Ford & Milosky: Inferring Emotional Reactions 21Journal of Speech, Language, and Hearing Research • Vol. 46 • 21–30 • February 2003 • ©American Speech-Language-Hearing Association1092-4388/03/4601-0021

Janet A. FordLinda M. MiloskySyracuse University

Syracuse, NY

Inferring Emotional Reactionsin Social Situations: Differencesin Children With LanguageImpairment

Anticipating and responding to a partner’s emotional reactions are key compo-nents in the comprehension of daily social discourse. Kindergarten children withlanguage impairment (LI) and age-matched controls (CA) were asked to labelfacial expressions depicting 1 of 4 emotions (happy, surprised, sad, mad) and toidentify those expressions when given a verbal label. Children then chose amongthese facial expressions when asked to infer emotional reactions from stories (3-sentence scenarios) presented in 1 of 3 modalities: verbal, visual, and combined.Although all children were able to identify and label the facial expressions,children with LI had difficulty integrating emotion knowledge with event context inorder to infer a character’s feelings. When these inferencing errors occurred,children in the LI group were more likely to provide emotions of a differentvalence (e.g., substituting happy for mad) than were children in the CA group.Inferencing ability was related to language comprehension performance on astandardized test. The findings suggest that inferencing errors made by childrenwith LI occur during the early stages of social processing and may contribute tosocial difficulties often experienced by this group of children.

KEY WORDS: comprehension, language impairment, emotion, inferences,children

Comprehension deficits experienced by children with language im-pairment (LI) are generally accounted for by assessing literalknowledge of semantics, morphology, and syntax. However, re-

search has demonstrated that impairments in making inferences con-tribute to discourse comprehension difficulties experienced by this groupof children (Bishop & Adams, 1992; Crais & Chapman, 1987; EllisWeismer, 1985). A key component in comprehension of daily social dis-course is the listener’s ability to infer a partner’s emotional reactions,which is the focus of the present study.

It has been suggested (Weaver & Kintsch, 1991) that up to 15 infer-ences must be made for successful comprehension of every statementthat we hear. In order to successfully understand discourse, childrenmust tie utterances together in some coherent fashion, and must alsoelaborate on this information, given their world knowledge. These basicinferencing processes allow individuals to develop a representation thatcan then be modified as more incoming information is received (e.g.,Yuill & Oakhill, 1991).

22 Journal of Speech, Language, and Hearing Research • Vol. 46 • 21–30 • February 2003

There are many different types of inferences thatmust be made in order to develop this representation(see Nicholas & Trabasso, 1980, for a complete tax-onomy). Of these different types, causal inferences (howone event enables another event or affects an individual’smental state) have received particular attention, espe-cially in research regarding narrative development andcomprehension. Causal connections affect what acomprehender judges as important in narrative text(Trabasso & Sperry, 1985). Children’s ability to retellstories is related to the cohesiveness of the causal eventsthat occur within the story (Trabasso, Secco, & van denBroek, 1984). That is, causal connections are mostreadily remembered in narrative retell tasks; they area basic building block for the mental representationsnecessary in the comprehension process.

Given that causal inferences may be based on physi-cal events (e.g., inferring that wet clothes were causedby rain) or mental states (e.g., inferring that forgettingan umbrella might cause someone to become angry), howmight these task demands differ? Physical events areobservable; a child sees someone get wet in the rain,allowing the child to easily make the link between rainand wet clothes in the future. On the other hand, men-tal states are not observable; therefore, it may be moredifficult for the child to link events with mental states.Even when a person directly expresses how he or shefeels, successful comprehension may require the listenerto infer an alternate mental state (e.g., a friend maystate that she is happy to do a favor, but the listenermay need to infer that she feels otherwise). It is not sur-prising that there is evidence to indicate that the abil-ity to make causal inferences based on (unobservable)mental states might be more difficult than those basedon (observable) physical events (Miller & Aloise, 1989).

The presence of a facial expression may facilitatethe ability to infer another’s mental state (Wiggers &van Lieshout, 1985). Facial expressions are an observ-able window into the mental states of others, and read-ing emotions in faces is a basic task involved in the per-ception of people (Walden & Field, 1982). Long beforechildren are able to verbally label emotions, they areresponsive to expressions around them.

Infants as young as 2 months of age can respond tofacial expressions of surprise and happiness by return-ing smiles and participating in vocal play (Flin &Dziurawiec, 1989). By the end of the first year, childrenrely on facial expressions to assist in determining ap-propriate responses (e.g., Klinnert, Emde, Butterfield,& Campos, 1986). Children as young as 2 are able toidentify some facial expressions (sad, happy) through averbal labeling task (Bretherton, McNew, & Beeghly-Smith, 1981). Preschool children are able to use facialexpressions to make inferences regarding basic emotions

(Denham & Couchoud, 1990). When given an event sce-nario, they also are able to predict a character’s emo-tional response by choosing the appropriate facial ex-pression (Denham, 1986).

However, some facial expressions are easier to in-terpret than others. Facial expressions that depict bipo-lar emotions (e.g., happy/sad) may be easier to discrimi-nate than those that share common features (e.g., sad/mad are both considered unpleasant). Additionally,whether emotions refer to an individual’s beliefs or de-sires may contribute to correct interpretation (Wellman& Banerjee, 1991). Emotions such as “happy, sad, andmad” describe a relation between desire and reality. Ifsomeone wants something (desire) and gets it (reality),that person is happy; if not, the person is sad or mad.The emotion “surprise” describes a relation between aperson’s belief or expectation about a situation and thereality of that situation. For example, a person who livesalone, arriving home in the evening, would typically ex-pect no one to be there (belief). If someone is in the house(reality), the homeowner is surprised (mismatch betweenbelief and reality). There is evidence that understand-ing another’s expectations, and violations of expectations,may be more difficult than understanding another’s de-sires (Hadwin & Perner, 1991). These factors may ac-count for the consistent findings (e.g., Denham, 1986;Field & Walden, 1982) that when children are asked tomake inferences regarding the emotions “happy, sad,mad,” they can infer “happy” emotions by age 3, but “sad”versus “mad” and “surprised” cannot be consistently in-ferred until after 5 years of age (Hadwin & Perner, 1991).

Despite the difficulties that one may encounter wheninterpreting facial expressions to make a social infer-ence, most comprehenders seem to accomplish this taskwith little or no conscious effort. According to Crick andDodge’s (1994) social processing model, individuals comeinto a situation with both biologically predeterminedabilities and previous experiences (world knowledge).These predetermined and experientially based skillsallow an individual to complete a series of at least sixinterrelated mental steps in order to make a social in-ference. The first step involves perceiving social cuesand recognizing which cues are relevant (i.e., attendingto those cues that are appropriate). Secondly, these cuesare integrated with prior knowledge in order to makean inference. Third, possible responses are consideredand (fourth) constructed. During the fifth step, a re-sponse selection is made, and finally (sixth step), theresponse is executed. Efficient processing requires com-pleting these steps in a timely manner.

Research has suggested that children with LI mayhave difficulty making inferences based on physicalevents (Bishop & Adams, 1992; Crais & Chapman, 1987;Ellis Weismer, 1985), as well as those based on mental

Ford & Milosky: Inferring Emotional Reactions 23

states (Axelrod, 1982; Dimitrovsky, Spector, Levy-Shiff,& Vakil, 1998; Holder & Kirkpatrick, 1991; SweetNichols, 1998). The ability to readily infer other people’sfeelings (mental state) may be a critical link betweenlanguage comprehension difficulties (specifically, mak-ing inferences) and the social difficulties experienced bychildren with LI (e.g., Bryan, 1977).

The negative social consequences based on LI be-come apparent during the early preschool years. Chil-dren as young as 3 tend to avoid conversing with theirlanguage-impaired peers (Rice, Sell, & Hadley, 1991).By 4 years of age, children with language comprehen-sion deficits are chosen as least liked by their typicalpeers (Gertner, Rice, & Hadley, 1994). These trends con-tinue throughout the school years, when children withlanguage-based learning disabilities continue to be lesspopular than typical peers (e.g., Horowitz, 1981). Onepotential source of these social difficulties may be re-lated to the ability to infer another person’s mental statebased on linguistic and/or nonlinguistic cues, and to thenmake an appropriate response.

According to Crick and Dodge’s (1994) model, the firststep in making an inference regarding an emotional stateinvolves a child’s attention to a facial expression, amongother relevant cues. As mentioned previously, typicallydeveloping children acquire this ability during the pre-school years. Research has suggested that children withLI may have difficulty quickly and accurately identify-ing emotions depicted by facial expressions. Dimitrovskyet al. (1998) and Holder and Kirkpatrick (1991) pre-sented children with facial expressions depicting sixemotions (anger, fear, sadness, surprise, happiness, anddisgust) and found that children with language/learn-ing impairments (age ranges of 9–12 years and 8–15years, respectively) were less accurate than their age-matched peers at labeling various emotions. In contrast,Trauner, Ballantyne, Chase, and Tallal (1993) found nodifferences in abilities of children with LI (9–14 years ofage) and their age-matched peers to identify facial ex-pressions. However, they asked children to identify onlythree expressions (happy, sad, angry) in a forced-choicesituation and, as a result, a ceiling effect was obtainedfor all emotions.

In addition to identifying a facial expression, a childmust also attend to and use contextual information,available either verbally or visually, to predict someone’semotional response to a series of events. Thus, even ifchildren with LI are able to identify facial expressions,they may not be able to use contextual information in alinguistically demanding task to make an inference(Worling, Humphries, & Tannock, 1999). The time re-quirements for making a successful inference, in addi-tion to the linguistic demands, may result in what Dodge(1986) described as process overload.

The first two steps of Crick and Dodge’s (1994) so-cial inferencing model were investigated in the presentstudy. Specifically, this study was designed to addressthe following three questions: (1) Do children with LIhave difficulty identifying facial expressions? (2) Canthey integrate facial expression knowledge with otherverbally and/or visually presented information in orderto make a social inference? (3) Are inferencing difficul-ties modality-specific? That is, do children with LI havedifficulty when presented with verbal information only,do these difficulties extend to visually presented infor-mation, and/or are the difficulties prevalent when vi-sual and verbal information is combined?

MethodParticipants

Two groups of children (N = 24) attending kinder-garten in the central New York area served as partici-pants in this study: 12 children with LI (LI group: meanage = 5;9 [years;months], SD = 4.7, range = 5;4–6;4)and 12 typically developing children (CA group: meanage = 5;8, SD = 3.9, range = 5;4–6;2). Three children inthe CA group and 4 children in the LI group were fromlow-income, inner-city schools. The remaining childrenattended suburban schools in middle-income neighbor-hoods. There were 6 boys and 6 girls in each group.There was 1 child originally identified as typically de-veloping who was disqualified from participation be-cause of a nonverbal IQ score below the cut off (below85). The next typically developing child who volunteeredwas included. All children were monolingual nativeEnglish speakers. A group of children matched on lan-guage ability was not included because of the concep-tual concerns that arise when comparing children withLI to younger peers (Plante, Swisher, Kiernan, &Restrepo, 1993).

MaterialsFour picture cards were drawn of facial expressions

depicting four emotions: happy, surprised, mad, and sad.Nine stories were generated for each of the four emo-tions. Each story consisted of two actions and an emo-tional response stem. Within the nine stories for eachemotion, three general types of stories were constructed.For example, in the happy condition, the three types ofstories included receiving a gift, going on an outing, andreceiving recognition (e.g., applause, hug). For each ofthe three story types, three different versions were con-structed. For example, the three versions for the outingtype of story included going swimming, going to the park,and going to a carnival. Thus, there were three types ofstories with three versions of each story, yielding a total

24 Journal of Speech, Language, and Hearing Research • Vol. 46 • 21–30 • February 2003

of nine stories per emotion, and a total of 36 stories.Stories were constructed in this manner to ensureequivalence of story topics across modalities, as modal-ity was a within-subjects factor. Therefore one childmight receive the swimming version of the outing storyin the visual modality, the park version in the verbalmodality, and the carnival version in the visual/verbalconcurrent modality.

Each story was depicted in three modalities: visualonly, verbal only, and visual/verbal concurrently. Thevisual stimuli consisted of a sequence of three blackand white line drawings on 6.5 × 9.5 in. cards. The firsttwo cards illustrated a common situation (e.g., a car-toon character gradually opening a present), and thethird showed a conclusion (e.g., the character holdingthe new teddy bear). The first two drawings showedonly profile or back views of the cartoon character, andthe last drawing showed the full front view of the char-acter, with the face missing. The facial expression pic-ture cards allowed children to choose and attach anexpression to complete the story. A cartoon character(Twinky) was used in all stories to minimize culturalidentity and gender as confounding variables. Verbalstimuli, three-sentence stories that corresponded to thevisual stimuli, were recorded. Thus, for the visual/ver-bal concurrent condition, stories were played on anaudiotape player while the examiner simultaneouslypresented picture cards. There were two to three com-munication units (Loban, 1976) per story in the verbalcondition, with the mean number of communicationunits equivalent across emotion conditions. There wereno complex sentences, and all communication units con-tained three to four clausal elements (Crystal, 1988). Astory example is illustrated in the Appendix. For thevisual versions, the character’s visual appearance andgender neutral name allowed each individual partici-pant to assume Twinky’s gender. For the verbal stimuli,two sets of tapes were recorded, varying pronouns thatdetermined Twinky’s gender. This allowed gender to bematched for each participant.

Selection of StimuliIn order to ensure the recognizibility of the facial

expressions used in the present study, 38 adults (gradu-ate students in speech-language pathology) were askedto label four emotion cards. All adults labeled the sadfacial expression correctly 100% of the time, and theremaining three expressions (mad, happy, surprised)were labeled correctly 97% of the time.

Prior to determining the 36 situational stories (9stories × 4 emotions) used in the study, 90 adults readwritten versions of the stories and were asked to writein the resulting emotion (e.g., Twinky wanted a teddybear for his/her birthday. S/he opened a present with a

big fluffy teddy bear. Twinky was _____ ). All stories se-lected were completed with the predicted emotion anaverage of 94% of the time (range = 86–100%). On thebasis of this information, three master lists of the 36stories were then constructed, varying by mode of pre-sentation for each story (e.g., one list contained the teddybear gift story through the verbal mode, the second con-tained the same story through the visual mode, and thethird contained the story with visual/verbal concurrentpresentation). To avoid possible ordering effects, storyorder was then randomized for each list.

ProcedureChildren were seen individually in a quiet room. To

determine eligibility for inclusion in the study, two stan-dardized tests were administered. The Clinical Evalua-tion of Language Fundamentals–Preschool (CELF-P;Wiig, Secord, & Semel, 1992) was administered to assesslanguage ability, and the nonverbal subtests of theKaufman Assessment Battery for Children (KABC;Kaufman & Kaufman, 1983) were administered to assesscognitive ability. Children in both groups received a per-formance composite score of at least 85 on the nonverbalsubtests of KABC. Children in the CA group received astandard score of at least 90 on the comprehensionsubtests of the CELF-P. Children in the LI group scoredmore than 1 SD below the mean (below 85) on the com-prehension subtests of the CELF-P. In addition, childrenin the LI group had been previously diagnosed by speech-language pathologists and were receiving language inter-vention services in their school districts. Mean scores forthe two groups are presented in Table 1.

Children were then asked to perform three experi-mental tasks: production of emotions depicted by facialexpressions, comprehension of emotion terms, andinferencing.

In the production task, the four facial expressionswere presented to each child one at a time (in randomorder). After presenting each picture, the examinerasked, “How does this child feel?” Responses were tran-scribed and audio recorded. The production task was

Table 1. Language and cognitive standardized test scores.

LI group CA group

CELF-P (Receptive) 81.5 103.4(6.9) (8.0)

KABC (Nonverbal) 98.1 113.1(10.4) (10.1)

Note. LI group = children with language impairment; CA = age-matched controls; CELF-P = Clinical Evaluation of Language Funda-mentals–Preschool; KABC = Kaufman Assessment Battery for Children.

Ford & Milosky: Inferring Emotional Reactions 25

always administered first to determine the labels chil-dren associated with the different expressions.

The purpose of the comprehension tasks was to de-termine understanding of the emotion labels used dur-ing the inferencing task (story). Children were shownthe same four drawings of the facial expressions used inthe production task, and asked, “point to happy (sad,etc.).” After each trial, the drawings were removed andshuffled in order to prevent a response set.

In the inferencing task, the examiner sat next tothe child in order to minimize unintentional facial cues.The examiner verbally stated the following instructions:“We are going to listen and look at some stories aboutmy friend Twinky. Things happen to Twinky that makehim (or her) mad, happy, sad or surprised. Listen andwatch carefully, because for each story, I want you toshow or tell me how Twinky feels. Put the right face onTwinky to finish the story. Let’s try one.” Children werethen given one practice story (concurrently presented)to familiarize them with the task. All children completedthe practice story without difficulty.

Data CollectionFor all experimental tasks, responses to each trial

were audio recorded, and a dichotomous correct/incor-rect score was given. When a child selected a visual pic-ture (without a verbal label), the examiner provided alabel (e.g., “you chose happy”) for purposes of obtaininginterrater agreement. To determine interrater agree-ment, one response from each child was randomly se-lected from the audiotapes for independent coding by anaïve judge. Point by point agreement was 100%.

ResultsProduction of Emotions

In the first experimental task, children were shownpictures of the emotional expressions (one at a time) andasked to verbally label the expression. Children in bothgroups (LI, CA) were able to spontaneously label happy,sad, and mad with 100% accuracy.

Children’s initial responses to the surprised expres-sion were slightly more varied. Within the CA group, 10of 12 children spontaneously labeled surprised (of the re-maining 2, 1 child initially labeled this expression as “re-ally happy” and 1 as “excited”). Within the LI group, 8 of12 children spontaneously reported surprised. Of the 4who provided different labels, 2 stated “happy”, 1 “oh my!”and 1 as “open.” In these cases, the child’s initial responsewas acknowledged but the child was then told “today wewill call this surprised.” The production task was thenrepeated, and all children used the label “surprised” onthe second trial.

Comprehension of EmotionsThe second experimental task was a picture point-

ing task to verify comprehension of all four emotions.Children in both groups (LI, CA) pointed to the correctfacial expression upon request with 100% accuracy.

Inferencing TaskBecause there were no significant list effects, F(2,

23) = 0.19, p = .83, data for the inferencing tasks werecombined across lists for analyses. Additionally, resultsfrom the Mauchly (1940) test revealed that the homo-geneity of variance assumption had not been violated,χ2 (2, N = 24) = 1.840, p = .3985. An alpha level of .05was used for all statistical tests. The size of effect forall significant results was measured using the standard-ized mean difference, d (Cohen, 1988).

Means and standard deviations of performance onthe inferencing tasks are presented in Table 2. Children’sscores on the inferencing task were analyzed with a 2(Group) × 3 (Modality) × 4 (Emotion) analysis of vari-ance (ANOVA) with repeated measures for the last twofactors. The analysis revealed three significant main ef-fects. First, as predicted, children with LI were not asproficient at inferring emotional reactions, compared totheir CA matched peers, F(1, 22) = 26.17, p < .0001, d =.88 (large effect). Second, there was a main effect due tomodality, F(2, 44) = 7.10, p = .0021. Post hoc compari-sons using Tukey’s HSD procedure showed that the vi-sual/verbal concurrent presentation resulted in morecorrect inferences than the visual only presentation,Tukey a, p < .05, d = .36 (modest effect) in both groups,with no statistically significant difference between au-ditory only and visual only presentations. There was nostatistically significant interaction between group andmodality, F(2, 44) = 0.14, p = .24. Children with LI hadmore difficulty than their CA matched peers, regard-less of modality of presentation.

The third main effect was type of emotion to be in-ferred, F(3, 66) = 13.65, p < .0001. Tukey HSD post hocanalyses revealed statistically significant differences be-tween happy and mad, Tukey a, p < .05, d = 1.03 (largeeffect), as well as differences between happy and surprised,Tukey a, p < .05, d = 1.06 (large effect). There was nostatistically significant interaction between modality andemotion, F(6, 132) = 1.27, p = .27, and no three-way inter-action, F(6, 132) = 1.43, p = .21. Thus, children in the LIgroup had more difficulty making appropriate inferences,regardless of emotion or mode of presentation.

Error AnalysisWhen children did make an error (i.e., chose an emo-

tion different from the one anticipated), we noted the

26 Journal of Speech, Language, and Hearing Research • Vol. 46 • 21–30 • February 2003

specific emotion selected. Were errors typically of thesame valence (e.g., happy/surprised, mad/sad) or differ-ent valence (e.g., happy/mad)? Errors that were of a dif-ferent valence were analyzed with a 2 (Group) × 4 (Emo-tion) ANOVA with repeated measures for emotion.Children with LI were more likely to infer emotions of adifferent valence, compared to their CA matched peers,F(1, 22) = 37.08, p < .0001, d = 1.5 (large effect). Therewas no significant main effect due to emotion, F(3, 66) =1.62, p = .19, and no significant interaction, F(3, 66) =0.42, p = .74. Table 3 provides an analysis of the errorsby valence, demonstrating that 53% of the LI group’serrors were of the wrong valence, compared to only 19%in the CA group. Three children in the typical groupproduced no errors of valence, with the remaining 9 pro-ducing between 1 and 5 confusions of valence. In con-trast, every child in the LI group produced valence er-rors, with 3 children producing between 2 and 5 errors,3 producing between 6 and 10 errors, and 6 producing11–15 errors. The confusion matrix presented in Fig-ure 1 provides information about which specific emo-tions were chosen for each targeted emotion.

Factors Influencing Inferencing AbilityAs anticipated, the two groups (LI, CA) differed sig-

nificantly on language ability as assessed by the com-prehension subtests of the CELF-P (M = 81 for LI; M =104 for CA), t(22) = 8.47, p < .0001. Although all partici-pants achieved a score within normal limits (>85) onthe measure of nonverbal intelligence as measured bythe nonverbal subtests of the KABC, there were signifi-cant differences between the two groups (M = 98 for LI;M = 113 for CA), t(22) = 3.58, p = .0017. In addition,there was a moderate correlation of KABC and languagescores (r = .70).

In order to investigate the extent to which languageand/or nonverbal cognition influenced performance onthe visual (nonverbal) condition of the inferencing tasks,regression analyses were performed. The effect of KABCscore on visual condition performance and the effect oflanguage score on visual condition performance initiallywere tested separately. Both KABC score, F(1, 22) = 5.83,p = .0245, and language score, F(1, 22) = 18.54, p = .0003,were found to significantly influence inferencing in thevisual condition. KABC score accounted for 21% of the

Table 2. Mean number of correct inferences by group and condition (maximum 3.0 per cell).

Modality

Emotion Verbal Visual Concurrent Across

CA group

Happy 2.833 2.750 2.750 2.778(.389) (.622) (.452) (.485)

Sad 2.833 2.167 2.750 2.583(.389) (.937) (.452) (.692)

Mad 1.500 1.833 1.917 1.750(1.807) (.718) (.996) (.937)

Surprised 1.917 1.417 2.167 1.833(.900) (1.165) (1.030) (1.056)

Across 2.271 2.041 2.399 2.236emotion (.939) (.988) (.844) (.931)

LI group

Happy 1.833 1.750 2.417 2.000(.835) (.866) (.669) (.828)

Sad 1.333 1.000 1.750 1.361(1.155) (.953) (1.138) (1.099)

Mad 1.333 1.250 1.167 1.250(1.073) (.866) (.718) (.874)

Surprised .917 .750 1.000 .889(1.084) (.965) (1.044) (1.099)

Across 1.354 1.188 1.583 1.375emotion (1.620) (.959) (1.480) (1.030)

Note. CA group = age-matched controls; LI group = children with language impairment.

Ford & Milosky: Inferring Emotional Reactions 27

variance (R2 = .21) in visual condition performance, andlanguage score accounted for 46% of the variance (R2 =.46) on this measure. In order to determine if these twoindependent variables accounted for more of the variancethan either variable alone, language and KABC scoreswere then entered simultaneously. Results of the test ofthe simultaneous model were significant, F(2, 21) = 8.87,p = .0016. The model continued to account for approxi-mately 46% of the variance in inferencing ability, R2 =.46, indicating no additional variance accounted for whenKABC and language were examined together in contrastwith language scores alone. Consistent with this, theregression slope significantly differed from zero for lan-guage ability when cognition was held constant, t(21) =3.104, p = .005, suggesting that language ability playeda major role in children’s ability to make inferences, evenwhen stimuli were only presented visually. The regres-sion slope did not significantly differ from zero for non-verbal intelligence measures when language was heldconstant, t(21) = –0.160, p = .874.

DiscussionThe current study suggests that children with LI

do differ from their typically developing peers in pro-cessing social information. Children in both groups (LIand CA) demonstrated the ability to identify (throughcomprehension and production tasks) the four emotionsdepicted in line-drawn facial expressions. However, chil-dren with LI were less proficient at integrating thisemotion knowledge with event context, in order to makean accurate social inference regarding a character’s feel-ings. When inferencing errors were made, children inthe LI group were likely to provide emotions of a differ-ent valence (e.g., substituting happy for mad), whereaschildren in the CA group tended to provide emotions ofa similar valence (e.g., substituting mad for sad).

The finding that all children were able to identifyand produce the lexical labels for the facial expressionsis not necessarily inconsistent with prior research, if oneconsiders methodological differences. Children in theDimitrovsky et al. (1998), Holder and Kirkpatrick (1991),and Trauner et al. (1993) studies all easily identifiedhappiness. Although children in those studies did notidentify sadness, anger (madness), or surprise as suc-cessfully as our children did, we attribute this to theuse of four line drawings rather than to the more com-plicated multiple photographic examples of adult facialexpressions. Limited information was provided in theprior studies, but it is unlikely that language differencesaccounted for better performance of children in the

Table 3. Number (and proportion) of errors by emotion andvalence.

Total Errors of Errors ofEmotion errors same valence different valence

CA group

Happy 8 5 3(.073) (.046) (.028)

Surprised 42 34 8(.385) (.312) (.073)

Mad 44 41 3(.404) (.376) (.028)

Sad 15 8 7(.138) (.073) (.064)

Total 109 88 21(.807) (.193)

LI group

Happy 36 11 25(.153) (.047) (.106)

Surprised 79 42 37(.335) (.178) (.157)

Mad 63 37 26(.267) (.157) (.110)

Sad 58 21 37(.246) (.089) (.157)

Total 236 111 125(.470) (.530)

Note. CA group = age-matched controls; LI group = children withlanguage impairment. Denominator for proportions is total number oferrors per group.

Figure 1. Confusion matrix illustrating the number of responses(actual) by predicted emotion. * indicates actual responses of theanticipated emotion (correct inference). ** indicates actualresponses of emotions of the same valence. CA group = age-matched controls; LI group = children with language impairment.

Correct emotion

Emotion chosen Happy Surprised Mad Sad

CA Group

Happy 100* 34** 1 7Surprised 5** 66* 2 0Mad 2 4 64* 8**Sad 1 4 41** 93*No response 0 0 0 0

LI Group

Happy 72* 42** 20 29Surprised 11** 29* 6 8Mad 12 16 45* 19**Sad 13 19 36** 50*No response 0 2 1 2

28 Journal of Speech, Language, and Hearing Research • Vol. 46 • 21–30 • February 2003

present study, as children in the prior studies were allseveral years older.

During the initial steps of social information pro-cessing proposed by Crick and Dodge (1994), individu-als must encode and interpret social cues (based on worldknowledge) in order to make an inference. Findings fromthe current study offer some insight into possible sourcesof difficulty children with LI experienced. The 5-year-old children in this study possessed the skills necessaryto recognize four facial expressions and to semanticallymap those emotions (Crick & Dodge’s step one). The dif-ficulty arose during information processing (Crick &Dodge’s step two), when this facial expression/emotionknowledge needed to be integrated with current con-text in order to arrive at an appropriate inference. Chil-dren with LI had significantly more difficulty than theirCA peers in choosing the appropriate emotional expres-sion, suggesting difficulty with information processing(Crick & Dodge’s step two).

Children in both groups made significantly morecorrect inferences in the happy condition than in themad or surprised conditions. The ease of inferring hap-piness is consistent across several studies (e.g., Denham,1986; Field & Walden, 1982), as is the greater difficultywith mad and surprised (Denham & Couchoud, 1990).Denham and Couchoud cited research demonstratingthat adults show happy expressions most often, and theysuggested that the emotions adults talk about most fre-quently may be learned at earlier ages. In addition,Denham (1986) indicated that facial expressions of hap-piness are likely to result in reciprocal positive reac-tions (one is likely to react to a smile with a smile).

When we examined the valence of errors childrenmade, we found striking differences between the LI andCA groups. The children in the LI group made signifi-cantly more errors that confused positive and negativeemotions than did children in the CA group. During a so-cial interaction, the consequence of predicting that a per-son may be sad instead of mad (inference of same valence)may or may not have a negative impact on the interac-tion. In fact, these slight inference confusions may occurfrequently in day to day interactions, because the personexperiencing the event may actually feel more than oneemotion (e.g., losing a pet may cause someone to feel sadand mad concurrently). However, the consequences ofmisreading the loss of a pet as making someone feel happycould have devastating social consequences. In otherwords, failure to correctly complete Crick and Dodge’s steptwo will result in errors in step three (choosing an appro-priate response to the situation). Further explorations ofpolarity confusions in emotional interpretation may pro-vide greater insight into the social difficulties experiencedby children with LI (e.g., Gertner et al., 1994; Horowitz,1981; Rice et al., 1991).

Another question addressed in this study waswhether the modality of presentation would affectinferencing ability in children with LI. Across groups,children performed more accurately during concurrentpresentation of the material than when presented withpictures only (visual condition). Both conditions utilizedthe same pictures, with the concurrent condition alsoproviding verbally encoded redundant information. Al-though the concurrent condition did not significantlydiffer from the auditory-only condition, the addition oflanguage to the visually presented material (concurrentcondition) did improve children’s ability to make anemotional inference.

Further evidence of the role of language in these emo-tional inferences is provided by the results of the regres-sion analyses. Children in the LI and CA groups differedon performance on the KABC. However, when we lookedspecifically at performance on the pictorially presentedstimuli (visual condition), we found that inferencing abil-ity was related to language ability when cognitive abili-ties were held constant, suggesting that children in bothgroups were verbally mediating visually presented mate-rial. We often observed children (in both groups) perform-ing verbal mediation as they told themselves the experi-mental story aloud during the visual condition. Themoderate correlation between KABC score and languagescore, coupled with the fact that addition of KABC scoreto language score did not increase the amount of explainedvariance, suggests that the nonverbal tasks on the KABCmay also be verbally mediated. This is consistent withevidence of language-based encoding of visually presentedstimuli on memory tasks, for example (Gillam, Cowan, &Marler, 1998; Penney, 1989). Determining the influenceof language versus other cognitive abilities on targetedtasks continues to present challenges to researchers andclinicians.

In a typical social interaction, the number, type, andmodality of cues to be identified (step one) and integrated(step two) will vary from situation to situation. Chil-dren with LI were able to identify facial expressions (stepone in the social processing model) but had difficultyinferring the appropriate emotion and choosing the cor-responding facial expression when given an event con-text. The speaker’s personal history, current motives,and varying facial expressions throughout the exchangeare examples of information that may need to be identi-fied and integrated concurrently in order to make anappropriate inference. We recognize that these types ofinformation were not considered in the design of thecurrent study. Future studies may consider these pos-sible sources of information in more contextually richsituations to determine if, with additional cues, childrenwith LI can accurately and efficiently identify and inte-grate those cues that are appropriate (based on theirworld knowledge). Measures that are more reflective of

Ford & Milosky: Inferring Emotional Reactions 29

moment-to-moment processing (e.g., response time mea-sures) may also help to validate the existence of stagesof processing.

Given that the ability to anticipate and respond to apartner’s emotional reactions are integral components inthe comprehension of social discourse, our study offerssome insight into the possible sources of difficulty chil-dren with LI may experience. The current study also sug-gests that at least some social inferencing difficulties arelanguage based, indicating that it is appropriate for clini-cians to address social inferencing in their assessmentand therapy of young children. Concurrent presentation(verbal and visual) of information should assist compre-hension in social situations. The findings further suggestthat it may be necessary to extend treatment beyond theidentification of facial expressions to include use of thosefacial expressions in making inferences about a speaker’sor character’s emotional state. Clearly the current studyhas demonstrated the value of further clinical and experi-mental exploration of this critical component of social pro-cessing in children with LI.

AcknowledgmentsThis study was supported in part by National Institute

on Deafness and Other Communication Disorders GrantDC01649 to the second author. The authors wish to thankthe Syracuse City and Liverpool school districts, as well asKynda Montessori in Syracuse, NY, for assistance withparticipant recruitment.

ReferencesAxelrod, L. (1982). Social perception in learning disabled

adolescents. Journal of Learning Disabilities, 15, 610–613.

Bishop, D.V. M., & Adams, C. (1992). Comprehensionproblems in children with specific language impairment:Literal and inferential meaning. Journal of Speech andHearing Research, 35, 119–129.

Bretherton, I., McNew, S., & Beeghly-Smith, M. (1981).Early person knowledge as expressed in gestural andverbal communication: When do infants acquire a “theoryof mind”? In M. Lamb & L. Sherrod (Eds.), Infant socialcognition (pp. 333–373). Hillsdale, NJ: Erlbaum.

Bryan, T. H. (1977). Learning disabled children’s compre-hension of nonverbal communication. Journal of LearningDisabilities, 10, 36–41.

Cohen, J. (1988). Statistical power analysis for the behav-ioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum.

Crais, E. R., & Chapman, R. S. (1987). Story recall andinferencing skills in language/learning-disabled andnondisabled children. Journal of Speech and HearingDisorders, 52, 50–55.

Crick, N. R., & Dodge, K. A. (1994). A review andreformulation of social information-processing mecha-nisms in children’s social adjustment. PsychologicalBulletin, 115, 74–101.

Crystal, D. (1988). Rediscover grammar. Essex, England:Longman Group.

Denham, S. A. (1986). Social cognition, prosocial behavior,and emotion in preschoolers: Contextual validation. ChildDevelopment, 57, 194–201.

Denham, S. A., & Couchoud, E. (1990). Young pre-schoolers’ ability to identify emotions in equivocal situa-tions. Child Study Journal, 20, 153–169

Dimitrovsky, L., Spector, H., Levy-Shiff, R., & Vakil, E.(1998). Interpretation of facial expressions of affect inchildren with learning disabilities with verbal or nonver-bal deficits. Journal of Learning Disabilities, 31, 286–312.

Dodge, K. A. (1986). A social information processing modelof social competence in children. In M. Perlmutt (Ed.),Cognitive perspectives on children’s social and behavioraldevelopment: The Minnesota Symposia on Child Psychol-ogy, 18. Hillsdale, NJ: Lawrence Erlbaum.

Ellis Weismer, S. (1985). Constructive comprehensionabilities exhibited by language-disordered children.Journal of Speech and Hearing Research, 28, 175–184.

Field, T. M., & Walden, T. A. (1982). Production anddiscrimination of facial expressions by preschool children.Child Development, 53, 1299–1311.

Flin, R., & Dziurawiec, S. (1989). Developmental factorsin fact processing. In A. W. Young & H. D. Ellis (Eds.),Handbook of research on face processing (pp. 335–378).New York: Elsevier.

Gertner, B. L., Rice, M. L., & Hadley, P. A. (1994).Influence of communicative competence on peer preferencein a preschool classroom. Journal of Speech and HearingResearch, 37, 913–923.

Gillam, R. B., Cowan, N., & Marler, J. A. (1998). Informa-tion processing by school-age children with specificlanguage impairment: Evidence from a modality effectparadigm. Journal of Speech, Language, and HearingResearch, 41, 913–926.

Hadwin, J., & Perner, J. (1991). Pleased and surprised:Children’s cognitive theory of emotion. British Journal ofDevelopmental Psychology, 9, 215–234.

Holder, H. B., & Kirkpatrick, S. W. (1991). Interpretationof emotion from facial expressions in children with andwithout learning disabilities. Journal of LearningDisabilities, 24, 170–177.

Horowitz, E. C. (1981). Popularity, decentering ability, androle-taking skills in learning disabled and normalchildren. Learning Disability Quarterly, 4, 23–30.

Kaufman, A. S., & Kaufman, N. L. (1983). KaufmanAssessment Battery for Children. Circle Pines, MN:American Guidance Service.

Klinnert, M. D., Emde, R. N., Butterfield, P., & Campos,J. J. (1986). Social referencing: The infant’s use ofemotional signals from a friendly adult with motherpresent. Developmental Psychology, 22, 427–432.

Loban, W. (1976). Language development: Kindergartenthrough Grade Twelve. Urbana, IL: National Council ofTeachers of English.

Mauchly, J. W. (1940). Significance test for sphericity of anormal n-variate distribution. Annals of MathematicalStatistics, 11, 204–209.

30 Journal of Speech, Language, and Hearing Research • Vol. 46 • 21–30 • February 2003

Miller, P. H., & Aloise, P. A. (1989). Young children’sunderstanding of the psychological causes of behavior: Areview. Child Development, 60, 257–283.

Nicholas, D. W., & Trabasso, T. (1980). Toward a taxonomyof inferences for story comprehension. In F. Wilkening, J.Becker, & T. Trabasso (Eds.), Information integration bychildren (pp. 243–265). Hillsdale, NJ: Erlbaum.

Penney, C. G. (1989). Modality effects and the structureof short-term verbal memory. Memory & Cognition, 17,398–422.

Plante, E., Swisher, L., Kiernan, B., & Restrepo, M.(1993). Language matches: Illuminating or confounding?Journal of Speech and Hearing Research, 36, 772–776.

Rice, M. L., Sell, M. A., & Hadley, P. A. (1991). Conversa-tional responsiveness of speech and language-impairedpreschoolers. Journal of Speech and Hearing Research, 34,1308–1317.

Sweet Nichols, C. (1998). Encoding of nonverbal social cuesby children with differing profiles of learning disabilities.Dissertation Abstracts International, 59(5-B), 2489.

Trabasso, T., Secco, T., & van den Broek, P. (1984).Causal cohesion and story coherence. In H. Mandl, N. L.Stein, & T. Trabasso (Eds.), Learning and comprehensionof text (pp. 83–111). Hillsdale, NJ: Erlbaum.

Trabasso, T., & Sperry, L. (1985). Causal relatedness andimportance of story events. Journal of Memory andLanguage, 24, 595–611.

Trauner, D. A., Ballantyne, A., Chase, C., & Tallal, P.(1993). Comprehension and expression of affect inlanguage-impaired children. Journal of PsycholinguisticResearch, 22, 445–452.

Walden, T. A., & Field, T. M. (1982). Discrimination offacial expressions by preschool children. Child Develop-ment, 53, 1312–1319.

Weaver, C. A., & Kintsch, W. (1991). Expository text. In M.Kamil (Ed.), Handbook of reading research (Vol. 2, pp.230–245). Mahwah, NJ: Erlbaum.

Wellman, H. M., & Banerjee, M. (1991). Mind andemotion: Children’s understanding of the emotionalconsequences of beliefs and desires. British Journal ofDevelopmental Psychology, 9, 191–214.

Wiggers, M., & van Lieshout, C. F. M. (1985). Develop-ment of recognition of emotions: Children’s reliance onsituational and facial expressive cues. DevelopmentalPsychology, 21, 338–349.

Wiig, E. H., Secord, W., & Semel, E. (1992). ClinicalEvaluation of Language Fundamentals–Preschool. SanDiego, CA: The Psychological Corporation.

Worling, D. E., Humphries, T., & Tannock, R. (1999).Spatial and emotional aspects of language inferencing innonverbal learning disabilities. Brain and Language, 70,220–239.

Yuill, N., & Oakhill, J. (1991). Children’s problems in textcomprehension: An experimental investigation. Cambridge,UK: Cambridge University Press.

Received January 3, 2002

Accepted July 25, 2002

DOI: 10.1044/1092-4388(2003/002)

Contact author: Janet Ford, Department of CommunicationSciences and Disorders, 805 South Crouse Avenue,Syracuse University, Syracuse, NY 13244.E-mail: jsford@syr.edu

Appendix. Story example: Inferring mad in visual and verbal conditions.

Visual Condition

Verbal Condition

Twinky was bouncing a ball.A bully took the ball.Twinky was……