Physical Attractiveness in Social Interactionjbnezl.people.wm.edu/Reprints/1980-JPSP-Physical-Attr-RIR.pdfphysical attractiveness correlated more highly with the attraction responses

Journal of Personality and Social Psychology1980, Vol. 38, No. 4, 604-617

Physical Attractiveness in Social Interaction

Harry T. Reis John NezlekUniversity of Rochester College of William and Mary

Ladd WheelerUniversity of Rochester

What impact does physical attractiveness have on social interaction? Althoughmany studies demonstrate that beautiful people are more highly regarded, thereare no reports about the effects of attractiveness on the course of people's actualeveryday social lives. Using a journal-style record, we constructed numerousindices characterizing subjects' socializing and then related these variables totheir independently rated physical attractiveness. The following major resultswere obtained: (a) Physical attractiveness was strongly related to the quantityof social interaction for males, positively with the opposite sex and negativelywith the same sex; no significant pattern emerged for females, (b) For bothsexes, particularly with opposite-sex interactions, satisfaction showed an increas-ing tendency over time to be positively correlated with attractiveness, (c) Fe-males with more variable attractiveness ratings were more likely to be satisfiedwith their socializing, (d) Physically attractive males tended to have moremutually initiated, and fewer self- or other-initiated, interactions with the op-posite sex. (e) Attractive males spent more of their interactions conversing andless in activities; attractive females also reported a lesser proportion of taskinteractions and more prevalent date/parties. Mediating mechanisms for theseeffects, notably including people's stereotypic beliefs about physical attractive-ness, are also discussed.

This article reports research on the relation- ticipation? (b) Are the effects the same forship between physical attractiveness and the females and males? (c) Do the effects changeeveryday social interaction of first-year college over time?students over an 8-month period. The data col- Adams (1977), outlining a "developmentallection technique used was that developed by social psychology of beauty," suggested fourWheeler and Nezlek (1977), which requires assumptions about the relationship betweensubjects to complete a short fixed-format inner behavioral processes and outer appear-record for every interaction of 10 minutes or ance. The first is that people have differentlonger that occurs during a specified interval, expectations about attractive and unattractiveThe major questions were: (a) Do normal others and that these expectations are con-levels of physical attractiveness affect quanti- sistent across numerous social situations. Thistative and qualitative aspects of social par- is the well-known physical attractiveness

stereotype that "what is beautiful is good"(Dion, Berscheid, & Walster, 1972). The sec-

We gratefully acknowledge the assistance of Randi ond assumption is that physically attractiveWeitzner, Audrey Shapiro, Alan Rubenstein, Richard i- rprp:vp „,„..- favnrahlp snrial pxrhanwsDeFrank, and John Huppertz in the conduct of this people receive more favorable social exchanges,research. The interaction records were collected in a The third assumption is that these moredissertation completed by the second author in par- favorable social exchanges create differentialtial fulfillment of the requirement for a PhD at the social images, self-expectations, and inter-University of Rochester. personal styles. The final assumption (Adams,

Requests for reprints should be sent to Harry T. ,„„„, . ,, . ,, ... .. , .„ ,Reis, Department of Psychology, University of 1977) ls that attractive people will be moreRochester, Rochester, New York 14627. likely to manifest confident interpersonal

Copyright 1980 by the American Psychological Association, Inc. 0022-3514/80/3804-0604$00.75

604

ATTRACTIVENESS IN INTERACTION 605

behavior patterns than [will] less attractiveindividuals" (p. 218).

Although evidence is provided for each ofthese assumptions, an implication derivedfrom all of them, namely that attractive indi-viduals should be more socially successfulacross a wide variety of social situations, hasnot been tested. Our first prediction, then,was that physically attractive people shouldhave more social encounters with the oppositesex and perhaps with the same sex and shouldfind these encounters more rewarding.

We based this prediction on two assump-tions. The first was that our culture teachesus that physical beauty is both important anddesirable, particularly in opposite-sex interac-tion. If attractive people are in greater de-mand as dating partners (e.g., Brislin & Lewis,1968; Tesser & Brodie, 1971; Walster, Aron-son, Abrahams, & Rottmann, 1966), then theyought to participate in more, and more pleas-ing, social events. The second assumption wasthat the more attractive are more sociallyskilled and confident or that such behavior iselicited from them by others. There is someevidence for both of these possibilities. Gold-man and Lewis (1977) used a blind, anony-mous telephone call to ascertain that moreattractive individuals of both sexes were infact rated as more socially skillful and likableby their conversants, who could not haveknown their appearance. Snyder, Tanke, andBerscheid (1977) reversed the procedure andtold male telephone callers that a female tar-get person was either unattractive or attrac-tive, when in fact all the targets were ofequal attractiveness. Independent judges' rat-ings of just the target's half of the conversa-tion revealed her to be more friendly, likable,and sociable when she was thought to beattractive than when she was believed to beunattractive. These studies suggest thatstereotypes about physical beauty may haveimportant consequences for people's socialbehavior and skills, necessitating furtherresearch on the relationship between attrac-tiveness and actual everyday social interaction(as distinguished from studies assessing at-tractiveness stereotypes or first encounters inthe lab).

Our second question was whether physicalattractiveness is more important to the social

behavior of males or females. Despite thepervasive cultural belief that it is more im-portant for a female to be attractive than fora male, the evidence is mixed. Walster et al.(1966) found that a self-report popularityindex correlated with attractiveness morehighly for females (.46) than for males (.31),and Berscheid, Dion, Walster, and Walster(1971) found that attractiveness and esti-mated number of dates in the last year corre-lated more highly for females (.61) than formales ( .25). On the other hand, Byrne, Ervin,and Lamberth (1970) found that a date'sphysical attractiveness correlated more highlywith the attraction responses of female sub-jects than with those of male subjects, al-though the males claimed that a date's attrac-tiveness was more important for them. Giventhese conflicting results, our second predic-tion was that physical attractiveness wouldaffect the social participation of females andmales equally. One qualification to this hy-pothesis is in order. Given Deaux's (1977)assertion that sex-stereotypic self-presenta-tions are more likely in more "social" situa-tions, we would expect attractive males tohave more self-initiated interactions, whereasattractive females would participate in moreother-initiated encounters.

Our third question concerned the lastingeffects of physical attractiveness. Most of thephysical attractiveness research has dealt withone-time interactions or with responses tophotographs. In both cases the information isvery limited, and we would expect attractive-ness to have strong effects in the absence ofother salient inputs. Many theorists, Ber-scheid and Walster (1974), for example, havespeculated that as we receive more informationabout a person, the effect of attractiveness maydiminish. The only evidence that we havebeen able to find is an experiment by Mathes(1975) in which attractiveness remained im-portant over five separate encounters, evenwith competing additional information. More-over, the theoretical position of Adams (1977)suggests that the attractiveness stereotypemay be largely true, in which case attractive-ness should remain socially important overtime. This "kernel of truth" may of course bedue to self-fulfilling prophecies (cf. Snyder etal., 1977). Our final prediction, then, was that

606 H. REIS, J. NEZLEK, AND L. WHEELER

the effects of physical attractiveness on socialparticipation would persist over the academicyear of the study.

The purpose of the present study is toinvestigate the nature and breadth of physi-cal attractiveness effects in people's ongoingsocial lives. Despite the apparent relevanceof appearance to social interaction, as dis-cussed by Adams (1977) and Berscheid andWalster (1974), there has been little researchrelating attractiveness to aspects of socialinteraction, and we are aware of none thatexamines and distinguishes the various param-eters of socializing as they relate to attrac-tiveness.

The technique developed by Wheeler andNezlek (1977) for studying social participa-tion makes investigation of this aspect of at-tractiveness possible. Their procedure requiressubjects to complete a short fixed-formatrecord for every interaction of 10 minutes orlonger that occurs during a specified interval.From their entries, indices of duration, satis-faction, intimacy, initiation, activity, location,and sex composition are assessed, both overalland after they have been broken down intovarious categories (same-, opposite-, andmixed-sex interactions, for example). Thesevariables were selected in their research be-cause they represented many of the essentialfeatures characterizing people's social encount-ers. For this reason, as well as for their suc-cess in Wheeler and Nezlek's study, they willbe utilized here also. Using this approach,we assessed the interaction pattern of a groupof first-year college students and then exam-ined the relationship of these numerous indiceswith physical attractiveness in order to explorethe three general predictions stated above.

Method

Subjects and General Overview

Subjects were 35 males and 36 females enrolled ina moderately sized, academically oriented north-eastern university. All were in their first year ofcollege and lived in dormitory rooms shared withone roommate. They completed the interaction rec-ords for four 10-day periods chosen to minimizeoverlap with exam periods and holidays: September27 to October 6, December 1 to 10, January 24 toFebruary 2, and April 5 to 14. Some subjects did notcomplete the records during all four intervals for

various reasons (e.g., illness, trips home, etc.),yielding the following subsamples: Time 1—36 males,35 females; Time 2—33 males, 34 females; Time 3—35 males, 35 females; Time 4—34 males, 34 females.Pictures were taken at the conclusion of the fourthtime period and were rated at a nearby university.

Procedure

Subjects were recruited during summer orientationsessions for a "research project on social interac-tion." During a brief meeting, the importance ofunderstanding interaction patterns was explained andthe students' role as collaborators in this naturalisticresearch was stressed. Although subjects were toldthat if grant support was obtained they wouldreceive $10 for each record-keeping period, theywere asked to volunteer only if the opportunity toparticipate in some interesting research was suffi-ciently rewarding. To emphasize this goal, it wasnoted that the probability of funding was small. Oursubjects' intrinsic motivation is evident in the factthat not one complained about the lack of paymentwhen informed that funding had not been secured.No other academic or extrinsic incentives wereinvolved.

Shortly after their arrival on campus in Septem-ber, another meeting was held with subjects who hadexpressed an interest in participating. At that time,the interaction record was explained more fully.The record, a sample of which is shown in Figure1, was to be completed for every interaction thatlasted 10 minutes or longer. An interaction wasdefined as any encounter with another person(s) inwhich the participants attended to one another andadjusted their behavior in response to one another.Examples were provided (e.g., sitting next to some-one in a lecture was not appropriate, whereas talkingduring the lecture for 10 minutes was), and thevarious categories were discussed until everyone feltcomfortable with the forms. A more detailed de-scription may be found in Wheeler and Nezlek(1977). It was suggested to subjects that they fillout the records at a uniform time, such as beforegoing to sleep. To encourage daily recording, sub-jects were asked to return their completed formsand pick up blank ones every few days. Throughoutthe study, a collaborative, nondeceptive atmospherewas maintained, which we believe aided the gatheringof valid data. Confidentiality of the records wasemphasized and closely guarded throughout. At theconclusion of each 10-day record period, a brief

am_ pm Length _

Sex;

Male Group Female GroupMixed Group

No

Initiator Self Other Mutual Unclear

Intimacy of Interaction Intimate 7 6 5 4 3 21 Not Intimate

Satisfaction Unpleasant 7 6 5 4 3 2 1 Pleasant

Location. Mine Theirs Ours Dining On Campus Off Campus

Nature Tosh Past-time Conversation Share Date Party Date/Parly Other

Figure 1, The interaction record form.


interview with one of our research assistants washeld. During that session, the interviewer probedfor difficulties, ambiguities, and potential sources ofinaccurate data. In particular, subjects were urgedto inform us of anything that might have impededtheir accuracy.

At the conclusion of the final interview, subjectswere informed that we wished to photograph themin order to investigate the effects of physical attrac-tiveness. They were told the slides would be evalu-ated at another university and would never be shownon this campus nor used for any other purposes.Further, they were allowed to reclaim their slides atany point. One subject declined to be photographed.

Uniform mid-thigh to over-the-head photographswere taken, with a beige background curtain. Allsubjects were asked to smile, and the most favorableof a minimum of 2 slides (as judged by the investi-gators) was used. Subjects had not been forewarnedthat they would be photographed; we sought photo-graphs that would reflect their "everyday appear-ance." The 71 final slides were then grouped by sexand were randomly arranged within sex. They werejudged by an introductory psychology class of 47males and 49 females at another university 75 milesaway. This university is essentially similar in itsorientation and in the type of students it attracts.Although a group rating session was used, the needfor independent ratings was highlighted, and thestudents remained silent throughout. They wereinstructed to use their own, personal standards ofphysical attractiveness. Each slide was judged onthe same 1-15 scale, with the high end indicatinggreater attractiveness. To provide a general orienta-tion, the entire set of slides was shown once. Theywere then rated on a second viewing, 25 sec perslide. All of the female slides were shown first, fol-lowed by the males.

Construction and Nomenclature ofInteraction Variables

From the raw interaction records, composite indiceswere created in the following manner: satisfaction—mean reported satisfaction over all interactions;intimacy—mean reported intimacy over all interac-tions; length—mean reported length of all interac-tions ; per day—mean reported number of interactionsper day; time per day—mean reported length summedacross all interactions per day; list—number of dif-ferent individuals interacted with during the entirerecord-keeping period; percentage—percentage of allinteractions falling into that category; initiation—proportion of all interactions that were self-, other,mutually or unclearly initiated (must sum to 1.00for each subject); nature—proportion of all inter-actions that were tasks, passing time, conversations,sharing thoughts and feelings, dates/parties, or otheractivities (must sum to 1.00 for each subject).

Each of these categories was then subdivided inaccordance with the sex composition of the encounter:same sex—interactions including up to three members

of the same sex; opposite sex—interactions includingup to three members of the opposite sex; mixed sex—interactions including three others, at least one ofeach sex; and group—interactions including morethan three other people. Overall measures incorporateall interactions. The same- and opposite-sex cate-gories were then further divided to distinguish theprocesses inherent in close and less close relation-ships. Interaction partners were first rank-orderedwithin each time period by their frequency of oc-currence. Where duplicate sets of initials appeared,subjects were asked to provide distinguishing middleinitials. Each of the interaction measures was thencomputed for subjects' three best friends (i.e., satis-faction, intimacy, etc., for those interactions inwhich the three most frequently reported partnersparticipated) and other friends (i.e., those interac-tions including friends ranked fourth through last).The appropriateness of frequency to define closenesshas been discussed earlier (Wheeler & Nezlek, 1977).In their sample, 93% of respondents named one ofthe three most frequent interactants as their bestfriend. It should be noted that some of the cate-gories listed above contained no observations forsome subjects. These entries were treated as missingdata in the analyses.

Results

Accuracy of the Interaction Records

All subjects were interviewed individuallyafter each data collection period. The recordwas examined for any peculiarities, subjectiveimpressions were elicited, and a standardizedinterview was given.

The following questions were analyzed witha 2 (Sex) X 4 (Time) unweighted analysis ofvariance. No sex differences were found, butsome variables exhibited significant lineartime trends that are reported below, alongwith their means, (a) Degree of difficultyrecording interactions (1 = no difficulty, 7 =very much difficulty): Time 1 = 2.90, Time2 = 2.84, Time 3 = 2.73, Time 4 = 2.52;F(l ,60) = 7.74, p < .01. (b) How accuratedid the student perceive his/her records tobe? (1 = very accurate, 7 = very inaccu-rate): Time 1 = 2.44, Time 2 = 2.42, Time3 = 2.85, Time 4 = 2.76; F(l, 60) = 27.72,p < .01. (c) What is the student's guess ofthe percentage of interactions not recorded?Time 1 = 10.94, Time 2 = 10.01, Time 3 =8.90, Time 4 = 9.11; F(l,60) < 1.00, ns.(d) How does the student rate the extent towhich keeping the record interfered with his/her interactions? (1 = no interference, 7 =


Table 1Mean Rating and Mean Variability ofStimulus Persons' Attractiveness

Stimulus persons

Mean

MalesMSD

FemalesMSD

Maleraters

attractiveness"

5.511.24

6.311.78

Femaleraters

S.151.52

6.641.88

Mean standard deviations of attractivenessacross raters

MalesFemales

5.055.83

6.466.81

• There are 36 female and 35 male stimulus personsand 49 female and 47 male raters.

a great deal of interference): Time 1 = 1.55,Time 2 = 1.37, Time 3 = 1.21, Time 4 =1.29; F(\, 60) = 6.88, p < .01. (e) Did suchinterference increase or decrease during therecord keeping period? (1 = decreased, 2 =no change, 3 = increased): Time 1 — 1.87,Time 2 = 1.94, Time 3 = 2.02, Time 4 =1.98; F(l, 60) = 4.69, p < .05. (f) Did thestudent consider his accuracy to have in-creased or decreased during the record-keepingperiod? (1 = decreased, 2 = no change, 3 =increased): Time 1 = 2.31, Time 2 = 2.27,Time 3 = 2.06, Time 4 = 2.05; F(l, 60) =8.15, p < .01.

Although these self-reports cannot be con-strued as objective measures of accuracy,taken together they indicate that subjects feltthe diaries were representative of their sociallives. The linear time trend suggests that therecord-keeping became easier over time, al-though the lack of effect on the percentage ofinteractions not recorded indicates that thiswas not due to differential reporting of inter-actions. Furthermore, mean differences weresmall when compared to the scale endpoints.When combined with the lack of sex differ-ences, these data imply that our results arenot likely to have been due to differentialrecording. It is useful to note further thatthese values are nearly identical to thosereported by Wheeler and Nezlek (1977). Al-

though the present study did not include suf-ficient pairs of interactants acquainted witheach other to compute reliabilities, theirs did.Intraclass correlations for their subjects,sorted by sex and time, ranged from .67 to.84. As the present measures and proceduresare highly similar, their data may be taken asevidence of adequate reliability.

Ratings of Physical Attractiveness

Before discussing the relationship of attrac-tiveness and interaction, it will be useful toexamine the nature of the ratings themselves.Male and female raters tended to agree onrelative attractiveness ratings. Their meanratings correlated .96 for male stimuli and .95for female stimuli. This is consistent withreports of other studies (cf. Berscheid &Walster, 1974). However, mean differencesappeared such that pictures of females wereseen as more attractive than pictures of males,F(l, 92) = 8.94, p < .01. This difference wasalso greater for female raters than it was formales; the Sex of Picture X Sex of Raterinteraction is also highly significant, F(l, 92)— 30.46, p < .001; the respective means areshown in Table 1 and are similar to thosefound in previous research (e.g., Morse, Reis,Gruzen, & Wolff, 1974). The extremely highcorrelations indicate that the sexes agreed intheir relative rankings. The variances of stim-ulus persons' level of attractiveness did notdiffer by sex of subject or rater, Fmax(4, 35)= 2.29, ns.

The bottom part of Table 1 presents theaverage standard deviation of the 96 judges'ratings of each stimulus person, that is, theextent to which judges agreed or varied as toeach person's attractiveness. Analysis ofthese ratings indicated that female raterstended to vary more than male raters, F(l,69) = 36.31, p < .001. No other effects ap-proached significance. Therefore it appearsthat the sex differences to be reported belowcannot be attributed to heterogeneity of vari-ance in females' and males' attractiveness. Asthe difference between male and female raters'overall mean judgments is small (1.14 units ona 15-point scale), it also seems unlikely thatthis accounts for sex differences, since bothgenders were rated in the mid-regions of


attractiveness. For all analyses, mean physicalattractiveness scores were obtained by averag-ing across all raters. Systematic sex-of-raterdifferences did not affect any of our results.1

Attractiveness and Social Interaction

Because of the large number of variablesand computations, our analytic strategy war-rants delineation. Unlike Wheeler and Nezlek(1977), who used analyses of variance toidentify sex differences in interaction, thepresent research concerned the relationship ofphysical attractiveness to interaction. A cor-relational approach is therefore more appro-priate. Since the major hypothesis posited asimple linear relation, the first step was tocompute Pearson product-moment correlationsbetween mean physical attractiveness on theone hand and the interaction indices on theother, separately for all four time periods. Thevarious social interaction parameters werebroken down for overall, same, opposite,mixed, and group composition; their correla-tions with attractiveness are presented inTables 2 through 4 and 6 through 8. Corre-lations for the three best and other friendsbreakdown are discussed only for those in-stances in which they deviated from the com-bined results and therefore shed additionallight on the phenomena. Results for theprimary dependent variables, interactionquantity and quality, will be presented first,followed by the secondary measures of initi-ation and nature of the interaction.

A brief note on the quantity of computa-tions is in order. Of course, computation ofsuch a massive set of correlations necessarilyproduces a substantial number of chance-generated significant correlations (using the.05 criterion, 1/20). The only possible controlfor spurious conclusion drawing will be in-ternal consistency across time periods andvariables. Consistent, repeated, and strongdata patterns will be treated as meaningfulresults; isolated significant correlations willbe noted but considered cautiously.

Interaction Quantity

Table 2 presents the correlations betweeninteraction quantity and attractiveness for

Table 2Correlations of Interaction Quantity andPhysical Attractiveness for Males

Composition/variable

Same listOpposite listSame percentOpposite percentMixed percentGroup percent

Overall per daySameOppositeMixedGroup

Overall lengthSameOppositeMixedGroup

Overall timeSameOppositeMixedGroup

Time1

-.03.59

-.62.57.32.16

.23-.25

.60

.44

.22

.21-.03

.17-.04

.14

.38-.15

.61

.49

.22

Time2

-.18.32

-.49.40.29.08

.14-.20

.35

.29

.12

.35

.07

.26

.47

.20

.35-.14

.40

.51

.26

Time3

-.33.52

-.68.63.54

-.07

.22-.46

.65

.59-.03

.44-.09

.64

.33

.35

.44-.40

.67

.65

.18

Time4

-.39.57

-.69.65.49

-.01

.13-.60

.56

.49-.00

.31

.21

.70

.00

.13

.30-.41

.60

.51

.12

Note, ns vary from 33 to 36. Two-tailed criticalvalues of r for 3 \df are as follows :r = .29 (p < .10);r = .34 (p < .05); r = .44 (p < .01). Length nsvary from 24 to 35 because some subjects had nointeractions in either the group or opposite-sexcategories.

males. These results are strong and striking.Physically attractive males socialized withmore females, more frequently, longer perinteraction and per.day, and as a greater pro-portion of their total social participation. Theyalso engaged in mixed-sex encounters moreoften and with a greater duration. By con-trast, attractive males reported interacting

1 To control for the possibility that the sex-of-rater differences might have affected these measures,mean attractiveness was also computed from standardscores, thereby equalizing the sex-of-rater means andstandard deviations. None of these correlationsdiffered from those presented by more than .01.Further, correlations calculated separately for sub-jects' ratings by male judges and by female judgesrevealed results virtually identical to those presented.Thus the sex-of-rater differences that emerged seemnot to be related to social interaction. Copies ofthese analyses are available from the senior authoron request.


Table 3Correlations of Interaction Quantity andPhysical Attractiveness for Females


Same listOpposite listSame percentOpposite percentMixed percentGroup percentOverall per day

SameOppositeMixedGroup

Overall lengthSameOppositeMixedGroup

Overall timeSameOppositeMixedGroup

Time1

.12-.10-.01-.15

.24

.22

.11

.11-.16

.24

.22-.02-.04-.05

.21

.11

.17

.04-.09

.35

.19

Time2

-.05.26

-.01-.03

.05

.10

.13

.09

.04

.14

.16-.12-.05-.10-.15-.17

.00

.07-.05-.03

.02

Time3

-.02.13.05

-.02-.05

.09-.04

.01-.01-.07

.06

-.11-.04-.18-.21-.06-.14-.06-.00-.15-.01

Time4

-.26.00

-.16.10.09.39

-.20-.23-.03-.08

.18

.21-.21

.32

.27-.11-.04-.28

.11

.11

.25

Note, ns vary from 34 to 35. Two-tailed criticalvalues of r for 32 df are as follows :r =• .29 (p < .10);r = .34 (p < .05); r = .43 (p < .01). Length nsvary from 31 to 35 because some subjects had nointeractions in the group category.

with fewer other males, less often, and as asmaller proportion of their daily socializing,although the length of these encounters wasunaffected. Interestingly, these decrements insame-sex involvement grew stronger over time,perhaps suggesting that as their first year ofcollege progressed and academic/extracurricu-lar pressures increased, social interaction nec-essarily became more selective. Early in theyear, interaction with females was essentiallyirrelevant to time spent with other males;later it precluded it. Group interactions showedno significant relationships with appearancefor males.

The corresponding correlations for femalesare shown in Table 3. Immediately apparentis the lack of significant relationships betweeninteraction quantity and physical appearance.Only one correlation, that with the proportionof group interactions during the fourth timeperiod, was significant at the p < .05 level.

Since this general paucity of significant corre-lations was surprising, the data were furtherscrutinized for nonlinear relationships. Nosignificant quadratic (i.e., either inverted orupright U-shaped bivariate frequency distri-butions) relationships were found. Another setof exploratory analyses investigated hypothe-ses 'positing a heteroscedastic relationship.Attractiveness did not have more of an effectat higher levels of beauty (as suggested byMiller, 1970) or at lower ones. That is, therewere no scatterplots resembling a funnelplaced on its side. This was confirmed bysplitting the subjects into three equal groupsof high, medium, and low attractiveness.These three groups did not significantly differin variance on any consistent set of inter-action variables. Finally, there was no evi-dence to support the contention that appear-ance most restricts the options of the leastattractive, these restrictions diminishing asattractiveness increases (i.e., what might bevisualized as the lower right-hand triangle ofa diagonally divided cube). Thus, the resultsfor interaction quantity may be summarizedsuccinctly: For males, beauty related posi-tively to interaction with women and nega-tively with men; for females, there were nodiscernible correlates.

Interaction Quality

Table 4 presents the Pearson correlationsbetween attractiveness and self-reported inti-macy and satisfaction for both sexes. Lookingat females and males simultaneously, per-ceived intimacy did not relate to appearance.Satisfaction revealed the more intriguing pat-tern of an increasingly positive correlationwith attractiveness over time, most apparentlyso in interactions including members of theopposite sex. All 10 satisfaction correlationsincreased in positivity from the first timeperiod to the third and fourth, and althoughonly 2 of the Time 4 correlations were sig-nificant at p < .05 (opposite-sex satisfactionfor females and males), the remainder were inthe appropriate direction approaching signifi-cance (for 7, p < .15). Examination of thesatisfaction-attractiveness correlations for bestand other friends separately revealed that thisincrement occurred primarily for other friend


relationships. For females, attractivenessmaintained a steady correlation with threebest opposite-sex friends satisfaction (.23,.07, .19, .12), whereas it increased steadilyfor other opposite-sex friends ( — .15, .18, .23,.39). For males, attractiveness showed simi-larly small changes with the three best oppo-site-sex friends (-.08, .18, .30, .17) and largegains with other opposite sex friends ( — .28,.07, .25, .58). Comparable results were foundfor females' same-sex relationships (threebest: .07, .08, .20, .19; others: .01, .12, .41,.33), although not for males. Thus, the in-creasingly positive relationship between at-tractiveness and satisfaction over time seemsto derive primarily from changes in moreperipheral opposite-sex interactions.

Table 4Correlations oj Interaction Quality andPhysical Attractiveness

Table 5Correlations of Attractiveness VarianceWith Satisfaction


Overall intimacySameOppositeMixedGroup

Overallsatisfaction


Overall intimacySameOppositeMixedGroup

Overallsatisfaction


Time1

Males

.20

.19

.03

.27

.12

.06-.03-.17-.00

.04

Females

.21

.24

.23

.04

.07

.03

.04

.24-.04-.11

Time2

.17

.16

.16

.21

.16

.10-.04

.04

.06

.08

-.05.00

-.20-.01-.05

.01

.07

.08

.01-.10

Time3

.16

.08

.32

.10-.02

.17

.03

.36

.14

.01

.25

.27

.37

.24

.34

.21

.29

.21

.21

.16

Time4

.23

.21

.25

.09

.12

.28

.17

.41

.15

.19

.15

.27

.31

.09

.22

.25

.22

.41

.26

.30


Overallsatisfaction


Overallsatisfaction


Time1

Males

-.02-.10-.22-.05

.04

Females

.17

.23

.16

.01

.05

Time2

-.04-.16

.10-.09-.04

.33

.37

.27

.34

.48

Time3

.07-.04

.38

.02-.03

.42

.47

.43

.37

.41

Time4

.16

.08

.35

.13

.13

.35

.29

.40

.37

.39

Note, ns vary from 33 to 36. Two-tailed criticalvalues of r for 31 d/are as follows: r = .29 (p < .10);r = .34 (p < .05); r = .44 (p < .01). Group ns andmale opposite sex ns vary from 24 to 35 becausesome subjects had no interactions in those categories.

Note, ns vary from 33 to 36. Two-tailed criticalvalues of r for 31 d/are as follows: r = .29 (p < .10);r = .34 (p < .05); r = .44 (p < .01). Group ns, andmale opposite sex ns vary from 24 to 35 becausesome subjects had no interactions in those categories.

An unexpected finding emerged at thispoint. In trying to account for the failure tosupport our original hypothesis regardingfemales' socializing, a number of potentialalternatives centered on variability. For somesubjects, judges readily agreed on the level ofattractiveness; for others, there was greaterdisagreement. Accordingly, the variance ofeach subject's attractiveness was computed(i.e., the extent to which judges differed on thesubject's attractiveness). As documented inTable 5, variance turned out to be a strongerpredictor of female satisfaction across allcomposition interactions and during the finalthree time periods. For males, these samecorrelations tended to be nonsignificant. Thepotential meaning of this result will be com-mented upon in the Discussion section. How-ever, none of the other interaction variablesshowed a consistent pattern of relationshipwith attractiveness variance.

Initiation of Interactions

Although Hypothesis 2 predicted equivalentrelationships between attractiveness and inter-action for both sexes, correlations for attrac-


Table 6Attractiveness as Correlated With Initiationof Opposite-Sex Interactions

Table 7Males' Attractiveness as Correlated With theNature of Their Interactions

Variable

Self-initiationOther initiationMutual initiationUnclear initiation

Self-initiationOther initiationMutual initiationUnclear initiation

Time Time Time Time1 2 3 4

Males

-.19 .08-.34 -.23

.50 .25

.07 -.14

Females

-.12 -.25.04 -.01.31 .11

-.35 .24

-.11 -.49.16 .05.26 .29

-.29 .30

.05 - .05-.22 -.10

.06 .09

.26 .11

Note, ns vary from 33 to 36. Two-tailed criticalvalues of r for 31 df are as follows :r = .29 (p < .10);r = .34 (p < .05); r = .44 (p < .01). Group ns, andmale opposite sex ns vary from 24 to 35 becausesome subjects had no interactions in those categories.

tiveness and initiation were expected to differby gender. Results did not support this pre-diction; however, they were consistent withthe data pattern that has emerged so far. Thatis, for males, initiation did relate to attrac-tiveness, chiefly in opposite-sex encounters.For females, no meaningful tendencies ap-peared. Table 6 lists the correlation betweenmale- and female-rated attractiveness and theproportions of opposite-sex interactions fallinginto each of the initiation categories. As canbe seen, beautiful males had a greater pre-ponderance of interactions that they perceivedas mutually initiated, and a lesser proportionof self- and other-initiated encounters.2 Al-though the overall, same, mixed, and groupinitiation data did not reveal any set of resultsbeyond chance expectation, it should be notedthat a number of individual correlations wereconsistent with this trend, and none contra-dicted it. The following significant correla-tions with attractiveness were found for males:Time 1, overall other initiation: r(33) =— .36; Time 4, overall unclear initiation:r (32) = .33.

Nature of Interactions

The final set of interaction variables sortedall social events into six distinct categories:


Tasks :Overall


Pastime :Overall


Converse :Overall


Share :Overall


Party /Date:Overall


Other:Overall


Time1

.10-.01

.21

.17

.12

-.27-.15

.19-.46-.36

-.03-.02

.03

.23

.15

.12

.11-.15-.03-.09

.16

.11-.17

.25

.18

.03

.02

.03

.01-.21

Time2

.02-.05

.24-.09-.11

-.38-.44

.09-.20-.23

-.00.22

-.19-.13

.00

.19

.10

.02

.42

.08

.27

.05-.05

.39

.21

.08

.06

.05-.28-.24

Time3

-.11.03

-.46-.28

.22

-.49-.30-.08-.15-.07

.30.22.41.14

-.16

.18

.15

.22-.13-.11

.23-.15

.07

.29

.31

.12

.29-.01-.04-.00

Time4

-.17.03

-.07-.00

.13

-.43-.44-.39

.16-.30

.39

.39

.25

.16

.06

.20

.08

.34

.11-.04

-.06-.19

.05-.23

.06

.03

.10-.17-.25

.14

Note, ns and significance criteria are the same as inTable 2.

tasks, pastimes, conversations, sharing feel-ings, dates/parties, and others. This was doneto examine qualitative differences in the mannerin which attractive and less attractive indi-viduals spend their social time. To control for

2 Because of skew in the distributions, Spearmanrank-order correlations were computed to ensurethat these findings were not attributable to a fewanomalous cases. All significant correlations re-mained significant in this analysis.


quantity differences, each score representedthe proportion of all interactions of that sexcomposition spent in that domain. Amongother things, these variables will allow us totest in our data the one finding most oftenreported in the literature: that attractiveindividuals date more frequently.

Table 7 presents these correlations formales. More attractive men appear to bespending a larger proportion of their interac-tions in conversations and a smaller propor-tion in pastimes and, to a lesser extent, intasks, the two activity-centered dimensions.This pattern was spread across all of the sexcomposition categories and time periods, al-though the results were most consistent foroverall and same-sex interactions, as well asfor the latter two time periods. There wasalso a less consistent tendency for attractivemales to date/party relatively more fre-quently in mixed-sex settings. Comparison ofthe three best friend and other friend corre-lations did not reveal any distinguishingtrends.

The corresponding correlations for femalesare listed in Table 8. As with their malecounterparts, attractive females tended tooccupy a lesser proportion of their interactionswith tasks. They also reported proportionallyfewer "other" activities and sharing feelings.In contrast, they experienced more prevalentdate/party interactions, notably includingmore than one other person, as most partyinteractions might. In the literature, attrac-tiveness typically reveals a positive correla-tion with the number of dates, and it does inour data as well. The number of mixed-sex -date/party interactions correlated .30, .10, .32,and .36 with attractiveness over the four timeperiods (all but the second figure are signifi-cant at p < .08). These dates/parties tendedto involve peripheral rather than their threebest male friends: the attractiveness - oppo-site-sex others - date/party correlations ap-proached significance at p < .11 for Time 3and Time 4, both ys(33) = .26. Thus, sup-port for the hypothesis that attractive womendate more often is contained in our data.

Table 8Females' Attractiveness as Correlated With theNature of Their Interactions


Tasks:Overall


Pastime :Overall


Converse :Overall


Share :Overall


Party/Date:Overall


Other:Overall


Time1

-.12-.19-.28-.08

.13

.04

.14

.02

.10-.05

-.13-.19

.18-.10-.09

.05

.15-.01

.11-.15

.29

.16

.05

.06

.23

.17

.06

.22

.01-.08

Time2

-.11-.16-.10-.04-.11

.05

.01

.16

.01-.08

.13

.23

.25-.00

.15

-.33-.35-.34-.13-.13

.13

.26-.09

.22

.20

.13

.18

.10

.06-.15

Time3

-.17-.10-.25-.43-.08

.16

.24-.00

.09-.03

-.16-.15

.12

.04-.03

.10

.05

.01

.27

.25

.29

.16

.14

.28

.14

-.05-.25

.20-.19-.30

Time4

-.12-.09

.06-.21-.33

.16

.08-.06

.09

.13

-.21-.11-.02-.17

.04

.11

.19-.02

.17

.21

.46

.10

.12

.48

.30

-.15-.38

.10-.20-.48

Note, ns and significance criteria are the same as inTable 3.

Discussion

We will begin discussing and interpretingour results by summarizing the predominanttrends that bear on our three initial hypothe-ses, (a) Physical attractiveness was stronglyrelated to the quantity of social interactionfor males, positively with the opposite sexand negatively with the same sex; no signifi-cant pattern emerged for females, (b) Forboth sexes, satisfaction, particularly with op-


posite-sex interactions, showed an increasingtendency over time to be positively correlatedwith attractiveness, (c) Females with morevariable attractiveness ratings were more likelyto be satisfied with their socializing, (d)Physically attractive males tended to havemore mutually initiated and fewer self orother initiated interactions with the oppositesex. (e) Attractive males spent more of theirinteractions conversing and less in activities;attractive females also reported a lesser pro-portion of task interactions and more preva-lent date/parties.

Thus, Hypothesis 1 was clearly supported,but only for males. Of course, this means thatHypothesis 2 was not supported, but the find-ings are interesting in that they challengefolk wisdom: Attractiveness was a more im-portant concomitant of social interaction formales than for females. Finally, Hypothesis3 was supported in that most correlations withattractiveness endured over time. Correlationsthat changed, such as those for satisfaction,grew stronger as the academic year progressed.

That male attractiveness was broadly andquite strongly related to the quantity of inter-action is consistent with other results re-ported by Berscheid et al. (1971), Krebs andAdinolfi (1975), and Walster et al. (1966).Prettier males had more and longer interac-tions for more time per day with more dif-ferent females, both alone and in mixed-sexcompany. In contrast, they had fewer inter-actions with fewer other males. In addition,interactions involving females occupied agreater proportion of their social lives. Ap-parently, these males expanded their opposite-sex contacts at the expense of, rather than as asupplement to, their same-sex socializing. Per-haps their beauty makes them more acceptableto female partners and therefore less likely orless able to seek companionship elsewhere.Attractive males may also have experiencedmore positive feedback in opposite-sex en-counters in the past, producing greater confi-dence, enjoyment, and attraction to relation-ships with women. This explanation is fa-vored by the finding that these interactionswere more likely to be conversations thantasks or pastimes and that they tended to bemutually initiated. A conversation focusesattention on the people involved. Anxiety or

general unease would lead one to seek "some-thing to do" in order to avoid awkward un-filled moments or self-focusing (objectiveself-awareness). Similarly, greater reliance onmutual initiation implies less of a seeking orbeing-sought orientation and more of a naturalprogression of chance contacts into interac-tions lasting at least as long as our criterionof 10 minutes. Along with Goldman andLewis's (1977) finding that attractive maleswere more socially skillful in anonymous tele-phone conversations, it appears that beauti-ful males' greater heterosexual contact reflectsgreater self-assurance and skill. Doubtless, so-cial anxiety stems from prior experiences.However, critical self-assessments may leadunattractive males to withdraw from hetero-sexual contact, thereby "proving" their ownhypothesis in the manner of a self-fulfillingprophecy. Skills and confidence require prac-tice, and one simply comes to prefer interac-tions with other males. By avoiding contactwith women, rejection, particularly withoutexternalizing justifications that maintain self-esteem, is not possible (Jones & Berglas,1978).

It is perplexing that our data revealed nosimilar relationships for females. Prior re-search and popular wisdom suggest moreprevalent effects for females, with the excep-tion of Byrne et al. (1970), who found thatattraction correlated more strongly with theirdates' physical beauty for female subjectsthan males. Supporting a "no relationship"conclusion over experimental artifact is ourreplication of the basic datum reported inprior research: Attractive females reportedmore date/party interactions. However, thisdid not extend to other quantitative indices oftheir social lives.

One potential explanation centers on male-female differences in orientation to social in-teraction, differences that Deaux (1977) hascharacterized in terms of self-presentationalstyle. Males, according to Deaux, are likely tochoose a status-asserting manner, seeking toestablish a more dominant position for them-selves. In contrast, females prefer an affilia-tive, or status-neutralizing, style that mini-mizes status differentials and instead buildsegalitarian bonds. If one thinks of beauty asa social asset, then more attractive males


would perceive their chances of acceptance ashigher (Huston, 1973) and would conse-quently seek interaction with females to en-hance their social status. Less attractive maleswould assess their lot less favorably and shuninteraction with females, since there was littlestature to be gained. (Interestingly, Morse etal., 1974, found that males who rated females'attractiveness more highly dated less.) Arelated alternative states that many malesfear rejection by attractive women. For fe-males, self-perceptions of beauty would be lessimportant, since the choice of an interactionpartner would not be based on developingstatus differentials, but instead on eliminat-ing them. The interaction and relationshipitself is the focus, rather than any gains insocial standing. The absence of a particulartype of interaction or partner then leadssimply to choosing a substitute, since there isgreater latitude in the scope of acceptableactivities and partners. Attractiveness maytherefore not be so relevant to whether, andin what manner, a female socializes.

As for our qualitative indices, two generaltrends were identified. That satisfactionshowed a tendency to be increasingly posi-tively correlated with attractiveness over timeis revealing, since it has often been presumedthat the importance of appearance diminishestemporally. However, the first year of collegeinvolves social exploration and experimenta-tion before stable, enjoyable socializing stylesare found and established. At first, the choiceof interaction partners may be of a "hit-or-miss" variety, satisfaction being somewhatrandomly determined. Presumably, attractivepersons have more options in the selection ofpartners, which would allow them greaterchoice in this process. It seems obvious thatgreater social assets would permit one toselect more desirable relationships. In addi-tion, the heightened sense of personal controlwould make for greater satisfaction as well.Herold (1979) also found that college stu-dents' social satisfaction correlated with theirattractiveness. Interestingly, he found thisrelationship to be somewhat stronger formales than for females.

An explanation of the fact that more varia-bly attractive women were generally moresatisfied must be more speculative, since this

result was not anticipated. Our best hunchresides in the nature of attractiveness, sexrole stereotypes, and the attribution process.In answering the question "Why is (isn't) thisperson interested in me?" it would seem thatone's attractiveness or the lack of it is asufficiently potent external cause to precludemore dispositional attributions about char-acter, intelligence, personability, wit, charm,or mystery. In Jones and Davis's (196S)terms, a socially desirable external factor pre-vails. Everyone is aware of the stereotypestipulating that males are interested in fe-males only for their looks, although mostwomen (and men) would probably prefer tobe liked for their dispositional qualities. Thestereotyped view of males' attraction is pre-sumably most compelling for women whoreceive rather uniform evaluations from othersand whose assessments of their own beautyare somewhat accurate. In other words, con-sistency, consensus, and distinctiveness arehigh for the low-variability individual ("thisman and all other men are always interested inattractive women like me but not other lessattractive women"; note that we are positingattractiveness as an entity factor). On theother hand, if one's attractiveness is variable,for example, by possessing features which are"in the eye of the beholder," then the patternfavoring an external explanation is not pres-ent. Attraction is less likely to be perceivedas covarying with beauty, since one has notreceived uniform feedback from others in thepast. Attributions to other presumably moreinternal factors are more likely ("this manand only this man likes something aboutme"). The argument is that a woman will bemore satisfied with being liked for her char-acter than for her looks. A related point isthat being liked by only some people maymake one feel individuated in that relation-ship, a personally gratifying idea as well.

The use of naturalistic journal records toprovide a comprehensive examination of on-going social participation is of course fraughtwith many of the deficiencies that nonlabora-tory data often entail. Of specific relevancehere are three methodological issues: bias inthe self-reports, uncontrolled and nonrandom"third variables" influencing and masking oureffects, and the probability of Type I errors


attendant in the large number of statisticscomputed. Regarding biased reporting, thehighly concrete format of the interactionrecord eliminated many of the errors that stemfrom variable memory and interpretation.Instructions were quite specific, and mostsubjects completed their entries daily. Becausethis procedure involves reporting more thanjudgment and therefore seems more amenableto accurate recall than general questions(e.g., "how many dates have you had in thelast month?"), we believe that the journalrecords are less affected by self-serving andself-denigrating biases than are other tech-niques using less concrete procedures. Thisfeeling is substantiated by the interview datapresented earlier.

As for uncontrolled third variables, they areobviously present. We have traded the tightcontrols of the laboratory for the ecologicalvalidity of a more naturalistic approach. Thisseems appropriate, as our goal was not toestablish causal linkages but rather to studyattractiveness as it relates to people's naturalsocial context. Laboratory research has yieldeda well-defined set of guiding principles. Manyfactors may intervene in real life, however,modifying the general tendencies observedunder stricter controls. Minimally, these mightinclude self-selection of partners, longer dura-tions of relationships, and the availability ofalternative partners. The usual admonitionthat causal statements are not possible seemsoverly conservative, however. Causality maybe inferred by extrapolation from the readybase of experimental literature. Social psy-chology may well be arriving at the luxuriousthreshold of articulately combining natural-istic and laboratory research into the coordi-nated study of social phenomena. Morenaturalistic research is clearly needed in thisarea. One important consideration would bereplication of this study in a noncollege sam-ple in which the subjects are older and arenot residing in a self-contained social com-munity.

Finally, a note on the large number of cor-relations computed is in order. Using a .OScutoff, we may of course expect 1 of every 20correlations to be significant, yielding thepossibility of substantial chance findings. Con-sistency has been our guide to counteract this

potential. The proportion of significant corre-lations in these data is well above 1/20. Addi-tionally, we have avoided any inferences what-soever about "significant" correlations thatwere well dispersed in the data and were notrepeated across categories and time periods.For the primary findings, we believe chancecauses may be ruled out.

From these data, nothing should be moreapparent than the fact that physical attrac-tiveness plays an important role in socialparticipation, the nature of this effect beingdecidedly more complex than was initiallysuspected. The "what is beautiful is good"stereotype may well be the origin of meaning-ful variations in one's style, extent, and feel-ings about socializing. These would includesocial skills, social confidence, the manner ofactivities and relationships one chooses, andcognitons about those relationships and one-self. What first figures as a superficial, mini-mally consequential trait is thereby imbuedwith substantial importance. Considerablymore research is therefore appropriate, payingattention not only to what people's implicittheories about physical attractiveness are butto the real-life ramifications of these beliefsas well. Such research entails the naturalisticstudy of people in a broad context as they goabout their lives.

References

Adams, G. R. Physical attractiveness research:Toward a developmental social psychology ofbeauty. Human Development, 1977, 20, 217-239.

Berscheid, E., Dion, K. K., Walster, E., & Walster,G. W. Physical attractiveness and dating choice:A test of the matching hypothesis. Journal ofExperimental Social Psychology, 1971, 7, 173-189.

Berscheid, E., & Walster, E. Physical attractiveness.In L. Berkowitz (Ed.), Advances in experimentalsocial psychology (Vol. 7). New York: AcademicPress, 1974.

Brislin, R. W., & Lewis, S. A. Dating and physicalattractiveness: Replication. Psychological Reports,1968, 22, 976.

Byrne, D., Ervin, C. R., & Lamberth, J. Continuitybetween the experimental study of attraction andreal-life computer dating. Journal of Personalityand Social Psychology, 1970, 16, 157-165.

Deaux, K. Sex differences in social behavior. In T.Blass (Ed.), Personality variables in social be-havior. Hillsdale, N.J.: Erlbaum, 1977.


Dion, K. K., Berscheid, E., & Walster, E. What isbeautiful is good. Journal of Personality and So-cial Psychology, 1972, 24, 285-290.

Goldman, W., & Lewis, P. Beautiful is good: Evi-dence that the physically attractive are moresocially skillful. Journal of Experimental SocialPsychology, 1977, 13, 125-130.

Herold, E. S. Variables influencing the dating ad-justment of university students. Journal of Youthand Adolescence, 1979, 8, 73-79.

Huston, T. L. Ambiguity of acceptance, social desira-bility, and dating choice. Journal of ExperimentalSocial Psychology, 1973, 9, 32-42.

Jones, E. E., & Berglas, S. Control of attributionsabout the self through self-handicapping strate-gies: The appeal of alcohol and the role of under-achievement. Personality and Social PsychologyBulletin, 1978, 4, 200-206.

Jones, E. E., & Davis, K. E. From acts to disposi-tions: The attribution process in person percep-tion. In L. Berkowitz (Ed.), Advances in experi-mental social psychology (Vol. 2). New York:Academic Press, 1965.

Krebs, D., & Adinolfi, A. A. Physical attractiveness,social relations, and personality style. Journal ofPersonality and Social Psychology, 1975, 31, 245-253.

Mathes, E. W. The effects of physical attractivenessand anxiety on heterosexual attraction over a

series of five encounters. Journal of Marriage andthe Family, 1975, 37, 769-774.

Miller, A. G. Role of physical attractiveness inimpression formation. Psychonomic Science, 1970,19, 241-243.

Morse, S. J., Reis, H. T., Gruzen, J., & Wolff, E.The "eye of the beholder": Determinants ofphysical attractiveness judgments in the U.S. andSouth Africa. Journal of Personality, 1974, 42,528-542.

Snyder, M., Tanke, E. D., & Berscheid, E. Socialperception and interpersonal behavior: On the self-fulfilling nature of social stereotypes. Journal ofPersonality and Social Psychology, 1977, 35, 656-666.

Tesser, A., & Brodie, M. A note on the evaluation ofa "computer date." Psychonomic Science, 1971, 23,300.

Walster, E., Aronson, V., Abrahams, D., & Rottmann,L. Importance of physical attractiveness in datingbehavior. Journal of Personality and Social Psy-chology, 1966, 4, 508-516.

Wheeler, L., & Nezlek, J. Sex differences in socialparticipation. Journal of Personality and SocialPsychology, 1977, 35, 742-754.

Received April 30, 1979

Documents

Physical Attractiveness in Social Interactionjbnezl.people.wm.edu/Reprints/1980-JPSP-Physical-Attr-RIR.pdfphysical attractiveness correlated more highly with the attraction responses