Recognition of facial expressions of emotions in school-age children:the intersection of perceptual and semantic categories

S Vicari1, J Snitzer Reilly2, P Pasqualetti3, A Vizzotto4 and C Caltagirone4,5

IRCCS, Ospedale Bambino Gesu`1, Santa Marinella, Rome, Italy; Department of Psychology2, College of Sciences, San Diego StateUniversity, San Diego, CA, USA; Associazione Fatebenefratelli per la Ricerca3, Rome, Italy; IRCCS, Clinica Santa Lucia4, Rome, Italy;II Universita’ di Roma “Tor Vergata”5, Rome, Italy

Vicari S, Reilly JS, Pasqualetti P, Vizzotto A, Caltagirone C. Recognition of facial expressions ofemotions in school-age children: the intersection of perceptual and semantic categories. ActaPædiatr 2000; 89: 836–45. Stockholm. ISSN 0803–5253

Recognition of the facial expressions of emotions is a critical communicative system early indevelopment and continues to play an important role throughout adulthood. In the past, the resultsof developmental studies of emotional facial recognition have often conflicted. The present studywas designed to examine the development of emotional facial recognition in a large sample ofschool-aged children (n = 120, ages 5–10 y). In particular, we investigate whether emotioncategories, i.e., those based on the visual spatial parameters of facial expression, develop in asimilar fashion to those that also recruit lexical knowledge of emotion terms. We have found twodistinct patterns of development and we suggest that these different profiles are a consequence ofthe very different cognitive abilities that they recruit.

Conclusion: Emotion cognition is a variegated domain which is differentially related to such areas ofcognition as visuo-spatial and lexical semantic abilities.

Key words: Children, emotions, facial expression

Stefano Vicari, Servizio di Neurologia e Riabilitazione, IRCCS, Ospedale Pediatrico BambinoGesu, Lungomare Guglielmo Marconi 36, I-00058 Santa Marinella, Rome, Italy (Tel.�39 0766524 4258, fax.�39 0766 524 4244, e-mail. vicari@opbg.net)

Recognition of emotional facial expression:the emergence

The ability to decode emotional facial expression hasimportant implications for emotional development andis particularly relevant during the infant’s first socialinteraction. According to the Darwinian perspective, theability to recognize emotional facial expression isinnate. Studies in a range of literate and pre-literatecultures independently conducted by Ekman & Friesen(1) and Izard (2) have provided convincing evidencethat facial expressions for the basic emotions areuniversal. Moreover, studies of neonates demonstratethat these specific configurations of facial muscles arepresent at birth (3, 4). Several authors studying visualpreference and habituation-recovery paradigms havedemonstrated that, from 3 mo on, infants can visuallydiscriminate adult facial expressions of happiness,anger, fear, surprise, sadness, and disgust when theyare shown pictures of these expressions in slides or films(5–7).

However, as several authors have noted (8), thesefindings should be interpreted with caution. There are,for example, limitations in the infant’s contrast sensi-

tivity and scanning ability that make it unlikely thatinfants much younger than 4 mo would have access tothe information in a face that would convey expression.Moreover, in these studies the different facial expres-sions were produced by a single model; it was thus notclear whether responses to each facial expression weredue to the perception of changes in the inner facialfeatures or to the perception of the different emotionsthat were expressed. In fact, using a single model, facialexpression was the only physical feature that variedfrom expression to expression. Using different modelsas stimuli, others have shown that infants from 4 to 9 mocan discriminate facial expressions of happiness andperhaps fear, sadness or surprise (8–10). Furthermore,Serrano et al. (11), using several adult female models,demonstrated that infants 4–6 mo old can discriminateand generalize facial expressions of anger, fear andsurprise across models. Although the results of thesestudies suggest that the infants became habituated to theexpressive attributes of the faces, they do not excludethe possibility that infants notice changes in physicalfacial characteristics per se. That is, the role of emotionand its significance is unclear in this context.

Reading emotions accurately on the faces of others is

2000 Taylor & Francis. ISSN 0803-5253

Acta Pñdiatr 89: 836±45. 2000

the result of a complex process developed over anextended period. Indeed, there are data to indicate thatthe ability of children to decode facial expressionsimproves with age during the period from 3 to 9 y(2, 12–14). Furthermore, the ability to recognize differ-ent emotional expressions does not emerge as acomplete package. Rather, recognition of some emo-tions precedes that of others. There are data, forinstance, to indicate that positive emotions are learnedfirst (12, 14, 15); specifically, happiness is the firstemotion to be accurately recognized, followed bysadness and, later, by anger and surprise (2, 5, 13).Although several researchers have broadly confirmedthese findings across the last 20–30 y, it is crucial tonote that developmental studies have differed greatly inmethodology. In particular, the types of responserequired from subjects have varied enormously. Insome investigations children were asked to generate alabel for a particular facial expression of emotion posedby an actor in a slide or a photograph (2). In otherstudies, children have had to choose from a list ofdifferent emotion-words to identify the expression inphotographs (16). With few exceptions (1), theseparadigms have conflated the perceptual ability ofchildren to recognize emotional expressions with theirsemantic knowledge of labels for emotions. Thus,performance on these tasks may reflect lexical abilitiesas much as, if not more than, a specific impairment inemotion recognition.

Acquisition of emotion wordsIn contrast to the very early emergence of emotionalfacial expression, naturalistic studies of languagedevelopment (17–19) have shown that children beginto acquire emotion words some time around theirsecond birthday. Smiley & Huttenlocher (20) proposethat the initial bases for these terms are behaviouralcues. For example, children first use the word “sad” inconjunction with a person or child crying; “happy” firstoccurs in conjunction with a person smiling. In addition,the data from deaf children of deaf parents acquiringASL as their first language, Reilly et al. (19) show adecentring in the use of emotion terms. Specifically, thetoddlers first use emotion labels to refer to their ownemotional state (at about 20–24 mo). Later, such labelsoccur in reference to others’ emotions, and finally, atabout 3 to 3.5 y of age, children will use emotion termsfor imaginary characters or in reference to those in astory. With respect to the order of acquisition foremotion terms in English, in their study of preschool-aged children, Smiley & Huttenlocher (20) found thefollowing sequence of emergence: happy, sad, mad,scared and surprised.

As noted above, the methodological approaches fortesting emotion recognition in children vary widely withrespect to what subjects must know to perform the tasks

successfully. Briefly, the production of verbal labels foremotion expressions requires an adequate lexicon ofemotion terms, understanding their meaning and asso-ciating the label with a particular facial expression. Incontrast, matching facial expressions to descriptions ofsituations with a high emotional content does notnecessarily require specific lexical knowledge but,instead, specific visuo-perceptual abilities, such as thediscrimination and generalizability noted in infancyplus an understanding of the expression’s affectivemeaning (12).

Broadly supporting a Piagetian perspective in whichlabels would be acquired following the development ofits referent concept, previous studies (21) have shownthat the ability to recognize affective facial expressionsprecedes the ability to produce verbal labels for theseemotions. However, it is not yet clear if the difference inthese two developmental timelines is solely temporal,with parallel developmental sequences in which lan-guage follows cognitive/perceptual development. Itcould also be the case that the two systems differqualitatitvely in their developmental sequences. Forexample, studies of children’s developing lexicons (22–24) indicate that lexical cateogories are initially con-structed along a variety of different parameters, e.g.,form and perceptual features or alternatively by func-tional similarities. Moreover, they often do not neatlyoverlap those of adults. This is evidenced in the earlyovergeneralizations of toddlers (for example, when achild labels any four-legged, even cows and dinosaurs,“doggie”). With respect to the categorization of facialexpression, Etkoff & Magee (25) have demonstratedthat, at least in adults, emotional expressions areperceived categorically and not as a direct reflectionof their continuous physical properties. It may thus bethat the perceptual and semantic categories for emotionsfollow significantly different paths of development. Inthis case, perceptual categories would be those cogni-tive classificatory systems that allow an individual torecognize, identify and generalize from one stimulus toanother independently of linguistic knowledge. Incontrast, semantic categories would be those categorieswhich are associated with a lexical or linguistic label.These classes may or may not overlap.

To tease apart the development of the visuo-perceptual categories of emotion from their role inlabeling emotional expressions, the present studyexamined the recognition of emotional facial expres-sions and emotion labels by 120 children (ages 5–10).Subjects were given four tests to assess their ability torecognize facial expressions of the six basic emotions:happiness, surprise, sadness, disgust, fear and anger.The first two tasks were designed to evaluate perceptualabilities: Task IA involved choosing a target expressionfrom four alternatives and Task IB was matching facialexpressions in a memory condition. The second set oftasks assessed verbal labels for emotional expressions;Task IIA involved matching facial expressions to verbal

ACTA PÆDIATR 89 (2000) Facial expressions of emotions in school-age children837

labels after hearing a short vignette and Task IIBconsisted of producing verbal labels for photographs ofdifferent facial expressions. With these data, we hope toclarify the relationship of perceptual and semanticcategories in the development of emotion recognition.

MethodParticipantsParticipants were 120 children (63 boys and 57 girls)attending a public kindergarten and elementary schoolin Rome. Children included in the study were perform-ing at grade level and were predominantly middle class.Table 1 gives age and gender information.

ProceduresChildren were first given a screening battery to verifytheir ability to recognize faces and to discriminate non-emotional facial expressions; then, having met inclu-sionary criteria, four measures to assess recognition ofemotional expressions were administered.

Screening batteryThe screening battery consisted of two tasks:

1. Benton Facial Recognition Test. We used areduced form of the test developed by Benton et al.(26) to assess basic facial recognition skills. Childrenwere required to perform at age level according to agenorms provided with the test.

2. Facial Miming Matching Test. Developed for thisstudy, the task consists of 18 items each with a targetshowing a child with a particular non-emotional facialgrimace (e.g. a child with mouth wide open) as well asfour photographs of other children of the same sex. Thesubject selected the picture of the only child who wasdemonstrating the same expression. To be included inthe study, children were required to be accurate on all18 items.

These two tasks were introduced to screen outchildren with difficulties in visual perceptual andrecognition abilities. All the children tested performedadequately on these two tasks and were included in thestudy.

Emotional facial expression measuresThe ability to recognize facial expressions of the six

basic emotions—happiness, surprise, sadness, disgust,fear and anger—was investigated by means of fourdifferent tests using sets of black and white photographs(4.5� 6.5 cm) developed and normed by Ekman &Friesen (27). The first two tasks (Task IA:EmotionMatching Taskand Task IB:Emotion Memory Task)were designed to evaluate perception abilities, whereasthe second set of tasks was designed to assess mappingof semantic categories to the different facial expres-sions. Task IIA:Emotion Recognition Taskinvolvessupplying verbal labels for emotional expressions in astory context and Task IIB:Emotion Labelling Taskrequires producing verbal labels for photographs ofdifferent facial expressions. A short description of thesetests is presented below.

Task IA.The Emotion Matching Task. In this task, thechild is asked to choose a match for a target expressionfrom four alternative photographs of expressions posedby actors of the same sex. The target was placedcentrally on a 95� 210 mm paper and the four alter-natives were arranged vertically on a 210� 297 mmsingle white sheet. The picture sequence was rando-mized. The test was presented by placing the target andthe sheet with the alternatives in front of the subject andsaying, “look at this man\woman (point the target),show me that man\woman from these four who feels thesame way (point the alternatives). Don’t look for thesame person but show me who is feeling the same”. Thetest had four initial training items. In the first two, theexaminer (E) showed the correct alternative to the child,without naming the emotion. In the second two trainingtrials, E did not model the correct answer, but if thesubject’s response was wrong, E repeated the instruc-tions. After the fourth item, no prompts were provided.There were 18 scored items, three for each basicemotion.

Task IB. The Emotion Memory Task. The materialswere identical to those of Task IA, but with the order ofitems randomized. The tester presented the target for 5sec, and said “Look at this man\woman”. Then, cover-ing the target and showing the alternative sheet to thesubject: “show me that man\woman from these fourwho is feeling the same as the man\woman I justshowed to you. Don’t look for the same person but showme the one who is feeling the same way”. This task alsoincluded four initial training items. In the first two, theexaminer (E) showed the correct alternative to the child,without naming the emotion. In training trials three andfour, E did not supply the right answer, but if thesubject’s response was wrong, E repeated the instruc-tions. Beginning with the fifth item, no prompts wereprovided. Again, there were 18 scored items, three foreach basic emotion.

Task IIA.The Emotion Recognition Task. Two sets ofsix black and white photographs showing the six basicemotions from the previous tests were used. Subjectswere asked to match each of the emotional expressionswith a short illustrative story of the following sort: “One

Table 1.

Age group (mean; SD) Male Female Total

5 (5.3; 0.3) 11 9 206 (6.6; 0.2) 14 8 227 (7.5; 0.3) 11 9 208 (8.5; 0.3) 10 9 199 (9.3; 0.3) 8 10 18

10 (10.4; 0.2) 9 12 21

838 S Vicari et al. ACTA PÆDIATR 89 (2000)

day Mrs. Patrizia went out and saw a big fierce dogrunning toward her. This made Mrs. Patrizia feelscared.” “One day Mr. Roberto opened his windowand saw an elephant fly by. He was really surprised.”The sex of the character in the story was matched to sexof the model in the set of photographs.

The subject was then asked to pick out the appro-priate photograph: “Can you show me which face Mrs.Patrizia (Mr. Roberto) made when she (he) saw the bigfierce dog (the elephant to fly)?”

Each subject was tested on the 24 vignettes.Task II B.The Emotion Labelling Task. The two sets

of six black and white photographs showing the sixbasic emotions employed in the previous test were usedfor this task. The 24 photographs were shown one at atime and the subject was asked: “How is this personfeeling? What is he\she doing?” Responses were scoredand we considered synonyms for the target word to becorrect, e.g., content for happy. Responses for all taskswere scored as correct or incorrect. For incorrectresponses, the child’s answer was recorded for subse-quent error analysis.

The children were volunteers whose parents gaveparental consent. Participants were individually testedin a quiet room outside the classroom. Each subject wastold that the experimenter was going to show him somepictures, to tell him some stories and ask him what hethought about the stories and pictures. The child, seatedat a table facing the examiner, was presented with thematerials and the tasks were always presented in thesame order (the Benton Facial Recognition test, theFacial Mimic Matching test, the Emotion Matching test,the Emotion Memory test, the Emotion Recognition testand, finally, the Emotion Labelling test). All testingtook place in the same session, with breaks to avoidfatigue and boredom.

ResultsIn order to evaluate the effects ofAGE (a between-subjects factor), type ofTEST(TASK) andEMOTION (bothwithin subjects factors) on the accuracy of responses, ananalysis of variance (ANOVA) for repeated measureswas conducted with three age groups (young, middleand old)� six emotions (surprise, happiness, sadness,fear, anger and disgust)� four tasks (matching, mem-ory, labelling & recognition). Greenhouse-Geissercorrection of degrees of freedom was applied. Maineffects and interactions are reported in Table 2, as areassociated graphics. To better understand these inter-actions, post-hoc analyses were conducted to examinecontrasts between levels of factors, and the Tukeycorrection procedure was applied to protect alpha.Statistical analyses and graphics were executed withSPSS for Windows (7.0).

As seen in Table 2, there are main effects forAGE,TEST and EMOTION. In addition, the interactions

betweenAGE� EMOTION (Fig. 1) andTEST� EMOTION(Fig. 2) were significant. However,AGE� TEST did notreach significance, suggesting that if we collapse acrossemotions there are parallel age-related improvementsacross the different tests. Importantly,AGE� TEST� EMOTION was also significant. To tease apart theseinteractions and to test our hypotheses, we performedseparate analyses for those tasks tapping visual-spatialabilities (Group 1: Matching & Memory) and thoserecruiting lexical abilities (Group 2: Recognition &Labelling).

Results for the Group 1 tasks: Matching & MemoryIn the Group 1 tasks, the three-way interaction(AGE� TEST� EMOTION) is still significant, F = 3.39;G-G df = 9.2/540.8;p< 0.001; this is primarily due tothe ceiling effect for happiness, disgust and surprise in

Table 2.

Source F df Sig.

Tests of within-subjects effectsMeasure:

Greenhouse-GeisserTEST 36.61 2.46 0.000TEST� AGE 1.78 4.93 0.117Error (TEST) 288.40EMOTION 48.55 4.56 0.000EMOTION� AGE 4.32 9.11 0.000Error (EMOTION) 533.20TEST� EMOTION 17.95 10.93 0.000TEST� EMOTION� AGE 4.02 21.86 0.000Error (TEST� EMOTION) 1279.04

Tests of between-subjects effectsMeasure:

Transformed variableIntercept 7017.89 1.00 0.000AGE 37.46 2.00 0.000Error 117.00

Fig. 1. Children’s ability to recognize each emotion at three agelevels considered, collapsing across tasks (* = Happiness,& = An-ger,◊ = Fear,~ = Disgust,* = Sadness,& = Surprise).

ACTA PÆDIATR 89 (2000) Facial expressions of emotions in school-age children839

the matching task, whereas performance on theseemotions is significantly lower in the memory task,especially for the youngest group of children (Fig. 3).

The significant AGE� TEST interaction, F = 3.71;df = 2,117;p = 0.029, indicates that memory improvessignificantly with age; the difference in matching/memory scores of the younger group is twice that ofthe older group (Fig. 4).

Fig. 5 demonstrates the significant effect of memoryon the recognition of all emotions and the strikingparallelism in performance across emotions. The sig-nificance ofEMOTION� TEST interaction, F = 2.6; G-Gdf = 4.6/540.8;p = 0.029 appears to be due specificallyto the lower performance on the memory task ondisgust, sadness and surprise (Fig. 5). In sum, we haveseen similar profiles of performance for the two taskswith a significantly lower performance in the memorycondition.

Results for the Group 2 tasks: Recognition andLabellingIn the second group of tasks, the three-way interaction isnot significant, indicating a similar relation between ageand emotions on these two tests (Fig. 6). However, wedid find a significant AGE� EMOTION interaction,F = 7.91, G-G df = 8.6/501.7;p< 0.001, produced bythe different developmental trajectories for each emo-tion. Specifically, the performance for happiness is flatand virtually at ceiling, and a lower, but almost parallelpath characterizes the responses for sadness. However,steeper developmental trajectories reflect performancefor the other emotions, especially disgust and surprise,and, to a lesser extent, fear, as seen in Fig. 7.

In addition, there was a significantEMOTION� TESTinteraction, F = 19.71; G-G df = 4.3/501.7;p< 0.001

Fig. 2. Children’s ability to recognize the six emotions in thedifferent tasks considered, collapsing across ages (* = Matching,& = Memory,◊ = Recognition,~ = Labelling).

Fig. 3. Children’s performance profiles in Matching and Memory task for each emotion at different ages. (* = Matching,& = Memory).

840 S Vicari et al. ACTA PÆDIATR 89 (2000)

(Fig. 8), which is due to the similar performance onhappiness, sadness and surprise compared to thedivergent performance on Anger and Fear and Disgust.Moreover, in contrast to all other emotions, perfor-mance for Disgust is higher on Recognition than forLabelling.

Finally, comparing the pattern of performance on thevisuo-spatial tasks (Fig. 5) with the pattern of perfor-mance on the lexically based tasks (Fig. 8), we found asignificant interaction betweenKIND OF TEST (Group 1versus Group 2)� EMOTION, F = 24.9; df = 5/595,p< 0.001.

DiscussionThis study was designed to examine the development offacial recognition in school-aged children and toinvestigate specifically whether emotional categories,i.e., those based on the visual spatial parameters of

facial expression, develop in a similar fashion to thosethat also recruit lexical knowledge of emotional terms.Overall, a number of general trends emerged. First ofall, the ability to recognize and label different facialexpressions continues to develop well into the schoolyears and overall performance improves at each agelevel. Although the greatest strides are taken between 5/6 and 7/8, performance on recognition, labelling andmemory have not reached asymptote even in the oldestgroup. Moreover, these emotional abilities do notemerge as a coherent package. Specifically, happinesswas the highest across all tasks, followed by sadness,which was significantly better than anger, fear anddisgust. Finally, recognition of surprise surpassedrecognition of fear. These results reflect the overallpattern for emotions. However, as noted above, if welook at performance within individual tasks, or acrossgroups of tasks, we find that recognition of emotionsvaries widely according to the task. For example,whereas happiness and sadness show generally consis-tent performance across tasks, the three other negativeemotions (fear, anger and disgust) show differentialscores according to the task at hand.

Visuo-spatial tasks: Matching and MemoryIn reviewing the data from the individual tasks, as wesee in Fig. 3, the rank order of emotions for theMatching task is: happiness, disgust, surprise, sadness,anger and fear. Generally, children’s performanceimproves with age, and this is especially clear forsadness, anger and fear. It may well be that therelatively flat developmental profiles for happiness,disgust and surprise reflect ceiling effects for theseparticular facial expressions. Let us for a momentconsider these three emotional facial expressions(happiness, disgust and surprise) against those wherewe see clear development (sadness, anger and fear).Using criteria from Ekman (28), each of the first groupcan be distinguished by its unique mouth pattern.Specifically, a smile is sufficient to distinguish happi-ness, a wide-open mouth with a dropped jaw isdistinctive of surprise and the raised upper lip is uniqueto disgust. In contrast, each facial expression in thesecond group of emotions (fear, anger, and sadness)requires the integration of both the upper and lowerface, that is, the pattern made by contracting the browsand the shape of the mouth. This is not to say that theupper face is not involved in the first group of facialexpressions, but rather that the lower face pattern maybe sufficient to distinguish each of these emotionalexpressions.

Moving on to the Memory task, if we consider therank order of emotions, we see that the patterneffectively mirrors that of Matching, but at a lowerlevel of performance (Figs 3, 5). Also note thesignificant jump in performance from the younger tothe middle age group. We suggest that the consistent

Fig. 4. Children’s performance profiles in Matching and Memorytask at different ages, collapsing across emotions (* = Matching,& = Memory).

Fig. 5. Children’s ability to recognize emotions in the Matching andMemory task, collapsing across ages (* = Matching,& = Memory).

ACTA PÆDIATR 89 (2000) Facial expressions of emotions in school-age children841

delay in performance in the second of these tasks is dueto the effect of the memorial component that improvesconsiderably in this developmental period (29). Thesetwo tasks primarily rely on visuo-spatial discriminationand appear to share the same general developmentalprofile.

Lexically based tasks: Recognition and Labelling

And now, let us consider the Recognition and Labellingtasks, both of which require recruiting lexical labels foremotions. For the Recognition task, overall, we findimproved performance with age (Figs 6, 8). This isespecially true for surprise, which shows markedimprovement between the younger and middle agegroups. However, for anger and fear, although perfor-mance improves, it is quite low across all age groups. Itis important to note that even at age 10, children do notachieve mastery for all the emotions. This is especiallyapparent in the responses for anger and fear. Withrespect to order, happiness and sadness reflect topperformance followed by surprise, disgust, anger andfear.

As for Labelling, again, children perform at ceilingfor happiness and sadness, and this is consistent with theliterature on the acquisition of emotional labels, wherehappy and sad are the first productive labels foremotions to emerge (20). Labelling for the remainingemotions improves with age with the sharpest rises forfear, surprise and disgust. Interestingly, labelling fordisgust is markedly below that of all the other emotions,achieving only 70% correct by age 10. With theexception of the singular performance for disgust, thepatterns of performance for these two lexically driven

Fig. 6.Children’s performance profiles in Recognition and Labelling task for each emotion at different ages (◊ = Recognition,~ = Labelling).

Fig. 7. Children’s ability to recognize each emotion at the three agelevels considered, collapsing across Recognition and Labelling task(* = Happiness,& = Anger, ◊ = Fear,~ = Disgust,* = Sadness,& = Surprise).

842 S Vicari et al. ACTA PÆDIATR 89 (2000)

tasks are virtually identical and, importantly, theirprofiles differ significantly from those of Group 1, thematching and memory tasks (compare Figs 5 and 8).These distinctively different profiles in performanceacross tasks raise an important issue: the relativedifficulty of recognizing emotions at the various agesdepends critically on the task we use to detect them.This notion was also confirmed by the significantinteraction we found between Task and Emotion. Arecent article by Markham & Adams (13) also notedthat children performed differentially on different typesof tasks assessing emotional facial expression, and theyresponded by proposing a hierarchy of difficulty. Theytested children at ages 4, 6 and 8 on four types of tasks:two recognition tasks, one matching discrimination andone situational discrimination, and two labelling tasks,one forced choice and one free labelling task. Similar toour results, they found that accuracy of performanceincreased with age, and they also found a significantinteraction between task and emotion. However, instriking contrast to our results, Markham & Adams (13)state that “the pattern of responses to different emotionswere somewhat similar over tasks” (p. 37). Nonetheless,they found only partial support for their proposedhierarchical scheme.

Qualitative versus quantitative differencesSuch a hierarchy, as Markham & Adams (13) propose,implies that the different tasks line up on a continuum ofdifficulty and therefore the differences in performanceare solely quantitative. In contrast, our findings suggestthat the differential performance according to taskreflects both qualitative and quantitative differences.The difference in emotional profiles for Matching andMemory is quantitative. Although the scores for theMemory Task are significantly lower than for Matching,the patterns for the different emotions are identical, and

the rank order of emotions for both tasks is: happiness,disgust, surprise, sadness, anger and fear, as seen in Fig.5. In contrast, the rank order of performance on theRecognition and Labelling is different: happiness,sadness, surprise, anger, fear and disgust, as seen inFig. 8.

Juxtaposing these two sets of patterns, we see adistinct qualitative difference. These observations areconfirmed by the statistically significant differencebetween the two patterns reflected in the performanceof Group 1 and Group 2 tasks. We propose that thesetwo patterns are due to the recruitment of different kindsof cognition: visuo-spatial and lexical-semantic. More-over, we suggest that these two types of categories candevelop in heterogeneous ways. To explore andexemplify this notion, we present in detail the resultsfor two emotions: sadness and disgust.

First, looking at the profile for sadness (Fig. 9), it isclear that performance improves across all tasks as thechildren get older. Moreover, development is parallelfor all tasks across age. From these data, it appears thatthe differences in tasks are quantitative, that is, some aremore difficult than others, but by about age 10 thechildren have mastered them. However, if we lookinstead at the results for disgust, on a task by task basis(Fig. 10), we see very different patterns of developmentdepending on the type of task.

For this emotion, children do very well on theMatching task, and, as we know from Ekman &Friesen’s work (1, 30), the facial configuration forDisgust is very distinctive. The upper lip is raised, thenose is slightly wrinkled; no other emotional facialexpression shares these features. When we turn to theperformance on labelling and recognition, however, wesee that performance for disgust, especially for theyounger children, is poor. We propose that this widedifference in performance is a direct reflection of the

Fig. 8. Children’s ability to recognize emotions in the Recognitionand Labelling task, collapsing across ages (◊ = Recognition,~ = Labelling).

Fig. 9. Children’s performance profiles to recognize Sadness in thefour different tasks at the three age levels considered (* = Matching,& = Memory,◊ = Recognition,~ = Labelling).

ACTA PÆDIATR 89 (2000) Facial expressions of emotions in school-age children843

very different cognitive bases that these tasks require.These findings strongly suggest that, in this case, lexico-semantic categories do not map directly onto thecomparable visuo-spatial category, but in fact lagseveral years behind. For sadness, it appears that thetwo types of category have significant developmentaloverlap; however, in the case of disgust they show quitedifferent developmental profiles. In sum, these datasuggest that the perception and lexical categories canand, in some cases, do follow quite different routesbefore reaching the integrated adult model.

ConclusionsFacial expression for emotion represents a criticalcommunicative system early in development andcontinues to play a significant communicative rolethroughout adulthood. In the past, in developmentalstudies of emotional facial expression a variety ofmethodologies have been used, often reporting conflict-ing results. The present study included a large sample ofchildren and explored the development of facialexpression recognition using four different tasks. Wefound two distinct patterns of development and wesuggest that these different profiles are a consequence ofthe very different cognitive abilities that they recruit.Finally, our findings raise questions regarding thecommonly used term “emotional cognition”, andsuggest that, rather than a unidimensional construct,emotional cognition represents a variegated domainwhich is differentially related to other areas of cogni-tion, such as visuo-spatial and lexical semantic abilities.

Acknowledgements.—Preparation of this article was supported by a grant(97/02/N010) to Stefano Vicari, M.D. and partially by NIH\NINDS,P50NS22343 and NIH\NIDCD, P50DC01289-0351. A special thanks toall the children and families who participated in this study.

References1. Ekman P, Friesen WV. Constants across cultures in the face and

emotion. J Pers Soc Psychol 1971; 17: 124–92. Izard CE. The face of emotions. New York: Appleton-Century-

Crofts; 19713. Field TM, Woodson RW, Greenberg R, Cohen C. Discrimination

and imitation of facial expressions by neonates. Science 1982;218: 179–81

4. Oster H, Ekman P. Facial behavior in child development. MinnSymp Child Psychol 1978; 11: 231–76

5. Labarbera JD, Izard CE, Vietze P, Parisi SA. Four- and six-month-old infants’ visual responses to joy, anger, and neutralexpressions. Child Dev 1976; 47: 535–8

6. Young-Browne G, Rosenfeld HM, Horowitz FD. Infant dis-crimination of facial expression. Child Dev 1977; 48: 555–62

7. Schwartz GM, Izard CE, Ansul SE. The 5-month-old’s ability todiscriminate facial expressions of emotion. Infant Behav Dev1985; 8: 65–77

8. Nelson CA. The recognition of facial expressions in the first twoyears of life: mechanisms of development. Child Dev 1987; 58:889–909

9. Caron RF, Caron AJ, MacLean DJ. Infant discrimination ofnaturalistic emotional expressions: the role of face and voice.Child Dev 1988; 59: 604–16

10. Ludemann P, Nelson C. Categorical representation of facialexpressions by 7-month-old infants. Dev Psychol 1988; 24: 492–501

11. Serrano JM, Iglesias J, Loeches A. Visual discrimination andrecognition of facial expressions of anger, fear, surprise in four-to six-month-old infants. Dev Psychobiol 1992; 25: 411–25

12. Camras L, Allison K. Children’s understanding of emotionalfacial expressions and verbal labels. J Nonverbal Behav 1985; 9:84–94

13. Markham R, Adams K. The effect of type of task on children’sidentification of facial expressions. J Nonverbal Behav 1992; 16:21–39

14. Boyatzis CJ, Chazan E, Ting CZ. Preschool children’s decodingof facial emotions. J Genet Psychol 1993; 154: 375–82

15. Michalson L, Lewis M. What do the children know aboutemotions and when do they know it? In: Lewis M, Saarnl C,editors. The socialization of emotions. New York: Plenum; 1985:117–39

16. Felleman ES, Barden RC, Carlson CR, Rosenberg L, Masters J.Children’s and adults’ recognition of spontaneous and posedemotional expressions in young children. Dev Psychol 1983; 19:405–13

17. Dunn J, Bretherton I, Munn P. Conversations about feeling statesbetween mothers and their young children. Dev Psychol 1987;23: 132–9

18. Ridgeway D, Waters E, Kuczaj S. Acquisition of emotion-descriptive language: receptive and productive vocabulary normsfor ages 18 months to 6 years. Dev Psychol 1985; 21: 901–8

19. Reilly J, McIntire M, Bellugi U. Faces: the relationship oflanguage and affect. In: Volterra V, Erting C, editors. Fromgesture to language in hearing and deaf children. Berlin, NewYork: Springer-Verlag. Reissued in 1994 by Gallaudet Uni-versity Press, Washington, D.C. 1990

20. Smiley P, Huttenlocher J. Young children’s acquisition ofemotion concepts. In: Saarni C, editor. Children’s understandingof emotions. Cambridge: United Press; 1989: 27–49

21. Klinnert M, Campos J, Sorce J, Emde R, Svejda M. Emotions asbehavior regulators: social referencing in infancy. In: PlutchnikR, Kellerman H, editors. Emotion in early development, vol. 2.The emotions. New York: Academic Press; 1983

22. Nelson K. Structure and strategy in learning to talk. Monographsin the Society for Research in Child Development, 38 (Serial No.149); 1973

23. Fenson L, Camerson SM, Kennedy M. Role of perceptual and

Fig. 10.Children’s performance profiles to recognize Disgust in thefour different tasks at the three age levels considered (* = Matching,& = Memory,◊ = Recognition,~ = Labelling).

844 S Vicari et al. ACTA PÆDIATR 89 (2000)

conceptual similarity in category matching at age two years.Child Dev 1988; 59: 897–907

24. Clark E. The lexicon in acquisition. Cambridge: CambridgeUniversity Press; 1993

25. Etkoff NL, Magee JJ. Categorical perception of facial expres-sions. Cognition 1992; 44: 227–40

26. Benton AL, Hamsher K, Varney N, Spreen O. Test diriconoscimento di volti ignoti. Firenze: Organizzazioni Speciali;1983

27. Ekman P, Friesen WV. Measuring facial movement. EnvironPsychol Nonverbal Behav 1976; 1: 56–75

28. Ekman P. Universal and cultural differences in facial expressionsof emotion. In: Cole J, editor. Nebraska Symposium onMotivation, vol. 9. Lincoln: University of Nebraska Press; 1972

29. Hulme C, Thompson N, Muir C, Lawrence A. Speech rate andthe development of short-term memory span. J Exp ChildPsychol 1984; 38: 241–53

30. Ekman P, Friesen WV. Facial action coding system (FACS). PaloAlto, CA: Consulting Psychologists Press; 1978

Received July 26, 1999; revision received Feb. 7, 2000; acceptedFeb. 16, 2000

ACTA PÆDIATR 89 (2000) Facial expressions of emotions in school-age children845