the components of young children's emotion knowledge: which are enhanced by adult emotion talk?

17
The Components of Young Children’s Emotion Knowledge: Which Are Enhanced by Adult Emotion Talk? Karen Salmon 1 , Ian M. Evans 2 , Sophie Moskowitz 1 , Melissa Grouden 1 , Fiona Parkes 1 and Emily Miller 1 1 Victoria University of Wellington 2 Massey University Abstract This research adopted observational and experimental paradigms to investigate the relationships between components of emotion knowledge in three- to four-year-old children. In Study 1, 88 children were assessed on the Emotion Matching Task (Morgan, Izard, & King), and two tasks requiring the generation of emotion labels and causes. Most tasks were significantly associated with age and language ability, and similar tasks were significantly but moderately correlated. In Study 2, 58 of these children were allocated to one of three conditions: emotion cause talk, in which they received four sessions of training focusing on emotion labels and causes; non-emotion cause talk, focusing on causal relationships in general; or were allocated to a no-training control. Children in the emotion causes condition showed improvement in their use of emotion labels but not other components of emotion knowledge. The findings suggest that, in three- to four-year-old children, emotion knowledge is con- stituted by related but separable components and that training in emotion language targets separate rather than all aspects of emotion knowledge. Keywords: emotion knowledge; emotion language; training study; preschool-aged children Introduction Children’s emotion competence, that is, their ability to regulate their emotional expres- siveness and experience when required, is critical for their academic and social success. An essential building block is the development of a body of knowledge about emotion that enables children to understand their own and others’ emotional responses (Denham, Bassett, & Zinsser, 2012). Across the preschool years, children establish a solid base of emotion knowledge, manifested in their acquisition of the key skills of recognizing and labelling facial expressions and emotions, and specifying appro- priate emotion causes (Fabes, Eisenberg, Nyman, & Michealieu, 1991; Pons, Harris, Correspondence should be addressed to Karen Salmon, School of Psychology,Victoria University of Wellington, PO Box 600, Wellington, New Zealand. Email: [email protected] Social Development Vol 22 No. 1 94–110 February 2013 doi: 10.1111/sode.12004 © Blackwell Publishing Ltd. 2012. Published by Blackwell Publishing, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA.

Upload: independent

Post on 21-Nov-2023

0 views

Category:

Documents


0 download

TRANSCRIPT

The Components of Young Children’sEmotion Knowledge: Which Are Enhancedby Adult Emotion Talk?Karen Salmon1, Ian M. Evans2, Sophie Moskowitz1, Melissa Grouden1,Fiona Parkes1 and Emily Miller1

1Victoria University of Wellington2Massey University

Abstract

This research adopted observational and experimental paradigms to investigate therelationships between components of emotion knowledge in three- to four-year-oldchildren. In Study 1, 88 children were assessed on the Emotion Matching Task(Morgan, Izard, & King), and two tasks requiring the generation of emotion labels andcauses. Most tasks were significantly associated with age and language ability, andsimilar tasks were significantly but moderately correlated. In Study 2, 58 of thesechildren were allocated to one of three conditions: emotion cause talk, in which theyreceived four sessions of training focusing on emotion labels and causes; non-emotioncause talk, focusing on causal relationships in general; or were allocated to ano-training control. Children in the emotion causes condition showed improvement intheir use of emotion labels but not other components of emotion knowledge. Thefindings suggest that, in three- to four-year-old children, emotion knowledge is con-stituted by related but separable components and that training in emotion languagetargets separate rather than all aspects of emotion knowledge.

Keywords: emotion knowledge; emotion language; training study; preschool-agedchildren

Introduction

Children’s emotion competence, that is, their ability to regulate their emotional expres-siveness and experience when required, is critical for their academic and social success.An essential building block is the development of a body of knowledge about emotionthat enables children to understand their own and others’ emotional responses(Denham, Bassett, & Zinsser, 2012). Across the preschool years, children establish asolid base of emotion knowledge, manifested in their acquisition of the key skillsof recognizing and labelling facial expressions and emotions, and specifying appro-priate emotion causes (Fabes, Eisenberg, Nyman, & Michealieu, 1991; Pons, Harris,

Correspondence should be addressed to Karen Salmon, School of Psychology, Victoria Universityof Wellington, PO Box 600, Wellington, New Zealand. Email: [email protected]

Social Development Vol 22 No. 1 94–110 February 2013doi: 10.1111/sode.12004

© Blackwell Publishing Ltd. 2012. Published by Blackwell Publishing, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street,Malden, MA 02148, USA.

& De Rosnay, 2004). Emotion knowledge predicts children’s social competence,including their peer relationships, successful adjustment to school, and academicachievement. By contrast, limited emotion knowledge is associated with internalizingand externalizing problems throughout childhood (Denham et al., 2002; Ensor,Spencer, & Hughes, 2011). The well-established association of emotion knowledgewith children’s psychological adjustment, and its rapid development in early child-hood, underscores its importance as a target in clinical interventions (e.g., Salmon,Dadds, Allen, & Hawes, 2009). Yet the extent to which effective interventions can bedeveloped is potentially constrained by a lack of clarity with respect to the construct ofemotion knowledge in the preschool years and the factors shaping its development. Wereport findings of two studies investigating these issues.

With respect to the construct of emotion knowledge, the various measures tend toidentify slightly different aspects or components (e.g., labelling one’s own and others’emotions and expressions, understanding emotion causes) (Morgan, Izard, & King,2010). The strength of the associations among these components is unclear, however.The very limited research also tends to include a relatively broad age range, which isproblematic given the rapid developmental changes across early childhood. It is unsur-prising, therefore, that findings with respect to the associations among measures ofemotion knowledge are mixed; Morgan et al. (2010) reported that a range of testscorrelated highly across the age of three to six years (M = 56 months), yet youngerchildren (M = 50 months) do not perform consistently when required, for example, tolabel emotions and identify their causal antecedents (Laible & Thompson, 1998). InStudy 1, therefore, we investigated the associations among tasks assessing severalcomponents of emotion knowledge. We focused on age three to four years given theimportance of this period for the development of emotion knowledge and the devel-opmental changes that might be expected in a broader age range.

The need for clarity in relation to the emotion knowledge construct assumes par-ticular importance in the context of research investigating factors associated with itsdevelopment. Evidence suggests that children’s emotion knowledge is associated withtheir language skill and develops within the context of parent–child conversations. Thereasons for the well-established relation between language ability and emotion knowl-edge are likely to be multiple, including the linguistic component of the assessmenttasks, and the essential role of symbolic and representational skill in reporting anddifferentiating internal experience (Beck, Kumschick, Eid, & Klann-Delius, 2012;Izard et al., 2011). Further, families who talk more about emotion have children whoalso tend to do so, and parent emotion talk predicts children’s later scores on tests ofemotion knowledge (Garner, Carlson Jones, Gaddy, & Rennie, 1997; Van Bergen &Salmon, 2010b). For example, using an observational design, Brown and Dunn (1996)found that children from families who engaged in high frequencies of feeling state andcausal talk manifested superior emotion knowledge seven months later, but only causaltalk (and not feeling state talk) predicted children’s understanding of ambivalentfeelings at the age of six. Unclear, however, is whether causal talk alone or causal talkin the context of emotional experience is critical for the development of emotionknowledge.

Experimental studies also demonstrate that the way in which emotion talk occurs hasconsiderable implications for children’s skills. Gavazzi and Ornaghi (2011) found thatfollowing a two-month period of training conducted in twice-weekly groups, preschoolchildren’s (M = 52 months) emotion knowledge was enhanced only when, afterlistening to a story, they participated in conversation that encouraged their use of

Components of Children’s Emotion Knowledge 95

© Blackwell Publishing Ltd. 2012 Social Development, 22, 1, 2013

emotion language and not when they engaged in free play following the story. Themeasure of emotion knowledge, the test of emotion comprehension (Pons et al., 2004),assessed the children’s recognition of faces, and external and mentalistic causes ofemotion. Van Bergen, Salmon, Dadds, and Allen (2009) encouraged mothers to labeltheir child’s (M = 46 months) emotions and discuss emotion causes during reminiscingconversations. At the six-month delay, their children produced a greater number ofcauses on an emotion cause knowledge task than did children in the control condition.Although this research suggests specificity of training effects, the most effective formsof emotion talk are not well established.

The second aim of the current research, tested in Study 2, was to further under-standing of the critical aspects of adult–child emotion talk for children’s emotionknowledge. We aimed to establish whether the components of emotion knowledgeassessed in Study 1 were differentially affected when children were engaged in trainingthat emphasized the causal relationship between emotions or situations, relative to ano-intervention control condition.

Study 1

Little research has compared young children’s performance on tests assessing emotionknowledge. An exception is the study by Morgan et al. (2010), who found moderate tostrong correlations among three measures for three- to six-year-old children [affectknowledge task, Denham, 1986; Kusche emotional inventory, Kusche, 1984, citedin Morgan et al., 2010; Emotion Matching Task (EMT), Morgan et al., 2010]. Severalquestions remained, however. Firstly, as all measures were strongly correlated with ageand language, it was unclear whether associations between tasks remained if thesevariables were controlled. Further, subtests assessing apparently similar skills (suchas receptive emotion knowledge components of the EMT and affect knowledge test)did not correlate significantly. One relevant factor may be the nature of the tasks andtheir associated cognitive demands (Widen & Russell, 2010a). For example, the tasksdiffered with respect to their reliance on pictorial stimuli (photographs of faces, puppetfacial expressions) and the extent to which they required the child to generate verbalinformation. According to Gelman (1982), preschool children’s performance on cog-nitive tasks is characterized by ‘pockets’ of competence, which they can demonstrategiven supportive experimental conditions (e.g., prompting, meaningful tasks) butwhich do not generalize readily from one task to another (p. 218).

In Study 1, we compared three- to four-year-old children’s (M = 42.83 months)performance on the EMT with their use of emotion labels and causes. We selected theEMT given its comprehensiveness and recent data relating to its psychometric prop-erties. The EMT comprises four subtests assessing children’s production of emotionlabels for photographs of emotion expressions, and their ability to match emotionexpressions with those of the same category, with situations or causes, and with spokenemotion labels (Morgan et al., 2010). Children’s use of emotion language (labels,causes) is arguably the most observable component of EK in a social setting, and wewere keen to assess the associations between emotion language and the EMT. Whereasmost research has investigated emotion language in naturalistic settings (Dunn, Brown,& Beardsall, 1991; Fabes, et al., 1991), to retain a degree of experimental control, weadopted semi-structured tasks (narration of the emotional aspects of a wordless picturebook, generation of emotion causes; Van Bergen et al., 2009; Wang, Hutt, Kulkofsky,McDermott, & Wei, 2006). Our aim, therefore, was to evaluate the performance of

96 Karen Salmon, Ian M. Evans, Sophie Moskowitz et al.

© Blackwell Publishing Ltd. 2012 Social Development, 22, 1, 2013

children within a limited age range on a relatively broad range of tasks, differing withrespect to the degree of structure and mode of presentation.

We expected that all tasks would be significantly associated with the child’s lan-guage ability. Moreover, we hypothesized that there would be low to moderate asso-ciations among the skills purported to assess components of emotion knowledge(Cohen, 1988), and further, that the pattern of associations would differ depending onboth the specific aspect of emotion knowledge under consideration (e.g., labellingemotion, identifying emotion causes), and possibly the mode of assessment (pictorial,verbal).

Study 1: Method

Participants

Participants were 88 children at the age of three to four years (M = 42.83 months, SD= 3.72 months) recruited from nine preschools in diverse sociodemographic areas inWellington, New Zealand. Information letters and consent forms were disseminated toparents via teachers. The sample comprised 51 girls and 37 boys, most of whom wereEuropean New Zealanders. Across kindergartens, the mean participation rate rangedfrom 40 to 75 percent.

Overview of Procedure

Children were tested in one session at preschool, and the experimenter initiallyengaged the child in conversation about a preferred activity. The order of testing wasthe same for all children. Firstly, to familiarize them with the impending activities, thechildren were asked to describe the contents of four frames in the wordless picturebook Zou (Gay, 2009). For the test materials, the Emotion Narrative Task was presentedfirst, followed by version A of the EMT, the Emotion Cause Production Task (ECPT),and finally the vocabulary subtest of the Wechsler Preschool and Primary Scale ofIntelligence – Third Edition (WPPSI-III; Wechsler, 2004). The children were praisedfor effort (e.g., ‘you’re trying really hard’), and upon completion were offered acolourful sticker in appreciation of their participation.

Measures

Emotion Narrative Task. The wordless picture book Frog, Where Are You? (Mayer,1969) was used to assess children’s spontaneous and prompted production of emotionlanguage. We reduced the original 24-picture story to 18 pictures, reproduced asA4-sized storyboards. The episodes central to the global structure of the story werepreserved. The booklet was presented twice. During the first presentation, the childrenwere encouraged to look at the pictures. During the second, open-ended prompts wereprovided (‘tell me everything you can about this picture’, ‘what’s happening here?’),and subsequent to a significant point in the story (the boy being tossed into the water,the 10th frame), the child was prompted to discuss emotion in the pictures. Thestandardized prompts were as follows: ‘How is everyone feeling on this page?’ If thechild labelled an emotion but the character was unclear, she/he was asked ‘Who’sfeeling ___?’ Additional prompts included ‘How is everyone feeling?’ Prompts foremotion causes and behavioural manifestations of emotion were as follows: ‘Howcome they are feeling (_)?’ and ‘How do you know that they are feeling (_)?’ The taskduration was 10–20 minutes.

Components of Children’s Emotion Knowledge 97

© Blackwell Publishing Ltd. 2012 Social Development, 22, 1, 2013

Emotion Matching Task (Morgan et al., 2010). The EMT contains 48 items, with twosplit half (equivalent) alternate forms (EMT-A and EMT-B). We used the EMT-A,which is presented as a booklet of colour photographs of ethnically diverse children(African-American, Hispanic, White, biracial) whose faces display expressions ofhappiness, sadness, anger, fear/surprise, and neutrality. The test is divided into fourparts (subtests), each containing six items. Part 1 (matching expressions to sample)requires participants to choose which face (from four) displays the same emotionexpression as that shown in a target photograph. Part 2 (expression–situation matching,requiring emotion cause understanding) involves choosing which face (from four)displays an emotion expression that matches the situation described by the experi-menter (e.g., ‘show me the one whose puppy got lost and did not come home’). Part 3(expression labelling) requires participants to name the expression shown in a singlephotograph. Part 4 (expression–label matching) involves choosing which face (fromfour) displays the emotion labelled by the experimenter. Each correct response receivesone point.

Emotion Cause Production Task (Adapted From Denham, Zoller, & Couchod, 1994;Dunn & Hughes, 1998; Van Bergen et al., 2009). The ECPT assesses children’sability to generate plausible causes of their emotions, and has been utilized in researchwith children at the age of three to six years. We were confident that our participantscould complete this task given Wang et al.’s (2006) findings that children (M age =35.49 months) reported M = 6.28 (SD = 4.41) valid and reliably coded causes. Acrossstudies, the task shows good interrater reliability (Cohen’s kappa range .66–1.00). Inthe current study, four drawings of faces displaying the emotions happy, angry, sad, andscared were presented successively in counterbalanced order. The child was told ‘Thisface is (emotion). Tell me what makes you feel (_)’. If the child responded ‘I don’tknow’ to the first prompt, the question was phrased differently [e.g., ‘Tell me one thingthat makes you feel (_)’]. Prompting continued [e.g., ‘What else makes you feel (_)?’,minimal encouragers] until the child produced no further examples. This process wasrepeated for all four emotions. The ECPT was audio-recorded for transcribing andcoding.

Language Ability. The receptive vocabulary subtest of the WPPSI-III (Australian)(Wechsler, 2004), standardized for children between the age of 2:6 and 3:11 years,assessed language ability. The WPPSI-III has excellent psychometric properties (reli-ability coefficients: a = .83–.95; test–retest reliability for verbal scores: a = .90). Thissubtest consists of 38 items, each containing four pictures. Children are asked to pointto the picture that matches the word spoken by the experimenter. Each correct answerreceives one point. The test is terminated following a 0 score on five consecutive items.

Coding

Emotion Narrative Task. Children’s responses were coded for emotion language (fol-lowing Garner et al., 1997; Marin, Bohanek, & Fivush, 2008), which were differenti-ated according to whether they were spontaneous (uttered in response to an open-endedprompt prior to specific emotion prompting from Frame 10, e.g., ‘tell me everythingthat is happening on this page’) or prompted (uttered in response to the more specificemotion prompts from Frame 10). Categories coded were emotion labels (e.g., ‘theyare all happy’); emotion causes, which were required to be plausible (e.g., ‘He’s crying.

98 Karen Salmon, Ian M. Evans, Sophie Moskowitz et al.

© Blackwell Publishing Ltd. 2012 Social Development, 22, 1, 2013

she won’t give him the ball’); and behavioural manifestations of emotion, in whichemotions were inferred from behaviour but not stated (e.g., ‘he was crying/ laughing’).Each mention of emotion language received one credit, whether it was producedspontaneously or following prompting. Quasi-emotion, terms such as ‘feeling funny/bad/’, was excluded. Reliability, determined when two coders independently coded 25percent of children’s responses, was kappa = .83.

Emotion Cause Production Task. The number of unique, plausible causes was talliedfor each emotion and summed across emotions. There was no upper limit for scoresand no time limit. Reliability, determined when two coders independently coded 25percent of children’s responses, was kappa = .82.

Study 1: Results

There were no significant associations of any measure with gender, which wasexcluded from analyses. As preliminary analyses revealed positive skewness for theECPT, correlations were rerun with the data for that variable having undergonelogarithmic transformation (Field, 2009). In reporting correlations, we use Cohen’s(1988) convention: strong (+ or -) .50–1.00; moderate (+ or -) .30–.49; and weak(+ or -) .10–.29. Descriptive statistics are presented in Table 1. The pattern of scoresreplicated those of Morgan et al. (2010) for the EMT-A, in that the children achievedtheir highest scores on Part 4 (receptive emotion knowledge) and their lowest scores onPart 2 (emotion situation knowledge). Given the low means for spontaneous emotionlanguage (spontaneous labels, M = .24, SD = .51; spontaneous causes, M = .00, SD =.00; prompted labels, M = 3.11, SD = 2.87; prompted causes, M = 1.29, SD = 1.58), andin light of our interest in both labels and causes, two collapsed variables were created:emotion causes (spontaneous and prompted mentions of emotion causes, summed) andemotion labels (spontaneous and prompted mentions of emotion labels, summed): the

Table 1. Descriptive Data

Variable

Name Mean SD Range

Age (months) 42.83 3.72 34–50Language (WPPSI) 11.74 2.59 4–17ECPT: Unique causes 4.43 3.38 0–18Emotion Narrative Task: Causes 1.29 1.57 0–5Emotion Narrative Task: Labels 3.35 3.12 0–15EMT-A Part 1 3.35 1.55 0–6EMT-A Part 2 2.66 1.45 0–6EMT-A Part 3 3.25 1.92 0–6EMT-A Part 4 4.43 1.04 0–6EMT-A total score 13.69 4.07 1–21

Note: WPPSI-III score is age-standardized. ECPT = Emotion Cause Production Task; EMT =Emotion Matching Task; WPPSI = Wechsler Preschool and Primary Intelligence Scale.

Components of Children’s Emotion Knowledge 99

© Blackwell Publishing Ltd. 2012 Social Development, 22, 1, 2013

findings did not change when we analyzed prompted information only. As the mean for‘behavioural manifestation of emotion’ was <1, it was excluded.

Table 2 presents the zero-order correlations among the variables. Most correlationswere weak to moderate. Several findings are of note. Firstly, there were significantassociations with age and language skill for most EMT-A subtests and total score.Secondly, focusing on the associations among EMT tasks, the strongest correlationswere between the subtests and the total score on that test. Of note, the correlationsamong EMT-A subtests were somewhat lower than those reported by Morgan et al.(2010), possibly due to the younger and more constricted age range of our participants.Thirdly, with respect to associations among similar tasks, there were significant cor-relations among labelling tasks (total labels in the Emotion Narrative Task and Part 3of the EMT-A) and some associations among causes tasks [Part 2 (emotion–situationmatching) of the EMT-A was associated significantly with the ECPT but not with thetotal causes reported in the Emotion Narrative Task]. Fourthly, the ability to labelemotions was significantly correlated with most other tasks, whether assessed bymeans of Part 3 (emotion expression labelling) of the EMT-A or production of emotionlabels on the Emotion Narrative Task (albeit inconsistently for the latter task).

Given their significant associations with the EMT-A in particular, analyses werererun partialling the effect of both age and language. The pattern of significant findingswas similar (see Table 3), albeit weaker. For example, Part 2 of the EMT-A (emotionsituation knowledge) was no longer related to either emotion cause task (ECPT, totalemotion causes in the Emotion Narrative Task), and Part 4 of the EMT-A (emotionlabelling) was no longer related to emotion labelling in the Emotion Narrative Task.

To explore further the relations among the tasks, an exploratory factor analysis wasconducted. Three factors were found, explaining 61.57 percent of the variance. Whenexamining the obliquely rotated factors, a two dimensional solution was the mosttheoretically meaningful (see Table 4 for the factor pattern matrix for this solution). Thefirst factor (named ‘picture labelling and emotion identification’) accounted for 31.75percent of the variance, and comprised tasks that appeared to have in common the useof pictorial stimuli that were either labelled or matched (all EMT-A tasks and the

Table 2. Zero-order Correlations Among Key Variables

1 2 3 4 5 6 7 8 9 10

1. Age 12. WPPSI stand. .11 13. ECPT .03 .27* 14. ENT labels .11 .18 .23* 15. ENT causes .11 .14 .29* .36** 16. EMT-A Part 1 .23* .29** .19 -.10 -.05 17. EMT-A Part 2 .43** .35** .22* .19 -.02 .10 18. EMT-A Part 3 .22* .23* .39** .41** .26* .34** .25* 19. EMT-A Part 4 .26* .08 .12 .26* .01 .30** .34** .38** 1

10. EMT-A total .41** .41** .38** .30** .10 .65** .59** .79** .67** 1

Notes: WPPSI is Wechsler Preschool and Primary Intelligence Scale – Third Edition age standardized score forthe Receptive Vocabulary subtest. ECPT is Emotion Cause Production Task. ENT labels and causes are EmotionNarrative Task labels and causes. Different numbers for each variable relate to incomplete datasets (WPPSI-IIIand EMT-A: n = 88; ECPT: n = 87; ENT: n = 75). EMT = Emotion Matching Task; WPPSI = Wechsler Preschooland Primary Intelligence Scale.

100 Karen Salmon, Ian M. Evans, Sophie Moskowitz et al.

© Blackwell Publishing Ltd. 2012 Social Development, 22, 1, 2013

WPPSI-III vocabulary subtest).The second factor (named ‘emotion labelling and causalunderstanding’) accounted for 17.26 percent of the variance, and comprised total causesand labels from the Emotion Narrative Task and the total unique causes produced onthe ECPT. Part 3 of the EMT-A (expression labelling) loaded on both factors.

Study 1: Discussion

The aim of Study 1 was to investigate the relations among various tasks considered toreflect aspects of three- to four-year-old children’s emotion knowledge. All subtests ofthe EMT-A were significantly associated with age whereas other tasks were not. Mosttasks were also significantly correlated with language ability, consistent with findings

Table 3. Partial Correlations, Controlling for Age and Language, Among KeyVariables

1 2 3 4 5 6 7 8

1. ECPT 12. ENT labels .19 13. ENT.causes .26* .34** 14. EMT-A Part 1 .01 -.16 -.12 15. EMT-A Part 2 .19 .12 -.11 -.11 16. EMT-A Part 3 .34** .38** .23 .28* .11 17. EMT-A Part 4 .10 .21 -.04 .15 .21 .34** 18. EMT-A total .32** .23* .01 .58** .44** .79** .60** 1

Notes: ECPT is Emotion Cause Production Task. ENT labels and causes are Emotion NarrativeTask labels and causes. EMT = Emotion Matching Task.* p < .05, ** p < .01.

Table 4. Pattern Matrix From Principal Components Analysis

Factor label

Picture labelling andemotion identification

Emotion labelling andcausal understanding

WPPSI-III .60 —EMT-A Part 1 .72 —EMT-A Part 2 .60 —EMT-A Part 3 .50 .45EMT-A Part 4 .72 —ECPT — .84ENT labels — .74ENT causes — .52

Note: ECPT is Emotion Cause Production Task. ENT labels and causes are Emotion NarrativeTask labels and causes. EMT = Emotion Matching Task; WPPSI-III = Wechsler Preschool andPrimary Intelligence Scale – Third Edition.

Components of Children’s Emotion Knowledge 101

© Blackwell Publishing Ltd. 2012 Social Development, 22, 1, 2013

of much other research (Beck et al., 2012; Izard et al., 2011; Van Bergen & Salmon,2010a,b). The EMT-A subtests and the total score of that test were strongly correlated,but the associations among the other tasks were of weak to moderate strength only, andthe amount of shared variance peaked at around 15 percent. The strength of associa-tions was weakened further among all tasks, including the EMT-A and its subtests,following removal of the influence of age and language ability, yet overall the generalpattern of associations among the various tasks remained remarkably similar. It seems,therefore, that there was relatively limited overlap among the tests (other than theEMT-A subtests with its total), and that some of this overlap, but by no means all, wasaccounted for by age and language ability.

Our findings suggested that labelling emotion expressions was relevant to otheraspects of emotion knowledge. This was demonstrated, first, by the moderate associa-tions between children’s ability to label an emotion expression (in a photograph orpicture book) and their performance on tasks assessing their generation of emotioncauses, and second by the finding that the Part 3 (emotion expression labelling) of theEMT-A was the only task to load on both factors yielded by exploratory factor analysis.This pattern is not surprising; to generate an appropriate emotion cause necessarilyentails an emotion label, although the reverse is not the case. It is also likely that theyoung children in our study were more easily able to generate labels than causes ofemotions, as language is used to label emotion at a younger age than to expresscausality (Bretherton, Fritz, Zahn-Waxler, & Ridgeway, 1986).

On the whole, significant associations were found among similar tasks, whether withrespect to the emotion skill involved (labelling, generating causes) or the mode ofassessment (pictorial, verbal/narrative stimuli). Moreover, whereas tasks involvingpictorial stimuli (including the vocabulary subtest of the WPPSI-III) loaded on onefactor yielded by exploratory factor analysis, those entailing more abstract stimuli andtasks (verbal/narrative and causal understanding) tended to load on the other (with theexception of emotion labelling). This pattern may reflect the method of elicitation; thepictorial stimuli tended to serve as their own cues whereas, to a greater extent, the moreabstract tasks required the child to generate a response. It seems, therefore, that boththe content and mode of presentation influenced the children’s performance (Gelman,1982; Widen & Russell, 2010a,b).

Study 2

The first study suggested that when evaluating the potential benefits of increasingemotion talk between adults and young children, we need to ensure that the children areassessed with multiple measures of the likely different components of emotion talk. Ifwe intervene with an experimental manipulation (training as an analogue for parent–child engagement in emotion talk), we can investigate further whether a particular typeof emotion talk enhances only similar components of the child’s emotion knowledge,or whether there is an effect spreading across other components.

An experimental training paradigm enables us to disentangle a confound in natu-ralistic research investigating the association between adult–child talk and children’semotion knowledge. That is, unclear from current findings is whether causal talk aloneor causal talk in the context of emotional experience is critical. One possibility is thatcausal talk alone is sufficient. Causal understanding plays a key role in the acquisitionof knowledge, and findings highlight the centrality of causal discourse in children’slives (Bartsch, Campbell, & Troseth, 2007). Analyses of parent–child conversations

102 Karen Salmon, Ian M. Evans, Sophie Moskowitz et al.

© Blackwell Publishing Ltd. 2012 Social Development, 22, 1, 2013

show that causal-explanatory utterances occur frequently from the age of two to threeyears. Moreover, young children adopt specific strategies to obtain causal informationin conversations with adults (Frazier, Gelman, & Wellman, 2009; Hickling & Wellman,2001). Children’s involvement in conversations of causality, not necessarily restrictedto emotion, may reflect an increasingly comprehensive understanding of human behav-iour (Brown & Dunn, 1996). Nonetheless, causal-explanatory conversations thatexplicitly link emotion words to events in the child’s life might be particularly impor-tant with respect to labelling emotional experience, and making causal relationsexplicit and salient (Nelson, 2007). Indeed, maternal explanations of the causes ofemotions are a stronger predictor of emotion knowledge than the overall frequency ofemotion talk or general talkativeness (Brown & Dunn, 1996; Van Bergen & Salmon,2010b). The second study compared the effect of training based on those two processeswith a no-training control condition.

We expected that emotion causal training would have greater impact on emotionknowledge than non-emotion relevant causal training, but that both would have greaterbenefit for emotion knowledge than no training. Moreover, as emotion labels andcauses are both implicated in emotion cause talk, but emotion labelling is an easier taskthan emotion cause understanding, the impact of training may be most evident inchildren’s increased use of emotion labels.

Study 2: Method

Participants and Design

From the 88, three- to four-year-old children involved in Study 1, 53 children wereselected. An important criterion was to ensure that the children’s scores were notat ceiling with respect to their overall emotion knowledge or emotion situationknowledge. Inclusion criteria, therefore, required that the child received an overallscore on one measure of emotion knowledge, the Emotion Matching Task (EMT-A)(Morgan et al., 2010), of less than 16 (out of a possible 24), or 3 or less (out of apossible 6) on Part 2 (the subtest of emotion situation knowledge). Exclusion criteriawere also that English was a second language, although this was the case for only twochildren. There were no differences between the included and excluded children ineither age or WPPSI-III score (both ps > .75).

The children were randomly assigned to one of the three experimental groups, butassignment was balanced within preschools and for gender. As much as possible,different experimenters (of four) conducted assessments and interventions, and werebalanced across training conditions. The three conditions were: emotion cause n = 18;7 boys (M = 42, SD = 3.8 months; 11 girls, M = 42, SD = 3.0 months); non-emotioncause n = 18; 8 boys (M = 44, SD = 3.3 months; 10 girls, M = 43, SD = 3.6 months);and no-training control n = 17; 6 boys (M = 44, SD = 4.8 months; 11 girls, M = 43,SD = 4.0 months).

Measures and Procedure

All pretest measures have been described in Study 1. We did not repeat the ECPT posttraining because our prior research has shown that even for four-year-old children,improvements are not evident until six months after a training intervention (VanBergen et al., 2009). Several other changes were made for the post-test. Firstly, in orderto reduce familiarity and practice effects (and given the high correlations between the

Components of Children’s Emotion Knowledge 103

© Blackwell Publishing Ltd. 2012 Social Development, 22, 1, 2013

measures), the EMT-B was substituted for the EMT-A (Morgan et al., 2010). Secondly,to increase the number of post-test measures, an abbreviated version of the affectiveknowledge task (Denham, 1986) was administered.

Abbreviated Affect Knowledge Test (Denham, 1986). The full Affect Knowledge Test(AKT) has good reliability (Cronbach’s alpha = .95; Denham, 1986) and is usedfrequently with children at the age of two to five years (Denham et al., 1994). Aschildren had completed the EMT-B, we omitted the affective labelling component, andhalved the number of items in the two subtests assessing emotion situation knowledge.The shortened version contained four vignettes in which the puppets displayed ‘stere-otypical’ emotional responses (e.g., scared when alone in the dark), and six vignettesdisplaying ‘non-stereotypical’ emotional responses, determined by a parent question-naire. The child is asked how the puppet is feeling and is encouraged to place one offour faces (happy/sad/angry/scared) on the puppet. Following Denham (1986), chil-dren received two points for a correct answer, one point for a response of the correctemotional valence and no points for incorrect valence.

Training

Children in the two experimental groups received four individual training sessions overa two-week period, during which the research assistant and child read two storiesfollowing standardized scripts matched in word length (one from the Frog series, andone from the Little Critters series, print removed). The same two books per sessionwere read in each condition (eight in total). For children in the emotion cause condi-tion, the scripts emphasized emotions, the labels, and their causes whereas for those inthe non-emotion cause condition, they emphasized non-emotional actions and theircauses. For example, with respect to the following section from A Boy, a Dog, and aFrog (Mayer, 1967), the script for the emotion cause condition was as follows: ‘Tommystarts to run down the hill with his net in his hand. He is looking at the frog and he hasa big smile on his face. How is Tommy feeling here? (Yes, that could be it or . . .). Hefeels happy, doesn’t he? How come Tommy’s feeling happy? (Yes, that could be it or. . .) Tommy feels happy because he’s going to catch a frog’. For the same section, thescript for the non-emotion cause condition was as follows: ‘This is just what Tommywas looking for! Tommy lifts up his bucket and his net and starts racing towards thefrog. What is Tommy doing now? (Yes, that could be it or . . .) He’s running down thehill. How come Tommy is running downhill? (Yes, that could be it or . . .) Tommy isrunning downhill because he is going to catch the frog’.

In both training conditions, the children’s participation was encouraged by thestandardized prompting and questioning, and by including four personal questions.For example, with respect to the book Just Me and My Little Brother (Mayer, 1991),in the emotion cause condition, having established that the characters were scared,the child was asked: ‘Tell me about a time when you were feeling scared? How comeyou felt scared when _? Sometimes you might feel scared because you see some-thing big and frightening. Ok, so when _ happened you felt scared because _’. Forchildren in the non-emotion cause condition, having established that the characterswere watching a movie, the child was asked: . . . Tell me about a time when youwatched a movie. How come you were watching a movie? Sometimes you mightwatch a movie because it’s raining and you can’t play outside. Ok, so you watched_ movie because _’.

104 Karen Salmon, Ian M. Evans, Sophie Moskowitz et al.

© Blackwell Publishing Ltd. 2012 Social Development, 22, 1, 2013

Study 2: Results

Preliminary Analyses

There were no pre-training differences between conditions in age or language skill(WPPSI-III vocabulary scores) nor any interactions with gender, all Fs < 1.67, allps > .20. There were also no differences between the conditions in the number of daysbetween pre- and post-testing (M = 66, SD = 17 days); between the two trainingconditions in the number of days between the child’s first training session and post-test(M = 11, SD = 3.6 days); or the number of days between the last training session andpost-test session (M = 3, SD = 2 days), all Fs < 1.51, all ps > .23.

Effects of Training

Table 5 presents the mean scores (and standard deviations) for all tasks before and aftertraining, by condition. Repeated measures analysis of variance, with time (pre-, post-training) as the within-participants factor, and condition as the between-participants

Table 5. Means (and Standard Deviations) of Key Variables Before and AfterTraining

Pre-training

No trainingcontrol Causal talk

Emotion +causal talk

Age (months) 43.47 (4.16) 43.11 (3.41) 41.89 (3.29)Language (WPPSI-III) 11.18 (2.49) 12.11 (2.19) 11.28 (2.61)EMT-A total 12.59 (3.74) 12.94 (1.15) 12.50 (2.96)

Part 2 1.88 (1.05) 2.50 (1.15) 2.94 (.83)Part 3 2.88 (2.21) 3.00 (2.09) 2.53 (1.91)

Emotion Narrative TaskSpontaneous causes + labels .06 (.24) .12 (.33) .56 (.81)Prompted causes + labels 5.47 (3.30) 7.59 (1.97) 7.56 (4.26)

Post-training

EMT-B total 14.53 (4.26) 13.39 (4.71) 14.06 (4.67)Part 2 3.24 (1.60) 3.56 (1.15) 3.12 (1.58)Part 3 3.29 (2.11) 2.83 (1.82) 3.82 (1.33)

Emotion Narrative TaskSpontaneous causes + labels 2.53 (2.67) 1.47 (2.13) 6.38 (7.20)Prompted causes + labels 13.41(9.77) 12.53 (9.93) 13.75 (6.48)

AKT: total score 14.24 (2.73) 16.17 (2.57) 15.76 (3.27)

Notes: For EMT-B and Denham’s AKT, ns = 17 (emotion cause); 18 (non-emotion cause);17 (control). For Emotion Narrative Task, ns = 16 (emotion cause); 17 (non-emotion cause);17 (control). AKT = Affect Knowledge Test; EMT = Emotion Matching Task; WPPSI = WechslerPreschool and Primary Scale – Third Edition Receptive Vocabulary Scale (age standardized).

Components of Children’s Emotion Knowledge 105

© Blackwell Publishing Ltd. 2012 Social Development, 22, 1, 2013

factor, assessed the influence of training on children’s performance on the emotionnarrative task. For spontaneous emotion labels and causes (combined), there weresignificant main effects of time, F(1, 47) = 27.76, p < .001, h2 = .37, and of condition,F(2, 47) = 5.75, p < .01, h2 = .20, modified by a significant interaction, F(2, 47) = 4.75,p < .05, h2 = .19. Follow-up tests indicated that, although spontaneous labels andcauses (combined) increased across conditions between pre- and post-test, children inthe emotion cause condition reported more than those in the non-emotion cause andcontrol conditions, who did not differ (see Table 5).

To establish whether the same patterns were found for labels and causes, thesevariables were analyzed separately. Follow-up analyses showed that the above patternwas essentially replicated for emotion labels [main effects of time, F(1, 47) = 25.35,p < .001, h2 = .35; condition, F(2, 47) = 5.45, p < .01, h2 = .19; interaction, F(2, 47) =4.72, p < .05, h2 = .17] [Ms (and SDs) as follows: control pre .06 (.55), post 2.24 (4.76);non-emotion cause pre .12 (.33), post 1.35 (1.97); emotion cause pre .50 (.82), post6.00 (6.97)]. For emotion causes, although the main effect of time was significant, F(1,47) = 9.22, p < .01, h2 = .16, both the main effects of condition and the interaction werenot (both Fs < 2.44, both ps > .09). At pretest across conditions, children made nospontaneous references to emotion causes, and even at post-test, the means remainedlower than .36 in all conditions. For prompted emotion language (causes and labels),only the effect of time was significant, F(1, 47) = 26.27, p < .001, h2 = .36, both otherFs < 1, ps > .60 (see Table 1).

Given the high correlation between EMT (A and B) at pre- and post-test for the controlcondition, r(17) = .81, p < .01, the impact of training on the EMT-B was investigated bymeans of parallel repeated measure analyses. The main effect of time was significant,F(1, 49) = 6.55, p < .05, h2 = .12, but the effects of intervention condition and theinteraction were not, both Fs < 1, both ps > .47 (see Table 5). Subsidiary analysesinvestigated the impact of the intervention on scores on Parts 2 (emotion situationmatching) and 3 (emotion labelling) of the EMT-B. For Part 2, there was a significantmain effect of time, F(1, 42) = 22.13, p < .01, h2 = .35, but no other significant effects(both Fs < 2.08, both ps > .1). With respect to Part 3, the main effect of time wassignificant, F(1, 49) = 4.29, p < .05, h2 = .11, and although the main effect of conditionwas not, F < 1, p > .90, the interaction between time and condition approachedsignificance, F(2, 49) = 2.96, p = .06, h2 = .11. Given our a priori hypotheses and thesignificant difference in emotion labelling between conditions at pretest, F(2, 52) = 3.09,p < .05, we conducted paired-samples t tests investigating change between pre- andpost-test. For children in the emotion cause condition, the increase in scores on Part 3between pre- and post-test was significant, t(16) = -2.81, p < .05, whereas this was notso for children in the non-emotion cause and the control conditions, both ps > .38.

Finally, we examined the influence of the training on the abbreviated version ofDenham’s AKT. The variance in both training conditions is likely to have beenconstrained for the Stereotyped Causes Task, with 11/17 children in the emotion- andnon-emotion cause conditions scoring 7 of a total possible 8. A one-way analysis ofvariance, with condition as the between-participants factor, revealed no significantdifferences between conditions on any AKT subtest (all Fs < 2.19, all ps > .12).

Study 2: Discussion

There were three major findings. Firstly, training brought about some but only a few ofthe possible changes in components of emotion knowledge. In particular, the children

106 Karen Salmon, Ian M. Evans, Sophie Moskowitz et al.

© Blackwell Publishing Ltd. 2012 Social Development, 22, 1, 2013

in the emotion cause condition, relative to those in both other conditions, showedimprovement in their emotion labelling skill. At post-test, they provided more emotionlabels with less prompting on a version of the training materials. Moreover, theymanifested improvement on Part 3 of the EMT, which required them to produceemotion labels for photographs of emotion expressions (Morgan et al., 2010), suggest-ing some generalization of these skills. By contrast, there was no differential impact oftraining on children’s production of causal language during wordless picture bookreading, nor on their overall or subtest scores of the EMT-B and AKT.

Secondly, although our findings suggested that talking about emotions and theircauses facilitated children’s ability to label emotion faces, the influence of training waslimited to this one outcome. One likely factor is that our ‘dosage’ was marginallysufficient (Hoffman, 2009). Although even relatively brief training interventions haveimproved preschool children’s understanding of minds, a key aspect of their socioe-motional functioning (e.g., Lu, Su, & Wang, 2008), the two studies showing positiveeffects of training on young children’s emotion knowledge likely involved morepowerful interventions than our own. Van Bergen et al. (2009) provided training toparents rather than directly to the children, with the result that increased emotion talkis likely to have been woven into the children’s daily lives. Moreover, the benefits oftraining on children’s understanding of emotion causes were found only after sixmonths. Gavazzi and Ornaghi (2011) developed a group-based training over atwo-month period with children who were, on average, 10 months older than ourparticipants.

Also, possibly contributing to the lower power of our intervention is the nature of ourtraining task. The advantage of experimental standardization provided by wordlesspicture books may have been offset by the demands of storytelling task, despite ourattempts to minimize these by previewing the story, scaffolding the children’s languagevia prompting, and relating the stories to the their own lives (Nicolopoulou & Richner,2007). Although storybook reading occurs frequently in children’s lives, theory andresearch indicate that children’s performance on a range of tasks is optimal when theyare personally relevant (Nelson, 2007). We may have seen a greater impact of trainingif we had engaged children in discussion about everyday interactions occurring at theirpreschool centres. The combined effect of these two factors, the dosage and trainingmaterials, appears to have posed problems for our young participants, for whom thegeneralization of abilities across tasks is difficult.

Third, consistent with predictions, the impact of training depended on the task underconsideration. Improvements shown at post-test were on ‘easier tasks’ and those wherethe children’s skills could be manifested under scaffolded conditions. That we found aboost to children’s emotion labelling skill whereas there was no effect on their pro-duction of emotion causes is consistent with findings suggesting that children uselanguage at a younger age to label emotion than to express causality (Bretherton et al.,1986). Augmenting the likelihood that the training would improve emotion labelling,children in the emotion cause condition were repeatedly exposed to the same emotionlabels (receiving a higher dosage) whereas the causal linkages varied across frames ofthe training materials for children in the emotion cause and non-emotion cause con-ditions. This repetition of emotion labels may have enabled the generalization ofchildren’s labelling ability across tasks, regardless of the stimuli.

Limitations additional to those discussed above include the use of an abbreviatedversion of Denham’s (1986) AKT to reduce the demands at post-test. The resultantreduced variance in scores may have masked the effects of the intervention. Adopting

Components of Children’s Emotion Knowledge 107

© Blackwell Publishing Ltd. 2012 Social Development, 22, 1, 2013

differing, although highly related, versions of the pre-training materials for the post-training assessment also rendered the task of detecting change more challenging.Finally, we cannot address the possibility that any language- based interaction mayimprove emotion knowledge (Pons et al., 2004). Nonetheless, our findings suggest thatthe effectiveness of emotion knowledge training depends on the convergence of a rangeof factors, including the developmental stage of the child, duration and intensity oftraining, and the materials and tasks adopted.

General Discussion

The current findings suggest that, for three- to four-year-old children, several compo-nents involving different cognitive–affective processes underlie the broad construct ofemotion knowledge. Overall, in Study 1, a modest amount of variance was shared bya range of tasks reflecting aspects of children’s emotion knowledge, with strongestassociations being between the components of the EMT-A and the total score on thattest. Overall, our findings suggest that whether consistency is found with respect topreschool children’s performance on tests of emotion knowledge depends somewhaton the materials used to assess it. This was indicated, too, by our finding that the twofactors yielded by exploratory factor analysis differed, in part, according to the modeof assessment (the nature of the materials and associated cognitive demands). It isnoteworthy that, consistent with our findings, Widen and Russell (2010a,b) havereported differences in children’s age (2.5–4 years) to label an emotion correctly froma story depending on both the stimulus (verbal or pictorial) and emotion type.

Both studies point to the importance of emotion labelling for three- to four-year-oldchildren. Study 1 suggested that emotion labelling is particularly relevant to otheraspects of emotion knowledge. Study 2 suggested that emotion labelling is the easiestskill to alter, and shows some generalization across tasks. These findings likely reflectthe development of emotion knowledge. Labelling emerges when infants are around18–20 months of age, and develops in systematic order thereafter (Widen & Russell,2010b). By contrast, the foundations of causal understanding are evident in children’slanguage at around 28 months, and plausible antecedents and consequences for a rangeof emotions are provided from three years onwards (Hickling & Wellman, 2001).

The effects of training were specific to one outcome (labelling) and not readilygeneralizeable to other skills. In a novel situation, young children may most easilydemonstrate and generalize their strongest skill, emotion labelling. That is, three- tofour-year-old children’s demonstration of emotion knowledge may be relatively boundto the context in which it is acquired (with respect to content and mode of presenta-tion), and may generalize only slowly beyond this. With greater cognitive maturity,emotion knowledge may be less influenced by the form of the assessment task,although the limited extant research suggests that even in middle childhood, correla-tions among measures of emotion competence are modest (Beck et al., 2012). Longi-tudinal research is necessary to establish changes in the relationships betweencomponents of emotion knowledge and their association with language ability. None-theless, our findings underscore the close relationship between language and emotionknowledge. Children’s performance on most tasks was significantly associated withtheir language skill, and our language-based intervention influenced a critical compo-nent of emotion knowledge. Indeed, language ability continues to have clear andsignificant associations with emotion competence for children at the age of seven tonine years (Beck et al., 2012).

108 Karen Salmon, Ian M. Evans, Sophie Moskowitz et al.

© Blackwell Publishing Ltd. 2012 Social Development, 22, 1, 2013

There are two implications of our findings. Firstly, comprehensive assessment ofemotion knowledge drawing on differing tasks using a range of modes of presentationis necessary to fully understand three- to four-year-old children’s socioemotionalfunctioning. It is conceivable, for example, that some children might incorrectly labelfacial expressions whereas others might attribute incorrect causes for correctly iden-tified expressions with different implications for remediation (Leist & Dadds, 2009).Secondly, with respect to interventions aiming to improve young children’s emotionknowledge, it appears that all key aspects of emotion knowledge need to be includedin conversational interactions. Using a range of emotion talk tasks can enhanceemotion knowledge in young children, but the effects may not necessarily generalizeacross all components of emotion knowledge.

References

Bartsch, K., Campbell, M. D., & Troseth, G. (2007). Why else does Jenny run?Young children’sextended psychological explanations. Journal of Cognition and Development, 8, 33–61.

Beck, L., Kumschick, I. R., Eid, M., & Klann-Delius, G. (2012). Relationship between languagecompetence and emotional competence in middle childhood. Emotion, 12, 503–514.

Bretherton, I., Fritz, J., Zahn-Waxler, C., & Ridgeway, D. (1986). Learning to talk aboutemotions: A functionalist perspective. Child Development, 57, 529–548.

Brown, J. R., & Dunn, J. (1996). Continuities in emotion understanding from 3–6 years. ChildDevelopment, 67, 789–802.

Cohen, J. (1988). Statistical power analysis for the behavioural sciences (2nd ed.). Hillsdale,NJ: Lawrence Erlbaum Associates.

Denham, S., Zoller, D., & Couchod, E. (1994). Socialization of preschoolers’ emotion under-standing. Developmental Psychology, 30, 928–936.

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

Denham, S. A., Bassett, H. H., & Zinsser, K. (2012). Early childhood teachers as socializersof young children’s emotional competence. Early Childhood Education Journal, 40, 137–142.

Denham, S. A., Caverly, S., Schmidt, M., Blair, K., DeMulder, E., Caal, S., et al. (2002).Preschool understanding of emotions: Contributions to classroom anger and aggression.Journal of Child Psychology and Psychiatry, 43, 901–916.

Dunn, J., Brown, J., & Beardsall, L. (1991). Family talk about feeling states and children’s laterunderstanding of others’ emotions. Developmental Psychology, 27, 448–455.

Dunn, J., & Hughes, C. (1998). Young children’s understanding of emotions within closerelationships. Cognition and Emotion, 12, 171–190.

Ensor, R., Spencer, D., & Hughes, C. (2011). ‘You feel sad?’ Emotion understanding mediateseffects of verbal ability and mother-child mutuality on prosocial behaviors: Findings from 2years to 4 years. Social Development, 20, 93–110.

Fabes, R. A., Eisenberg, N., Nyman, M., & Michealieu, Q. (1991). Young children’s appraisalsof others’ spontaneous emotional reactions. Developmental Psychology, 27, 858–866.

Field, A. (2009). Discovering statistics using SPSS (3rd ed.). London: Sage.Frazier, B. N., Gelman, S. A., & Wellman, H. M. (2009). Preschoolers’ search for explanatory

information within adult-child conversation. Child Development, 80, 1592–1611.Garner, P. W., Carlson Jones, D., Gaddy, G., & Rennie, K. M. (1997). Low-income mothers’

conversations about emotions and their children’s emotional competence. Social Develop-ment, 6, 37–52.

Gavazzi, I. G., & Ornaghi, V. (2011). Emotional state talk and emotion understanding: Atraining study with preschool children. Journal of Child Language, 38, 1–16.

Gay, M. (2009). Zou. Wellington: Gecko Press.Gelman, R. (1982). Accessing one-to-one correspondence: Still another paper about conserva-

tion. British Journal of Psychology, 73, 209–220.Hickling, A., & Wellman, H. (2001). The emergence of children’s causal explanations and

theories: Evidence from everyday conversation. Developmental Psychology, 37, 668–683.

Components of Children’s Emotion Knowledge 109

© Blackwell Publishing Ltd. 2012 Social Development, 22, 1, 2013

Hoffman, L. M. (2009). Narrative language intervention intensity and dosage: Telling the wholestory. Topics in Language Disorders, 29, 329–343.

Izard, C. E., Woodburn, E. M., Finlon, K. J., Krauthamer-Ewing, S., Grossman, S. R., &Seidenfeld, A. (2011). Emotion knowledge, emotion utilization, and emotion regulation.Emotion Review, 3, 45–54.

Laible, D., & Thompson, R. A. (1998). Attachment and emotion understanding in preschoolchildren. Developmental Psychology, 34, 1038–1045.

Leist, T., & Dadds, M. R. (2009). Adolescents’ ability to read different emotional faces relatesto their history of maltreatment and type of psychopathology. Clinical Child Psychology andPsychiatry, 14, 237–250.

Lu, H., Su, Y., & Wang, Q. (2008). Talking about others facilitates theory of mind in Chinesepreschoolers. Developmental Psychology, 44, 1726–1736.

Marin, K. A., Bohanek, J. G., & Fivush, R. (2008). Positive effects of talking about the negative:Family narratives of negative experiences and preadolescents’ perceived competence. Journalof Research on Adolescence, 18, 573–593.

Mayer, M. (1967). A boy, a dog, and a frog. New York: Dial Books for Young Readers.Mayer, M. (1969). Frog, where are you? New York: Dial Books for Young Readers.Mayer, M. (1991). Just me and my little brother. New York: Random House.Morgan, J. K., Izard, C. E., & King, K. A. (2010). Construct validity of the emotion matching

task: Preliminary evidence for convergent and criterion validity of a new emotion knowledgemeasure for young children. Social Development, 19, 52–70.

Nelson, K. (2007). Young minds in social worlds: Experience, meaning, memory. Cambridge,MA: Harvard University Press.

Nicolopoulou, A., & Richner, E. S. (2007). From actors to persons: The development ofcharacter representation in young children’s narratives. Child Development, 78, 412–429.

Pons, F., Harris, P. L., & de Rosnay, M. (2004). Emotion comprehension between 3 and 11 years:Developmental periods and hierarchical organization. European Journal of DevelopmentalPsychology, 1, 127–152.

Salmon, K., Dadds, M. R., Allen, J., & Hawes, D. (2009). Can emotion language be taughtduring parent training for conduct problem children? Child Psychiatry and Human Develop-ment, 40, 485–498.

Van Bergen, P., & Salmon, K. (2010a). Emotion-oriented reminiscing and children’s recall of anovel event. Cognition and Emotion, 24, 991–1007.

Van Bergen, P., & Salmon, K. (2010b). Parent-child reminiscing: Associations between anelaborative style and emotion knowledge. New Zealand Journal of Psychology, 38, 51–56.

Van Bergen, P., Salmon, K., Dadds, M. R., & Allen, J. (2009). Training mothers in emotion-richelaborative reminiscing: Facilitating children’s autobiographical memory and emotion knowl-edge. Journal of Cognition and Development, 10, 162–187.

Wang, Q., Hutt, R., Kulkofsky, S., McDermott, M., & Wei, R. (2006). Emotion situationknowledge and autobiographical memory in Chinese, immigrant Chinese, and EuropeanAmerican 3-year-olds. Journal of Cognition and Development, 7, 95–118.

Wechsler, D. (2004). Wechsler preschool and primary scale for children (3rd ed., AustralianStandardised Edition). Pearson: PsychCorp.

Widen, S. C., & Russell, J. A. (2010a). Children’s scripts for social emotions: Causes andconsequences are more central than facial expressions. British Journal of DevelopmentalPsychology, 28, 565–581.

Widen, S. C., & Russell, J. A. (2010b). Differentiation in preschooler’s categories of emotion.Emotion, 10, 651–661.

Acknowledgments

This research was funded by a Victoria University Research Fund grant. We are grateful to theteachers, parents, and children of the nine preschools, and also thank the senior teachers fromthe Wellington Kindergarten Association. Thanks also to Dr. Ron Fischer for statistical advice.

110 Karen Salmon, Ian M. Evans, Sophie Moskowitz et al.

© Blackwell Publishing Ltd. 2012 Social Development, 22, 1, 2013