12
Journal of Personality and Social Psychology 1982, Vol. 43, No. 2, 236-247 Copyright 1982 by the American Psychological Association, Inc. 0022-3514/82/4302-0236$00.75 Impact of Salient Vocal Qualities on Causal Attribution for a Speaker's Behavior Janet Robinson and Leslie Zebrowitz McArthur Brandeis University Considerable research has demonstrated an illusory causation effect in which visually salient people are perceived as more causal of events in a social inter- action than their nonsalient counterparts. The present studies extended this work to the realm of auditory salience. Two determinants of auditory salience were manipulated—the intensity of a speaker's voice and the sex of a speaker's voice. As predicted, subjects attended more to a 75-dB (A) than to a 70-dB voice on a binaural listening test, and subjects attributed more causality to an actor in a two-person conversation when his voice was 75 dB in intensity than when it was 70 dB. Contrary to expectation, subjects did not attend more to the voice whose sex matched their own on the binaural listening test. Rather, all subjects listened more to the actor with the male voice. Consistent with this tendency for the male voice to be more salient, subjects attributed more causality to an actor when the voice was-male than when it had been electronically converted to a female voice of the same intensity and intonation. Vocal salience also influenced subjects' impressions of the actors, but it had no impact on recall of the actors' verbalizations. Theoretical and practical implications of these findings are dis- cussed. Considerable research evidence has re- vealed that salient stimuli—those that draw perceivers' attention—tend to be perceived as causing the events in a social interaction. This illusory causation effect (McArthur, 1980) has been obtained in a number of stud- ies that employed salience manipulations derived from Gestalt laws of "figural em- phasis." For example, Arkin and Duval (1975) found that an actor's artwork choices were attributed more to situational causes when the situation was figural or salient by virtue of a moving slide show than when it contained still photographs. McArthur and Post (1977) conducted a series of studies The research reported in this article was supported in part by National Institute of Mental Health Grant MH 26621. We would like to express appreciation to Doug Paul of Lincoln Laboratories, Massachusetts In- stitute of Technology, for all 6f his help in preparing the stimulus tapes. We would also like to thank Terry Moore and 'Andy Newman for contributing their acting talents to the creation of the stimulus material, Anne Sandoval for the use of her stereo and recording equip- ment, and Stuart Isaacs for his help with the data anal- ysis. Requests for reprints should be sent to Leslie Z. McArthur, Department of Psychology, Brandeis Uni- versity, Waltham, Massachusetts 02254. that generalized this effect to interpersonal interactions in which the salient situational stimuli were other people. When an actor conversed with someone who was salient by virtue of being brightly lit, moving, wearing a patterned shirt, or forming a physically similar unit with others present, then the actor's behavior was attributed more to sit- uational causes than when he or she con- versed with someone whose physical attri- butes were less salient. Presumably these "situational" causes are causes in the atten- tion-drawing conversational partner. Re- search by McArthur and Solomon (1978) explicitly tested this and found that an ag- gressor's antisocial behavior was attributed more to causes in her partner when the part- ner was salient by virtue of a novel physical attribute—either a leg brace or red hair— than when the novel attribute was not pres- ent. Research by Taylor and her associates (e.g., Crocker, Fiske, Sprinzen, & Winkler, 1979; Taylor & Fiske, 1975) has also doc- umented the impact of visual salience on perceptions of the extent to which one person causes the behavior of another. Actors who are salient by virtue of facing the perceiver are perceived as playing a greater causal role 236

Impact of Salient Vocal Qualities on Causal Attribution ... · vealed that salient stimuli those that draw ... the stimulus tapes. We would also like to thank Terry Moore and 'Andy

  • Upload
    lytuong

  • View
    213

  • Download
    0

Embed Size (px)

Citation preview

Journal of Personality and Social Psychology1982, Vol. 43, No. 2, 236-247

Copyright 1982 by the American Psychological Association, Inc.0022-3514/82/4302-0236$00.75

Impact of Salient Vocal Qualities on Causal Attributionfor a Speaker's Behavior

Janet Robinson and Leslie Zebrowitz McArthurBrandeis University

Considerable research has demonstrated an illusory causation effect in whichvisually salient people are perceived as more causal of events in a social inter-action than their nonsalient counterparts. The present studies extended this workto the realm of auditory salience. Two determinants of auditory salience weremanipulated—the intensity of a speaker's voice and the sex of a speaker's voice.As predicted, subjects attended more to a 75-dB (A) than to a 70-dB voice ona binaural listening test, and subjects attributed more causality to an actor ina two-person conversation when his voice was 75 dB in intensity than when itwas 70 dB. Contrary to expectation, subjects did not attend more to the voicewhose sex matched their own on the binaural listening test. Rather, all subjectslistened more to the actor with the male voice. Consistent with this tendency forthe male voice to be more salient, subjects attributed more causality to an actorwhen the voice was-male than when it had been electronically converted to afemale voice of the same intensity and intonation. Vocal salience also influencedsubjects' impressions of the actors, but it had no impact on recall of the actors'verbalizations. Theoretical and practical implications of these findings are dis-cussed.

Considerable research evidence has re-vealed that salient stimuli—those that drawperceivers' attention—tend to be perceivedas causing the events in a social interaction.This illusory causation effect (McArthur,1980) has been obtained in a number of stud-ies that employed salience manipulationsderived from Gestalt laws of "figural em-phasis." For example, Arkin and Duval(1975) found that an actor's artwork choiceswere attributed more to situational causeswhen the situation was figural or salient byvirtue of a moving slide show than when itcontained still photographs. McArthur andPost (1977) conducted a series of studies

The research reported in this article was supportedin part by National Institute of Mental Health GrantMH 26621. We would like to express appreciation toDoug Paul of Lincoln Laboratories, Massachusetts In-stitute of Technology, for all 6f his help in preparingthe stimulus tapes. We would also like to thank TerryMoore and 'Andy Newman for contributing their actingtalents to the creation of the stimulus material, AnneSandoval for the use of her stereo and recording equip-ment, and Stuart Isaacs for his help with the data anal-ysis.

Requests for reprints should be sent to Leslie Z.McArthur, Department of Psychology, Brandeis Uni-versity, Waltham, Massachusetts 02254.

that generalized this effect to interpersonalinteractions in which the salient situationalstimuli were other people. When an actorconversed with someone who was salient byvirtue of being brightly lit, moving, wearinga patterned shirt, or forming a physicallysimilar unit with others present, then theactor's behavior was attributed more to sit-uational causes than when he or she con-versed with someone whose physical attri-butes were less salient. Presumably these"situational" causes are causes in the atten-tion-drawing conversational partner. Re-search by McArthur and Solomon (1978)explicitly tested this and found that an ag-gressor's antisocial behavior was attributedmore to causes in her partner when the part-ner was salient by virtue of a novel physicalattribute—either a leg brace or red hair—than when the novel attribute was not pres-ent. Research by Taylor and her associates(e.g., Crocker, Fiske, Sprinzen, & Winkler,1979; Taylor & Fiske, 1975) has also doc-umented the impact of visual salience onperceptions of the extent to which one personcauses the behavior of another. Actors whoare salient by virtue of facing the perceiverare perceived as playing a greater causal role

236

SALIENT VOCAL QUALITIES 237

in the conversation in which they are en-gaged than actors who are seated with theirback to the perceiver. (See McArthur, 1981,and Taylor and Fiske, 1978, for more ex-tensive reviews of this literature.)

Although there is a great deal of evidenceto indicate that visually salient people tendto be perceived as causal agents, there hasbeen no research investigating the impact ofa person's vocal salience on causal attribu-tions. Such research would seem importantfor at least two reasons. First, research hasrevealed that people consider voice to be amore vital cue for personality judgment thanphotographs, videotapes of a person walking,or a personal essay (Dzida & Kiener, 1978),and vocal qualities have been shown tostrongly influence such social judgments (eg.Scherer, 1978). Because we are so dependenton voice cues in our perceptions of otherpeople, research investigating the impact ofa salient voice on causal attributions has im-portant practical significance. It also hastheoretical value. Despite all of the evidencedocumenting causal attribution to visuallysalient people, the question still remains asto why this occurs. Although the present re-search will not provide a definitive answerto that question, the demonstration of illu-sory causation effects within the auditorydomain will help to narrow the range of plau-sible explanations for the effect.

Based on a comparison of those paradigmsthat have failed to find illusory causationeffects and those that have demonstratedsuch effects, McArthur (1980) concludedthat the tendency to attribute causality tovisually salient actors requires the simulta-neous presence of the salient and the non-salient actor. There are two possible reasonswhy this might be so, and each suggests adifferent theoretical explanation for illusorycausation effects. One possibility is thatcausal attribution to visually salient personsrequires visual scanning activity that regis-ters an interchange in units that begin withthe visually salient person and end with theless salient one. A second possibility is thatcausal attribution to visually salient personsrequires direct competition for the perceiv-ers' attention by salient and nonsalient stim-uli. This explanation suggests that prefer-ential processing of the salient stimuli (e.g.,

greater recall, more thought) accounts forillusory causation effects, whereas the firstexplanation suggests that the perceptual or-ganization of events involving salient andnonsalient stimuli accounts for illusory caus-ation effects. (See McArthur, 1980, for amore extended discussion.) Manipulating/vocal salience provides a means for sortingout these two possibilities. Unlike manipu-lations of visual salience, the manipulationof vocal salience can be achieved in a par-adigm that allows the perceiver to registerevents in units that begin with the salientperson and end with the nonsalient one with-out at the same time dividing perceivers'attention between salient and nonsalientstimuli. More specifically, when we listen toa conversation between a vocally salient anda nonsalient person, we can register this in-terchange in units that begin with the salientvocalizations and end with the less salientones. We can also process just as much ofthe salient person's behavior as we can of thenonsalient person's behavior, since the vo-calizations occur sequentially. Thus, thedemonstration of illusory causation effectswithin the auditory modality would indicatethat direct competition for perceivers' atten-tion by .salient and nonsalient stimuli is nota necessary condition for the effect.

The first step in investigating the impactof a salient voice on causal attributions is todetermine what vocal qualities are salient.As noted above, research investigating theimpact of visual salience on person percep-tion has drawn on Gestalt principles of fig-ural emphasis as well as on theories of se-lective attention to identify those visual at-tributes that will be salient to perceivers;Although these theories have typically fo-cused on stimuli in the visual modality, someof the principles are equally applicable to theauditory domain. For example, the generalprinciple that "intense" stimuli are salientsuggests that just as a brightly lit person ismore salient than a dimly lit one (McArthur& Post, 1977), so should a louder speakingperson be more salient than a quieter one.Similarly, just as a person with a novel ap-pearance is salient, so should a person witha novel voice be salient.

Although the impact of stimulus intensityand novelty can be generalized across mo-

238 JANET ROBINSON AND LESLIE ZEBROWITZ McARTHUR

dalities, the impact of other stimulus qual-ities is less easily translated from the visualto the auditory domain. For example, it isnot clear what would be the auditory ana-logue to the salience of a moving visual stim-ulus or a complexly patterned visual stim-ulus, both of which have been found to beperceived as causal (MeArthur & Post,1977). One qualitative feature that doestranscend modalities is a person's "sex,"which is manifested in both visual and vocalqualities. Moreover, there is some evidencethat people who are the same sex as the per-ceiver are more salient than those of the op-posite sex. For example, it has been foundthat children recall more of the behaviors ofa same-sex model presented on videotape orin a storybook (McArthur & Eisen, 1976a;1976b). More direct evidence for the greatersalience of people whose sex matches theperceivers' has been reported by Maccoby,Wilson, and Burton (1958), who recordedperceivers' eye movements while theywatched scenes in which only the male andfemale leads of entertainment films were onthe screen. The results revealed that womenspent proportionately more time than menwatching the female lead, whereas menspent more time watching the male lead.

Based on the foregoing consideration ofwhat vocal qualities are likely to be salientto perceivers, two studies were designed toinvestigate the impact of an actor's voice onperceiver's attention and causal attributions.Study 1 varied the volume of a speaker'svoice, and Study 2 varied the sex of aspeaker's voice. Because cultural assump-tions about "loud" versus "quiet" peoplecould provide an alternative to the salienceexplanation for greater causal attribution tothe person with the louder voice, a very sub-tle manipulation of volume was employed inStudy 1. Specifically, there was only a 5-dB(A) difference in the intensity of the twovoices, one of which was set at 75 dB andone of which was set at 70 dB. This differ-ence is just barely detectable, and it does notrender one speaker loud and the other quiet.Rather, both speakers had voices of averageintensity. The manipulation of voice sex thatwas employed in Study 2 was also designedto rule out certain alternative explanations.Specifically, a computer conversion of a male

voice to a female voice was employed. Sincethis treatment introduces a slight buzzinessinto the speaker's voice, it was also per-formed on the male voice without the sexconversion feature, so that the only differ-ence between the male and female versionof a voice was its apparent sex: There wereno differences in acoustical quality, rhythm,intonation, pauses, and so forth. Althoughthis manipulation ensures that it is the sexof the speaker's voice per se that is influ-encing perceivers' judgments, one could nev-ertheless argue that this influence derivesfrom perceivers' stereotypes about "males"and "females" rather than from the, salienceof male and female voices. However, if theresults of Studies 1 and 2 parallel one an-other as well as replicate the effects obtainedin past research employing visual saliencemanipulations, this would render a salienceinterpretation of the data the most parsi-monious.

Both Study 1 and Study 2 employed abinaural listening task as a manipulationcheck designed to determine whether thetheoretically salient voice did indeed drawmore attention. The binaural listening testwas based on research that has revealed thatsubjects who listen to two short messages atthe same time are able to attend to only oneof them (e.g., Broadbent, 1958). It was as-sumed that when the two messages were spo-ken by different voices, perceivers would lis-ten to the most salient voice just as perceiverslook at the most salient people (eg. Mc-Arthur & Ginsberg, 1981). Thus, it was pre-dicted that the more intense voice or thevoice that matched the sex of the perceiverwould receive greater attention on the bin-aural listening task and would be perceivedas more causal of events in a tape-recordedconversation.

In addition to measuring attention andcausal attribution to people with salientvoices, recall of their verbalizations was alsoassessed. As noted above, one explanationthat has been offered to account for illusorycausation effects is the suggestion that theyare mediated by preferential recall about thebehavior of salient people (Taylor & Fiske,1978). Although the existing research lit-erature has provided little evidence that thetotal quantity of information recalled about

SALIENT VOCAL QUALITIES 239

salient versus nonsalient people serves as amediator, it has been suggested that this lackof evidence may be due to the fact that re-searchers have typically assessed total recallof verbal information, which may not beparticularly sensitive to manipulations ofvisual salience. However, such recall shouldbe sensitive to manipulations of vocal sa-lience, and the present study, therefore,tested the hypothesis that there would bemore total recall of a vocally salient actor'sverbalizations and that this recall would cor-relate with causal attributions.

Method

SubjectsVolunteer subjects were recruited from the introduc-

tory psychology class at Brandeis University and re-ceived 1 hour of experimental credit for their partici-pation in an experiment on person perception. Sixteenmales and 16 females were recruited for the volume-salience study, and 32 males and 32 females were re-cruited for the sex-salience study. Subjects in each ofthe two studies were randomly assigned to one of twosalience conditions (Actor A salient or Actor B salient)and to one of two rating orders (salient actor rated firstor salient actor rated last).

MaterialsGetting-acquainted conversation. A "getting-ac-

quainted" conversation between two male actors wasrecorded on tape. Each actor was recorded on a separateband of the same tape. The conversation of approxi-mately 5 min. duration was read from a script to ensurethat each actor contributed the same amount of infor-mation and initiated the same number of topics of dis-cussion. The names that were given to the actors, Patand Gerry, and the topics they discussed did not providethe listeners with any information that would allow themto determine the actor's sex.

Listening tape. The binaural listening test proce-dures described by Broadbent (1958) were used as amodel for the design of a binaural listening test. A taperecording was made by the two actors who were re-corded in the getting-acquainted conversation. The twoactors simultaneously listed seven sets of three digits.In a given set, the three digits spoken by Actor A dif-fered from the three digits spoken by Actor B. For ex-ample, Actor A was heard to recite "three-five-two,"while Actor B recited at the same time, "nine-six-eight."Each actor was recorded on a separate band of the sametape.

Salience manipulations. Manipulation of the volumeof voice or the sex of voice on the getting-acquaintedconversation tape and on the binaural listening tape wascarried out in the following manner. In the volume sa-lience study, the volume of one speaker was set at 70dB, and the other was set at 75 dB. As noted above, this

volume difference is detectable, but it is not greater thanwhat might occur in a normal, everyday conversation. .Moreover, the difference in intensity of the two voicesis sufficiently small so that when both are played si-multaneously, a perceiver can voluntarily attend to ei-ther one—that is, the 75-dB voice is not so loud thatit makes it impossible to listen to the 70-dB voice. InCondition 1, the speaker on the east wall was turned upso that Actor A was louder. In Condition 2, the speakeron the west wall was turned up so that Actor B waslouder. In the sex-salience study, a modification of thetape was made to produce a sex change in the voice ofeach actor (Paul, Note 1). First, the voices of both actorswere put through a low bit-rate vocoder. A very slightbuzziness was heard from both speakers as a result ofthe vocoderization. The parameters of each actor's voicewere then adjusted to produce a highly realistic imita-tion of a female voice. This technique allowed us to varythe sex of each actor without changing the expressionor tone or volume of each one's voice during the con-versation. In Condition 1, Actor A was male and ActorB was female. In Condition 2, Actor A was female andActor B was male.

ProcedureIn both studies the getting-acquainted-conversation

tape and the binaural-listening tape were heard by sub-jects through two stereo speakers. One speaker emittedthe voice of Actor A, and the other emitted the voiceof Actor B. The speakers were placed against the eastand west walls of a square room and faced toward thecenter of the room. A maximum of four subjects par-ticipated in each run of the experiment. Two subjectswere seated one behind the other, with their backs facingthe north wall of the room while the other two subjectssat with their backs toward the south wall of the room(see Figure 1). This arrangement allowed subjects to bepositioned at an equal distance from each speaker.1

Seating arrangements were planned so that half of themale subjects and half of the female subjects heard thevoice of Actor A from the speaker on their left, and theother half of the subjects heard the voice of Actor Afrom the speaker on their right. The sex of the subjectseated in the centermost or outermost chairs was alsocounterbalanced.

On entering the lab, subjects were seated and giventhe following instructions:

This is a study in an area of Social Psychology calledinterpersonal dynamics. More specifically, what we'reinterested in is something called the getting-ac-quainted process, and what happens when two peoplemeet for the first time. I'm going to play a tape re-cording so that you can listen to two people who aregetting acquainted for the first time, although theywere asked in advance to think about what they weregoing to say and, of course, they knew they were beingtaped.

1 The two centermost chairs were 2 feet (.61m) closerto the speakers than the outermost seats; however, thisdistance was not great enough to bring about a signif-icant difference in sound intensity.

240 JANET ROBINSON AND LESLIE ZEBROWITZ McARTHUR

NORTH WALL

ransi

ISUBJE

SOUTH WALL

Figure 1. Seating arrangement of subjects: bird's eyeview.

After listening to the getting-acquainted-conversationtape, subjects were asked to record their impression ofeach stimulus person on a written questionnaire. Whenthe questionnaires were completed, subjects were giventhe following instructions for the binaural listening test:

The next part of the experiment is basically a hearingtest. You may have noticed that one speaker is di-rected at each ear. We are interested in rinding outwhether you process the information going into yourright ear differently than you process the informationgoing into your left. I'm going to play a recording oftwo people speaking at the same time. They're eachgoing to recite three digits. For example, the speakeron your right might say "3-5-9" while the speaker onyour left is saying "2-6-8." We want you to listen toboth speakers, and write down all six numbers, or asmany of the numbers as you can. There will be eightsets of digits in all. Set one will be a practice set, setstwo through eight will be counted.

Subjects listened to the practice set of three digit num-bers and recorded the numbers that they heard. Next,subjects were asked whether they understood the in-structions and if they had any questions. Subjects thenlistened to and recorded the next seven sets of three digitnumbers. Subjects were thanked for their participationand thoroughly debriefed as to the purpose and goalsof the experiment.

Dependent Measures

Causal attributions. Subjects rated how much eachactor set the tone of the conversation, determined the

kind of information exchanged, and caused the otherperson to behave as he or she did. Subjects also madecausal attributions for specific behaviors. They ratedhow friendly, competent, nervous, and assertive eachactor's behavior was and the extent to which each ofthese behaviors was caused by the actor's personality,by his or her conversational partner, and by the situa-tion. Personality causes were defined for subjects as"the actor's traits, abilities, personal style, etc." Partnercauses were described as "the conversational partner'sbehavior and personality, i.e., traits, abilities, personalstyle, etc." Situational causes were defined as "being inan experiment, being tape-recorded, etc."2 All ratingswere made on 9-point scales.

Impressions. In addition to rating the extent towhich each actor manifested the four behaviors forwhich causal attributions were made, subjects wereasked to rate each actor on six trait scales. Three of thetraits were stereotypically feminine, according to Brov-erman, Vogel, Broverman, Clarkson, and Rosenkrantz(1972), and included influenceability, sensitivity to thefeelings of others, and emotionality. The remainingtraits were stereotypically masculine and included com-petence, confidence, and logic. All ratings were madeon 9-point scales.

Recall, Subjects were asked nine specific questionsabout what was said by each actor. The questions werethe same for Actor A and Actor B. An example of thequestions asked is as follows: "Where does he/she attendschool?" The number of correct answers was determinedto give a recall score for each actor.

Listening test. Subjects recorded digits heard throughboth speakers. The number of correctly recalled digitsfrom Actor A and the number of correctly recalled digitsfrom Actor B were tallied for each of the seven sets ofdigits.

Results

Volume Study

The dependent measures were subjectedto an analysis of variance utilizing saliencecondition (2) and sex of subject (2) as be-

2 Although attributions to personality disposition andto situational causes were measured, as has been tra-ditionally done in research on salience effects, thesemeasures will not be reported in the Results section.Although it might be expected that attributions of anactor's behavior to his own personality would increaseas his voice becomes more salient, such salience effectshave not been obtained in previous research, nor werethey obtained in the present studies (see McArthur,1981, for a discussion of this issue). Also, althoughMcArthur and Post (1977) found that making one actorphysically salient in a two-person conversation increasedsituational attributions for the behavior of the nonsalientactor, McArthur and Solomon (1978) found that whena partner attribution measure was included, it usurpedthis effect, and salience influenced only the more specificand appropriate partner attribution measure.

SALIENT VOCAL QUALITIES 241

tween-subject factors. Since each subjectrated both speakers, actor (2) was employedas a repeated measure.

Listening. Salience of the voices was as-sessed by means of the binaural listeningtests. The number of digits that were cor-rectly recorded for each actor was deter-mined for each of the seven sets of the threedigits. For each actor, a score between 0(none correct) and 3 (all correct) could beobtained for each set. The number of correctresponses were summed to produce a totallistening score for each actor that couldrange from 0 to 21. The analysis of varianceperformed on the total listening scores yieldeda significant Salience Condition X Actor in-teraction effect that revealed that as pre-dicted, subjects listened to the actors morewhen their voices were the louder of the twothan when they were the softer, F(l, 28) =11.02,p = .002 (see Table 1).

Causal attributions. Following the pro-cedure employed in past research (e.g., Tay-lor & Fiske, 1975), an index of each actor'sgeneral causal role was created by summingsubjects' ratings of the extent to which theactor set the tone of the conversation, de-termined the information exchanged, andcaused the behavior of the other actor. Sinceeach of the individual ratings was made on9-point scales, scores on the causal role indexcould range from 3 to 27. The analysis ofvariance performed on this index yielded asignificant Salience Condition X Actor in-teraction effect, F(\, 28) = 5.27, p = .03,which revealed that as predicted, both actorswere seen as more causal when their voiceswere 75 dB than when they were 70 dB (seeTable 1).

The extent to which each actor's specificbehaviors (e.g., competence, nervousness,etc.) were seen as caused by the actor's part-ner was summed to form a partner attri-bution index. This composite measure wasemployed, as had been done in previous re-search (e.g., McArthur & Solomon, 1978),because the hypothesis concerned the rela-tive strength of partner attributions andmade no distinction among the behavioraldimensions. Since partner attributions on 9-point scales were summed across four be-haviors, scores on the partner attribution in-dex could range from 4 to 36. The analysis

Table 1Listening. Causal Attributions, Recall, andImpressions as a Function of the Intensityof the Actors' Voices

Dependentmeasure

Listening index

Causal roleindex

Partner attributionindex

Recall

Friendliness

Logic

Emotionality

Sensitivity

Saliencecondition

(dB)

A + 5B + 5A + 5B + 5A + 5B + 5A + 5B + 5A + 5B + 5A + 5B + 5A + 5B + 5A + 5B + 5

Actor

A

18.9414.4420.7516.8122.5622.44

5.885.256.696.446.816.195.505.885.695.75

B

12.8819.9411.8115.0622.8121.886.065.445.756.945.446.385.444.815:815.94

Note. n= 16 in each cell.

of variance performed on this index did notyield a significant Salience Condition X Ac-tor interaction effect (F < 1).

The parallel effects of vocal salience onlistening and causal attributions raises thequestion of the relationship between the two.The degree to which subjects listened moreto the 75-dB than to the 70-dB actor on thebinaural listening task was not significantlyrelated to their causal attributions for eitheractor's behavior during the getting-ac-quainted conversation. The average inter-correlation between relative listening to the75-dB actor and perceptions of the actors'causal role was -.23 for the 75-dB actorsand .14 for the 70-dB actors. The averageintercorrelation between relative listening tothe 75-dB actor and perceptions of the extentto which the actors' behavior was caused bytheir partner was .00 for the 75-dB actorsand .18 for the 70-dB actors.

Impressions. With hindsight it becameapparent that several of the impression mea-sures were conceptually redundant with thecausal attribution indices, and correlationsamong the two sets of measures revealed thisto be the case. Only ratings of the actors'friendliness, emotionality, sensitivity, andlogic were independent of causal attributions

242 JANET ROBINSON AND LESLIE ZEBROWITZ McARTHUR

20-

8too

zut-(/)

15

<s

10FEMALE MALE

SEX OF VOICE

o FEMALE SUBJECTS - ACTOR A* MALE SUBJECTS-ACTOR A* FEMALE SUBJECTS-ACTOR B* MALE SUBJECTS-ACTOR B

Figure 2. Mean listening scores in sex study as a func-tion of the sex of an actor's voice.

for the actors' behavior. Of these four mea-sures, ratings of friendliness and logic weresignificantly influenced by the volume ma-nipulation: The actors' behavior during theconversation was rated as more friendly, andthe actors were perceived as generally morelogical when their voices were 75 dB thanwhen they were 70 dB, Fs(l, 28) = 4.80 and3.7'1, ps = .04 and .06, respectively. The vol-ume of the actors' voices had no significanteffect on ratings of sensitivity (F < 1) oremotionality, F(l, 28) = 1.98, p> .153 (seeTable 1).

Recall. Recall was measured by tabu-lating the number of correct responses tonine specific questions about each actor'sstatements during the getting-acquaintedconversation. The Salience Condition X Ac-tor effect did not approach significance (F <1), and recall of the actors' verbalizationswas not significantly correlated with causalattribution to them in either intensity con-dition (see Table 1).

Sex Study

As in the volume study, the dependentmeasures were subjected to an analysis ofvariance utilizing salience condition (2) andsex of subject (2) as between-subjects factorsand actor (2) as a repeated measure.

Listening. The analysis of variance per-formed on the total listening scores yieldeda significant Salience Condition X Actor in-teraction effect, which indicated that all sub-jects listened to the actors more when theirvoices were male than when they were fe-male, F(l, 60) = 37.79, p < .001. The pre-dicted triple-order interaction between sub-ject sex, salience condition, and actor wasnot significant, F(\,60)= l.55,p = .22, andplanned comparisons to test the impact ofvoice sex on the listening of male and femalesubjects considered separately revealed thatthe tendency to listen more to a male voicewas highly significant for subjects of bothsexes, Fs(l, 28) = 27.31 and 12.02, bothps < .002, for male and female subjects, re-spectively. Although both male and femalesubjects listened significantly more to themale than to the female voices, the patternof means revealed a tendency for female sub-jects to listen more to the female voices thanmale subjects did (see Figure 2). Pooling themeans across actors, female subjects listenedmore to female voices than did male subjects,/(62) = 2.28, p < .03, whereas there was nosignificant sex difference in listening to malevoices (t < 1).

Causal attributions. The analysis of vari-ance performed on the causal role indexyielded a significant Salience Condition XActor interaction effect, which indicatedthat both actors were seen as more causalwhen their voices were male than when theywere female, F(l, 60) = 19.30, /x.OOl.Although the tendency to attribute causalityto the actor with a male voice was less pro-nounced for female subjects than for malesubjects, the predicted triple-order SubjectSex X Salience Condition X Actor interac-tion was not significant (F < I), and plannedcomparisons performed to test the impact ofan actor's sex on perceptions of an actor'scausal role for male and female subjects con-

3 In the volume study, four of the five measures thatwere highly correlated with the causal role index (com-petent, nervous, assertive, and confident) yielded sig-nificant effects for volume, paralleling those obtainedon the causal role index. Only ratings of influenceabilityfailed to yield a significant volume effect. In the sexstudy, all four measures that were highly correlated withthe causal role index (competent, assertive, influence-able, and confident) yielded significant effects for sexparalleling those obtained on the causal role index.

SALIENT VOCAL QUALITIES 243

Table 2Causal Attributions, Recall, and Impressions as a Function of the Sex of the Actors' Voices

Subject group

Dependentmeasure

Causal roleindex

Partner attributionindex

Recall

Nervousness

Friendliness

Logic

Emotionality

Sensitivity

Saliencecondition

(male)

ABABABABABABABAB

All

Actor A

20.5615.7522.7224.196.255.723.315.317.066.666.815.725.285.726.096.22

subjects

Actor B

13.0616.7822.7221.386.726.316.004.386.626.815.256.256.255.066.195.59

Males

Actor A Actor B

20.6215.4421.0624.626.195.442.945.007.196.886.816.124.815.885.816.12

12.0015.9422.2522.506.386.386.314.626.756.564.566.196.384.816.315.31

Females

Actor A

20.5016.0624.3823.756.316.003.695.626.946.446.815.315.755.566.386.31

Actor B

14.1217.6223.1920.257.066.255.694.126.507.065.946.316.125.316.065.88

Note, n = 32 for means based on all subjects, n = 16 for means based on subjects of one sex.

sidered Separately revealed that subjects ofboth sexes rated the male voices as signifi-cantly more causal, F(l, 28) = 11.04, p =.002 for male subjects, and F(l, 28) = 8.36,p - .007 for female subjects (see Table 2).

The analysis of variance performed on thepartner attribution index yielded results par-allel to those for the causal role index. Asignificant Salience Condition X Actor in-teraction effect, F(l, 60) = 4.63, p = .03,revealed that the behavior of each actor wasattributed more to the partner when thepartner's voice was a male voice than whenit was a female voice. Although the predictedSubject Sex X Salience Condition X Actortriple-order interaction was not significant(F < 1), the pattern of means once againindicated that the tendency to attribute cau-sality to the partner with the male voice wasstronger for male subjects than for females.Furthermore, planned comparisons revealedthat simple second-order interaction be-tween salience condition and actor was mar-ginally significant for male subjects consid-ered alone, F(l, 28) = 3.21, p = .08, but notfor female subjects considered alone, F(l,28) = 1.58, p = .22. Thus, the overall ten-dency to attribute causality to the partnerwith the male voice was largely due to malesubjects (see Table 2).

As was true in the volume study, the de-gree to which subjects listened more to thesalient than the nonsalient actor on the bin-aural listening task was not significantly re-lated to their causal attributions for eitheractor's behavior during the getting-ac-quainted conversation. The average inter-correlation between relative listening to themale actor and perceptions of the the actors'causal role was -.12 for the male actors and.01 for the female actors. The average in-tercorrelation between relative listening tothe male actor and perceptions of the extentto which the actors' behavior was caused bytheir partner was .00 for the male actors and.06 for the female actors.

Impressions. As in the volume study, rat-ings of the actors' competence, assertiveness,influenceability, and confidence were signif-icantly correlated with the causal attributionindices.3 Unlike the volume study, ratings ofthe actors' nervousness were not signifi-cantly correlated with causal attributions.Thus, five rating measures were independentof attributions: nervousness, friendliness,emotionality, sensitivity, and logic. Three ofthese yielded significant effects: The actors'behavior during the conversation was ratedas more nervous, and the actors were per-ceived as generally more emotional and less

244 JANET ROBINSON AND LESLIE ZEBROWITZ McARTHUR

logical when their voices were female thanwhen they were male, Fs(l, 60) = 27.82,8.75, and 18.81; all/>s < .01. The sex of theactors' voices had no significant effect onratings of friendly behavior or general sen-sitivity, F < 1 and F = 1.36, both ps > .20(see Table 2).

Recall. The Subject Sex X Salience Con-dition X Actor triple-order interaction effecton recall was not significant (F < 1), andthere was no significant Salience Condi-tion X Actor second-order interaction asthere was for the attribution and impressionsmeasures (F < 1; see Table 2). In addition,recall of the actors' verbalizations was notsignificantly correlated with causal attribu-tions to them either -when their voice wasmale or when it was female.

Discussion

The listening index in the volume studyconfirmed the hypothesis that the voice ofan actor will be attended more when it ishigher in intensity. In contrast, the listeningindex in the sex study did not confirm pre-dictions. Although it had been expected thatan actor whose sex matched that of the lis-tener would be attended most, male voiceswere attended more than female voices bysubjects of both sexes. However, female sub-jects did listen more to female voices thandid male subjects. This finding might be ex-plained by considering the possibility thatthere are two forces controlling the salienceof voices of a particular sex. One force maybe a general tendency to attend more tomales than females. This possibility couldresult from the fact that males play a dom-inant role in our society. It might also be theresult of differences in physiological auditorysensitivity to the acoustical characteristicsof male versus female voices. It is conceiv-able that the frequencies of the laryngealfundamental and vocal tract resonances thatcharacterize a male voice'(Coleman, 1976)are picked up more readily than the fre-quencies that characterize a female voice.4

A second force may be to attend to a personof the same sex because, as noted earlier,research on visual attention has revealed thatsubjects do attend more to same-sex actors.In males, these two forces would act in thesame direction and combine to produce a

strong tendency for male perceivers to listento males. In females, these two forces wouldact in opposite directions. If the tendency toattend to males were stronger than the ten-dency to attend to females, the two forcescould interact to yield the pattern of resultsobtained for females in the present study: astronger tendency to attend to a male thanto a female voice coupled with a strongertendency to attend to the female voice thanis shown by male subjects.

Consistent with past research that hasmanipulated visual salience, auditory sa-lience yielded an illusory causation effect inwhich actors with salient voices were per-ceived as more causal than those whosevoices drew less attention. Actors were per-ceived as more causal when they had themore salient 75-dB voice than the less salient70-dB one. Similarly, actors were perceivedas more causal when their voice was a sa-lient, male voice than when it was convertedto a female voice of the same intonation andintensity. The latter effect was stronger formale subjects than it was for females. Onepossible explanation for this trend is a gen-eral tendency for male subjects to attributemore causality to a salient actor of any kind.However, this explanation is not borne outby the data from the volume study in whichmale subjects did not attribute any morecausality to actors with the higher intensityvoice than did female subjects.5 Another ex-

4 Another conceivable explanation for the greater sa-lience of the male voice is that some feature of the com-puter sex conversion rendered the female voice less sa-lient. As noted earlier, the male voice as well as itsfemale conversion were processed through the vocoderso that there would be no difference in the clarity of thetwo. Furthermore, the sex-stereotyped differences inimpressions of the male and female voice provide con-vincing evidence that the sex conversion produced anauthentic female voice. Nevertheless, the conclusionthat male voices are more salient would be strengthenedby future research that demonstrates such salience whenit is a female voice that has been converted to malerather than vice versa.

5 On the contrary, simple effects analyses performedto test the impact of voice volume on perceptions of anactor's causal role for male and female subjects consid-ered separately revealed that the Salience Condition XActor interaction was significant for female subjects,F(l, 12) = 6.31, p = .03, but not for males (F < 1).Thus, the overall tendency to perceive the louder actoras exercising a greater causal role was largely due tofemale subjects.

SALIENT VOCAL QUALITIES 245

planation for female subjects' weaker ten-dency to attribute causality to actors withmale voices is consistent with the originalpredictions as well as with the data from thebinaural listening task. More specifically,the tendency for a female voice to be moresalient for female than for male subjects mayhave yielded a weaker tendency for femalesto attribute causality to the male actor.

Although the voices that proved more sa-lient on the binaural listening task were alsoperceived as more causal during the getting-acquainted conversation, the degree to whichsubjects listened more to the salient voicewas not correlated with their causal attri-butions for either actor's behavior. This find-ing is consistent with data reported byMcArthur and Ginsberg (1981), who foundthat the extent to which subjects looked atvisually salient actors was not positively cor-related with their causal attributions, eventhough both looking time and causal attri-butions were higher for salient actors. De-spite the similarity of the results, an impor-tant difference between the listening mea-sure in the present research and the lookingmeasure in the McArthur and Ginsberg re-search should be noted. The listening mea-sure was intended as a manipulation check—a means to ascertain which voice was indeedmore salient. There is no reason to expectthat the degree to which subjects listenedmore to the salient actor when the voiceswere presented simultaneously on the bin-aural task will be related to the degree towhich they attributed causality to that actorwhen the voices were presented sequentiallyduring the earlier getting-acquainted con-versation. Indeed, the sequential presenta-tion during the getting-acquainted conver-sation made it possible for subjects to listento 100% of what each actor said, regardlessof the salience of his or her voice. Althoughthe data from the binaural listening task arethus inappropriate for testing mediationalhypotheses, the causal attribution datathemselves argue against the hypothesis thatpreferential processing of salient people'sbehavior mediates greater causal attribu-tions to them: The illusory causation effectsdocumented in the present research wereobtained under conditions that did not dividesubjects' attention between the salient andnonsalient actors and that, therefore, al-

lowed subjects to process as much of thenonsalient as the salient person's behavior.

In addition to demonstrating that illusorycausation effects do not require direct com-petition for the perceiver's attention amongsalient and nonsalient stimuli, the presentresearch also revealed that these effects donot depend on greater recall of a salient per-son's behavior. Moreover, since the recall ofa vocally salient actor's verbalizations wasassessed, the lack of significant effects can-not be attributed to measurement of recallin the wrong modality as it can be in muchof the research manipulating visual salience.Although there was no greater total recallof a vocally salient actor's verbalizations,some recent research on visual salience sug-gests that there may be greater recall of par-ticular verbal behaviors by the vocally sa-lient actor. More specifically, Fiske, Kenny,and Taylor (1982) found preferential recallfor salient visual stimuli, which were judgedby the perceivers to be representative ofcausal influence. Although the recall mea-sure employed in the present research wasnot designed to assess verbal recall analo-gous to this more specific visual recall, itshould be noted that such recall would notprovide a satisfactory explanation for illu-sory causation effects. Specifically, it failsto explain how perceivers come to recallmore information representative of a salientactor's causal influence rather than othertypes of information about the salient actor.In essence, this explanation seems to requirethat we selectively perceive the causal effi-cacy of a salient actor, which is just whatthe representative recall mediator is at-tempting to explain.

Although the present findings rule out onespecific explanation for illusory causationeffects—namely, that they derive fromgreater total recall about the behavior ofsalient people within the salient modality—and although they further reveal that illu-sory causation effects do not require directcompetition for the perceiver's attentionamong salient and nonsalient stimuli, a de-finitive explanation for illusory causationeffects remains to be found. Although thepresent research does not provide this expla-nation, it is worth noting that the percep-tual organization hypothesis suggested byMcArthur (1980) is compatible with the

246 JANET ROBINSON AND LESLIE ZEBROWITZ McARTHUR

present findings. This explanation holds thatthe power of certain stimuli to draw atten-tion may cause perceivers to register inter-actions between salient and nonsalient peo-ple in units that capture the impact of thesalient person on the nonsalient person ratherthan vice versa. Thus, a conversation be-tween a soft-spoken person and a louder per-son (soft-LOUD-soft-LOUD-soft) may be reg-istered in units reflecting the causal influ-ence of the louder actor on the quieter one(LOUD —> soft) rather than in units reflectingthe causal influence of the quieter actor(soft —» LOUD). Some evidence consistentwith this hypothesis is provided in the psy-cholinguistic literature in which it has beendemonstrated that as one of two syllables ina two-syllable word increases in intensity, itis more likely to be perceived as the stressedsyllable (e.g., Fry, 1955). Thus, for example,the word object is more apt to be perceivedas a noun the louder the ob syllable is relativeto the ject syllable. An even closer analogueto the perceptual organization explanationfor illusory causation effects would be pro-vided by an experiment that presents sub-jects with a continuous string of syllablesthat can be segmented in one of two ways(e.g. gun shot gun shot) and determineswhether the intensity of each word affectsthe perceptual organization of this reversibleseries (e.g., GUN shot, SHOT gun).

Although the primary focus of the presentresearch was to assess the impact of vocalqualities on causal attributions for a speaker'sbehavior rather than their impact on impres-sions of the speaker, several of the impres-sion effects were independent of the attri-bution effects and are quite interesting. Ac-tors were rated as significantly less nervousand less emotional when their voice was malethan when it was converted to a female voiceof the same intonation and intensity; theywere rated as less friendly when their voicewas 70 dB than when it was 5 dB higher inintensity. In addition, actors with a malevoice and those with a 75-dB voice wererated as more logical than those with a fe-male or 70-dB voice. Most of these effectsare consistent with past research that hasdemonstrated more extreme ratings of sa-lient than nonsalient actors on a variety oftrait dimensions and for a variety of salience

manipulations (see McArthur, 1981, for areview of this literature). The higher ratingsof the 75-dB actor's friendliness and logic,the lower ratings of the male actors' ner-vousness, and the higher ratings of their logicall represent ratings that deviate more fromthe midpoint of the scale than do ratings ofthe less salient 70 dB or female actors. How-ever, the lower ratings of the male actors'emotionality represent a less rather than amore extreme rating of salient actors.

The impact of vocal qualities on bothcausal attributions and impressions has im-portant practical significance in addition tothe aforementioned theoretical implications.Not only do these findings corroborate pastresearch, which cites the general importanceof vocal qualities in person perception, butthey also provide more specific informationregarding what perceptions are elicited byvoices that vary in intensity and sex. Thefindings suggest, for example, that malesmay be more effective in a debate than theirfemale adversaries, particularly when theyare judged by a male audience. They alsosuggest that a very slight increase in vocalintensity may significantly augment percep-tions of a person's friendliness and logic aswell as the extent to which he or she is per-ceived to be influencing the behavior of an-other. Whether increased vocal intensity canoffset the disadvantages of a female voice isa question worth pursuing in future research.

Conclusion

This research has demonstrated that a 75-dB voice is more salient than a 70-dB oneand that a male voice is more salient thana female one. Regardless of whether vocalsalience was manipulated by volume or bysex, actors were perceived as more causalwhen their voices were salient than whenthey were not. Salience also influencedimpressions of the actors' personality on di-mensions unrelated to their causal role. Thepresent findings have implications for theimpact of a person's vocal qualities on var-ious social and occupational endeavors aswell as for our attempts to develop a satis-factory theoretical explanation for illusorycausation effects.

SALIENT VOCAL QUALITIES 247

Reference Note

1, Paul, D. B. A robust vocoder with pitch-adaptivespectral envelope estimation and an integrated max-imum-likelihood pitch estimation. Unpublishedmanuscript, 1981. (Available from Lincoln Labo-ratories, Lexington, Massachusetts).

References

Arkin, R. M., & Duval, S. Focus of attention and causalattributions of actors and observers. Journal of Ex-perimental Social Psychology, 1975, //, 427-438.

Broadbent, D. E. Perception and communication. NewYork: Pergamon Press, 1958.

Broverman, I., Vogel, S. R., Broverman, S. M., Clark-son, F. E., & Rosenkrantz, P. S. Sex-role stereotypes:A current appraisal. Journal of Social Issues, 1972,28, 59-79.

Coleman, R. O. A comparison of the contribution oftwo voice quality characteristics to the perception ofmaleness and femaleness in the voice. Journal ofSpeech and Hearing Research, 1976, 19, 168-180.

Dzida, W., & Kiener, F. Strategies in using nonverbalinformation in judging personality characteristics.Zeitschrift fur experimental^ und Angewandte Psy-chologie, 1978, 25, 552-563.

Fiske, S. T., Kenny, D, A., & Taylor, S. E. Structuralmodels for the mediation of salience effects on attri-bution. Journal of Experimental Social Psychology,1982, 18, 105-127.

Fry, D. B. Duration and intensity as physical correlatesof linguistic stress. Journal of the Acoustical Societyof America, 1955, 27, 765-768.

Maccoby, E. E., Wilson, W. C., & Burton, R. V. Dif-ferential movie-viewing behavior of male and femaleviewers. Journal of Personality, 1958, 26, 259-267.

McArthur, L. Z. Illusory causation and illusory corre-lation: Two epistemological accounts. Personality andSocial Psychology Bulletin, 1980, 6, 507-519.

McArthur, L. Z. What grabs you: The role of attention

in impression formation and causal attribution. InE. T. Higgins, C. P. Herman, & M. P. Zanna (Eds.),Social cognition: The Ontario Symposium (Vol. 1).Hillsdale, N.J.: Erlbaum, 1981.

McArthur, L. Z., & Eisen, S. V. Achievements of maleand female storybook characters as determinants ofachievement behavior of boys and girls. Journal ofPersonality and Social Psychology, 1976, 33, 467-473. (a)

McArthur, L. Z., & Eisen, S. V. Television and sex-rolestereotyping. Journal of Applied Social Psychology,1976, 6, 329-351. (b)

McArthur, L. Z., & Ginsberg, E. Causal attribution tosalient stimuli: An investigation of visual fixationmediators. Personality and Social Psychology Bul-letin, 1981, 7, 547-553.

McArthur, L. Z., & Post, D. L. Figural emphasis andperson perception. Journal of Experimental SocialPsychology, 1977, 13, 520-535.

McArthur, L. Z., & Solomon, L. K. Perceptions of anaggressive encounter as a function of the victim's sa-lience and the perceiver's arousal. Journal of Per-sonality and Social Psychology, 1978, 36, 1278-1290.

Scherer, K. R. Personality inference from voice quality:The loud voice of extroversion. European Journal ofSocial Psychology, 1978, 8, 467-487.

Taylor, S. E., Crocker, J., Fiske, S. T., Sprinzen, M.,& Winkler, J. D. The generalizability of salience ef-fects. Journal of Personality and Social Psychology,1979, 37, 357-368.

Taylor, S. E., & Fiske, S. T. Point of view and percep-tions of causality. Journal of Personality and SocialPsychology, 1975, 32, 439-445.

Taylor, S. E., & Fiske, S. T. Salience, attention, andattribution: Top of the head phenomena. In L, Ber-kowitz (Ed.), Advances in experimental social psy-chology, (Vol. 11). New York: Academic Press, 1978.

Received June 10, 1981Revision received November 10, 1981 •