Psychological Bulletin1978, Vol. 85, No. 4, 845-857

Gender Effects in Decoding Nonverbal Cues

Judith A. HallThe Johns Hopkins University

This article summarizes results of 7S studies that reported accuracy for malesand females at decoding nonverbal communication. The following attributes of thestudies were coded: year, sample size, age of judges, sex of stimulus person(s),age of stimulus person(s), and the medium and channel of communication (e.g.,photos of facial expressions, filtered speech). These attributes were examined inrelation to three outcome indices: direction of effect, effect size (in SD units),and significance level. Results showed that more studies showed female advan-tage than would occur by chance, the average effect was of moderate magnitudeand was significantly larger than zero, and more studies reached a conventionallevel of significance than would be expected by chance. The gender effect forvisual-plus-auditory studies was significantly larger than for visual-only andauditory-only studies. The magnitude of the gender effect did not vary reliablywith sample size, age of judges, sex of stimulus person(s), or age of stimulusperson(s).

The study of people's ability to judge themeanings of nonverbal cues of emotion has along history in social psychology, dating tothe second decade of this century. The firstquestion asked was whether people couldrecognize nonverbally expressed emotions atall, and this was followed by the search forcorrelates of judging ability. Gender was oneof the first attributes of judges to be exam-ined in relation to judging ability.

Gender has not always been an importantvariable in psychological research. In thestudy of nonverbal communication, however,gender was considered important from thestart, because of the predictions that could bemade based on gender role stereotypes andfolk beliefs about "woman's intuition." Re-searchers of nonverbal communication haveclearly felt that the comparison of males'and females' performances is theoretically

The author was formerly named Judith HallKoivumaki.

The author extends many thanks to RobertRosenthal for his encouragement and advice, and tohim, Bert F. Green, Jr., and Stephen M. Kosslynfor criticisms of earlier versions of this article.

Requests for reprints should be sent to Judith A.Hall, Department of Psychology, The Johns Hop-kins University, Baltimore, Maryland 21218.

interesting and not just a more exhaustiveway of reporting the data.

Despite the intuitive and theoretical im-portance of learning about differences be-tween the sexes in ability to judge nonverbalcommunication, there exists considerable un-certainty over the facts. Coleman (1949, p.1) said the existing literature showed "nomajor sex differences . . . in the identificationof emotions," but went on to say that"women tend to be slightly superior to themen in identification of facial expressions."Weisgerber (1956) said the results in theliterature were not clear, and Davitz (1964,p. 20) said that "the research evidence aboutsex differences in sensitivity to facial expres-sion is consistently contradictory." Tagiuri(1969) said the literature was "somewhatcontradictory" (p. 406) but that Jenness's(1932b, p. 339) conclusion that "womenslightly excel men as judges of facial expres-sions of emotion" might be "barely right."Tagiuri also suggested (as did Davitz, 1964)that the gender effect may be moderated bythe sex of the communicator. In the contextof examining judgments of facial expressionsamong preliterate people, Ekman and Friesen(1971, p. 126) said, "No differences betweenmale and female subjects were expected, and

Copyright 1978 by the American Psychological Association, Inc. 0033-2909/78/8504-0845$00.75

845

846 JUDITH A. HALL

no such differences had been found in theliterate culture data." Here, unfortunately, itis unclear precisely what literate culture datawere being referred to. Finally, Gitter, Kozel,and Mostofsky (1972) said the results ofsex analyses in the literature were mixed.

Maccoby and Jacklin (1974) reviewed sexdifferences in "empathy" and concluded that,despite the stereotype that girls are more"tuned in" to what other people are thinkingand feeling than boys are, there is "no cleartendency" for females to be more sensitiveto social cues (p. 211), and "neither sex hasgreater ability to judge the reactions and in-tentions of others in any generalized sense"(p. 214). Maccoby and Jacklin's definition ofempathy allowed for the inclusion of studiesusing stimuli other than nonverbal cues anddependent measures other than accuracy atjudging. In fact, of the more than 20 studiesthey reviewed, only two occur in the data setanalyzed in the present article. The point isnot that Maccoby and Jacklin drew the wrongconclusion1 but that their conclusion wasphrased in language that further clouds ourunderstanding of gender differences in abilityto decode nonverbal cues. Further, as has beennoted by Block (1976), Maccoby and Jack-lin's use of probability level as an index ofoutcome precludes a full understanding of thedirections and magnitudes of the effects.

In a recent article, Hoffman (1977) drewattention to the heterogeneous nature of thestudies included by Maccoby and Jacklinunder their heading of empathy. In his arti-cle, Hoffman made the important distinctionbetween empathy in the sense of vicariousresponding and empathy in the sense of rec-ognition of affect. He defined recognition ofaffect as affective perspective taking andunder that heading included studies measur-ing a variety of more cognitive abilities, in-cluding verbal report of how a story char-acter feels (e.g., Deutsch, 1975), set toperceive anger in faces (not specificallyaccuracy at perceiving such cues; Hebda,Peterson, & Miller, 1972), and decoding ofnonverbal cues of emotion (e.g., Gitter, Mos-tofsky, & Quincy, 1971). Thus, his recognitionof affect category, like Maccoby and Jack-lin's, was heterogeneous and also more in-clusive than the category of decoding non-

verbal cues that is the subject of the presentarticle. Hoffman, on the basis of his recogni-tion of affect category, concluded that "whenencountering someone in an emotional situa-tion, both sexes are equally adept at assessinghow that person feels" (p. 716), though sev-eral of his studies did not involve directjudging of persons but a general understand-ing of emotions or social situations, and onedid not measure judging accuracy of any kind(Hebda et al., 1972). Finally, Hoffman'sconclusion of "no difference" was based on avery incomplete review of the relevant litera-ture, especially in the area of decoding non-verbal cues: Only two of the many publishedstudies on decoding are to be found in Hoff-man's review (Gitter et al., 1971; Walton,1936).

The present article represents an attemptto summarize all the available studies bearingon the question of gender differences in abil-ity to decode nonverbal cues of emotion. Theusual interpretative verbal summary is sup-planted here, as much as possible, by quanti-tative analyses of the relationship of direc-tion of effects, effect sizes, and significancelevels to attributes of the studies in the dataset. The use of effect sizes in summarizingresearch has recently been demonstrated inreviews by Rosenthal and Rosnow (1975) ofattributes of volunteer subjects, by Rosenthal(1976) of outcome of studies on self-fulfillingprophecy, and by Smith and Glass (1977) ofpsychotherapy outcome studies. In each ofthese cases, the calculation of effect sizesadded appreciably to the author's ability todetect consistencies in the literature.

Method

Description of Data Set

The data set consists of studies that were eitherpublished or known to be submitted for publicationwhen this article was written. All the studies mea-sured ability to make correct judgments of emo-tions or states that were communicated or indicatedby nonverbal cues conveyed via face, body, or voicetone. The following were not included: (a) unpub-lished data, (b) studies that assessed attention or

1 It is true, however, that Maccoby and Jacklindid not examine all the studies that would havemet their definition of empathy.

GENDER EFFECTS 847

interest instead of accuracy of judging, (c) studiesthat assessed accuracy of judging personality, (d)studies in which informative verbal cues were in-cluded among the stimuli, (e) studies that did notpresent any results according to gender, and (f)studies in which a describable and quantifiable cri-terion of judging accuracy was not available. De-scribaible criteria of accuracy were researcher opinionabout the feeling expressed in each stimulus, thecommunicator's intention as to what feeling wasexpressed in each stimulus, or modal response (insome studies involving open-ended responses) .2

In all of the studies in the data set, groups ofsubjects served as judges of nonverbal expressivestimuli. These stimuli consisted of face, body, orvocal cues (alone or in combination) presented in avariety of media: drawings, photos, films, video-tapes, content-standard speech, randomized-splicedspeech, and electronically filtered speech. In themethod of content-standard speech, the speakersrecite meaningless or affectively ambiguous material,varying the expression to suit the intended emotion.In the randomized-spliced method (Scherer, 1971),the audiotape is cut into small pieces and randomlyreassembled, creating scrambled speech. Electronicfiltering (Rogers, Scherer, & Rosenthal, 1971) re-moves critical bands of frequencies, making thevoice muffled and the words unintelligible.

The expressive stimuli were generated in a varietyof ways. By far the most common method was toask the sender (s) to express various emotions whilethese expressions were photographed, audiotaped,and so forth. The early studies often employedstandard sets of facial-expression stimuli generatedin this way. "Spontaneous" expressions were elicitedin some studies by presenting subjects with emotion-laden stimuli and recording their unrehearsed re-sponses. For example, Coleman (1949, Studies 21-24in Table 1 below) used stimuli such as a snake anda crushed snail; Buck (1976, Studies 43-49) andZuckerman, Hall, DeFrank, and Rosenthal (1976,Study S3) used affectively arousing slides or video-tapes. Other stimuli were photos cut out of maga-zines (e.g., Vinacke, 1949, Study 25) , voices ofpeople who were either rested or tired (Fay & Mid-dleton, 1940, Study 57), voices of people who wereeither lying or telling the truth (Fay & Middleton,1941, Study 58), and drawings of facial expressions(e.g., Moss, 1929, Study 10).

The stimuli were presented individually and sub-jects made a judgment immediately following thepresentation of each stimulus. Exposure times andinterstimulus intervals were often not reported inthis literature. Judgments were typically multiplechoice, with the response alternatives describing thepossible names of the emotion (e.g., happiness, fear,disgust), the interpersonal situation represented inthe stimulus (e.g., ordering food in a restaurant,telling the truth or lying), or the nature of the eventeliciting the nonverbal expression in the stimulus(e.g., car accident, pleasant adult-child interaction).The number of response alternatives ranged from 2to over 25. Other judging formats included pointingto a photograph after hearing a story (Ekman &

Friesen, 1971, Studies 36 & 37), picking from anarray the two faces that expressed the most similaremotions (Fields, 1953, Study 26), open-ended re-sponses (Gates, 1923, Study 1), and rating degreesof friendliness (Zaidel & Mehrabian, 1969, Study69).

In all but one of these studies, gender effects werereported or commented on. The exception was thestudy by Zuckerman and Przewuzman (Note 1,Study 54); in that study, the data-analytic modelmade it clear that gender effects were examined, andsince no significant effect was reported, it was in-ferred that no significant effect was obtained.

Many studies that were examined did not reportor comment on gender effects. There is no way ofknowing if such effects were never looked at or ifthey were looked at, found to be nonsignificant, andthen forgotten. If the latter is the case, then it isconceivable that the data set is biased in having adisproportionate number of studies showing largeand/or significant effects. Whether there might bea bias to overrepresent significant effects favoringone sex over the other is a different question. Sucha bias seems unlikely, but if present would, ofcourse, create a greater problem for inference inthis article than would a bias to include significanteffects ignoring direction.

Each entry in the data set represents an inde-pendent sample or group of samples, except forStudies 16 and 56 (Dusenbury & Knower, 1938,1939), Studies 50 and 64 (Rosenthal, Hall, DiMat-teo, Rogers, & Archer, in press), and Studies 65 and66 (Rosenthal et al., in press). In these cases, thesame samples were given two different tests andwere entered separately in the data set in order notto sacrifice results for the different tests. In addi-tion, in the book by Rosenthal et al. (in press),gender effects were reported for many independentsamples. In these cases, the reported results of allsamples of the same age group that were admin-istered the same test were pooled and not entered asindividual samples, as is explained in Footnotes f-iin Table 1.

Attributes Recorded

For all of the studies in the data set, the followingattributes were recorded and later coded into thecategories explained in Table 1: year, sample size, age

- One study (Walton, 1936) involved four tests,only one of which measured decoding of nonverbalcues; the other three involved understanding theaffective connotations of lines and colors. To thisauthor's reading, Walton's statement of "no sex dif-ferences" was made in the context of one of theseother tests, though he was very imprecise about thisand there is room for disagreement about what hemeant. Hoffman (1977) included the study in hisreview of "recognition of affect," citing the "nodifference" result, but it is not included in thepresent review for the reason stated above.

848 JUDITH A. HALL

Table 1List of Studies, Their Attributes, and Results

Studynumber Study

Age ofN subjects

Sex ofsender (s)

Age ofsender (s) Channel

Direc- Effecttion (in SD)

Visual mode123456789

101112

1314

15

16

17

18

19

20

2122232425262728293031

32

33

34

35

363738

3940

Gates (1923)Allport (1924)Buzby (1924)Buzby (1924)Fernberger (1927)Fernberger (1927)Fernberger (1927)Hunt (1928)Guilford (1929)Moss (1929)Kanner (1931)Kellogg & Eagleson

(1931)Jenness (1932a)Carmichael, Roberts, &

Wessell (1937)Carmichael, Roberts, &

Wessell (1937)Dusenbury & Knower

(1938)Dickey &

Knower (1941)Dickey &

Knower (1941)Dickey &

Knower (1941)Dickey &

Knower (1941)Coleman (1949)Coleman (1949)Coleman (1949)Coleman (1949)Vinacke (1949)Fields (1953)Frijda (1953)Vinacke & Fong (1955)Weisgerber (1956)Weisgerber (1956)N. Levy &

Schlosberg (1960)Staffieri &

Bassett (1970)Staffieri &

Bassett (1970)Staffieri &

Bassett (1970)Staffieri &

Bassett (1970)Ekman & Friesen (1971)Ekman & Friesen (1971)Gitter, Mostofsky, &

Quincy (1971)Izard (1971)Gitter, Black, &

Mostofsky (1972a)

458—207501770743724430

151,000

409332

536348

292

388

233

70

389

238

1337769

100600"20040

330"95

100390

60

49

60

59

13018980

26848

1»44444444441'

44

4

4

2

2

3

3

44444444444

1

1

1

1

1"41

44

211111131322

11

1

3

3

3

3

3

22113323212

1

2

2

1

332

33

222222232322

22

2

2

2

2

2

2

22222"223222

2

1

2

1

332

22

221111112122

24

4

3

2

2

2

2

33332282242

2°

2°

2«

2°

222

22

113311131333

33

3

3

3

3

3

3

32223212333

3

3

2

1

331

33

————————.00

—.14.53

.48*

.02

.05

1.02*

—

—

—

—

.10-.31—

-.12.14

-.15.00

-.02.67*.40*.11

.40*

.69*

-.60*

—

———

.30*—

GENDER EFFECTS 849

Table 1 (continued)

Studynumber Study N

Age of Sex of Age ofsubjects sender (s) sender (s) Channel

Direc- Effecttion (in SD)

Visual mode (continued)

41

42

43444546474849SO

51

5253

54

55

Gitter, Black, &Mostofsky (1972b)

Buck, Miller, &Caul (1974)

Buck (1976)Buck (1976)Buck (1976)Buck (1976)Buck (1976)Buck (1976)Buck (1976)Rosenthal, Hall,

DiMatteo, Rogers, &Archer (in press)

Rosenthal et al.(in press)

Sweeney & Cottle (1976)Zuckerman, Hall,

DeFrank, & Rosenthal(1976)

Zuckerman &Przewuzman (Note 1)

Zuckerman &Przewuzman (Note 1)

160

64

6010

723531010

275

92

10060

77

92

4

4

43444441

4

44

1

4

3

3

33333332

2

33

3

3

2

2

22222222

2

22

2

1

2

3

333333

. 33

3

93

2

2

3

3

31121333

3

33

1

3

.26

—

.54*——

——1.86*—.03

.30

1.31*.65*

—.60

Auditory mode

56

5758596061626364

65

66

67

68

Dusenbury &Knower (1939)

Fay & Middleton (1940)Fay & Middleton (1941)Pfaff (1954)Pfaff (1954)Beldoch (1964)Dimitrovsky (1964)P. K. Levy (1964)Rosenthal et al.

(in press)Rosenthal et al.

(in press)Rosenthal et al.

(in press)Rosenthal et al.

(in press)Rosenthal et al.

(in press)

388

4647

20110389

22474

275

119

119

80

131

4

44424141

3

3

4

4

Visual-plus-auditory

69

70

71

Zaidel &Mehrabian (1969)

Gitter, Kozel, &Mostofsky (1972)

Zuckerman, Lipets,Koivumaki, &Rosenthal (1975)

72

183

30

4

4

4

3

33113332

2

1

1

1

mode

3

2

3

2

22222221

2

2

2

2

2

2

2

5

55°555556

6

6

6

6

7

7

7

3

23312323

3

3

3

3

3

3

3

.34*

-.17

—*—

-.21.38*

-.14.31*

.03

.29

.67*

.32

3.28*

.46*

1.12*

(table continued)

850

Table 1

Studynumber

(continued)

Study

JUDITH A.

Age ofN subjects

HALL

Sex ofsender (s)

Age ofsender (s) Channel

Direc- Effecttion (in SD)

Visual-plus-auditory mode (continued)

72

73

74

75

Rosenthal et(in press)

Rosenthal et(in press)

Rosenthal et(in press)

Rosenthal et(in press)

al.

al.

al.

al.

200

581

1,725

237

1

3

4

2

2

2

2

2

2

2

2

2

7

7

7

7

3

3

3

3

.78'*

.568*

,36h*

.611*

Note. The coding in the table columns is as follows: For Age of subjects, 1 = preschool and grade school,2 = junior high school, 3 = high school, 4 = college and older; for Sex of sender(s), 1 = male, 2 = female,3 = both male and female; for Age of sender(s), 1 = children, 2 = college and older, 3 = both childrenand college and older; for Channel, 1 = drawings of faces, 2 = photos of faces, 3 = moving face or faceand body (film or videotape), 4 = photos or film of hands and arms, 5 = standard content speech, 6 = elec-tronically filtered and randomized-spliced (scrambled) speech, 7 = both 5 or 6 and 1, 2, 3, or 4, 8 = both2 and 3, 9 = photos of faces, postures, hands and arms, legs and feet; for Direction, 1 = no difference,2 = male advantage, 3 = female advantage; for Effect, positive values indicate female advantage, negativevalues indicate male advantage." Ages 3-14 years.b Best estimate.0 "Pictures."d Children whose ages were unspecified.6 Vocal responses indicating truth telling or lying.' Mdn of 10 samples; N is the sum of all 10 samples; test of significance performed with N — 2 df.* Mdn of 12 samples; N is the sum of all 12 samples; test of significance performed with TV — 2 df.h Mdn of 34 samples; N is the sum of all 34 samples; test of significance performed with N — 2 df.1 Mdn of 2 satnples; TV is the sum of the 2 samples; test of significance performed with N — 2 df.* p < .05, two-tailed.

group of judges, sex of stimulus person(s), agegroup of stimulus person(s), and the channel ofcommunication and medium (e.g., facial expres-sions shown on videotape). The following outcomeswere recorded and coded: direction of the gendereffect,3 magnitude of the effect (if reported orcalculable), and significance level (if known).

Definition and Computation oj Effect Size

Effect size tells how large an effect is, disregardingits significance level and its A'. The effect size esti-mate used here is d. It is a measure, in standarddeviation units, of the degree of departure from thenull hypothesis and is defined as the differencebetween the means of the two groups divided bytheir common standard deviation (Cohen, 1969).

A few studies reported effect sizes in terms of d(SD units). In others, it was possible to compute d,using formulas given by Friedman (1968) and Cohen(1969), if any of the following were provided: F,i, U, critical ratio, point-biserial r, or z; means, Ns,and standard deviations for males and females; orproportions of males and females who correctlyjudged each stimulus. In each case the effect sizeestimate was signed after computation to show thedirection of the effect, with positive values indicating

female advantage and negative values indicatingmale advantage.

Since d is affected by the overall amount ofvariability in the data, in poorly controlled or lessprecise studies the signal (mean difference) may beundetectable because of too much noise (variabil-ity). Thus d may tell us as much about the natureof the study itself as about the true magnitude ofthe effect under examination, since it is affectedboth by variation between groups and by variationamong members. Trends in the obtained magnitudeof effects over time thus might be attributableeither to a reduction in variability due to method-ology (reduced measurement error, sampling frommore homogeneous populations) or to an actualchange in the magnitude of the effect due possiblyto some sociological or educational changes thatcould differentially affect the groups' performanceon such tasks.

3 As Table 1 shows, for two studies, the phrase"no difference" was literally true (d=00 SD). Inall other studies showing "no difference," the authormerely stated there was no difference or little dif-ference without providing information on direc-tion or magnitude.

GENDER EFFECTS 851

Table 2Summary of Attributes of Studies

Modality

Attribute

Median yearMedian NAge of subjects"

Sex of sender (s)a

Age of sender (s)a

Visual(N = 55)

1955115

10/2/3/40(18/4/5/73)

15/13/27(27/24/49)

3/47/5(5/85/9)

Auditory(N = 13)

1964119

2/1/2/8(15/8/15/62)

5/2/6(38/15/46)

1/12/—(8/92/-)

Visual plusauditory(tf = 7)

1976200

1/1/1/4(14/14/14/57)

— /5/2(-/71/28)—111—

(— /100/—)

All(N = 75)

1964124

13/4/6/52(17/5/8/69)

20/20/35(27/27/47)

4/66/5(5/88/7)

0 Entries are counts in each of the coding categories described in Table 1. Percentages are given inparentheses.

Results uli consisting of college-age or older stimulusTable 1 lists all the studies in the data set, persons (mainly college age). More studies

with attributes and outcomes numerically used b°th male and female stimulus personscoded. Codes for the numeric categories are t^an either sex alone.given in the note to Table 1. Table 2 pro- Table 3 shows the proportion of studies invides frequencies for each category of each which females were better judges of nonverbalcoded attribute, separately for each of the cues than males were. These proportions arethree communication modalities (visual, audi- defined in two ways: (a) including and (b)tory, and visual plus auditory). Studies using excluding studies showing "no difference."only auditory cues were conducted more These proportions are given both for all stud-recently than studies using only visual cues, ies and for only those studies for which theand studies involving both modalities tended effect size was known. It can be seen that theto be done much more recently. For all three inclusion of the 14 studies showing "no dif-modalities together, studies were most fre- ference" when only the direction of the genderquently performed with college-age and older effect was known resulted in the lowest pro-subjects (mainly college age) and used stim- portion of studies showing female advantage

Table 3Proportion of Studies Showing Female Advantage

Direction of Size ofeffect known effect known

Definition % Proportion % Proportion

Female advantagevs. male advantageand no difference 68 51/75" 78 36/46b

Female advantagevs. male advantage,excluding no difference 84 51/61° 82 36/44d

" x2(l) = 9.01, p < .01 (expected values set at .50 of N).b x"(l) = 13.59, p < .001 (expected values set at .50 of N).« x2(l) = 26.23, p < .001 (expected values set at .50 of N).d X2(l) = 16.57, p < .001 (expected values set at .50 of N).

852 JUDITH A. HALL

Table 4Magnitude of Effect (in SD units) Favoring Females

Modality

Measure

M (unweighted)95% confidence

interval (SD units)MdnN of subjects

Visual(N = 29)

.32

.14-51.26

5,671

Auditory(N = 10)

.18

-.02-.39.30

1,545

Visual plusauditory(N = 7)

1.02

.08-1.97.61

3,028

Total(N = 46)

.40"

.22-.S8.32

10,244

• Weighted M = .34 SD.

(68%). These 14 studies reporting "no dif-ference" probably did in fact show one sex toexceed the other in accuracy; that is, inmost cases the difference was probably notexactly zero. Based on the other proportionsshown in Table 3, the figure of 68% is prob-ably quite an underestimate of the true pro-portion. Regardless of the means of figuringthe proportion, Table 3 reveals that the pro-portion of studies showing female advantagesignificantly exceeded chance.

In Table 4, effect sizes are shown for the46 studies for which an effect size estimatewas available or could be computed. Belowthe mean effect size, the table shows the95% confidence interval for the mean gen-der effect, in SD units. These confidence

Table SNumbers of Studies Reaching ConventionalLevel of Significance in Favor of Females(p < .05, two-tailed)

Definition % Proportion

All studiesStudies for which

effect size was knownStatistically

significant studies

31

48

96

23/75

22/46

23/24

65.62'*

61.09b*

18.38°*

8 Based on expected values of 5 and 70, or .07and .93 of N, to meet minimum cell-size require-ment of chi-square test; expected values wouldotherwise have been .025 and .975 of N.

b Based on expected values of 5 and 41, or .11and .89 of N; expected values would otherwise havebeen .025 and .975 of N.

0 Expected values set at .50 of N.* p < .001.

intervals generally put the population meanabove .00, indicating that the true effect is infavor of females. The average effect size(.40 SD) is of moderate magnitude (Cohen,1969) and is very significantly greater than.00, *(45) = 4.38, p < .0001. The effect sizefor the visual-plus-auditory category is sig-nificantly greater than the mean of the twoother modes, *(43) = 3.23, p - .002. Thesestudies tended to be very recent and prob-ably employed more precise measuring in-struments than did earlier studies. Thenumber of stimuli used in these visual-plus-auditory studies tended to be large (oftenover 100), and the greater reliability of suchsets of stimuli would allow for larger effectsto be detected. In addition, it can be notedthat the visual-plus-auditory stimuli probablyhad more ecological validity than did thevisual or auditory stimuli alone. The greaterrealism of these stimuli may have resulted ineffect sizes that are more representative ofgender effects in everyday judgments of non-verbal cues than are the effects obtained forthe single modalities.

Turning to significance level, we see inTable 5 that the proportion of studies reach-ing significance (p <. .05) in favor of femalesfar exceeded chance in three alternativeanalyses. When significance level was exam-ined separately for the three communicationmodalities, the percentages (proportions)reaching significance in favor of femaleswere as follows: visual—20% (11/55), audi-tory—38% (5/13), visual plus auditory—100% (7/7). For all 75 studies, the propor-tion showing significant female advantage was

GENDER EFFECTS 853

Table 6Correlations of Results of Studies with Study Attributes

Result of study

Attributeof study

YearNSex of

sender (s)Age of

sender (s)ModalityAge of

subjects

Size ofeffect

.28 (44)*-.07 (44)

.02 (26)

-.OS (43)-.14 (37)

.06 (44)

Direction ofoutcome(Index 1)

.05 (59)

.21 (59)

.05 (28)

-.10 (54)-.09 (52)

-.13 (59)

Direction ofoutcome(Index 2)

.15 (73)

.12 (72) '

.21 (38)

-.04 (68).05 (66)

.02 (73)

Significanceof effect

.34 (73)**

.04 (72)

.21 (38)

-.08 (68).15 (66)

-.10 (63)

Note. The coding in the table is as follows: For Size of effect, higher values indicate greater female advantage;for Direction of outcome (Index 1), 1 = male advantage, 2 = female advantage; for Direction of outcome(Index 2), 1 = male advantage and no difference, 2 = female advantage; for Significance of effect, 1 = p> .05, 2 = p < .05; for Sex of sender (s), 1 = male sender (s) only, 2 = female sender (s) only; for Age ofsender(s), 1 = child(ren), 2 = adult(s); for Modality, 1 = visual modality, 2 = auditory modality; forAge of subjects, 1 = preschool and grade school, 2 = junior high school, 3 = high school, 4 = college andolder. Degrees of freedom for the correlations are in parentheses.*p < .10.

** p < .005.

12 times what would be expected by chance,and the combined probability was .00003,two-tailed.

As suggested above, the 75 studies mayoverrepresent significant effects if studiesshowing such effects were more likely to bepublished than studies yielding nonsignificanteffects. A quick calculation reveals that over300 more studies showing nonsignificant gen-der effects (p > .05) would have to existsomewhere in order for x2 to fall to .00. Thisseems unlikely. Nevertheless, there may stillbe some bias toward overrepresentation ofsignificant effects in the 75 studies reviewedhere. However, such a bias cannot explainwhy so many of the published studies haveshown females to outperform males.

Table 6 displays correlations of study at-tributes (dichotomously coded in some in-stances, as explained in the table's notes)with results of studies. What is most appar-ent about this table is the generally smallmagnitude of the correlations, which atteststo the robustness of the female advantage.More recent studies tended to show largereffects (and to be more significant, despite anegative r between year and sample size,r (72) = -.38, p = .001). However, the mag-

nitude of the gender effect was not reliablypredicted by sample size, sex of sender (s),age of sender(s), communication modality(visual vs. auditory), or age of subjects. Thelack of relationship between effect size andage of subjects (r = .06) is supported by thefailure of internal analyses reported in someof the studies to detect appreciable interac-tions of sex of subject and age of subject(Dimitrovsky, 1964, Study 62; Rosenthal etal., in press, Studies 72-75).

The slight tendency shown in Table ,6 forthe gender effect to be larger for visual thanauditory stimuli (r = -.14) occurred muchmore powerfully in the study by Zuckerman,Lipets, Koivumaki, and Rosenthal (1975,Study 71), and in the one by Rosenthal etal. (in press, Study 73), where significantinteractions of modality (visual versus audi-tory) and sex of subjects were reported, ineach case showing the gender effect to begreater for visual stimuli than for auditorystimuli.

The correlations in Table 6 showing thatsex of sender made little difference are con-sistent with internal analyses reporting nointeractions of sender's and judge's sex insome of the studies in the data set (Dimitrov-

854 JUDITH A. HALL

sky, 1964, Study 62; Dusenbury & Knower,1938, Study 16; Ekman & Friesen, 1971,Study 37; Gitter, Black, & Mostofsky,1972a, 1972b, Studies 40 & 41; Rosenthal etal., in press, Studies 65-68; Vinacke & Fong,1955, Study 28).

Discussion

The analyses revealed that more studiesshowed female advantage than would occurby chance, the average effect size was signifi-cantly larger than ,00, and more studiesreached the p < .05 level of significance infavor of females than would be expected bychance. The magnitude of the gender effecthas increased with time, most likely due to acombination of more precise measuring in-struments and more powerful data analysis.The effect size was appreciably greater forvisual-plus-auditory studies than for thesingle modalities. The magnitude of the gen-der effect did not vary reliably with samplesize, sex of sender(s), age of sender(s), orage of judges. That the gender effect did notvary appreciably with the sex of the sender isinconsistent with hypotheses put forth byDavitz (1964) and Tagiuri (1969).

The most theoretically interesting findingwas the absence of an interaction of sex ofjudge and age of judge, as shown by the cor-relation of .06 in Table 6 between effect sizeand age of subjects. If the effect grew eithersmaller or larger with age, we would beguided toward one or another set of hypothe-ses regarding the origins of this gender dif-ference. As it is, we can only say that cross-sectionally, the female advantage at judgingnonverbal cues is stable.

The direction of the effect is consistentwith gender role stereotypes (e.g., Broverman,Vogel, Broverman, Clarkson, & Rosenkrantz,1972), and it is possible that females learnearly (very early, judging from the resultsreported here) "how a girl ought to act."This learning would probably not directlyproduce a performance advantage in judgingnonverbal cues, but over time the addedmotivation to relate to others expressivelyand practice at attending to interpersonalexpression might result in females' superiorjudging ability. It is also possible that non-

verbal sensitivity is especially socially adap-tive for females; English (Note 2) andWeitz (1974) have both suggested that theoppressed status of women might make sensi-tivity to nonverbal communication particu-larly important if it allows them to readbetter the wishes of more powerful others. Aperson who has less than optimal socialpower may learn to act on and employ subtlecues in order to effect more social control.However, that young girls should be betterjudges of nonverbal communication thanyoung boys are is inconsistent with this hy-pothesis, unless one seriously believes thatvery young girls as well as women are op-pressed in our society. If either of the social-learning hypotheses is true—the gender rolesocialization hypothesis or the dealing-with-oppression hypothesis—then the gender effectshould diminish if our society is able gradu-ally to shed its ingrained gender role stereo-types and patterns of male dominance. Thisreview should therefore be updated from timeto time.

Another kind of explanation, in its simplestform, would hold that females are "wired"from birth to be especially sensitive to non-verbal cues or to be especially quick learnersof such cues. This would make evolutionarysense, because nonverbal sensitivity on amother's part might enable her to detect dis-tress in her youngsters or threatening signalsfrom other adults, thus enhancing the sur-vival chances of her offspring. Relevant tothis idea, Rosenthal et al. (in press) reportedthat females in grade school, high school, andcollege samples were especially superior atjudging negative affective cues. It would beinteresting, though methodologically difficult,to learn whether other primate females mightalso show superior ability to judge nonverbalcues and to see if their special sensitivity isalso for angry or sad cues.

The present review clarifies a previouslyconfused issue and establishes that on a vari-ety of conceptually similar tasks of decodingnonverbal cues females are reliably more ac-curate than males. The mean difference of.40 SD is not, however, a large difference. Anaverage effect of this size means that theupper 50% of the female distribution ex-ceeds about 65% of the male distribution, or

GENDER EFFECTS 855

that about 21% of the combined area cov-ered by the two populations is not over-lapped (Cohen, 1969). Less than 4% of thevariance in decoding scores is accounted forby gender. Even if only the 23 studies con-ducted since 1970 are considered, the averageeffect of .52 SD is equivalent to an rz of onlyabout .06, However, an effect of this size issaid by Cohen (1969) to be "large enoughto be visible to the naked eye" (p. 24).

Questions for further study are whetherthe difference is ameliorable, and at what agethe difference emerges. It is also important tokeep in mind that every gender difference isa problem in construct validity. Because wedo not experimentally manipulate gender, wehave to deal with many variables that may becorrelated (confounded) with gender, such aspersonality, socialization, and physical char-acteristics. Along these lines, research is cur-rently underway that examines whether sex-role-related personality differences may ac-count for the gender difference in decodingnonverbal cues.

Recent research has indicated that sensi-tivity to nonverbal cues is positively associ-ated with, among other things, interpersonaladequacy, maturity, and achievement poten-tial as measured by the California Psycho-logical Inventory, teaching skill as measuredby supervisors' ratings, and clinical skill asmeasured by peer and supervisor ratings andpatient outcomes (Burruss, 1977; Rosen thaiet al., in press). Since nonverbal sensitivitythus appears to be an asset, perhaps our edu-cational programs should include training innonverbal as well as verbal communicationskills. The results of the present review wouldsuggest that in such a program, special at-tention should be given to improving boys'nonverbal decoding skills.

Reference Notes

1. Zuckerman, M., & Przewuzman, S. J. Decodingand encoding facial expressions in pre-school agechildren. Unpublished manuscript, University ofRochester, 1976.

2. English, P. W. Behavioral concomitants of de-pendent and subservient roles. Unpublished man-uscript, Harvard University, 1972.

ReferencesAllport, F. H. Social psychology. Boston: Houghton

Mifflin, 1924.Beldoch, M. Sensitivity to expression of emotional

meaning in three modes of communication. In J.R. Davitz (Ed.), The communication of emo-tional meaning. New York: McGraw-Hill, 1964.

Block, J. H. Issues, problems, and pitfalls in as-sessing sex differences: A critical review of Thepsychology of sex differences. Merrill-PalmerQuarterly, 1976, 22, 283-308.

Broverman, I. K., Vogel, S. R., Broverman, D. M.,Clarkson, F. E., & Rosenkrantz, P. S. Sex-rolestereotypes: A current appraisal. Journal of SocialIssues, 1972, 28, S9-78.

Buck, R. A test of nonverbal receiving ability: Pre-liminary studies. Human Communication Research,1976, 2, 162-171.

Buck, R., Miller, R. E., & Caul, W. F. Sex, person-ality, and physiological variables in the com-munication of affect via facial expression. Journalof Personality and Social Psychology, 1974, 30,587-S96.

Burruss, J. A. Evaluating therapists' effectiveness.Unpublished doctoral dissertation, Harvard Uni-versity, 1977.

Buzby, D. E. The interpretation of facial expres-sion. American Journal of Psychology, 1924, 35,602-604.

Carmichael, L., Roberts, S. O., & Wessell, N. Y. Astudy of the judgment of manual expression aspresented in still and motion pictures. Journal ofSocial Psychology, 1937, 8, 115-142.

Cohen, J. Statistical power analysis for the behav-ioral sciences. New York: Academic Press, 1969.

Coleman, J. C. Facial expressions of emotion. Psy-chological Monographs, 1949, 63(1, Whole No.296).

Davitz, J. R. A review of research concerned withfacial and vocal expressions of emotion. In J. R.Davitz (Ed.), The communication of emotionalmeaning. New York: McGraw-Hill, 1964.

Deutsch, F. Effects of sex of subject and storycharacter on preschoolers' perceptions of affectiveresponses and intrapersonal behavior in storysequences. Developmental Psychology, 1975, 11,112-113.

Dickey, E. C., & Knower, F. H. A note on someethnological differences in recognition of simu-lated expressions of the emotions. American Jour-nal of Sociology, 1941, 47, 190-193.

Dimitrovsky, L. The ability to identify the emo-tional meaning of vocal expressions at successiveage levels. In J. R. Davitz (Ed.), The communi-cation of emotional meaning. New York: Mc-Graw-Hill, 1964.

Dusenbury, D., & Knower, F. H. Experimentalstudies of the symbolism of action and voice: I.A study of the specificity of meaning in facialexpression. Quarterly Journal of Speech, 1938, 24,424-435.

Dusenbury, D., & Knower, F. H. Experimentalstudies of the symbolism of action and voice: II.

856 JUDITH A. HALL

A study of the specificity of meaning in abstracttonal symbols. Quarterly Journal of Speech, 1939,25, 67-75.

Ekman, P., & Friesen, W. V. Constants across cul-tures in the face and emotion. Journal of Person-ality and Social Psychology, 1971, 17, 124-129.

Fay, P. J., & Middleton, W. C. The ability tojudge the rested or tired condition of a speakerfrom his voice as transmitted over a public ad-dress system. Journal of Applied Psychology, 1940,24, 645-650.

Fay, P. J., & Middleton, W. C. The ability tojudge truth-telling, or lying, from the voice astransmitted over a public address system. Journalof General Psychology, 1941, 24, 211-215.

Fernberger, S, W. Six more Piderit faces. AmericanJournal of Psychology, 1927, 39, 162-166.

Fields, S. J. Discrimination of facial expression andits relation to personal adjustment. Journal ofSocial Psychology, 1953, 38, 63-71.

Friedman, H. Magnitude of experimental effect anda table for its rapid estimation. PsychologicalBulletin, 1968, 70, 245-251.

Frijda, N. H. The understanding of facial expres-sion of emotion. Acta Psychologica, 1953, 9, 294-362.

Gates, G. S. An experimental study of the growth ofsocial perception. Journal of Educational Psychol-ogy, 1923, 14, 449-461.

Gitter, A. G., Black, H., & Mostofsky, D. Race andsex in the communication of emotion. Journal ofSocial Psychology, 1972, 88, 273-276. (a)

Gitter, A. G., Black, H., & Mostofsky, D. Race andsex in the perception of emotion. Journal of So-cial Issues, 1972, 28W, 63-78. (b)

Gitter, A. G., Kozel, N. J., & Mostofsky, D. I. Per-ception of emotion: The role of race, sex, andpresentation mode. Journal of Social Psychology,1972, 88, 213-222.

Gitter, A. G., Mostofsky, D. I., & Quincy, A. J., Jr.Race and sex differences in the child's perceptionof emotion. Child Development, 1971, 42, 2071-2075.

Guilford, J. P. An experiment in learning to readfacial expression. Journal of Abnormal and SocialPsychology, 1929, 24, 191-202.

Hebda, M. E., Peterson, R. A., & Miller, L. K. Ag-gression anxiety, perception of aggressive cues,and expected retaliation. Developmental Psychol-ogy, 1972, 7, 85.

Hoffman, M. L. Sex differences in empathy andrelated behaviors. Psychological Bulletin, 1977, 84,712-722.

Hunt, T. The measurement of social intelligence.Journal of Applied Psychology, 1928, 12, 317-334.

Izard, C. E. The face of emotion. New York: Ap-pleton-Century-Crofts, 1971.

Jenness, A. Differences in the recognition of facialexpression of emotion. Journal of General Psy-chology, 1932, 7, 192-196. (a)

Jenness, A. The recognition of facial expressions ofemotion. Psychological Bulletin, 1932, 29, 324-350.(b)

Kanner, L. Judging emotions from facial expres-sions. Psychological Monographs, 1931, 41(3,Whole No. 186).

Kellogg, W. N., & Eagleson, B. M. The growth ofsocial perception in different racial groups. Jour-nal of Educational Psychology, 1931, 22, 367-375.

Levy, N., & Schlosberg, H. Woodworth Scale valuesof the Lightfoot pictures of facial expression.Journal of Experimental Psychology, 1960, 60,121-125.

Levy, P. K. The ability to express and perceivevocal communications of feeling. In J. R. Davitz(Ed.), The communication of emotional meaning.New York: McGraw-Hill, 1964.

Maccoby, E. E., & Jacklin, C. N. The psychologyof sex differences. Stanford, Calif.: Stanford Uni-versity Press, 1974.

Moss, F. A. Your mind in action: Applications ofpsychology. Boston: Houghton Mifflin, 1929.

Pfaff, P. L. An experimental study of the communi-cation of feeling without contextual material.Speech Monographs, 1954, 21, 155-156. (Abstract)

Rogers, P. L., Scherer, K. R., & Rosenthal, R. Con-tent filtering human speech: A simple electronicsystem. Behavior Research Methods and In-strumentation, 1971, 3, 16-18.

Rosenthal, R. Experimenter effects in behavioralresearch (Enlarged ed.). New York: Irvington,1976.

Rosenthal, R., Hall, J. A., DiMatteo, M. R., Rogers,P. L., & Archer, D. Sensitivity to nonverbal com-munication: The PONS Test. Baltimore, Md.:Johns Hopkins University Press, in press.

Rosenthal, R., & Rosnow, R. L. The volunteer sub-ject. New York: Wiley, 1975.

Scherer, K. R. Randomized splicing: A note on asimple technique for masking speech content.Journal of Experimental Research in Personality,1971, S, 155-159.

Smith, M. L., & Glass, G. V. Meta-analysis of psy-chotherapy outcome studies. American Psycholo-gist, 1977, 32, 752-760.

Staffieri, J. R., & Bassett, J. E. Birth order andperception of facal expressions. Perceptual andMotor Skills, 1970, 30, 606.

Sweeney, M. A., & Cottle, W. C. Nonverbal acuity:A comparison of counselors and noncounselors.Journal of Counseling Psychology, 1976, 23, 394-397.

Tagiuri, R. Person perception. In G. Lindzey & E.Aronson (Eds.), Handbook of social psychology(2nd ed., Vol. 3). Reading, Mass.: Addison-Wes-ley, 1969.

Vinacke, W. E. The judgment of facial expressionsby three national-racial groups in Hawaii: I.Caucasian faces. Journal of Personality, 1949, 17,407-429.

Vinacke, W. E., & Fong, R. W. The judgment offacial expressions by three national-racial groupsin Hawaii: II. Oriental faces. Journal of SocialPsychology, 1955, 41, 185-195.

GENDER EFFECTS 857

Walton, W. E. Empathic responses in children.Psychological Monographs, 1936, 48(1, Whole No.213).

Weisgerber, C. A. Accuracy in judging emotionalexpressions as related to college entrance testscores. Journal of Social Psychology, 1956, 44,233-239.

Weitz, S. (Ed.). Nonverbal communication: Read-ings with commentary. New York: Oxford Uni-versity Press, 1974.

Zaidel, S. F., & Mehrabian, A. The ability to com-municate and infer positive and negative atti-

tudes facially and vocally. Journal oj Experi-mental Research in Personality, 1969, 3, 233-241.

Zuckerman, M., Hall, J. A., DeFrank, R. S., &Rosen thai, R. Encoding and decoding of spon-taneous and posed facial expressions. Journal ojPersonality and Social Psychology, 1976, 34, 966-977.

Zuckerman, M., Lipets, M. S., Koivumaki, J. H., &Rosenthal, R. Encoding and decoding nonverbalcues of emotion. Journal of Personality and SocialPsychology, 1975, 32, 1068-1076.

Received May 12, 1977 •

Mark I. AppelbaumElliot AronsonPierce BarkerHerbert BensonAllen E. BerglnT. Berry BrazeltonAnthony BrykCharles M. ButterRobert R. CarkhuffPeter L. CarltonRussell M. ChurchGerald L. CloreBarry E. CollinsC. Keith ConnersAnne ConstantinopleRichard DarlingtonBarbara DohrenwendDonald D. DorfmanWard EdwardsLeon EisenbergAllen FleishmanJosef E. GaralDaniel GlaserPaul C. ClickHarold Goodglass

Editorial Consultants for This Issue

Richard L. GorsuchJohn M. GottmanJ. P. GullfordLawrence V. HarperRichard J. HarrisJohn L HornPaul J. HummelDouglas N. JacksonGideon KerenMelvin KonnerMichael J. LambertGeorge LevingerPeter M. LewinsohnMichael LewisJoseph LoPiccoloPaul ManowitzJames Leslie McCaryStanley MilgramJerome L. MyersMelvin NovickJum C. NunnallyJ. C. OgilvieEllis PageAllan PalvloRobert M. Pruzek

Robert A. RescorlaAlexander RichRoss RlzleyAryeh RouttenbergZick RubinBarry R. SchlenkerFrank L. SchmidtDavid J. SchneiderJohn W. SchneiderDean Keith SimontonPaul SlovicDonald P. SpenceRoss StagnerRichard M. SteersJune Lorin TappShelley E. TaylorB. J.WeinerHerbert I. WeisbergJay M. WeissDavid E. WileyLeland WilkinsonMarian WinterbottomKenneth WitteAubrey J.YatesRobert Thomas Zentall