINTENSITY I.-OR Tin? THREE PERIODS

Occasions

Statistic 1

M" 79.50SD 4.49

2

81.654.46

3

84.024.83

a Mean sound pressure level in decibels.

The results with each subject can be treatedas a separate experiment, and his three meanscan be evaluated by an F test. Such an anal-ysis is presented in Table 2. When these 21individual F tests were computed, the signifi-cance level across the three means was .001for 10 subjects, .01 for 3 subjects, .05 for 4subjects, with the remaining 4 subjects failingto reach significance. A conservative F testyields similar results with the significancelevel at .001 for 7 subjects, .01 for 5 subjects,and .OS for 4 subjects. Thus it appears thatthe planned change in the vocal intensity ofthe interviewer has a rather marked effect onthe speech behavior of the subject, whetherthe latter is taken individually or as a group.In reference to the analysis presented inTable 2, it should be noted that the order ofthe three periods was not the same for allsubjects, the order being counterbalancedacross subjects in a Latin square design. Thiseliminated the necessity of a control group,which has been used by previous research(Matarazzo et al., 1963; Matarazzo & Wiens,1967) working within the framework of thestandardized interview to examine the match-ing or convergence of noncontent speech be-haviors.

Upon being informally interviewed postex-perimentally, none of the subjects were awarethat different levels of vocal intensity werepresented to them, which suggests that noneof the subjects guessed the purpose of theexperiment.

EXPERIMENT 2

Evidence from the first experiment sup-ported the proposition that convergence ofvocal intensity occurs in a dialogue situation.The second experiment used free conversa-tional exchanges with the intent of generaliz-

ing the occurrence of vocal intensity conver-gence as a natural paralinguistic phenomenonto the unstructured conversational situationin which a speaker would have to perceiveand respond to a full range of speech behav-ior, not restricted to the 4-clb. range as inthe first experiment. It was hypothesized inthis experiment that a person's (A or B) andthe dyad's (A X B) social desirability, asmeasured by the Marlowe-Crowne scale,would significantly account for the occur-rence of mean vocal intensity in a dyad. Itwas also expected that convergence of vocalintensity would be a function of the lengthof interpersonal contact, in other words, afunction of time.

MethodSubjects

Fifty same-sex dyads (25 females and 2S males)were chosen from the same subject pool describedin Experiment 1. Subjects in the dyad were ran-domly paired; it was also a necessary condition thatthe subjects had never been acquainted with eachother before and that they not be smokers, thelatter condition being necessary to avoid nonspecchbehavior from entering the data domain. The ex-clusion of smokers who would voluntarily refrain

TABLE 2

SUBJECT'S MEAN VOCAL INTENSITY ACROSS THREEPERIODS OF A DIALOGUE

Subject

s,S2S;,S4S6SeS7S8s.Su,SuSi!

S,3

Su

S,6

S,«SnSl8

Si»SaoSn

Period I :80-8>

76.0675.4579.S681.2981.4779.8469.8982.8275.7285.1781.6782.9482.1974.5083.7179.1481.6080.6284.1880.9069.10

Jntcrviewt

Period 2 :86-89»

77.9576.5482.5681.7181.5080.5073.7583.9477.5087.0386.3187.0087.5777.3986.6080.2585.3783.0981.9085.1972.13

;r's vocal intensity

Period 3 :92-95"

80.3377.7082.4683.2583.7581.6772.4487.2178.6789.7186.7587.7891.5279.6389.5785.8086.3587.7587.4586.3778.30

P

.001

.01

.05

.01

.05

.01

.001

.001

.001

.001

.001

.001

.001

.001

.05

.001

.05

Conser-vative1' test

.01

.05

.01

.05

.01

.001

.01

.001

.001

.001

.001

.001

.01

.05

.001

.05

a dh. SPL.

Page 7: Convergence of Mean Vocal Intensity in Dyadic ......CONVERGENCE OF MEAN VOCAL INTENSITY 793 social desirability, as measured by the Marlowe-Crowne scale (Crowne & Marlowe, 1960). In

796 MICHAEL NATALE

was deemed necessary because requiring a smokerto abstain for a lengthy period of time might induceanxiety in the person, and Feldstcin (1964) has in-dicated that nervousness and anxiety have substan-tial effects on an individual's vocal intensity.

Assessment oj Social DesirabilityA measure of the need for approval was obtained

by using the Marlowe-Crowne scale (Crownc &Marlowe, 1960). The scale was given to the subjectsto fill out before the experimental task (conversa-tions) .

ApparatusThe conversation between the pair of subjects took

place in a room (10 X 18 feet, or 3 X 5 m) withthe participants listening naturally to each otherwithout earphones or speakers. For recording pur-poses, each subject spoke into a lavalier unidirec-tional microphone set at a distance equal for bothmembers of a dyad. The room was acousticallydampened by putting cloth and egg cartons on thewalls in order to prevent the spilling over of oneperson's speech into his partner's microphone. As inthe first experiment, the subjects were not in viewof each other. This was accomplished by a hangingcurtain that divided the room in half. A GeneralRadio 1521-A graphic level recorder was used tomeasure the mean vocal intensity of the subjects.The writing and paper speed of the graphic levelrecorder was the same as in the first experiment.The subject's mean vocal intensity was defined bythe average of the peak-recorded amplitude at ap-proximately 10-sec intervals. The peak level closestto the 10-scc mark was used as the datum. Thetapes were monitored as they were analyzed by thegraphic level recorder, thus eliminating nonspecchbehavior from the data domain. Approximately 3%of the data were eliminated as nonspeech behavior;it should be noted that simultaneous speech wasnot deleted.

ProcedurePrior to the conversational task, the subjects were

asked to fill out the Marlowc-Crownc Social De-sirability scale; then some preliminary instructionswere given to ensure proper tape recording. As inthe previous experiment, the subjects were told thatthis was merely an experiment on interpersonal com-munication, and the only experimental task was totalk freely with their partner. Each session lasted60 minutes, the dyad meeting once a week for threeconsecutive weeks. The dyad participants were re-quested to make no attempts to see or meet witheach other, at least until the experiment was com-pleted.

The dialogues were unobtrusively divided intothree 10-minute periods. Period 1 was the first 10minutes of Dialogue 1; Period 2 was the middle 10minutes of the second dialogue; Period 3 was thelast 10 minutes of the final dialogue. The magnitudeof the convergence of the vocal intensity was de-

fined as the absolute change (increase or decrease)of the absolute difference of the mean vocal in-tensities in a dyad. The initial difference of thespeech parameter was defined as the difference ofthe mean values of the first period and the finaldifference defined as the difference of the mean vocalintensities in the last period. The convergence ofvocal intensity was observed over three 1-hour ses-sions because the only previous research (Welkowitzet al,, 1972) on the phenomenon indicated increas-ing similarity of vocal intensity between speakingpartners across three 1-hr dialogues. Evidence ofconvergence across other units of time, for example.10 or 20 minutes, has not been previously examined.

In an effort to gain further knowledge of the pro-cess of noncontcnt speech convergence, the subjectswere asked postexperimentally the degree to whichthey were conscious, if at all, of any self-regulationof the speech behavior being examined.

Results

In the second experiment, interspeaker in-fluence was examined in an unstructuredsituation. Multiple regression analysis wasused to evaluate the contribution of socialdesirability to the convergence of mean vocalamplitude. The social desirability scores forthe participants recorded on Channel A andfor those recorded on Channel B and the mul-tiplicative interaction of the two sets of scoreswere used as independent variables. Theamount of convergence of vocal intensity inthe dyad was the dependent variable. Theaverage social desirability score of the Mar-lowe-Crowne scale for the sample was 15.4with a standard deviation of 4.49, corre-sponding well with population norms of theMarlowe-Crowne Social Desirability scale:X= 1S.99, SD = 5.54.

The initial multiple correlations betweenthe vocal intensity convergence and the so-cial desirability variable are shown in Table3.

In performing the multiple regression, astepwise procedure was used that adds onevariable at a time to the regression until nofurther improvement can be made. The sen-sitivity of this stepwise procedure is both itsadvantage and disadvantage, since it mayeasily capitalize on accidental and unreli-able relationships to the criterion. Thereforea cross-validation is always necessary, whichin this case meant dividing the data into two25-dyad samples (12 male and 13 femaledyads in the first sample; 13 male and 12

Page 8: Convergence of Mean Vocal Intensity in Dyadic ......CONVERGENCE OF MEAN VOCAL INTENSITY 793 social desirability, as measured by the Marlowe-Crowne scale (Crowne & Marlowe, 1960). In

CONVERGENCE OF MEAN VOCAL INTENSITY 797

female dyads in the second sample). Thecross-validated multiple correlations are alsoreported in Table 3.

The social desirability of the individual,for Channel A, /?(1,23) = 1S.9S, p < .001;for Channel B, F(2, 22) = 7.63, p < .01, waspositively related to the convergence of meanvocal intensity; the subjects evaluated ashaving more "social desirability" contributedmore to the convergence of vocal intensitythan did the subjects with low "social desir-ability."

It should be noted that the separate regres-sion analyses for the dyadic partners wereredundant in terms of the hypothesis; bothanalyses tested the relationship of the indi-vidual's social desirability to the convergenceof vocal intensity in the dyad.

As was also predicted, a significant inter-action variable demonstrated that the socialdesirability of the dyad is positively relatedto the vocal intensity convergence, F(3, 21)= 4.61, p < .01.

A trend analysis across occasions on theabsolute differences between dyad memberson mean vocal intensity was used to evaluatewhether increasing similarities of vocal am-plitudes occurred as a function of time.Table 4 presents the means and standarddeviations of the dyads' absolute differencesof mean vocal intensity in the three 10-minute periods that were separated by 1-weekintervals.

Simple observation of the means indicatesthat the subjects converged increasingly overtime in the loudness of their speech behavior.This is verified in the trend analysis, which

TABLE 3

MULTIPLE CORRELATIONS BETWEEN VOCAL INTENSITYCONVERGENCE AND SOCIAL DESIRABILITY VARIABLES

Item

Multiple correlations

Cross-Inilial validated

.78

.79

.79

.64**

.64**

.63*

Note. SDn refers to the social desirability score for partici-pants on Channel A, SDb refers to the social desirability scorefor those on Channel B, and SD»ij refers to the interaction of thetwo sets of scores.

* p < .01.**]> <.0()1.

TABLE 4

MEAN AND STANDARD DEVIATIONS 01? DIITERENCKSCORKS OVER TlISKE PERIODS

Occasions

Item

MSD

4.672.80

4.383.14

2.962.58

demonstrates a highly significant, F(l, 147)= 8.94, p < .001, negative slope with a non-significant quadratic component, F(l, 147) =3.12, p> .05.

The postexperimental interview revealedthat none of the subjects were aware of thepurpose of the experiment; also, the subjectswere unaware of any self-regulation of vocalintensity.

Discussion

As was suggested by previous research, theresults of the present study indicate thatvocal intensity is a paralinguistic behaviorthat is modifiable in reference to the vocallevel of one's speaking partner. The resultsof Black's studies (1949a, 1949b), whichfound that subjects respond with greater in-tensity as they hear more intense signalswhen repeating or giving one-word answersto five-syllable phrases, can now be general-ized to normal dyadic communication that isunstructured. This finding is in completeagreement with Lane and Tranel's (1971)theory, which postulates that speakers regu-late their vocal level and, specifically, thatthey match their speaking partner's vocal in-tensity when choosing the optimum level fortheir own vocalizations. The present studyalso demonstrated that the social desirabil-ity score of the individual is positively re-lated to the extent that he will match hisspeaking partner's vocal intensity; this resultis consonant with Lane and Tranel's well-supported theory that the autoregulation ofvocal intensity is an attempt by the speakerto be communicable and intelligible. In otherwords, the autoregulation of vocal intensity is"other directed" or for social purposes. Be-fore discussing all of the theoretical implica-tions of the present findings, however, it isnecessary to interpret the results of this study

Page 9: Convergence of Mean Vocal Intensity in Dyadic ......CONVERGENCE OF MEAN VOCAL INTENSITY 793 social desirability, as measured by the Marlowe-Crowne scale (Crowne & Marlowe, 1960). In

798 MICHAEL NATALE

in reference to previous research on the con-vergence of noncontcnt speech behavior.

Convergence oj Noncontent Speech Behavior

The phenomenon of convergence was firstnoticed in early studies that examined thefrequency and distribution characteristics ofvocal parameters. Hargreaves (1960) in hisearly work on the duration of utterancesrealized that the mean duration changes fora person if that person changes speakingpartners. Matarazzo, Hess, and Saslow (1962)in a study on the duration of speech latencynoted that the interviewee's behavior is af-fected by the mean duration of the latencyof the interviewer. Matarazzo (1962), insummarizing previous work, pointed out thateven small changes in the interviewer's dura-tion of speech will significantly affect theduration of the interviewee's speech. Changesin the interviewee's speech corresponding tothat of the interviewer's verbal behavior hadbeen casually noted in other studies (Mata-razzo, Saslow, Matarazzo, & Phillips, 19S8;Saslow & Matarazzo, 1959; Saslow, Mata-razzo, & Guze, 19SS). In view of all of theempirical evidence, Matarazzo then pro-ceeded to demonstrate experimentally thatthe interviewee does converge to the noncon-tent speech behavior of the interviewer inlength of utterance (Matarazzo et al., 1963),frequency of interruptions (Wiens, Saslow,& Matarazzo, 1966), and speech latencyduration (Matarazzo & Wiens, 1967). But itis interesting to note that in both of Mata-razzo's studies, the interviewer and the in-terviewee were in view of each other. How-ever, Matarazzo, Wiens, Saslow, Dunham,and Voas (1964) did find a positive rela-tionship between the mean duration of ut-terances of astronauts and ground communi-cators.

Consistent with previous findings, Ray andWebb (1966) demonstrated that a similarconcordance of mean duration of utteranceoccurs in a dyadic communication with mul-tiple speakers at one end. Confirmation ofsynchrony or "tracking" of noncontent speechbehaviors has been demonstrated in analogousstudies: length of pause (Jaffe, 1967), dura-tion of utterances (Simpkins, 1967), and re-action time latency (Jackson & Pepenski,

1972; Cassotta, Feldstein, & Jaffe, Note 5).Concerning the fluency of speech, it shouldbe noted that Webb (1967), using a stan-dardized interview, demonstrated that theinterviewee will change his speech rate tothat of the interviewer's when the latter'sspeech is experimentally controlled. Inter-estingly, Lauver, Kelley, and Froehle (1971)have shown that interviewee verbal behaviorcan cause synchrony on the part of the coun-selor on reaction time latency and initiativetime latency, which supports the contentionof Jaffe and Feldstein (1970) and Matarazzoand Wiens (1972) that convergence is a two-channel phenomenon. Extending the syn-chrony model, Matarazzo, Wiens, Matarazzo,and Saslow (1968) have demonstrated theconvergence of interruption behavior and re-action time latency in psychotherapy sessions.

In light of the previously cited evidence,the convergence of mean vocal intensity dem-onstrated in the present study can be easilyunderstood as a constituent of the paralin-guistic "synchrony" phenomenon.

Theoretical Implications

At this point, an important question strikesthe investigator of noncontent speech con-vergence: Even though the results of thepresent experiment can be explained in thecontext of previous research, one is still leftwith the problem of defining the underlyingmechanism by which vocal intensity con-vergence occurs and the further question ofwhether this mechanism operates on the con-vergence of other noncontent speech be-haviors.

A long history of research in the speechsciences points toward a possible explanationof the vocal intensity convergence phenome-non. Lombard noted in 1911 that a speakerincreases his voice level when ambient noiselevel increases. Since then, many researchershave verified the existence of the "Lombardreflex" under different conditions: communi-cating words reproduced by listeners (Web-ster & Klump, 1962), reading word lists(Kryter, 1946), and shouting sentences un-der high noise levels (Hanley & Steer,3949). As Lane and Tranel (1971) havepointer! out in summarizing previous research,the speaker increases his vocal intensity about

Page 10: Convergence of Mean Vocal Intensity in Dyadic ......CONVERGENCE OF MEAN VOCAL INTENSITY 793 social desirability, as measured by the Marlowe-Crowne scale (Crowne & Marlowe, 1960). In

CONVERGENCE OF MEAN VOCAL INTENSITY 799

50% for each unit increase in noise. It wasalso discovered by Lombard that the voicelevel of the speaker increases as the level atwhich he hears his own voice (sidetone) de-creases. This has also been verified both inresearch under binaural conditions (Black,1951; Lightfoot & Merrill, Note 6) andmonaural conditions (Lane et al., 1961).

It was long thought that noise and side-tone compensation resulted from a loss inaudiofeedback control (Azzi, 19S6; Becker,1932; Charlip & Burk, 1969; Kcrrison,1918). But this does not explain why thereis a decrease in voice level when noise leveldecreases or why sidetone compensation oc-curs under monaural conditions (Noll, 1964a,1964b). Lane and Tranel (1971) pointed outthat noise and sidetone compensation arcdifferent aspects of a unitary process; that is,both types of autoregulation of vocal inten-sity by the speaker are attempts to adjustfor changes in the signal-to-noise ratio in or-der to preserve intelligibility. This "communi-cative" explanation by Lane and Tranel ofthe Lombard reflex and sidetone compensa-tion is supported by research (Gardner,1966) demonstrating that noise compensa-tion will increase in a speaker if the taskputs a premium on intelligibility. Also, Black(1949c) has shown that the extent of theLombard reflex is positively related to themessage type, with the Lombard reflex in-creasing with messages perceived by thespeaker as more meaningful and importantto communicate.

But the strongest evidence for the "com-munication" function of the autoregulationof vocal intensity is provided by Lane andTranel (1971), who illustrated that if noiseand sidetone compensation are a unitaryprocess to increase intelligibility, then a "sub-ject told to imitate changes in the level of anoise or speech signal will match a four-fold increase with a twofold increase, justas he does in noise and sidetone compensa-tion" (p. 688). In other words, both noiseand sidetone compensation are a "matching"(convergence?) function. Previous researchindeed supports this critical test of the "com-munication" model of autoregulation of vocalintensity (Lane, 1962; Lane et al., 1961;Black, Note 1).

But given that the autoregulation of vocalintensity in sidetone and noise compensa-tion is an attempt by the speaker to beunderstood and to be intelligible, how doesthis apply to the matching of an individual'svocal level and that of the person with whomhe is speaking? This becomes understandableif one pays heed to Lane and Tranel's (1971)observation that public feedback loops, inaddition to private feedback loops, are usedby the speaker to choose the optimum levelof speech. To put it more simply, the vocallevel at which a conversational partner di-rects his speech is an excellent indicant of"guideline" for choosing the appropriate vocallevel of one's own speech. The phenomenonof vocal intensity convergence can then beseen to be the result of a public feedback sys-tem used by speakers to ensure intelligibilityof their speech. Evidence for this can befound in Black's (1949a, 1949b) studies andthat of Baird and Tice (1969), which dem-onstrated that speakers answering questionsof repeating statements will adjust their vocalintensity to that of the vocal level of the ex-perimenter. The present study generalizesthis phenomenon to the natural dyadic com-munication situation.

The strength of this "communication"model of vocal intensity convergence is thatit can account for the "matching" phenome-non of other noncontent speech behaviors.Convergence of speech rates (Webb, 1967;Lightfoot, Note 7) , precision of articulation(Tolhurst, Note 8), duration of utterance(Matarazzo et al., 1963), and speech latencyduration (Matarazzo & Wiens, 1967) haveall been demonstrated and are easily under-stood through the "communication model,"which posits that matching or convergence ofnoncontent speech behavior is an automaticprocess used by the speaker to attain the op-timal "format" of his speech behavior so thathis message will be intelligible.

Jakobson (1968), in talking about phono-logical development in children noted that"there arises and grows by degree in childrena desire for communication. We witness thefirst expression of his social life; the childseeks to respond to and adapt himself to theperson with whom he is speaking in everyway, even in changes of volume" (p. 24).

Page 11: Convergence of Mean Vocal Intensity in Dyadic ......CONVERGENCE OF MEAN VOCAL INTENSITY 793 social desirability, as measured by the Marlowe-Crowne scale (Crowne & Marlowe, 1960). In

800 MICHAEL NATALE

Jakobson, citing the work by Gregoire,van Grinken, & Buhlcr & Helzer, interest-ingly pointed out that linguists had long ob-served that the infant, in his apparent desireto communicate, will match the vocal inten-sity of the person he is attempting to com-municate with. It then seems that anecdotalevidence from early developmental linguisticresearch is strong support for the propositionthat the desire for communication and intel-ligibility in one's speech is the mechanism bywhich the convergence of nonconlent speechbehavior occurs.

The ability of the communication modelto explain convergence of noncontent speechbehavior is indeed important, since in review-ing previous research, Webb (1970) notedthat Matarazzo's (196S) "synchrony" modelIs the only previous attempt to explain thematching of vocal behavior but that thereare numerous faults in the theory. The syn-chrony model posits that the greater activityby the interviewer is perceived by the inter-viewee as indicating greater interest by theinterviewer and is, hence, imitated. Webb(1970) aptly pointed out that the synchronymodel is especially deficient in accounting forconvergence of speech rates, since previousresearch (Rucsch & Priestwood, 1949) hasindicated that high speech rates are perceivedas unpleasant, hence not positively reward-ing; yet matching of speech rates is found.This line of reasoning is also applicable tovocal intensity convergence, since it has beenshown that higher vocal intensity is asso-ciated with persuasiveness (Packwood, 1974;Scherer, Rosenthal, & Koivumaki, 1972) andthat very high vocal intensity is perceived asan expression of hostility (Constanzo, Mar-kcl, & Constanzo, 1969). In other words,there seems to be curvilinear relationshipbetween persuasiveness and vocal intensity,yet vocal intensity convergence has beendemonstrated to occur at several levels ofvocal intensity in the present experiment,and imitative modeling of vocal intensity(Baird & Tice, 1969) has been shown tooccur at very high and very low levels ofvocal intensity. It therefore seems that areinforcement paradigm as put forth in thesynchrony model is not able to account for

the convergence of mean vocal amplitude indyads.

Webb (1970) mentioned that "visual, butnon-behavioral, cues instigate synchrony" (p.14). Specifically, it has been demonstratedby Bender and Brister (Note 9) that there isa positive correlation between the number ofstimulus words and the number of responsewords in a sentence-completion test. This ex-ample of congruent behavior cannot be ex-plained by Matarazzo's synchrony modelbut can easily be understood within theframework of the communication model,which defines the format or structure of in-coming input as the criterion by which asource structures the format of this output.

The communication model can also ex-plain the convergence of an individual'sspeech behavior to general environmentalstimuli, which has been demonstrated byHeckel, Wiggen, and Salzberg (1963) andBrister (Note 10), whose studies illustratedthat convergence of speech rates to the tempoof background music occurs. If the speakeruses external stimuli, such as the conversa-tional partner's speech, as the feedbackmechanism for establishing his own vocal be-havior, then it is possible that leakage intothis public feedback loop from environmentalstimuli can affect the vocal behavior of aspeaker.

Matarazzo and Wiens (1967) suggestedthat noncontent speech convergence might berelated to Bandura's (1965) modeling frame-work. Baird and Tice (1969), in discussingimitative modeling of vocal intensity, alsostated that a modeling process similar to thatdescribed by Bandura might provide an ex-planation for noncontent speech convergence.But Bandura's modeling theory would havedifficulty explaining the results of Black's(1949b) study in which, after demonstratingthe convergence of vocal intensity, the ex-perimenter informed the subjects of the "con-vergence" phenomenon and then requestedthe subjects to prevent its occurrence in thefollowing experimental session. It was foundthat the subjects converged in vocal intensityin spite of the experimental instructions toprevent the imitative behavior. The social re-ward paradigm of Black's study becomes

Page 12: Convergence of Mean Vocal Intensity in Dyadic ......CONVERGENCE OF MEAN VOCAL INTENSITY 793 social desirability, as measured by the Marlowe-Crowne scale (Crowne & Marlowe, 1960). In

CONVERGENCE OF MEAN VOCAL INTENSITY 801

complex, and with it, a Bandurian modelingtheory of noncontent speech behavior becomesdifficult to interpret.

The efficacy of a Bandurian model as ap-plied to the modifiability of noncontentspeech characteristics is also discounted byLauver et al. (1971), who demonstrated thatthe timing of a counselor's speech behaviorcan be "influenced" by the client's analogousspeech behavior.

Communication Model

As stated thus far, the communicationmodel merely posits that convergence of non-content speech behavior is a "social" func-tion; that is, it represents an attempt by thespeaker to attain optimal format of hisspeech, via a public feedback loop, so thathis speech will be intelligible. Tt should benoted that the role of a "public feedbacksystem" in regulating nonverbal behavior hasbeen touched upon by Ekman and Friesen(1969); these authors, in formulating amodel to account for nonverbal leakage,stated that an individual uses both internaland external nonverbal feedback to programhis own actions (both verbal and nonverbal).However, the posited communication modelin this study can be expanded to encompassthe following constructs: (a) that personcharacteristics are related to noncontentspeech "synchrony" in dyadic communica-tion, (b) that dyadic communication is trans-actional in nature, with certain social con-straints modifying the particular dialogue,and (c) that Factors a and b can interact.

Relevant to the first proposition are thefindings of the present study. Specifically, 4out of 21 subjects in the first experiment didnot converge in mean vocal intensity, andthat in the second experiment, 4 subjects withlow social desirability had a correspondingnegative dyadic synchrony. This finding issimilar to Matarazzo et al.'s (1968) obser-vation that some therapist-client pairs con-verge over time (reaction time latency andinterruptions), whereas others do not. Theseresults indicate that differences in an indi-vidual's empathy, rapport, social desirability,or other relevant personality characteristicsfire related to the degree of noncontent speech

convergence in various dyads. Directly re-lated to this proposition, Welkowitz and Kuc(1973) have demonstrated a relationship be-tween the degree of dyadic convergence ofresponse latency and the rated warmth of anindividual. This is congruent with Mata-razzo's (1973) stated belief that speech andsilence indexes (i.e., s5Oichrony) have the po-tential to reveal an individual's mood, atti-tude, or motivational state. Further researchshould evaluate the relationship of therapisttype and therapist experience level to thephenomenon of synchrony.

Concerning the second assumption of thecommunication model, Heller (1971) in sum-marizing previous research, stated that theinterview participants are seen as a partialcause as well as a partial result of the behav-ior of the other. The effect of "social con-straints" on the synchrony phenomenon isexemplified by Matarazzo et al.'s (1968)finding of a lack of positive synchrony ofduration of utterance. In analyzing the data,Matarazzo et al. found some evidence sug-gesting that therapist-client productivity ofspeech was inversely related. The investi-gators interpreted this negative synchrony aspossibly being the result of a "therapeuticset," described as the tendency to allow apatient to talk productively by means oflow activity on the part of the therapist orto elicit talk from reticent patients by meansof high therapist activity. It then appearsthat patient verbalization in psychotherapyis a clear example of the relationship ofsocial-situational factors to the synchronyphenomenon. Further confirmation of this no-tion can be found in a recent study by Pope,Nudler, Vonkorff, and McGhee (1974),which demonstrated that dyadic synchrony ofspeech productivity was inversely related tothe experienced interviewer and positively re-lated for a fellow-student interviewer; thesame interview under different social "con-straints" produced different convergence be-havior. It should be noted in the Pope et al.study that both the experienced and thenovice interviewer were rated as approxi-mately equal in empathy, indicating that thesocial-perceptual set (status of interviewer)of the particular dialogue determined the

Page 13: Convergence of Mean Vocal Intensity in Dyadic ......CONVERGENCE OF MEAN VOCAL INTENSITY 793 social desirability, as measured by the Marlowe-Crowne scale (Crowne & Marlowe, 1960). In

802 MICHAEL NATALE

synchronous behavior. The influence of per-ceptual set on speech convergence behaviorwas also demonstrated in an earlier study byWelkowitz et al. (1969), who found thatinduced perceived similarity was positivelyrelated to the synchrony of response latenciesand pause lengths.

Directly related to this research is thethird assumption of the communication model,which emphasizes the interrelationships ofperson characteristics, social constraints ofthe dialogue, and noncontent speech conver-gence. This can best be understood in termsof interview research. In summarizing previ-ous research, Matarazzo and Wiens (1972)pointed out that concerning certain speechvariables, the activity level of the therapistand the client in a psychotherapy situa-tion usually covaried. However, research byHeller, Myers, and Kline (1963) has shownthat a direct convergence relationship is ob-scured in the naturalistic interview by thepersonality characteristics of the conversation-alists. Specifically, passive interviewees pullactivity from interviewers, while dominantinterviewees induce interviewer passivity.This research implies that the variables ofperson characteristics and dialogue situationprobably interact concerning convergence ofnoncontent speech. A representative design,as suggested by Brunswick (19S6) might bethe only suitable approach toward investi-gating the complex convergence phenomenon.

In summary, a strong argument has beenmade for a communication model of non-content speech convergence. The evidence forthis model is primarily that it successfullypredicted the convergence of vocal intensityin the present study and that it offers asatisfactory explanation of the convergenceof vocal characteristics of speech. However,the communication model serves a primarilyheuristic purpose in that it emphasizes theimportance of both person characteristics andenvironmental press (i.e., therapeutic set)toward the convergence phenomenon. Theauthor acknowledges, along with Matarazzoet al. (1968), that delineation of the myriadof environmental determinants and individualtrait and state factors relevant to the silenceand speech characteristics of the dialoguemay very well be beyond the scope of an

individual investigator, and that massive"cross-research" comparisons may be the onlymethod of investigation.

REFERENCE NOTES

1. Black, J . W. Control of the soimd measure ofthe voice (Joint Project No. 42, U.S. NavyProject No. NM 001 104 500.42). Bureau ofMedicine and Surgery, 1955.

2. Marcus, E. S., Welkowitz, J., Feldstein, S., &Jaffc,, J. Psychological differentiation and thecongruence of temporal speech patterns. Paperpresented at the meeting of the Eastern Psycho-logical Association, Atlantic City, April 1970.

3. Crutchfickl, R. S., Woodworlh, D. G., &Albrccht, R. E. Perceptual performance and theeffective person (Personal Laboratory ReportWADC-TN-S860 ASTIA, Document No. 151-039). Lackland Air Force Base, Texas, 1958.

4. Salman, A. R. The need for approval, improvisa-tion and attitude change. Unpublished master'sthesis, Ohio State University, 1962.

5. Cassotta, L., Feldstein, S., & Jaffc, J. Stability,influence, and congruence in dyadic vocal inter-action. Paper presented at the meeting of theEastern Psychological Association, Boston, April1967.

6. Lightfoot, C., & Morrill, S. W. Loudness ofspeaking: The effect of the intensity of sidetoneupon the intensity of the speaker (Joint ProjectReport, No. 4, Bureau of Medicine and Surgery,U.S. Navy Project No. 001.053, 1949).

7. Lightfoot, C. Rate of speaking: I. Relationsbetween original and repealed phrases (JointProject NR 782 004, Report No. 1). Pensacola,Fla.: Kenyon College and Naval School of Avia-tion Medicine and Research, 1949.

8. Tolhurst, G. C. Some effects of changing, timepatterns and articulation upon intelligibility andword perception (Project No. 145-993, ContractN6 onw-22525; NMRI Project NM 001 104 50040). Pensacola, Fla.: U.S. Naval School ofAviation Medicine Naval Air Station; andColumbus: Ohio State University ResearchFoundation, 1955.

9. Bender, L., & Bristcr, U. M. Sex, synchrony andsentence completion. Unpublished manuscript,University of Alabama, 1968.

10. Brislcr, D. M. The effects of music on verbalrate in an interview situation. Unpublishedmanuscript, University of Alabama, 1968.

REFERENCES

Asch, S. E. Social psychology. New York: Prentice-Tlall, 1952.

Azzi, A. II controllo auditive dclla fonazione. RivistaIlaliana di Audiologia Practicare, 1956, 6, 65-68.

Baird, J. C., & Ticc, M. Imitative modelling ofvocal intensity. Psychonomic Science, 1969, 19,219-220,

Page 14: Convergence of Mean Vocal Intensity in Dyadic ......CONVERGENCE OF MEAN VOCAL INTENSITY 793 social desirability, as measured by the Marlowe-Crowne scale (Crowne & Marlowe, 1960). In

CONVERGENCE OF MEAN VOCAL INTENSITY 803

Bandura, A. Behavioral modification throughmodelling procedures. In L. Krasner & L. P.Ullman (Eds.), Research in behavior modification.New York: Holt, Rinehart & Winston, 1965.

Becker, B. M. The Barany noise apparatus. TheAnnals of Otology, Rhinology and Laryngology,1932, 41, 222-226.

Bicri, J. W., Bradburn, W. M., & Galinski, M. D.Sex differences in perceptual behavior. Journal ofPersonality, 1958, 26, 1-12.

Black, J. W. The intensity of oral responses tostimulus words. Journal of Speech and HearingDisorders, 1949, 14, 16-22. (a)

Black, J. W. Loudncss of speaking: The effect ofheard stimuli on spoken responses. Journal ofExperimental Psychology, 1949, 39, 311-315. (b)

Black, J. W. The relation between message typeand vocal rate and intensity. Speech Monographs,1949, 18, 74-84. (c)

Black, J. W. The effect of noise induced temporarydeafness upon vocal intensity. Speech Monographs,1951, 18, 74-77.

Brunswik, E. Perception and representative designof psychological experiments. Berkeley: Universityof California Press, 19S6.

Charlip, W. S., & Burk, K. W. Effects of noise onselected speech parameters. Journal of Communi-cation Disorders, 1969, 2, 212-219.

Constanzo, F. S., Markcl, N., & Conslanzo, P. R.Voice quality profile and perceived emotion. Jour-nal of Counseling Psychology, 1969, 16, 267-270.

Crownc, D. P., & Marlowe, D. A new scale ofsocial desirability independent of psychopalhology.Journal of Consulting Psychology, 1960, 24, 349-354.

Crownc, D. P., & Strickland, B. R. The condition-ing of verbal behavior as a function of the needfor approval. Journal of Abnormal and SocialPsychology, 1962, 63, 395-401.

Egan, J. J. Psychoacoustics of the Lombard reflex.Unpublished doctoral dissertation, Western ReserveUniversity, 1967.

Ekman, P., & Friesen, W. Nonverbal leakage andclues to deception. Psychiatry, 1969, 32, 88-106,

Fcldstein, S. Vocal patterning of emotional expres-sion. In J. H. Masserman (Ed.), Science andpsychoanalysis (Vol. 3). New York: Grune &Straiten, 1964.

Gardner, M. B. Effect of noise, system gain, andassigned task on talking level in loud-speakercommunication. Journal of the Acoustical Societyof America, 1966, 40, 955-965.

Greenhouse, S. W., & Geiscr, S. On methods in theanalysis of profile data. Psychometrika, 1959, 24,95-112.

Hanley, T. D., & Steer, M. Effect of level of dis-tracting noise upon speaking rate, duration, andintensity. Journal of Speech and Hearing Dis-orders, 1949, 14, 363-368.

Hargreaves, W. A. A model for speech unit duration.Language and Speech, 1960, 3, 164-173.

Heckcl, R. V., Wiggcn, S. L., & Salzburg, H, C.The effect of musical tempo in varying operant

speech levels in group psychotherapy. Journal ofClinical Psychology, 1963, 19, 129.

Heller, K. Laboratory interview research as analogueto treatment. In A. E. Bergin & S. L. Garficld(Eds.), Handbook of psychotherapy and behaviorchange. New York: Wiley, 1971.

Heller, K., Myers, R., & Kline, L. Interviewer be-havior as a function of standardized client role.Journal of Counseling Psychology, 1966, 30, 501-508.

Irwin, R. J., & Mills, A. W. Matching loudncss andvocal level: An experiment requiring no apparatus.British Journal of Psychology, 1965, 56, 143-146.

Jackson, R., & Pcpinski, H. Interviewer activity andstatus effects upon rcvealingness in the initialinterview. Journal of Clinical Psychology, 1972,28, 440-404.

Jaffc, J. Computer analysis of verbal behavior inpsychiatric interviews. In D. Rioch (Ed.), Dis-orders in communication: Proceedings of the Asso-ciation for Research in Nervous and Mental Dis-eases (Vol. 42). Baltimore: Williams & Wilkins,1967.

Jaffe, J., & Fcldstein, S. Rhythm.s of dialogue. NewYork: Academic Press, 1970.

Jakobson, R. Child, language, aphasia and phono-logical universals. The Hague: Mouton, 1968.

Kerrison, P. D. The tests for malingering in defec-tive hearing. Larnygoscope, 1918, 28, 662-665.

Krytcr, K. D. Effects of ear protective devices onthe intelligibility of speech in noise. Journal of theAcoustical Society of America, 1946, 18, 412-417.

Lane, H. L. Psychophysical parameters of vowelperception. Psychological Monographs, 1962, 76(No. 44, Whole No. 563).

Lane, H. L., Catania, A. C., & Stevens, S. S.Voice level: Autophonic scale, perceived loudncss,and effects of sidctone. Journal of the AcousticalSociety of America, 1961, 33, 160-167.

Lane, H. L., & Trancl, B. The Lombard reflex andthe role of hearing in speech. Journal of Speechand Hearing Research, 1971, 14, 677-709.

Lauvcr, P., Kellcy, J., & Froehle, T. Client reactiontime and counselor verbal behavior in an interviewsetting. Journal of Coimseling Psychology, 1971,18, 26-30.

Lombard, E. Le signe dc ['elevation de la voix.Annales de Maladie Oreille et Laryngologie, 1911,37, 101-119.

Marlowe, D. Need for approval and the operantconditioning of meaningful verbal behavior. Jour-nal of Consulting and Clinical Psychology, 1962,26, 79-83.

Marlowe, D., Si Crowne, D. P. Social desirabilityand responses to perceived situalional demands.Journal of Consulting and Clinical Psychology,1961, 25, 109-115.

Matarazzo, J. D. Prescribed behavior therapy:Suggestions from interview research. In A. J.Bachrach (Ed.), Experimental foundations ofclinical psychology. New York: Basic Books,1962.

Matarazzo, J. D. The interview. In B. B. Wolman

Page 15: Convergence of Mean Vocal Intensity in Dyadic ......CONVERGENCE OF MEAN VOCAL INTENSITY 793 social desirability, as measured by the Marlowe-Crowne scale (Crowne & Marlowe, 1960). In

804 MICHAEL NATALE

(Ed.), Handbook of clinical psychology. NewYork: McGraw-Hill, 196S.

Matarazzo, J. D. A speech interaction system. InIX J. Kieslcr (Ed.), The process of psycho-therapy: Empirical foundations and systems ofanalysis. Chicago: Aldinc, 1973.

Matarazzo, J. IX, Hess, II. F., & Saslow, G. Fre-quency and duration characteristics of speech andsilence behavior during interviews. Journal ofClinical Psychology, 1962, 18, 41 6-426.

Matarazzo, J. D., Saslow, G., Matarazzo, R. G., &Phillips, J. Stability and modifiabilily of person-ality patterns during the standardized interview.In P. A. Hoch & J. Zubin (Eds.), Psychopathol-ogy of communication. New York: Grune &Stratton, 1964.

Matarazzo, J. IX, Weitman, M., Saslow, G., &Wicns, A. N. Interviewer influence on duration ofinterviewee speech. Journal of Verbal LearninKand Verbal Behavior, 1963, 1, 451-458.

Matarazzo, J. D., & Wicns, A. N. Interviewer influ-ence on duration of interviewee silence. Journalof Experimental Research in Personality, 1967,2, 56-59.

Matarazzo, J. D., & Wiens, A. W. The interview:Research on its anatomy and structure. Chicago:Aldine-Athcrton, 1972.

Matarazzo, J. D., Wicns, A. N., Matarazzo, R., &Saslow, G. Speech and silence behavior in clinicalpsychotherapy and its laboratory correlates. InJ. M. Schlein, H. Flunt, J. Matarazzo, & G.Saslow (Eds.), Research in psychotherapy (Vol.3). Washington, D.C.: American PsychologicalAssociation, 1968.

Matarazzo, J. D., Wiens, A. N., Saslow, G.,Dunham, R. M., & Voas, R. Speech duration ofastronauts and ground communicator. Science,1964, 143, 148-150.

Noll, A. M. Effects of head and air-leakage sidetoneduring monaural telephone speaking. Journal ofthe Acoustical Society of America, 1964, 36, 598-599. (a)

Noll, A. M. Subjective effects of sidelonc duringtelephone conversations. IEEE Transactions onCommunications and Electronics, 1964, S3, 228-231. (b)

Packwood, W. T. Loudncss as a variable in per-suasion. Journal of Counseling Psychology, 1974,21, 1-2.

Pope, B., Nudler, S., Vonkorff, M., & McGhce, J.The experienced professional interviewer and thecomplete novice. Journal of Consulting and Clin-ical Psychology, 1974, 42, 680-690.

Ray, M. L., & Webb, E. J. Speech duration effectsin the Kennedy news conferences. Science, 1966,153, 899-901.

Ruesch, J., & Pricstwood, A. R. Anxiety: Its initia-tion, communication and interpersonal manage-

ment. Archives of Neurology and Psychiatry,1949, 62, 527-550.

Saslow, G., & Matarazzo, J. D. A technicrue forstudying changes in interview behavior. In E. A.Rubinstein & M. D. Parloff (Eds.), Research inpsychotherapy (Vol. 1). Washington, B.C.: Amer-ican Psychological Association, 1959.

Saslow, G., Matarazzo, J. D., & Guze, S. B. Validityof interaction chronograph patterns in psychiatricinterview. Journal of Counseling Psychology,1955, 19, 417-430.

Schercr, K. R., Roscnthal, R., & Koivumaki, J.Mediating expectancies via vocal cues: Differentialspeech intensity as a means of social influence.European Journal of Social Psychology, 1972, 2,163-176.

Simpkins, L. The effects of utterance duration onverbal conditioning in small groups. Journal ofSocial Psychology, 1967, 71, 69-78.

Starkweather, J. A. Content-free speech as a sourceof information about the speaker. Journal of Ab-normal and Social Psychology, 1956, 52, 394-402.

Webb, J. T. An analysis of two speech rate mea-sures in an automated standardized interview.Unpublished doctoral dissertation, University ofAlabama, 1967.

Webb, J. T. Interview synchrony: An investigationof two speech rate measures in an automatedstandardized interview. In A. W. Siegman & B.Pope (Eds.), Studies in dyadic communication:Proceedings of a research conference on the inter-view, New York: Pergamon, 1970.

Webster, J. C., & Klump, R. G. Effects of ambientnoise and near-by talkers on a face-to-face com-munication task. Journal of the Acoustical Societyof America, 1962, 34, 936-941.

Welkowitz, J., & Feldstein, S. Dyadic interaction andinduced differences in perceived similarity. Pro-ceedings of the 77th Annual Convention of theAmerican Psychological Association, 1969, 4, 343.(Summary)

Welkowitz, J., Feldstein, S., Finkelstein, M., SiAyelsworth, L. Changes in vocal intensity as afunction of inlerspcaker influence. Perceptual andMotor Skills, 1972, 35, 715-718.

Wclkowitz, J., & Kuc, M. Interrelationships amongwarmth, genuineness, empathy and temporal speechpatterns in interpersonal interaction. Journal ofConsulting and Clinical Psychology, 1973, 41, 472-473.

Wicns, A. N., Saslow, G., & Matarazzo, J. IXSpeech interruption behavior during interviews.Psychotherapy: Theory, Research, and Practice,1966, 3, 153-158.

Winer, B. J. Statistical principles in experimentaldesign. New York: McGraw-Hill, 1962.

(Received August 16, 1974)