origin of the happy face advantage | My Assignment Tutor

actapsychologicaELSEVIER Acta Psychologica89 (1995) 149-163Happy face advantage in recognizing facial expressionsT a k a h i r o Kirita a, *, M i t s u o E n d o ba Shokei Women’s Junior College, 4-1O-1 Yurigaoka, Natori, Miyagi, 981-12 Japanb Hachinohe Junior College, 13-384 Mihono, Hachinohe, Aornori, 031, JapanReceived 3 December 1993;revised 6 April 1994;accepted 14 April 1994AbstractIn this study, the origin of the happy face advantage was examined with respect to themode of processing, as well as the spatial characteristics, of happy faces. In Experiments 1and 2, happy and sad schematic faces of equal form deviations from a neutral face wereadopted as stimuli, and the task of categorizing these two facial expressions was given to thesubjects. Experiment 1 demonstrated that happy faces were recognized faster than sad facesonly when they were presented as in an upright position: conversely the sad face advantagewas observed to be slight when the stimuli were presented as inverted. The results ofExperiment 2 showed that the pattern of recognition for happy faces differed in the rightand left visual fields: the affect of inversion was much more pronounced when the happyfaces were presented in the right visual field than when they were presented in the leftvisual field. From these results, it was assumed that while happy faces were likely to berecognized holistically, sad faces were likely to be recognized by analytic mode. Thisassumption was not rejected by the data from Experiment 3 where the uncontrolled happyand sad expressions on real faces were used as stimuli.1. IntroductionIt has been repeatedly observed that happy or smiling faces are recognizedmore quickly and more accurately than any other facial expression (Ekman et al.,1982; Kirouac and Dord, 1983; Ladavas et al., 1980). However, few researchershave attempted to examine the origin of this happy face advantage in therecognition of facial expressions.Exceptionally, Feyereisen et al. (1986) focused their attention on the happy faceadvantage (emotion category effect in their terms) and conducted a few experi-* Corresponding author.0001-6918/95/$09.50 © 1995 Elsevier Science B.V. All rights reservedSSDI 0001-6918(94)00021-8150 T. lO’rita, M. Endo / Acta Psychologica 89 (1995) 149-163ments to explore its origin. The results of their experiments showed that theadvantage of happiness was not limited to the case of facial expressions: this effectwas also found in the categorization task of words expressing happy and sad states.But it was also found that the advantage was greater when happy and sad facialexpressions were used as stimuli. Based on these results, they suggested that thehappy face advantage depended partly on the internal organization of the emotional field, and partly on the spatial characteristics of happy faces. Unfortunately,they did not clarify what aspects of the spatial characteristics of happy facesyielded such an advantage. In this study, we will examine this problem.Primary consideration is likely to be given to the possibility that the happy faceadvantage depends exclusively on a specific local feature. For example, a U-shapedmouth might be responsible for the recognition of happy faces, because only happyfaces have such a salient feature. Our preliminary experiment, however, showedthat there was no difference in reaction times needed for categorizing simpleU-shaped and inverted U-shaped line segments. Rather, this result suggests thatthe U-shaped mouth by itself might have no effect on the happy face advantage.Alternatively, the global characteristics, instead of local feature, of happy facesmay give rise to such an advantage. In this case, however, the question arises as tohow facial expressions should be characterized globally. Although all facial expressions can be described in terms of total deviations from a neutral face, it is unlikelythat only happy faces should deviate so much. Furthermore, such a definitionwould be meaningless unless it is verified that our brain systems actually useinformation concerning global deviations in classifying facial expressions.Yet little has been discovered about the process for recognizing facial expressions. For example, it is still uncertain whether all facial expressions are processedin the same way; or if the way of processing changes depending on the facialexpression. For the latter case, happy faces are likely to be recognized by quite adifferent strategy. Interestingly, the recognition of happy faces has been said toplay an important role in infant-mother communications (Bowlby, 1969). In fact,regardless of their insufficient visual acuities, young infants are known to firstdiscriminate happy faces from all other facial expressions (LaBarbera et al., 1976;Young-Brown et al., 1977; Barrera and Maurer, 1981). Thus, starting with infancy,there is good reason to believe that happy faces are recognized in a different way.So far, we have discussed two possible explanations for the happy face advantage. One is concerned with the spatial factors of happy faces such as local orglobal deviations from a neutral face. The other assumes specific way of processingfor happy faces. These two explanations seem to offer quite different predictionsfor this phenomenon. While the former view would predict that the happy faceadvantage disappears or, at least, is reduced when the spatial deviations arecontrolled, the latter view would predict that this advantage should be influencednot by spatial factors but by the presentation variables such as stimulus orientations, durations and visual fields, which have been assumed to affect the mode ofprocessing; that is, the mode would shift from holistic to analytic processing or viceversa depending on those variables (Sergent and Bindra, 1981).In this study, besides controlling the spatial deviations of happy and sadT. Kirita, M. Endo /Acta Psychologica 89 (1995) 149-163 151schematic faces from a neutral face, we examined the effects of stimulus durations(Experiment 1) and visual fields (Experiment 2), as well as the effect of stimulusorientations on the happy face advantage. Experiment 3 was carried out toreplicate the results of Experiment 2 using uncontrolled happy and sad expressionson real faces.2. Experiment IIn this experiment, several happy and sad schematic faces were used. Theirlocal, as well as global, form deviations from a neutral face were approximatelyequal. The task of categorizing happy and sad faces was given to each subject withthe orientations and the durations of the stimuli being varied.2.1. MethodSubjectsFourteen students of Tohoku University (7 males and 7 females) participated inthe experiment. All subjects were right-handed.Stimuli and apparatusHappy and sad schematic faces, which had been evaluated by Kirita (1994),were used as stimuli. The process of stimulus evaluation was as follows. In hisresearch, Kirita made a standard (neutral) schematic face composed of eyebrows,eyes with pupils, a nose and a mouth (see Fig. 1). To examine the contributions ofthe angle of eyebrows, the location of pupils and the shape of mouth, to theperceived emotions, 22 sample schematic faces were made by systematicallyvarying these variables and printed on separate sheets with a list of 12 emotioncategories. Sixty-six independent subjects were instructed to select one appropriateemotion category for each schematic face from the list.Frequency data matrix was first analyzed by Dual Scaling (Nishisato, 1980).Three orthogonal solutions were obtained, by which both 22 schematic faces and12 emotion categories were weighted. These three solutions could be interpretedas dimensions of emotion i.e. access-avoidance, pleasantness- unpleasantness andtension-relaxation. Then by using a categorical multiple regression analysis(Hayashi, 1952), he examined whether the locations of schematic faces in threedimensional space of emotion could be predicted by their components. For eachdimension, multiple correlation coefficient exceeding 0.9 was obtained, indicatingthat the predictions were successful. From this analysis, it was found that combinations of horizontal or medially downturned eyebrows and U-shaped mouth contributed to happiness. Likewise, mediallyupturned eyebrows and inverted U-shapedmouth made the schematic faces look sad. The location of pupils had little effectfor both emotions.In this experiment, a combination of medially downturned eyebrows and Ushaped mouth was chosen for making happy faces so that their local, as well as152 T. Kirita, M. Endo /Acta Psychologica89 (1995) 149-163J JFig. 1. Standard schematicface(left) and expamplesof happy(middle)and sad (right)faces.global deviations from the neutral face should be approximately equal to those ofsad faces. Five happy and five sad faces were made by mainly varying the locationof pupils. Examples of the stimuli are shown in Fig. 1.The stimuli were presented on the screen of a TV-type tachistoscope (IwatsuISEL: IS701-A). The subjects, with their heads immobilized by a chin rest,observed them at a distance of 100cm. The size of the stimuli was 3.4° x 3.4° . Tworesponse keys connected to the tachistoscope were used to measure reaction times.DesignA 2 x 2 x 2 factorial design was used. The three factors were Emotion (happyand sad), Orientation (upright and inverted), and Duration (100 and 300 msec).ProcedureThe experiment was carried out in a dark room. Before the experiment thesubjects were shown each schematic face for 3 sec to confirm the emotionexpressed by it. Each experimental trial was begun by presenting a small fixationpoint for 2 sec, which was immediately followed by one of the 10 schematic facespresented as either upright or inverted for either 100 or 300 msec. These stimulusdurations were determined in accordance with the implications found in thestudies of laterality that the critical duration dissociating two modes of processingshould be somewhere around 200 msec (see Sergent, 1982,1983). The subjects’ taskwas to decide whether the schematic face expressed happiness or sadness asquickly and accurately as possible. The decision was made by pressing one of tworesponse keys located beneath the index fingers of both hands. One second afterthe subject pressed one of the two response keys, the fixation point for the nexttrial was presented. Each schematic face appeared four times under each condition, thus 160 trials (5 faces x 4 times x 8 conditions) were performed. The orderof presentation was random. Before the experimental session, the subjects weregiven 20 practice trials using the same stimuli. The response keys were counterbalanced across the subjects.2.2. ResultsFor each subject, the mean correct RT and error rate were calculated for eachcondition. Eight correct responses which exceeded the 2 see limit were treated asT. Kirita, M. Endo /Acta Psychologica 89 (1995) 149-163 153,,,,- 20u JLLI< lOI,–zIJJIJJ500U JI–.” ‘ 450STIMULUS DURATION (msec)I00 300happyI |happysadI Ihappyi I I i PRI GHTINVERTED UPRI GHTINVERTEDST I MULUS OR I ENTAT I ON Fig. 2. Percentages of errors (upper panels) and reaction times of correct responses (lower panels) fortwo stimulus durations in Experiment 1.errors. The results are shown in Fig. 2. A three-way repeated measures ANOVA,by Emotion, Orientation and Duration, was performed on the RTs, and separatelyon the error rates.For the RT data, the main effects of Orientation and Duration were significant(F(1,13) = 33.54, p < 0.001, F(1,13) = 10.46, p < 0.01, respectively). Upright faceswere processed faster than inverted faces. Judgments were made faster at 100msec duration than at 300 msec duration. There were two significant interactions,Emotion by Orientation (F(1,13)= 5.58, p < 0.05), and Orientation by Duration(F(1,13) = 5.41, p < 0.05). For the interaction of Emotion by Orientation, ananalysis of the simple main effects of Emotion at two levels of Orientationrevealed that, while happy faces were judged faster than sad faces in an uprightcondition (F(1,13)= 4.41, p = 0.053), sad faces were judged slightly faster thanhappy faces in an inverted condition (F(1,13) = 3.19, 0.05 < p < 0.1). The alternative analysis of Orientation at two levels of Emotion indicated that a greaterinversion effect was found for the happy faces (F(1,13) = 28.59, p < 0.001) than forthe sad faces (F(1,13)= 6.96, p < 0.05). For the interaction of Orientation by154 T. lO’rita,M. Endo /Acta Psychologica 89 (1995) 149-163Duration, an analysis of the simple main effects of Orientation at two levels ofDuration showed that a greater inversion effect was found in the 100 mseccondition (F(1,13) — 78.48, p < 0.001), but this effect only reached marginal significance in the 300 msec condition (F(1,13) = 4.20, 0.05 0.1).Before the analysis, the error rates underwent an arcsine-transformation. Onlythe main effect of Orientation was significant (F(1,13)= 5.35, p < 0.05). Thesubjects made more errors under inverted conditions than in upright conditions.2.3. DiscussionThis experiment demonstrated that, even if the spatial deviations of happy andsad schematic faces were equal, the happy face advantage was found in the uprightcondition. Unexpectedly, when the faces were presented as inverted, a slight sadface advantage was found. These results were not consistent with the view that thehappy face advantage originated in the spatial deviations of happy faces. Instead,the results indicated the happy face advantage should be affected by the presentation modes. It is clearly shown in Fig. 2 that the interaction of Emotion byOrientation derived from the fact that the judgments for happy faces were muchmore influenced by the inversion of the stimuli than for sad faces. A similarpattern was observed for the interaction of Orientation by Duration: a greaterinversion effect was found during a shorter duration.What conclusions can be drawn from these two interactions? As mentionedearlier, the inversion of faces has been assumed to affect the mode of stimulusprocessing: while holistic processing operates predominantly for upright faces,analytic processing takes place for inverted faces (Carey and Diamond, 1977). Thedifference of the inversion effect on the two facial expressions can be explained ifthey were processed by different strategies: happy faces were judged by a holisticmode, but sad faces were judged by a analytic mode. As a consequence, happyfaces should undergo a much greater inversion effect than sad faces because theinversion of happy faces caused a drastic shift in the mode of processing.Likewise, the interaction of Orientation by Duration can be explained bypostulating that the mode of processing changed according to the stimulus durations. Namely, the holistic processing was taken for the stimuli of 100 msecduration whereas the analytic processing operated on the stimuli of 300 msecduration. In fact, Navon (1977) suggested that the global aspects of a visualstimulus would be processed prior to local components, which is quite consistentwith the results that the subject made faster judgments in shorter duration. Butthis is only true for the case of upright faces because the inversion of stimulishould interfere with the holistic processing.However, for the possibility that the mode of processing changes depending onT. Kirita, M. Endo /Acta Psychologica89 (1995) 149-163 155the facial expression, the opposite relation has been implied. Reuter-Lorenz andDavidson (1981) claimed that whereas happy faces were recognized predominantlyby the left hemisphere, sad faces were recognized predominantly by the righthemisphere. By taking into account the notion that the right hemisphere operatesholistically, whereas the left hemisphere works analytically (Bradshow and Nettleton, 1981), one may well think that happy faces are judged by a local feature, suchas a mouth, whereas sad faces are recognized as a global pattern (Ley and Strauss,1986).In any case, if the mode of processing changes according to the facial expression, the interaction of Emotion by Orientation should have been related to thefactor of Duration, which should influence the mode of processing. In thisexperiment, however, neither the two-way interaction of Emotion by Duration(F(1,13) = 1.51, p = 0,24) nor the three-way interaction of Emotion by Orientationby Duration (F(1,13) = 1.375, p = 0.261) were observed. For this point, we have nosatisfactory explanation.In the next experiment, the factor of Visual Field, which is also known toinfluence the mode of processing, and its interaction with Emotion and Orientation, were examined.3. Experiment 2In this experiment, the effect of the right and left visual fields on the categorization task of the two facial expressions was examined together with the factor ofOrientation. Given the hypothesis that happy faces are recognized by the holisticmode, and that sad faces are recognized by the analytic mode, the happy faceadvantage should dearly be found in the left visual field, where the holistic modewould be dominant, whereas in the right visual field, where the analytic modewould be dominant, such an effect should be reduced or absent.The effect of the subjects’ gender was also studied as a between-factor since ithas been claimed that the hemispheric asymmetry should vary between males andfemales in the recognition of facial expressions (Ladavas et al., 1980; Strauss andMoscovitch, 1981).3.1. MethodSubjectsTwenty-eight students and staff members of Tohoku University (14 males and14 females) participated in the experiment. All of them were right-handed.Stimuli and apparatusThe same as in Experiment 1. In this experiment, the schematic faces werepresented in either the right or the left visual fields. The distance between thefixation point and the center of each schematic face was 4.0° .156 T. Kirita, M. Endo / Acta Psychologica 89 (1995) 149-163DesignA 2 X 2 X 2 X 2 factorial design was used. Four factors were Gender (male andfemale) as a between-factor, Emotion (happy and sad), Orientation (upright andinverted), and Visual Field (right and left) as within-factors.ProcedureBefore the experiment, the subjects were shown each schematic face for 3 sec toconfirm the emotion expressed by it. In each experimental trial, first a fixationpoint appeared, then after 2 sec, one of the 10 schematic faces was presentedlaterally for 150 msec as either upright or inverted. This stimulus duration was setto ensure the visibilities of the stimuli in lateral presentation, as well as to preventeyemovements. The subjects’ task was to decide whether the schematic faceexpressed happiness or sadness by pressing one of two response keys as quicklyand accurately as possible. They were also instructed to stare at the fixation pointthroughout the trial. Each schematic face was presented four times in eachcondition, thus 160 trials (5 faces x 4 times x 8 conditions) were performed. Theorder of presentation was random. Response keys were counterbalanced across thesubjects. Before the experimental session, the subjects were given 20 practice trialsusing the same materials.3.2. ResultsFor each subject, the mean correct RT and error rate were calculated for eachcondition. In this experiment, only one correct response which exceeded 2 sec wastreated as an error. A four-way split plot repeated measures ANOVA, by Gender,Emotion, Orientation and Visual Field, was performed on the RTs, and separatelyon the error rates. In no case were there any reliable effects involving Gender sothe data were collapsed across this factor. The results are shown in Fig. 3. Athree-way repeated measures ANOVA was performed on both types of data.For the RT data, the main effect of Orientation was significant (F(1,27) = 42.2,p < 0.001). The interaction of Emotion by Orientation was also significant (F(1,27)= 17, p < 0.001). Further, the three-way interaction, Emotion by Orientation byVisual Field, was found to be significant (F(1,27) = 5.2, p < 0.05). To interpret thisinteraction, the simple interaction of Emotion by Orientation was analyzed for thetwo levels of Visual Field (Keppel, 1991). For the right visual field, the simpleinteraction of Emotion by Orientation was significant (F(1,27) = 22.2, p < 0.001).A further analysis of simple simple main effect revealed that, while happy faceswere judged faster than sad faces in the upright condition (F(1,27) = 4.7, p < 0.05),no emotional effect was found for the inverted condition (F(1,27 = 2.83, p > 0.1).The alternative analysis of the simple simple main effect of Orientation indicatedthat the inversion effect was significant only for happy faces (F(1,27)= 49.9,p < 0.001). For the left visual field, the simple interaction of Emotion by Orientation was not significant. Thus a two-way repeated measures ANOVA was performed on the RTs of the left visual field. The main effects of Emotion andOrientation were significant (F(1,27) = 4.9, p < 0.05, F(1,27) = 30.3, p < 0.001,T. Kirita, M. Endo /Acta Psychologica 89 (1995) 149-163 157r,,-LLILLII – –ZLLIr,,-LJJ03¢./’)I==vLLII–zI.–¢..;,LLIC¢:20VISUAL HEMIFIELDLEFT RIGHT10 happysad0 i !550 sad500t ~ P Y ,UPRIGHT INVERTEDhappysadI IUPRIGHT INVERTEDST I MULUS ORI ENTAT I ONFig.3. Percentagesoferrors(upperpanels)andreactiontimesofcorrectresponses(lowerpanels)fortwovisualfieldsinExperiment2.respectively). Happy faces were processed faster than sad faces irrespective of theirorientations, and inversion effects were found for both emotions.For the error rates, only the main effect of Orientation was significant (F(1,27)= 15.4, p < 0.001). Again, the subjects made more errors under inverted conditions than in upright conditions.3.3. DiscussionAs expected, a three-way interaction, by Emotion by Orientation by VisualField, was found. It is clearly shown in Fig. 3 that the interaction of Emotion byOrientation differed in the two visual fields. In the left visual field, there was nointeraction of Emotion by Orientation: happy faces were judged consistently fasterthan sad faces irrespective of their orientations. On the contrary, there was theinteraction of Emotion by Orientation in the right visual field: the happy faceadvantage was found only in the upright condition. This interaction resulted fromthe fact that while the judgments for happy faces were greatly affected by theinversion of stimuli, the judgments for sad faces were affected very little by it.Taken together, in the conditions which would favor the holistic processing,such as the upright or the left visual field, a happy face advantage could be found.158 T. lOrita,M. Endo /Acta Psychologica89 (1995) 149-163However, when the stimuli were presented upside down to the right visual field,which would favor the analytic processing, there was no emotion effect. Theseresults generally support the hypothesis mentioned above.4. Experiment 3Based on the data from Experiments 1 and 2, we propose that happy faces areprocessed holistically, and sad faces are processed by analytically. Can this proposal be generalized for the recognition of emotions from real faces, or is it onlylimited to the case of schematic faces? If the happy face advantage depends, noton the spatial factors of happy faces, but on the mode of the processing itself, thesame results should be obtained when the uncontrolled facial expressions on realfaces are used as stimuli. To explore this possibility, Experiment 2 was repeatedusing the facial expressions on real faces.4.1. MethodSubjectsTwenty students of Tohoku University (10 males and 10 females) participated inthe experiment. All of them were right-handed.Stimuli and apparatusEight female students were asked to be models and to show happy and sadfacial expressions. Two black-and-white photographs of happy and sad expressionswere taken of each model. Thus 16 slides of faces (8 happy and 8 sad) were made.The slides were back-projected onto a screen. In the center of the screen was afixation point. The size of the stimuli was 9.0° × 6.4°. Each face was presented ineither the right or left visual fields so that, irrespective of its orientation, thedistance between the fixation point and the center of the face should be 5.2° . Withthe chin rest, the subjects observed the stimuli at a distance of 57.3 cm. Tworesponse keys connected to a microcomputer (NEC: PC-9801 VX) were used tomeasure the RTs.DesignA 2 x 2 × 2 factorial design was used. The three factors were Emotion, Orientation, and Visual Field.ProcedureThe experiment was carried out in a dark room. In each trial a warning buzzerwas sounded for 500 msec, then one of the 16 faces was presented laterally aseither upright or inverted for 150 msec. The subjects’ task was to decide whetherthe face expressed happiness or sadness by pressing one of two response keys asquickly and accurately as possible. The subjects were also instructed to stare at thefixation point all through the trial. Each face appeared twice for each condition,T. Kirita, M. Endo /Acta Psychologica 89 (1995) 149-163 159P i .P PLLIl.uI-.zuJt.u2010VISUAL HEMIFIELD LEFTRIGHTsad~ happy(1)EvILlI’-,-ZI.L,Ie,-800 sad700I I IUPRIGHT INVERTED UPRIGHTST I MULUS OR I ENTAT I ONsad!INVERTEDFig. 4. Percentages of errors (upper panels) and reaction times of correct responses (lower panels) fortwo visual fields in Experiment 3.thus 128 trials (8 faces x 2 times × 8 conditions) were carried out. The order ofpresentation was quasi-randomized and response keys were counterbalanced acrossthe subjects. Before the experimental session, 30 practice trials were given to thesubjects using a different set of stimuli.4.2. ResultsFor each subject, the mean correct RT and error rate were calculated for eachcondition. There were 10 correct responses which exceeded 2 sec and they weretreated as errors. A three-way repeated measures A N O V A was performed on theRTs, and separately on the error rates. The results are shown in Fig. 4.For the R T data, the main effects of Emotion and Orientation were significant(F(1,19) = 8.29, p < 0.01, F(1,19) = 37.08, p < 0.001, respectively). No interactionwas found to be significant.For the error data, the main effect of Orientation was significant (F(1,19) = 8.08,p < 0.01) and the main effect of Emotion was marginally significant (F(1,19) = 3.70,0.05 < p < 0.1). Furthermore, the interaction of Emotion by Visual Field was alsomarginally significant (F(1,19) = 3.11, 0.05 < p < 0.1). An analysis of simple maineffects of Emotion at two levels of Visual Field indicated that more errors were160 T.Kirita, M. Endo /Acta Psychologica89 (1995) 149-163found for happy faces than for sad faces in the right visual field (F(1,19) = 4.83,p < 0.05), but no such emotion effect was found for the left visual field (F(1,19) =0.13, p > 0.1). The alternative analysis of Visual Field at two levels of Emotionrevealed that for the happy faces the subjects made slightly more errors in the rightvisual field than in the left visual field (F(1,19)= 3.11, 0.05

QUALITY: 100% ORIGINAL PAPER – NO PLAGIARISM – CUSTOM PAPER

Leave a Reply

Your email address will not be published. Required fields are marked *