Social and Emotional Messages of Smiling: An Ethological Approach

Journal of Personality and Social Psychology1979, Vol. 37, No. 9, 1539-1553

Social and Emotional Messages of Smiling:An Ethological Approach

Robert E. Kraut and Robert E. JohnstonCornell University

Did smiling evolve as an expression of happiness, friendliness, or both? Nat-uralistic observation at a bowling alley (N — 1,793 balls) shows that bowlersoften smile when socially engaged, looking at and talking to others, but notnecessarily after scoring a spare or a strike. In a second study, bowlers (N =166balls) rarely smiled while facing the pins but often smiled when facing theirfriends. At a hockey game, fans (N = 3,726 faces) smiled both when theywere socially involved and after events favorable to their team. Pedestrians(TV = 663) were much more likely to smile when talking but only slightlymore likely to smile in response to nice weather than to unpleasant weather.These four studies suggest a strong and robust association of smiling with asocial motivation and an erratic association with emotional experience.

Everyday experience suggests that smilingis one of the most common nonverbal signalsused in communication among humans. De-spite this, and despite more than 100 yearsof research on facial expressions, we still knowrelatively little about the causation of smilingand its social functions. In this article we at-tempt to provide evidence about the causationof smiling in social settings and to raise someneglected questions about the analysis of fa-cial expressions in general.

Research and thought on the facial expres-sion of emotion has had a checkered historysince the publication of Darwin's The Expres-sion of the Emotions in Man and Animals in1872, as has been documented by a number

This research was supported in part by NationalInstitutes of Health Grant MH 30041 to RobertKraut and National Science Foundation Grant BNS76-18681 to Robert Johnston. Peter Goldenthal, Vir-ginia Engelhard, David Eaton, Celia Littauer, Cyn-thia Morgan, Sherric Zweig, and Barbara Petersonprovided valuable discussion in early stages of theresearch and helped collect data. Scott Hamilton,manager of Ide's Bowling Lanes in Ithaca, NewYork, was both generous and patient about our re-search efforts.

Requests for reprints should be sent to Robert E.Kraut, Departments of Psychology and Sociology,Uris Hall, Cornell University, Ithaca, New York14853.

of excellent reviews (Ekman, 1973; Ekman,Friesen, & Ellsworth, 1972; Izard, 1971), butrecently several of the perennial questions inthis field have been settled. In a variety ofstudies researchers have shown that peoplecan consensually and accurately recognize atleast six emotional expressions from picturesof faces (e.g., Ekman, 1972; Ekman et al.,1972; Tomkins & McCarter, 1964) and thatthese abilities seem to be universal amonghumans (Eibl-Eibesfeldt, 1972 and 1973; Ek-man, 1973; Izard, 1971; Vinacke, 1949;Vinacke & Fong, 19SS). Ekman (1972) hasproposed a general model for the factors in-fluencing facial expression that we take asrepresentative of this "emotional expressionapproach" to facial expressions, and we ex-plicate this model in more detail below forthe particular case of smiling. Tomkins'(1962) and Izard's (1971) positions empha-size the influence of facial expressions on emo-tional experience more than Ekman's ap-proach does but are otherwise similar.

The Emotional Expression Approach:Smiling as the Expression of Happiness

According to the emotional expression view,a smile is the major component of a facial dis-play associated with and caused by feelings

Copyright 1979 by the American Psychological Association, Jnc, 0022-3514/79/3709-1539$00.7S

1S39

1540 ROBERT E. KRAUT AND ROBERT E. JOHNSTON

of happiness or joy. Anything that makes aperson feel good or happy should producesmiling unless the individual wants to maskor inhibit this display. Laughing is consideredto be the expression of either more intensehappiness (Darwin, 1872/1965) or a par-ticular type of happiness (Ekman & Friesen,1975). Cultural and individual differences in-fluence smiling both by determining the inter-pretation of events, which affects the cause ofhappiness, and by shaping display rules, whichdetermine when it is socially appropriate tosmile. But such differences do not alter peo-ple's innate and universal tendency to smilewhen they are happy. Thus, when a smile doesoccur, the message is usually happiness (Ek-man & Friesen, 1975), although this may bea false message if the sender is masking an-other emotion with a smile or if the sender issimulating happiness for some other reason.

It is important to note that althoughworkers in this tradition have emphasized theimportance of facial expression in communica-tion and social behavior, they have rarelystudied such communication in natural socialsettings by studying the causes and conse-quences of smiling; rather, they have focusedon the recognition and verbal labeling of emo-tions in facial expressions, generally in stillphotographs.

Ethological Studies:Smiling Indicates Friendliness

A different paradigm has of necessity beenused by ethologists studying nonhuman pri-mates. These workers have used naturalisticobservation as a research tool and have drawnmany of their hypotheses about humans fromcomparisons of humans with other primates;they have concentrated on the proximatecauses of smiling, its consequences for theimmediate interaction, and its evolutionaryfunctions.

Many nonhuman primates have a submis-sive facial display, called a grimace, a grin,or a silent bared-teeth face. The display re-sembles the human smile, and in all species inwhich it occurs, it seems to have the functionof deflecting hostile behavior of more dom-inant animals (Hooff, 1962). In a detailed

study of chimpanzees, Hooff (1972, 1973)showed that in addition to averting attacks,variations of this display were used to main-tain or increase afnliative behaviors betweenindividuals. In some circumstances dominantanimals used one variety of the bared-teethface to reassure subordinates of their nonhos-tile, affiliative intent. Hooff hypothesized thatthe human smile is evolutionarily related to thechimpanzee's bared-teeth displays and servesthe same functions of deflecting hostility andmaintaining friendly contact. On the otherhand, according to Hooff, laughter evolvedindependently and is related to the primate"play face."

If human smiling is a behavioral homologueof chimpanzee bared-teeth displays, onewould expect smiling to occur most in face-to-face interaction, especially where friendly in-tent is problematic or where social bonds arebeing established or renewed. The smiler'smotivation may be genuine friendliness or anintent to establish friendly relations. Re-searchers in this ethological tradition have notbeen concerned with the emotions or feelingsexperienced by those doing the smiling.

The two approaches outlined above are notnecessarily incompatible, but they have talkedpast each other by using different method-ologies on different species to ask differentquestions. The present article tries to comparepredictions based on smiling as the expressionof happiness with those based on smiling asan indication of friendliness, since the moststraightforward extrapolations from each posi-tion do lead to different predictions about thecauses of smiling in social settings. We askedabout the motivational state of the smiler andthe conditions under which smiling is pro-duced, and we used naturalistic observationas our methodology. We chose public settingsin which we could observe people's faces, inwhich the two theoretical orientations wouldpredict that smiling would occur frequently,and in which we could distinguish the twotheories. Thus, for a setting to be relevant tothe emotional orientation, emotionally arous-ing and happiness-producing events had to befrequent. To be relevant to the social orienta-tion, social interaction had to be frequent.

MESSAGES OF SMILING 1541

Study 1: Naturalistic Observations of Bowling

Watching bowlers is an excellent way todistinguish the social and emotional hypoth-eses about the causation of smiling. Observa-tions of the progress of the game and the ac-tions of friends and teammates provide evi-dence about external events that might causesmiles, and observations of the bowlers' be-haviors as they are smiling provide evidenceabout their motivational states.

According to the emotional hypothesis,bowlers should smile whenever they feelhappy, for example, immediately following aspare or strike. But according to the socialhypothesis, smiling should occur during socialinteraction, and the score obtained should beirrelevant.

To some extent, the entire game of bowlingis played in a social context. Bowlers gen-erally play with several friends or teammateswith whom they talk and drink between turnsand who shout encouragement, taunts, andinsults after the play (see White, 1955, for alively description). Yet during the game, theextent to which bowlers are engaged in socialinteraction varies greatly from moment tomoment. When bowlers are facing the pins,preparing to roll the ball and watching theoutcome, social interaction is minimal. It in-creases when they turn to face friends and isgreatest when they are engaged in face-to-faceinteractions in close proximity to them. Thesocial hypothesis predicts that smiling shouldoccur most during these bouts of more intensesocial involvement.

Method

The bowling alley in which we made observationshad 36 lanes. The lanes were set up in pairs with ascoring table centered about 4.5 meters behind thefoul line and a semicircular bench defining a pit forbowlers waiting their turn. Behind the pit and aguardrail was a gallery with small tables at whichspectators and bowlers might sit, watch the game,and drink.

Although there were many variations, most bowlersfollowed a predictable sequence. They would arisefrom a seat in the pit, select a ball from the ball re-turn located between the two lanes, approach andrelease the ball at the foul line, stand or back upwhile facing the pins to see the outcome of the roll,

turn 180° to face the pit, and walk toward the pit.After turning from the pins, they might show acommunicative display o.r engage in social behavior,and they would then go back to 'the ball return andstart the second roll. At the end of the second roll,bowlers would return to their seats or leave the pit,occasionally talking to people on the way.

We made observations of bowlers who appearedto be at least 18 years old. Lanes to observe werearbitrarily chosen from among those with an openspectator table located about 7 meters behind thefoul line. If no spectator tables were available, ob-servers arbitrarily selected a lane and observed whileleaning against a wall. An observer recorded behaviorfrom all bowlers using a selected lane for a maximumof 20 rolls of the ball from all bowlers. At the endof 20 rolls, the observer moved to another arbitrarilychosen lane. Most bowlers were unaware that theywere being observed by us. Those bowlers who askedwere told that we were watching their technique andtheir score.

The observer recorded verbal and nonverbal dataabout bowlers from the time they turned to face thepit after bowling until they turned away to pick upthe ball after the first roll or returned to thek seat,left the pit, or turned away from the observer afterthe second roll. The mean length of an observationperiod for a sample of 120 rolls was 5.S seconds. Apattern of body and facial behavior believed by theobserver to have occurred simultaneously was re-corded as a unit. If an element in the display changedor if the same pattern was held for what seemed avery long time to the observer, the observer recordedanother unit.

Our observational study of bowlers was replicatedthree times, using slightly different lists of behaviors,different observers, and different recording techniques.Over all replications we observed 1,793 rolls of theball, based on approximately 350 different bowlers.Observers were trained by watching and recordingfrom videotapes of people bowling. In the first rep-lication, observers were five students in a seminar onhuman ethology and the two authors. We were awareof the hypotheses of the study and recorded be-haviors using a brief notational scheme with penciland paper. In the second replication, two observersnaive to the hypotheses of the study watched 550rolls of the ball and recorded behaviors by speakingcode names of the behaviors into portable taperecorders, pausing between behaviors in order toindicate unit boundaries. Using this technique, ob-servers didn't have to remember behavioral sequencesas long and didn't have to look away from the re-search subjects to record data; they were thereforeable to record more behaviors and units per ob-servational period.

In the third replication we videotaped 155 rolls ofthe ball and made detailed analyses of these tapes.Recordings were taken with bowlers' awareness andpermission. We set up the lights and camera at thespectator table and pretended to film for 5 minutesprior to the actual filming to acclimate subjects to


the equipment. Two observers recorded data fromthese tapes. As soon as both eyes of the subject couldbe clearly seen, the observers stopped the tape, re-corded behaviors, hand-turned the tape one quarter-turn, recorded behaviors, and repeated this sequenceuntil the subject moved out of focus or turned sothat only one eye could be seen. Using this method,data were recorded at approximately two-thirds-of-a-sccond intervals. After this stop motion analysis,the tapes were played at least once at normal speedto validate the behaviors recorded.

We did no reliability checks on the live recording.On the analysis of the videotaped sequences the twoobservers agreed on 97% of the behaviors that eitherrecorded.

We recorded the following behaviors because priorresearch suggested that they might be communicativeand because pretesting showed that they had occurredwith sufficient frequency to allow meaningful statis-tical analyses:

Neutral face: blank expression; mouth relaxed;head straight forward; absence of all other codedbehaviors except talked to, groom, and good score.

Closed smile: corners of the mouth turned up; lipstogether; teeth together.

Open smile: corners of mouth turned up; lips areparted to show teeth. In Replication 1 only, smilingwas coded without distinguishing between closed- andopen-mouth smiles.

Laugh: mouth open, corners of mouth sometimesturned up; laughter accompanying open mouth.

Tight lips: lips compressed tightly; mouth instraight line; teeth probably clenched.

Look: gaze fixed on another person.Look down: gaze directed at floor.Look away: head turned off body axis, gaze not

directed at others in group.Talk: vocalization by subject to another person.Talked to: vocalization directed to the subject.Groom: subject preening, scratching, or rubbing

any part of his body or face.Face cover: one or both hands covering facial fea-

tures for several seconds. This was not .recorded inReplication 2.

Head shake: a continuous horizontal movement ofthe head, usually repeated several times.

Ham: a nonspontaneous, exaggerated facial or bodyexpression apparently intended to communicate; forexample, "funny" faces, sticking out tongue, wrinklingnose, jumping up and down, dancing a "jig."

Positive exclamations: spontaneous and often exag-gerated actions or words that the observer believedcorresponded to a pleasant experience or one duringwhich the individual felt plcasurably excited. Forexample, a leap and squeal of joy after a strike.

Negative exclamations: spontaneous and oftenexaggerated actions or words that the observer be-lieved corresponded to an unpleasant experience. Forexample, swinging the fist across one's body, stampingone's foot, or swearing after missing the pins. Ex-clamations were more spontaneous, less exaggerated,and shorter, and appeared to be less intentially com-

municative than hamming, with which they might beconfused.

Good Score: a strike or spare.

Sequential Analysis

In order to understand the causes of smil-ing, it is necessary to examine the events withwhich it is temporally associated. If smilingoften followed an external event in thesmiler's environment, it is possible and evenlikely that the event caused the smile. Ifsmilers often performed other behaviors whilesmiling, it is likely that the motivation under-lying the other behaviors is also underlyingthe smiling.

Whether we consider two events or be-haviors temporally associated depends on thetime unit we select; the meaning of simul-taneous depends on temporal resolution. InReplication 1 we have considered two be-haviors as temporally associated or co-occur-ring if the original observer thought that theyhad occurred simultaneously, that is, placedthem in the same behavioral unit, or if theobserver placed them in adjacent behavioralunits. Replication 2, using tape recorded datacollection, and Replication 3, using analysisof videotapes, had finer temporal resolution.Therefore, to make all data analyses com-parable, we considered two behaviors as co-occurring if they occurred within 3 behavioralunits of each other in Replication 2 andwithin 8 behavioral units of each other inReplication 3. This means that we consideredtwo behaviors as temporally associated if theyoccurred within about 4 seconds of each other.

Because the frequency of co-occurrence isbiased by the frequency of each behavior,Yules Q or gamma is the appropriate measureof association to use. We use this statisticrather than the more familiar phi correlationbecause we considered a positive or negativeassociation between two behaviors as perfectif the more frequent behavior always or neveroccurred with the less frequent behavior (Bla-lock, 1972, pp. 298-299).1

1 The formula for gamma in this research is asfollows. Let A = the frequency of co-occurrence. T =total behavioral units; this is the total number ofopportunities for co-occurrence. Ant = the frequency


Results

The data were stable and the co-occurrenceof smiling with other behaviors is similaracross the three replications. The mean Pear-son correlation between columns in Table 1is .80 (p < .001).

The social hypothesis leads us to expectthat bowlers will smile when they are sociallymotivated independently of emotional experi-ence. Table 1 provides evidence of this.Bowlers were more likely to smile when theywere engaged in social contact, for example,when looking at their friends (M y = .71),and less likely to smile when they were tem-porarily avoiding social contact for some rea-son and were looking at the ground or awayfrom their friends (My = —.56) . In addition,bowlers showed a tendency to smile morewhen they were talking to friends or beingtalked to by them (M y = .20).

Smiling, especially open-mouth smiling, alsooccurred with playful behaviors; laughter (My — .23 over all smiles; M y = .80 for open-mouth smiles), hamming (M y — .33 for allsmiles; M y = .47 for open-mouth smiles),and face-covering (M y = .34 for all smiles;M y — .93 for open-mouth smiling). At thebowling alleys, open-mouth smiling occurredwhen bowlers were being socially playful.Since nonsocial playfulness rarely occurredin this setting, we cannot tell if the social orthe playful motivation was a more importantdeterminant of open-mouth smiling.

The happiness hypothesis leads us to expectthat bowlers would smile more after playingwell and bowling a good score, but this wasnot the case. The association between smiling

of the antecedent behavior; this is the maximumnumber of times it could have co-occurred with asubsequent behavior. Sub = number of behavioralgroupings (i.e., two behavioral units in Replication 1,three units in Replication 2, and eight units in Rep-lication 3) in which a subsequent behavior occurred;this is the maximum number of times it could haveco-occurred with an antecedent behavior. Then B =A n t - A ; C = S u b ~ A ; D = T - A - B - C ; y~[(A X D ) - (B X C)1/[(A X D ) + (B X C) l . Sincethe gamma based on Behavior X preceding BehaviorY is not necessarily the same as Y preceding X, weused the mean of these two measures in all ouranalyses.

and scoring a spare or strike was weak (My - .13). Of the 1,793 rolls for which we col-lected data, 448 were spares or strikes.Bowlers smiled 30% of the time after goodscores and 23 % otherwise. Other support forthe happiness hypothesis is mixed. In general,smiling showed weak associations with thebehaviors that we had identified as subtle indi-cators of negative affect and stronger associa-tions with larger scale emotional displays.Thus, smiling had little association with thesubtle cues of grooming and head shaking.Smiling had a substantial negative associationwith the tight-lip display, which our observa-tions had led us to believe was an indicatorof anger, frustration, and perhaps determina-tion. It is a component in the traditional angerdisplay (Ekman & Friesen, 1975). However,the negative association could have beenpartly caused by the physical difficulty abowler would have smiling and lip pressingsimultaneously (and the difficulty coderswould have distinguishing two behaviors inthe mouth region) as well as by the emo-tional incompatibility of smiling and lippressing.

On the other hand, the spontaneous emo-tional displays we have termed positive andnegative exclamations showed substantialassociations with smiling. When bowlers werecommunicating surprise and glee, generallyafter getting a good score, they smiled (My = .49); when they swore and showed othersigns of disappointment and anger, generallyat not getting a good score, they failed tosmile (M y — —.49) . Since smiling showedno association with score, which is the likelycause of bowlers' emotional experience, thesedata suggest that when bowlers were attempt-ing to communicate their happiness throughpositive exclamations, they used smiles aspart of the communication; when they wereexperiencing positive emotions but were notattempting to communicate them, however,smiling did not covary with other subtle be-havioral indicators of their emotional state.

Several problems limit confidence in theconclusions one can draw from these co-occur-rence data. As we have suggested above, thedegree of temporal co-occurrence between be-haviors may reflect as much their physical


Table 1Temporal Associations (Gamma) of Smiling and Other Behaviors

Replication

Behavior

Neutral faceClosed smileOpen smileLaughHamLook atLook downLook awayTalked toTalkGroomTight lipFace coverHead shakePositive exclamationNegative exclamationGood score

1

Allsmiles

-77——-48

2070

1 59I

4237

-47-79

28-18

24-65

07

2

Closedsmile

-52—-19

070863

-48-56

050801

-52—

1723

-5937

Opensmile

-88-19—

914488

-82-51

4441

-26-68—-07

71-40

29

3MJ

Closedsmile

-70

—05-06

4353

-40-38

3707

-14-33-18

1272

-1601

Opensmile

-93OS

—705083

-66-56-26-02

04-63

931556

-64-09

J.VJ

allsmiles"

-76-07-07

233371

-59-52

2018

-16-59

340449

-4913

'

10.5——

.904.09

11.148.10

13.951.502.091.757.551.07.55

4.535.221.50

Note. Decimal points have been omitted." In this table significance tests have not been performed on the individual gammas, since the units onwhich they are based were not independent of each other. They are based on 5,527 overlapping 4-secondperiods in which two behaviors could co-occur, spread over 1,793 rolls of the ball and over approximately350 different bowlers. The / tests performed on the mean gammas are based on five observations and testwhether smiling is associated with the other behaviors reliably over the different replications and typesof smiles.

compatibility as their similarity in underlyingmotivation or external causation. For ex-ample, since smiling, laughing, talking, andcompressing one's lips are all behaviors doneby the mouth, their mutual co-occurrence islimited. Similarly, the neutral display, bydefinition, cannot co-occur with other be-haviors. A second problem could be a resultof observer errors. Faced with a rich and com-plex event, observers' errors tended to beomissions; our training procedures showedthat observers were more likely to ignore be-haviors that did occur than to record be-haviors that did not occur. What might betermed a climax error may be characteristic ofevent sequences in which observers under-report low intensity forms of a behavior thatgradually change to a climax form of the sameor another behavior. For example, it is pos-sible that smiling might be underreported inthe sequences leading to laughter, hamming,and positive exclamations, while tight lips and

grooming might be underreported in the se-quences leading to negative exclamations.

A third problem is more conceptual. Ourresearch strategy has been to infer the mes-sages of smiling by examining other behaviorswith which it occurs, that is, marker be-haviors. Our beliefs about the significance ofthese other behaviors were based on the priorliterature on nonverbal communication, oninformal observation in our research setting,and on intuition. More systematic analysis ofthe patterns of co-occurrences among markerbehaviors could provide direct evidence on thesignificance of marker behaviors and mightprovide deeper insight into the messages ofsmiling.

Principal Components Analysis

One partial solution to these problems is toanalyze the similarities in co-occurrence thatpairs of behaviors had with other behaviors


rather than analyzing only the temporal asso-ciation between pairs. For example, if liptightening and closed-mouth smiling wereequivalent but alternative behaviors, an anal-ysis of temporal association would show thatthey never co-occurred, but an analysis oftheir similarities in co-occurrence would showthat they occurred in exactly the same con-texts. Following Hooff's logic (1973), westarted with the assumption that to the extentthat a pair of behaviors occurred in the samecontexts, that is, to the extent that they hadsimilar co-occurrences with other behaviors,they also shared underlying motivations orexternal causes. The Pearson correlation be-tween rows in a matrix of gamma scores (ourmeasure of co-occurrence) is one measure ofthe extent to which the pair of behaviors rep-resented by the rows had similar co-occur-rences with each other behavior, includingco-occurrences with themselves. Because thethree replications of our basic study resultedin three estimates of the co-occurrences be-tween each pair of behaviors, we took themedian of these three as our best estimate ofthe co-occurrence. The Pearson correlation be-tween rows was conducted on this matrix.2

One can factor analyze this correlationmatrix. Table 2 is a varimax rotation of theprinciple components analysis of this matrix.The five factors that emerged prior to aneigenvalue falling below 1.0 represent 86%of the variance in the original matrix. An ex-amination of the factor structure providesevidence both about the pattern of co-occur-rences and the underlying motivations repre-sented by the marker behaviors and about theassociation of closed- and open-mouth smilingwith these underlying motivations.

On what factors did smiling load highly?Factor 3 seems to represent a social motiva-tion. Talking, being talked to, looking at an-other, and not looking down or away all loadhighly on this factor. Closed-mouth smilinghas its highest loading on this factor, and inaddition, open-mouth smiling and laughingboth loaded highly here. This result supportsthe previous analysis and clearly suggeststhe social motivation underlying both closed-and open-mouth smiling and laughing.

Table 2Component Analysis of Similarities inCo-occurrence

Behavior

Tight lipHead shakeLaughFace coverOpen smile

Good scoreNegative

exclamationPositive

exclamation

TalkTalked toClosed smileLook atLook down

HamGroom

Neutral faceLook away

1

-87-85

847560

18

-23

-08

13121031

-57

08-34-25-09

2

02-24-00

273292

-91

70

-10114441

-14

08-05

24-22

Factor

3

-341331OS4402

-09

30&2876463

-61

27-08-07-47

4

08231808

-24-23

-19

-54-04-19-16-20

30-89

75

1716

S

-03-01-30

4243

-19

-11

06

-11082947

-OS15

-14-SJ-67

Note. Decimal points have been omitted. Loadingsgreater than or equal to 60 are shown in italics.

Smiling shows a complex pattern of rela-tionships with the emotional displays. Factor2 represents the explosive emotional displaysassociated with extreme scores: good scoresand positive exclamation loaded highly, and

- In Replication 1, closed-mouth smiles were notdistinguished from open-mouth smiles and lookingdown was not distinguished from looking away. Incomputing the matrix of median gammas, the un-diffcrentiated categories of Replication 1 were gen-erally averaged with each differentiated category ofReplications 2 and 3. For example, the associationbetween smiling and grooming in Replication 1 con-tributed in the final matrix to both the associationof closed-mouth smiling and grooming and the asso-ciation of open-mouth smiling and grooming. How-ever, in a few cases closed and open smiles and look-ing down and away had substantially different asso-ciations with another behavior in both Replications2 and 3 (mean difference in gamma > .30). In thesecases, Replication 1 was ignored in computing thefinal measure of association. Specifically, Replication1 was ignored in computing the median relationshipof closed- and open-mouth smiling to laughter andto face cover.


negative exclamations loaded negatively. Theassociation of closed-mouth smiles with thisset of behaviors may reflect the role of closed-mouth smiling in the communication of emo-tion. Factor 1 seems to represent the contrastbetween affect and playfulness. The tight lip,head-shake, and looking-down displays clusterat one pole; and laughter, face covering, andopen-mouth smiling cluster at the other. Thisfactor may reflect the incompatibility of gen-uine playfulness with feeling bad.

Factors 4 and 5 are especially difficult tointerpret. Factor 4 primarily represents thecontrast between hamming and positive ex-clamations, on the one hand, and grooming,on the other. Factor 5 may indicate expres-siveness. The neutral and looking-away dis-plays are at one extreme, and open-mouthsmiling and face covering are at the other.

In summary, both the temporal co-occur-rence of smiling with other behaviors and thefactor analysis of similarities of co-occurrencesuggest that bowlers smiled when they werebeing social, when they were being playful, orwhen they were otherwise communicating anemotional statement to an audience. Bothclosed-mouth and open-mouth smiling shareda nonemotional, social motivation. In addition,open-mouth smiling was similar to laughter inadding a playful motivation; to the extentthat one cannot be playful and at the sametime distressed, angry, or disappointed, open-mouth smiling was incompatible with nega-tive affect.

Study 2: Bowlers Facing the Pins orTheir Companions

One might dismiss our failure to find astrong association between smiling and emo-tional experience in Study 1 with the claimthat bowlers were masking their emotional ex-periences; they hid the joy or disappointmentthey felt in order to appear modest or sports-manlike. To meet this objection we made ob-servations of people who were bowling aloneand were therefore under no pressure to masktheir emotions. Lone bowlers rarely showedany facial displays or other gestures but in-stead maintained a generally neutral face. Themost common expressions seen were relatively

Table 3Percentage of Bowlers Smiling According toBowling Score and Social Focus

Score

Social focus Good Not good Total

YesNo

Total

Note. N = 116."n = 26.b n = 90.

424

46»

283

31"

313

—

antisocial or negative—looking down, tightlips, and negative exclamations; they rarelysmiled. However, one may object to thesedata, since people who bowl alone may be thetype of person who is unexpressive in all cir-cumstances.

Therefore, to further examine the social andemotional messages of smiling, we looked atbowlers when they were facing the bowlingpins and reacting to their rolls, and at thesame bowlers when they turned to face theirfriends. The social hypothesis would lead usto expect very little smiling as bowlers facedthe pins, since this is a relatively nonsocialsetting. The emotional hypothesis would leadus to expect that bowlers would smile afterrolling a good score and would engage in be-haviors revealing negative affect after badscores, regardless of their social orientations.When bowlers roll a good score and remainfacing the pins, they should feel happy andshould not need to mask or hide the expressionof this emotion, since they believe that theyare not being observed.

Method

An observer knelt on a platform among the pinsetting equipment at the end of the bowling alleybehind the bowling pins and watched bowlersthrough binoculars as they finished their roll. Theobserver was 19.2 meters from the bowlers and ob-served through a narrow slit in the facade of the pinsetting equipment. The observer was invisible tobowlers. As bowlers finished their roll and steppedback while watching its outcome, the observer re-corded characteristics of the bowler and the behaviorslisted above. Data were recorded from the momentthe bowler stepped into view until he turned toward


the pit. Simultaneously, a second observer positionedas in Study 1 recorded behaviors in the standard wayafter the bowler turned to face friends in the pit.One hundred and sixteen rolls from 34 bowlers wereobserved from both positions (behind the pins andfacing the pit).

Results

As shown in Table 3, bowlers were gen-erally unexpressive while facing the pins, incomparison with their behavior when theyfaced their friends in the pit. Smiles were par-ticularly rare. In 116 observations, bowlerssmiled 36 times when facing friends but only4 times while facing the pins, t ( H 5 ) = 6.25,p < .001. Smiling was unrelated to how wellthey bowled. Only one of the pin-facing smileswas after a good score, although bowlers inthis sample rolled 26 strikes or spares. Thesedata clearly support the social hypothesis.People rarely smiled in nonsocial settings, re-gardless of emotional experience.

Study 3: Smiling by Fans at a Hockey Game

Hockey at Cornell University is probablythe most important school sport. Students,faculty, and townspeople line up overnight toget season tickets. In 1977, Cornell's team hadfinished its regular season first in the IvyLeague and was in the playoffs for the EasternCollege Athletic Conference Championships.On March 8, 1977 the team faced RensselaerPolytechnic Institute in the quarter-final atCornell. As might be expected, the game wassold out, mainly to enthusiastic Cornell fans.Given this situation, the emotional hypothesisabout the causation of smiling would predictthat Cornell fans would smile more when thehockey match was going well for Cornell thanwhen it was going well for the opposing team.On the other hand, the social hypothesiswould predict that fans would smile morewhen socially interacting with other fans thanwhen watching the game.

Method

A photographer sat in the stands during the hockeygame and, using a telephoto lens, took pictures ofthe spectators across the ice from him (about 30meters away) immediately following events favor-

Table 4Percentage of Spectators Smiling According toValence of Hockey and Social Involvement

Valence of play forhome fans

Social

mcnt

YesNo

Total

Good

%

271213

n

591,2581,317

Not good

%

2223

n

1652,2442,408

Total

%

2367

n

2243,5023,726

able, neutral, or unfavorable to Cornell's chances ofwinning the hockey game. About 220 people werephotographed at each exposure (M ~ 223; SD —21.8). The photographer used a prcfocused cameraon a tripod and took pictures while watching thehockey game in order to be uninfluenced by the be-havior of his subjects as he was photographing them.Photographs were taken after (a) goals for Cornellor the opposing team, (b) penalties called on Cornellor the opposing team, (c) face-offs, before the puckcame into play, and (d) time-killing passes of thepuck at center ice. Cornell goals and opposing-teampenalties were considered events favorable to Cornell,the opponent's goals and Cornell penalties were con-sidered unfavorable, and face-offs and center-icepasses were considered neutral. The section of thestands that was repeatedly photographed mainlycontained season ticket holders and was filled withCornell fans. There were few opposing fans in theentire arena, and they were segregated in anotherarea.

Each transparency was coded by a person naive toboth the hypothesis of the study and the events pre-ceding the photograph. Five randomly selected trans-parencies were then rccoded by a second person tocheck reliabilities. The transparencies were firstscanned for social units, which were defined as agroup of two or more spectators, at least one ofwhom was turned towards the other or others in thegroup. The two independent coders agreed on 73%of the social units that either identified. Each memberof a social unit was defined to be socially involvedwith others in that unit, and all other spectatorswere defined to be socially uninvolved. The trans-parencies were next coded for smiling. The two codersagreed on 73% of the smiles that either identified,69% for smiles within social units, and 74% forother smiles.

Results

Because the results involving neutral andbad events didn't differ substantially fromeach other, they have been combined in theanalyses that follow. Table 4 shows the per-


centage of fans smiling who were or were notpart of a social unit following events thatwere good or not good for the Cornell hockeyteam.

We coded a total of 3,726 faces in 16photographs. Over all photographs, the prob-ability of any one of these faces smiling atthe moment that the picture was taken was.069. Some data supported the emotional hy-pothesis. Spectators were more likely to smilefollowing events favorable to the home team(13%) than following neutral or bad events(4f/( ; y = .60; N = 3,726; p < .001). Somedata supported the social hypotheses. Regard-less of the valence of the events that precededthe photograph, spectators were more likely tosmile if they were members of social units(23'/f ) than if they were not (.6%; y = .66;N = 3,726; p < .001). The effects of beingsocially involved on smiling were strongerfollowing bad and neutral events (y — .85)than following good ones (y = .45, p < .01),perhaps because the base rate of smiling waslower following bad events.

Study 4: Smiling, Social Interactions,and the Weather

People feel happier on days when theweather is nice. Given this assumption, theemotional orientation would lead us to expectthat people walking outdoors would smilemore in pleasant than unpleasant weather. Asthey walk, they can either be socially in-volved with someone or not. The social orien-tation would lead us to expect that peoplewould smile more if they are socially involvedbut that the weather alone should not influ-ence the frequency of smiling.

Method

A single observer made observations of pedestrianstwice at each of four public walkways in Ithaca,New York in September and October 1977. Each sitewas observed once during pleasant weather, whenthe temperature was between SO and 70° F. and thesky was sunny or partly sunny, and once during un-pleasant rainy weather. Pairs of observations weremade at approximately the same time of day andwere made within 3 weeks of each other, A total of663 subjects were observed in the eight observationperiods.

Table 5Percentage of Pedestrians Smiling According tothe Weather and Social Interaction

Weather

Socialinteraction

Good Bad

% n % n

Total

% n

Yes 61 61 57 60 59 121No 12 264 5 288 8 552

Total 21 325 14 348 — —

At the site, the observer selected two referencemarks about 10 meters apart on the sidewalk toindicate the limits of observation. The next pedestrianapproaching the starting mark was selected as a sub-ject. If pedestrians were walking in pairs or largergroups, the person on the extreme left or the personon the extreme right was alternately selected as thesubject for that group, before the group passed hestarting mark. Subjects were observed for the 8 to12 seconds it took them to walk between the tworeference marks. If they talked to anyone, greetedanyone, or were talked to by anyone at any timeduring the observation period, they were classified associally interacting during that period. If theysmiled at any time during the observation period,they were classified as smiling during the observationperiod. Reliability of the smile classification, based ona second observer's judgments of 100 pedestrians,was high (</> = .86).

Results

Table S shows the cross classification ofsocial interaction and smiling among pedes-trians during days with good and badweather. These results show that pedestriansare no more likely to engage in social inter-action on nice days than on bad ones (19% vs.17%, <£ = .02). They were slightly morelikely to smile on nice days than on bad ones(21% vs. 14%, </> = .10, p < .01) and werevery much more likely to smile if they wereconversing with or greeting someone than ifthey were not (59% vs. 8%, <£ = .54, p <.001). The effect of social interaction on smil-ing was slightly stronger on bad weather daysthan on good days (<£ = .58 vs. .47, z =— 1.41, /> < .10), but this interaction wasminor compared to the main effect of socialinteraction and may not replicate.

We conducted an extensive pretest in which1,489 subjects were observed at six sites on


the Cornell campus in the summer of 1977.Good weather was defined as between 65° F.and 75° F. with low humidity, whereas badweather was over 85° F. with high humidity.Subjects were observed as long as they re-mained in sight, about 10-20 seconds, andobservation times were not standardizedacross sites. In this pretest, subjects wereagain slightly more likely to smile on nicedays than on unpleasant ones (<£ = .13, p <.01) and were much more likely to smilewhile socially engaged than when not (</> =.56, p < .001). Here, however, social inter-action had similar effects on good weather andbad weather days (</> = .58 vs. .54, z = .89,p> .10).

In both comparisons, social interaction wasthus a much more powerful predictor of smil-ing than was the weather. In the main study,variations in social interaction accounted for29% of the variance in smiling, whereasvariations in the weather and the positive andnegative emotions they may have producedaccounted for about 1% of the variance insmiling.

Discussion

The Social Hypothesis

Both the present and earlier research pro-vide strong evidence that social involvementis a major cause of smiling, independent ofthe smiler's emotional state. In each of thefour studies described above, smiling wasstrongly associated with social interaction:talking to and looking at others in Study 1,facing fellow bowlers in Study 2, orientingtoward other fans in Study 3, and talking toanother person in Study 4. Other researchershave also found that smiling often occurs ina social context. Almost from its first appear-ance, smiling is socially produced and hassocial consequences. Human infants, from theage of 1 to 5 months, smile most in responseto the human voice and the human face orabstractions of it (Sroufe and Waters, 1976).These smiles seem to be a major determinantof the bond between an infant and its care-taker (Fraiberg, 1977; Spitz & Wolf, 1946).Among nursery school children, smiles arelikely to occur in the context of other social

behaviors such as pointing, giving, receiving,and talking (Blurton-Jones, 1972). Whennursery school children approach a stranger,they often smile, and they are more likely toapproach when the stranger smiles and talksto them (Connolly & Smith, 1972). In addi-tion, the smile appears to be a universal com-ponent of greetings (Kendon & Ferber, 1973;Eibl-Eibesfeldt, 1972). Even when peoplesmile in response to humorous or other non-social stimuli, they smile more in the presenceof other people than when they are alone(Mackey, 1976; Leventhal & Mace, 1970).

It could be argued that positing a separatesocial cause for smiling is a theoretical extrav-agance. According to this view, the presenceof others is just one of the many events thatmake people happy, and apparent socialsmiling is mediated by the pleasant emotionsthe smiler feels in the presence of others. Thisis undoubtedly true on occasion.

However, there are several reasons to be-lieve that many smiles have purely socialcauses independent of happiness. The first isparsimony. We need not assume that hap-piness is a cause when we have evidence of astrong relationship between social stimuli andsmiling, but no evidence of a mediating emo-tion. A theory of friendliness displays, basedon independent comparative data, can ac-count for the empirical relationship. The com-ponent analysis of Study 1 showed a social,nonemotional motivation behind some smiling(Factor 3 in Table 2 ) : both closed- and open-mouth smiling occurred in the same contextsas each of the social behaviors we recorded ata bowling alley. These in turn were inde-pendent of both gross and subtle emotionaldisplays of happiness, disappointment, andanger. This evidence of purely social smilingexists even though smiling has an emotionalcomponent under some circumstances (Fac-tors 1 and 2 in Table 2).

The existence of smiles in uncomfortablesocial settings is further evidence against thehypothesis that social smiles are mediated byhappiness. Repeated viewings of some of ourbowling videotapes convinced us that somesmiling was done to apologize for an espe-cially clumsy performance or for poor bowl-ing, such as dropping the 'ball before bowling

1550 ROBKRT E. KRAUT AND ROBERT E. JOHNSTON

or bowling a gutter ball immediately afterbowling a strike. To examine this possibilitymore systematically, we have started lookingat facial expressions in uncomfortable socialsettings. If smiling evolved from primateappeasement displays, we would expect thatin humans some smiling should occur when aperson is trying to placate or appease an-other, for example when he or she has madea mistake or has violated a social norm andis apologizing for it. An exploratory field ex-periment suggests that subjects smile morein an appeasement than in a control condi-tion. In the appeasement condition, customersin a store were made to think that they hadmade a mistake when they interrupted a clerkbusy with paperwork who told them "I'm notworking here. She (another clerk) will help."In the control condition, customers were madeto think that no mistake had been made("Fine, I'll get it for you, and she will ring itup") . Customers apologized more in the ap-peasement (11%) than the control (0%)condition. More to the point, customerssmiled more in the appeasement condition(28%) , when they were presumably tryingto rectify a mistake, than in the control con-dition (5%, y = .77; « = 99; p < .001).8

Ekman (1972) also notes that people oftensmile when experiencing an unpleasant emo-tion in the presence of others, although heinterprets the smile as a mask for a sociallyinappropriate facial expression that the emo-tion would cause rather than as an appease-ment display. A question for future researchis to determine whether an appeasement or amasking hypothesis can 'better account for theoccurrence of smiling in tense or uncom-fortable social situations.

The detailed analysis of patterns of behaviorin which smiling is embedded provides addi-tional evidence that social smiling need notbe mediated by happiness. For example, Ken-don and Ferber (1973) have carefully de-scribed the behavior in a greeting. They di-vided the greeting into three stages, adistance salutation in which the participantsestablish their readiness and willingness forinteraction, an approach phase in which onewalks toward the other (or both do), and aclose salutation that is a prelude to conversa-tion or other social interaction. During the

distance salutation, the greeters orient theirbodies toward each other, look at each other,and show a greeting gesture such as a headtoss, an eyebrow flash, a wave, or a verbalgreeting. During the approach phase theydecrease the amount they look at each other,until immediately prior to the close saluta-tion, when all participants looked at and thentalked to each other. Clearly the greeterswere regulating their social contact duringthe greeting, first agreeing to it, then post-poning it, and finally engaging in it whentheir physical separation made it convenient.

Significantly for the present argument,smiling almost invariably occurred in both thedistant and close salutation, when greeterswere showing a willingness to greet each otherand establish or reaffirm their relationship.Smiling occurred much less frequently andintensely during the approach phase, whenthe greeters were showing a temporary with-drawal from social contact by looking awayfrom each other. Smiling thus varied with theintensity of social contact. Since emotions areoften regarded as diffuse with a gradualdecay, it is difficult to account for the rapidand asymmetrical shifts in smiling during a10-20 second greeting by referring to shiftsin happiness.

Through this literature review, we havetried to establish that social involvement is amajor cause of smiling and that happinessdoes not seem to be a necessary mediator.Why do people smile in the presence ofothers? Drawing on comparisons of humanswith other primates, Hooff (1972) argues thatmost human smiling is affinitive, used inthe expression of sympathy, reassurance, orappeasement, that is, that the smiler's motiva-tion is to insure the establishment and main-tenance of friendly interaction. The messagemight be paraphrased "I am friendly" or "Iwould like us to be friendly for a while." Thismay occur when friendliness is highly prob-able, as when two old friends greet each otherafter an absence, or when friendliness is prob-lematic, as when a client interrupts a con-versation between two professionals to askone of them a question. The smile is an evolu-

3 Peter Goldenthal, Lou Zambello, and DonnaBrown planned and conducted this study.


tionarily designed signal to smooth interac-tion among members of a species who mustcooperate in group living.

The Emotional Hypothesis

In our studies smiling had a weaker andmore erratic association with happiness thanit had with social interaction. Although peoplesmiled when shouting, jumping, and gestur-ing after bowling a spare or a strike or whencheering their hockey team's goals, they didnot smile much when they had gotten a goodbowling score and were alone or had not yetturned to face their friends, or when theywere walking down the street alone on a niceday. Given the very strong link between smil-ing and the experience of pleasant emotionsposited by both the popular culture and the100-year-old Darwinian tradition, our failureto document this link convincingly demandsexplanation. We will discuss three attempts toreconcile our results with the happiness hy-pothesis.

First, we may have found more smilingwhen people were social than when they werehappy because our independent variableswere not equated for strength. If bowling,hockey games, and walks on fine days do notproduce strong pleasant emotions, they couldnot be expected to produce much smilingeither. We believe that in general, smiling as adisplay of happiness is relatively infrequentin daily life, partly because the strong emo-tions that may be needed to elicit it are alsorare. However, this is not a plausible explana-tion for the present results. We picked set-tings and activities that were likely to pro-duce variations in positive and negative emo-tions. Our observations suggest that strongpositive and negative emotions were producedin these settings, but that they were expressedwithout a consistent relationship to smiling.The positive and negative exclamations dur-ing bowling were one indicator of the strengthof the emotions produced. Not getting a spareor a strike led to fewer positive exclamationsand more negative exclamations, head shakes,and tight-lipped displays. Thus variations inscore produced variations in emotional expres-sions. Although we did not systematicallycollect data on emotional displays among

hockey fans or among pedestrians, our casualobservations suggest that hockey fans ex-pressed their excitement, joy, and approval byjumping up and down, clapping, and scream-ing, whereas pedestrians celebrated spurts ofpleasant weather, after days of Ithaca'sdrizzly gloom or oppressive heat and humid-ity, by walking with a lilt, whistling, andhumming, not by smiling,

A second possibility is that although smil-ing did indeed have a strong association withhappiness, our subjects masked their emo-tional expressions according to cultural dis-play rules, to confuse their audiences, or tocomply with felt normative pressures (Ek-man, 1972). They refused to smile when theyfelt good and used smiling as a mask for nega-tive emotional expressions when they felt bad.However, this too is an unlikely explanationfor our results. In both the bowling andhockey settings, subjects were very expres-sive, probably much more so than in othersettings of daily life. For example, the fre-quency of positive and negative exclamationsamong bowlers and the strong association ofpositive exclamations with good scores attestto the freedom of this setting from constraintson emotional expression. In addition, as men-tioned earlier, bowlers smiled least when theywere alone or were facing away from fellowbowlers and were thus under the least pres-sure to use display rules.

A third reason why our results linking smil-ing and happiness were not as strong as wehad expected is that we may have been mis-led by the phrase emotional expression. Per-haps because they have underemphasized thefunctions of emotional displays, earlier com-mentators, including Darwin (1872/1965)and Ekman (1972), have used the phraseexpression to mean the outward manifestationof an internal state. While this definition iscorrect, it is incomplete. To the extent thatsmiling is linked with happiness, it is anevolutionarily adapted signal that informsother members of the species about thesender's emotional state in order to influencetheir behavior. Thus, we should expect smil-ing, like other primate emotional displayssuch as fear (appeasement) or anger(threat), to be shown to a recipient and to beless frequently seen in the absence of an audi-

1552 ROBERT K. KRAUT AND ROBERT E. JOHNSTON

ence. This interpretation is consistent withour data. Smiling had its strongest associa-tion with emotion-producing conditions whensubjects were communicating emotions in thepresence of others through additional displayslike positive and negative exclamations, but itwas only weakly associated with emotion-producing conditions when subjects were so-cially uninvolved. These tentative results sug-gest the hypothesis for further research thatemotional displays, in general, should bemore frequent and more intense in the pres-ence of others, although this trend could bemodified by display rules (Ekman, 1972).

A Comparison of the Social andEmotional Hypotheses

The social and emotional hypotheses aboutthe causation of smiling are not incom-patible with each other. As we have sug-gested, social contact may sometimes producehappiness, which in turn may lead to smiling.In addition, people experiencing happinessmay show it more in the presence of others.

It is also possible that emotional and socialmotivations both independently produce fa-cial displays involving what we have termeda smile, but the morphology of the displaysmay differ. One possibility is that the mouthregion has a different shape when expressinghappiness, friendliness, or appeasement. Bran-nigan and Humphries (1972) and Grant(1969) have described several differentsmiles. Our own attempts to distinguishclosed- from open-mouth smiles suggest thatthey have different underlying motivations,with the closed-mouth smile more purely so-cial and the open-mouth smile more playfuland possibly more emotional. Moreover, evenif the mouth is the same, other componentsof the face may differ. The whole face, ratherthan just the smile per se, carries informationabout emotion (Ekman et al., 1972; Ekman,Friesen, & Tomkins, 1971). Although thesmiling mouth may be the most salient com-ponent of a happiness display, other com-ponents may often co-occur and differentiatehappiness, friendliness, or appeasementsmiles. Further research on facial displaysneeds to move to a finer level of descriptionand categorization in an attempt to differ-

entiate the varieties of smiles and facial dis-plays. Ekman and Friesen's recent work(1978) is a step in this direction.

A Comparison oj the Ethological and theExpressive Traditions

The ethological approach to nonverbal com-munication, from which the social hypothesisderives, differs from other approaches in sev-eral ways. Most important, it has remainedfirmly wedded to evolutionary theory and,as a result, has stayed concerned with thefunctions of nonverbal displays and theirsocial consequences. This concern with func-tions and consequences was clearly presentin Darwin's original work (1872/1965; seehis discussion of the principle of antithesis),but was lost as the study of emotional expres-sion passed through experimental psychology.The expressive approach to nonverbal be-havior, from which the emotional hypothesisabout smiling derives, has focused on the cor-respondence between individuals' internalstates and their facial and other expression.As a result it has often embedded the studyof nonverbal behavior in individualistic psy-chology by treating individuals as sociallyencapsulated. It is true that the expressiveapproach has studied communication in thelimited sense of establishing that informationabout emotions can be transmitted throughfacial expression. With its almost exclusivereliance on the recognition experiment, how-ever, this approach has not shown that peopleuse information from facial expressions indaily life.

The effects of this neglect, while perhapsunintentional, are large. Ekman et al.(1972) , after reviewing more than 100studies on the facial expression of emotionpublished in the 100 years since Darwin'soriginal work (1872/1965), found only a fewthat investigated causation or production ofemotional displays and none on the effects ofemotional expression on subsequent socialinteraction.

In contrast to the expressive approach, theethological approach to human nonverbal be-havior treats the individual as part of a socialnetwork and examines the interactions be-tween people and the effects of nonverbal be-


havior on others. By studying usage, thisapproach guides investigators toward a care-ful, descriptive analysis of the situations inwhich smiling occurs. The types of situationwe studied were chosen to compare the emo-tional and social hypotheses about the causesof smiling, but we, along with others, havenot looked at its effects. Indeed, we know ofno research on human nonverbal communica-tion from any tradition that has simulta-neously studied the social and motivationalcauses, the morphology, and the social con-sequences of a human display, as Hooff hasdone with chimpanzee bared-teeth displays.This holistic approach is the direction, wethink, that future work on human nonverbalcommunication should take.

References

Blalock, H. M. Social statistics. New York: McGraw-Hill, 1972.

Blurton-Jones, N. G. Categories of child-child inter-action. In N. G. Blurtonjjones (Ed.), Ethologicalstudies of child behavior. Cambridge, England:Cambridge University Press, 1972.

Brannigan, C. R., & Humphries, D. A. Human non-verbal behavior, a means of communication. InN. G. Blurton-Jones (Ed.), Ethological studies ofchild behavior. Cambridge, England: CambridgeUniversity Press, 1972.

Connolly, K., & Smith, P. K. Reactions of pre-schoolchildren to a strange observer. In N. G. Blurton-Jones (Ed.), Ethological studies of child behavior.Cambridge, England: Cambridge University Press,1972.

Darwin, C. The expression of the emotions in manand animals. Chicago: University of Chicago Press,1965. (Originally published, 1872).

Eibl-Eibesfeldt, I. Similarities and differences betweencultures in expressive movements. In R. A. Hinde(Ed.), Non-verbal communication. Cambridge,England: Cambridge University Press, 1972.

Eibl-Eibesfeldt, I. The expressive behavior of thedeaf- and blind-born. In M. Von Cranach andI. Vine (Eds.), Social communication and move-ment. New York: Academic Press, 1973.

Ekman, P. Universal and cultural differences infacial expressions of emotions. In J. K. Cole (Ed.),Nebraska Symposium on Motivation (Vol. 19).Lincoln: University of Nebraska Press, 1972.

Ekman, P. Cross cultural studies of facial expression.In P. Ekman (Ed.), Darwin and facial expression.New York: Academic Press, 1973.

Ekman, P., & Friesen, W. V. Unmasking the face.Englewood Cliffs, N.J.: Prentice-Hall, 1975.

Ekman, P., & Friesen, W. V. The facial action coding

system. Palo Alto, Calif.: Consulting Psychologists'Press, 1978.

Ekman, P., Friesen, W. V., & Ellsworth, P. Emotionin the human face. New York: Pergamon Press,1972.

Ekman, P., Friesen, W. V., & Tomkins, S. S. Facialaffect scoring technique: A first validity study.Semiotica, 1971, 3, 37-38.

Fraibcrg, S. Insights from the blind. New York:Basic Books, 1977.

Grant, E. C. Human facial expression. Man, 1969, 4,52S-S36.

Hooff, J. A. R. A. M. van. Facial expressions inhigher primates. Symposia of the Zoological So-ciety of London, 1962, 8, 97-125.

Hooff, J. A. R. A. M. van. A comparative approachto the phylogeny of laughter and smiling. In R. A.Hinde (Ed.), Non-verbal communication. Cam-bridge, England: Cambridge University Press, 1972.

Hooff, J. A. R. A. M. van. A structural analysis ofthe social behavior of a semi-captive group ofchimpanzees. In M. Von Cranach and I. Vine(Eds.), Social communication and movement. NewYork: Academic Press, 1973.

Izard, C. The face of emotion. New York: Applcton-Century-Crofts, 1971.

Kendon, A., & Ferber, A. A description of somehuman greetings. In R. P. Michael and J. H. Crook(Eds.), Comparative ecology and behavior ofprimates. New York: Academic Press, 1973.

Levcnthal, H., & Mace, W. The effects of laughter onevaluation of a slapstick movie. Journal of Per-sonality, 1970, 38, 16-30.

Mackey, W. C. Parameters of the smile as a socialsignal. Journal of Genetic Psychology, 1976, 129,125-130.

Spitz, R. A., & Wolf, K. The smiling response: Acontribution to the ontogenesis of social relations.Genetic Psychology Monographs, 1946, 34, 57-125.

Sroufe, L. A., & Waters, E. The ontogenesis of smil-ing and laughter: A perspective on the organizationof development in infancy. Psychological Review,1976, 83, 173-189.

Tomkins, S. S. Affect, imagery and consciousness:Vol. 1. The positive affects. New York: Springer,1962.

Tomkins, S. S., and McCarter, R. What and whereare the primary affects? Perceptual and MotorSkills, 1964, IS, 119-158.

Vinacke, W. E. The judgment of facial expressions bythree national-racial groups in Hawaii: I. Cau-casian faces. Journal of Personality, 1949, 17, 407-429.

Vinacke, W. E., & Fong, R. W. The judgment offacial expressions by three national-racial groups inHawaii: II. Oriental faces. Journal of Social Psy-chology, 1955, 41, 184-195.

White, W. F. Street corner society (2nd ed.). Chi-cago: University of Chicago Press, 1955.

Received September 14, 1978

Documents

Social and Emotional Messages of Smiling: An Ethological Approach