Journal of Language and Social Psychology · PDF file262 Journal of Language and Social Psychology 32(3) emotion words: happiness, sadness, anger, fear, disgust, and surprise” ...

http://jls.sagepub.com/Psychology

Journal of Language and Social

http://jls.sagepub.com/content/32/3/261The online version of this article can be found at:

 DOI: 10.1177/0261927X12461004

October 2012 2013 32: 261 originally published online 8Journal of Language and Social Psychology

Mary H. Kayyal and James A. RussellEquivalent

Arabic Translations Are Not−Language and Emotion: Certain English  

Published by:

http://www.sagepublications.com

at: can be foundJournal of Language and Social PsychologyAdditional services and information for

  http://jls.sagepub.com/cgi/alertsEmail Alerts:

http://jls.sagepub.com/subscriptionsSubscriptions:  

http://www.sagepub.com/journalsReprints.navReprints:  

http://www.sagepub.com/journalsPermissions.navPermissions:  

http://jls.sagepub.com/content/32/3/261.refs.htmlCitations:  

What is This? 

- Oct 8, 2012OnlineFirst Version of Record  

- Feb 7, 2013OnlineFirst Version of Record  

- Aug 16, 2013Version of Record >>

at BOSTON COLLEGE on November 21, 2013jls.sagepub.comDownloaded from at BOSTON COLLEGE on November 21, 2013jls.sagepub.comDownloaded from

Journal of Language and Social Psychology32(3) 261 –271

© 2013 SAGE Publications DOI: 10.1177/0261927X12461004

http://jls.sagepub.com

462574 JLS32310.1177/0261927X12461004Journal of Language and Social PsychologyKayyal and Russell

1Boston College, Chestnut Hill, MA, USA

Corresponding Author:Mary H. Kayyal, Department of Psychology, Boston College, 301 McGuinn Hall, 140 Commonwealth Avenue, Chestnut Hill, MA 02467, USA.Email: [email protected]

Language and Emotion: Certain English–Arabic Translations Are Not Equivalent

Mary H. Kayyal1 and James A. Russell1

Abstract

Happiness, sadness, and anger are translated into Arabic as farah, huzn, and ghudub, respectively, by the translation–back translation method. But are these translations equivalent? To be equivalent, they must have the same referents, specifically, show a high correlation between profiles of endorsement and a similar breadth of endorsement when used to refer to emotions. Here, English-speaking Americans, English-speaking Palestinians, and Arabic-speaking Palestinians (N = 60, 60, and 42, respectively) rated the extent to which each of 12 words referred to the various emotions conveyed by 22 facial expressions. Only one translation (happiness–farah) passed both tests of equivalence. All others differed with culture or language.

Keywords

language, emotion, words, translation, Arabic

Much psychological theory and research on emotion presupposes that common Eng-lish words for emotions and feelings—such as happiness, sadness, anger, and fear—can be translated into other languages. In Spanish, they are translated as felicidad, tristeza, ira, and miedo, respectively, and in Arabic as (written in Latin script) farah, huzn, ghudub, and khof, respectively. Ekman (1973) expressed this presupposition of equivalence: “Regardless of the language, or whether the culture is Western or East-ern, industrialized or preliterate, these facial expressions are labeled with the same

Article

at BOSTON COLLEGE on November 21, 2013jls.sagepub.comDownloaded from

262 Journal of Language and Social Psychology 32(3)

emotion words: happiness, sadness, anger, fear, disgust, and surprise” (pp. 219-220). This claim could make sense only if emotion words have exact equivalents in all lan-guages. But do they?

Many English emotion words have been translated into other languages, and the translation has been verified through the back-translation test. One set of bilinguals translates an English word into, for example, Arabic, and another set of bilinguals translates the Arabic word back into English; equivalence is claimed if the process returns us to the original word. On the other hand, linguistic and ethnographic research has questioned whether English emotion words have translation equivalents in all other languages (Russell, 1991; Wierzbicka, 2009). There are now hints that the trans-lation may not be equivalent to the original, even when it passes the back translation test. For example, shame and its common translation into Spanish, vergüenza, do not refer to emotions with identical features (Hurtado de Mendoza, Fernández-Dols, Parrott, & Carrera, 2010). Thus, back translation may ensure that the translation is the best available but not that two are equivalent.

To be equivalent, two emotion words must refer to the same events. Following Russell and Sato (1995), we propose that two emotion words must refer to the same facial expressions, on the assumption that facial expressions are the prime candidates for universal signs of emotion. A word refers to the emotion conveyed by a face not in an all-or-nothing manner but to a certain degree (Russell & Bullock, 1986). Consider the American data in Figure 1 which shows the degree to which sadness refers to the

Figure 1. Mean percentage of observers in each sample who attributed the word sadness to each of the 22 facial expressions.Note. Each facial expression is identified by its plate number (in parentheses) and the emotion it was predicted to convey by Ekman (1980). The facial expressions are ranked along the x-axis in decreasing order (based on combined mean across the three samples.

at BOSTON COLLEGE on November 21, 2013jls.sagepub.comDownloaded from

Kayyal and Russell 263

emotion conveyed by each of 22 different facial expressions. There is no sharp divi-sion between facial expressions that convey sadness and those that do not. We there-fore characterize the relation of a word to the set of facial expressions as a profile. Equivalence of reference to facial expressions, then, entails two specific criteria. First, the two words must show a high correlation between their profiles. A word and a per-fect synonym or a perfect translation have the same profiles. Second, the two profiles must show the same breadth. That is, they must show an overall equivalent level of endorsement for the facial expressions. We quantified breadth as the mean endorse-ment across faces.

Study OverviewWe reanalyzed data from a study reported by Kayyal and Russell (2012) in which Americans responded in English and bilingual Palestinians responded in either English or Arabic by rating the extent to which each of 12 words referred to the emo-tion or feeling expressed in each of 22 facial expressions. Having these three groups of respondents allowed us to disentangle effects due to language from those due to culture. The facial expressions were published by Ekman (1980) to illustrate the abil-ity of faces to communicate different emotions and feelings universally. There were 12 English words—happiness, sadness, anger, fear, interest, relaxation, embarrass-ment, disgust, contempt, surprise, perplexity, and hesitancy—which were the emo-tions and feelings that Ekman said were expressed by these facial expressions. Translation–back translation was used to obtain the best available Arabic translation for each word.

MethodDetails of method are provided by Kayyal and Russell (2012).

ParticipantsThe American sample consisted of 60 Boston College undergraduates (18-25 years, 17 male) who completed the survey in English. The Palestinians-in-English sample consisted of 60 bilingual Palestinians (18-19 years, 17 male) who completed the same survey in English. The Palestinians-in-Arabic sample consisted of 42 bilingual Palestinians (18-19 years, 20 male) who completed the same survey in Arabic.

Photographs of Facial ExpressionsPhotographs of facial expressions were 22 black-and-white, approximately 2 inches × 3 inches, frontal views of faces (Ekman, 1980). Eighteen were said by Ekman to sig-nal a single emotion or feeling: happiness (Figure 1: Plates 37B and 8-furthest), sad-ness (14), anger (17), surprise (7, 8-middle), disgust (16), interest (22, 23, 26A, 26B),

at BOSTON COLLEGE on November 21, 2013jls.sagepub.comDownloaded from

264 Journal of Language and Social Psychology 32(3)

embarrassment (36, 33G), relaxation (45), perplexity (8-closest, 20, 21), and hesi-tancy (3). The remaining four were said to signal a blend of two emotions: fear/disgust (12), fear/surprise (9), interest/fear (25), and perplexity/anger (19).

Translation of Emotion WordsThe twelve English emotion or feeling words were translated into Arabic using a standard translation–back translation procedure. For each word, a list of possible Arabic translations taken from standard English–Arabic dictionaries and bilingual Palestinians was presented to five bilingual speakers, who indicated the single closest and most commonly used Arabic translation for each English word. The closest Arabic synonyms for each of the 12 English words were then presented to a different set of five bilinguals who were asked to back translate the Arabic word into English. The back translation that returned to the original English word by the most judges was chosen as the translation. At least three of the five judges had returned the original English word in all 12 cases.

ProcedureEach participant received a booklet with general instructions, followed by 22 separate sheets each showing a single facial expression and a rating form. There were four dif-ferent orders of presentation of the faces (two random and two reversed). For each face, participants indicated whether each of the 12 emotions was present or not by circling either yes or no. If the participant circled yes, he or she indicated “the intensity to which the emotion was present in the face” by circling a number from 1 (barely) to 6 (maximum intensity).

ResultsInterrater Reliability

With the exception of one word (perplexity) for one sample, interrater reliability for each word was adequate, ranging from moderate to high. Table 1 (columns 4-6) gives the results of a two-way random effects model of intraclass correlation coefficients. There were differences between samples: On average, Americans (mean = 0.94) used the emotion words more reliably than either the Palestinians-in-English (0.79) or the Palestinians-in-Arabic (0.77; t = 3.29 and 3.90, respectively; all ps < .01), who did not differ from each other (t = 0.29, p = .77).

Correlation of ProfilesFor each emotion word, a profile was created as the mean rating within a sample for each of the 22 faces. Table 1 (columns 7-9) shows the correlation between the profiles

at BOSTON COLLEGE on November 21, 2013jls.sagepub.comDownloaded from

Kayyal and Russell 265

for each word and its counterpart. For example, Table 1 shows that the profile for the English word happiness as used by Americans (a) correlated .99 with the profile for the English word happiness as used by Palestinians-in-English and (b) correlated .98 with the profile for its Arabic translation (farah) as used by Palestinians-in-Arabic. While not surprising, this result is encouraging, for it shows that even with different samples of observers, very high correlations are possible.

Five words (happiness, anger, disgust, relaxation, and sadness) reached the .80 criterion set by Russell and Sato (1995; mean r = .91, ps < .001; Table 1).

Three words (surprise, fear, and interest) fell slightly short of that criterion, reveal-ing a small problem in translation not attributable to a cultural difference. Despite their cultural differences, the two groups who responded in English used the three words similarly (rs = .93, .90, and .82, respectively; all ps < .001), but despite their sharing a culture, the two Palestinian groups produced correlations slightly less than .80 (rs = .78, .76, and .65, respectively; all ps < .001). The differences between surprise and mufajeh and between fear and khof were small, were not readily interpretable, and require further testing. A possible difference between interest and ihtimam is that the English word includes low arousal states of apprehension and uncertainty more often than does its Arabic translation. The largest differences between the attribution of

Table 1. Interrater Reliability and Between-Sample Correlations for Word Profiles.

Word Interrater reliability

Between-samples correlations

English Arabic AmericansPalestinians-

in-EnglishPalestinians-

in-Arabic

Americans Versus

Palestinians-in-English

Americans Versus

Palestinians-in-Arabic

Palestinians: English Versus Arabic

Happiness farah .99 .96 .99 .99 (.98) .98 (.99) .98 (.99)Anger ghudub .98 .76 .73 .94 (.94) .89 (.93) .89 (.89)Disgust ishmi’zaz .97 .85 .94 .92 (.96) .88 (.92) .87 (.93)Relaxation istirraha .97 .85 .83 .88 (.90) .88 (.80) .88 (.83)Sadness huzn .97 .80 .93 .90 (.90) .94 (.89) .88 (.90)Surprise mufajeh .95 .66 .76 .93 (.95) .83 (.86) .78 (.88)Fear khof .95 .80 .86 .90 (.91) .72 (.74) .76 (.73)Interest ihtimam .94 .83 .72 .82 (.79) .76 (.68) .65 (.54)Contempt yhad .86 .79 .61 .79 (.77) .40 (.45) .47 (.49)Perplexity irtibak .94 .40 .65 .60 (.58) .62 (.59) .21 (.16)Hesitancy tirradud .90 .91 .58 .38 (.53) .−02 (−.06) −.30 (−.40)Embarrassment ihraj .82 .88 .69 .59 (.56) .42 (.39) .65 (.61)Mean .94 .79 .77

Note. Columns 1 to 3 give the English word and its Arabic translation. Columns 4 to 6 give, for each of the three samples, the intraclass correlation reliability coefficients for word profiles from a two-way random effects model. Correlations in boldface are significant at p < .05. Columns 7 to 9 give, for each possible comparison pair, the between-sample cor-relations for word profiles. For each comparison, high correlation coefficients indicate that the two groups in that pair applied the emotion word to the facial expressions in similar percentages. High correlation coefficients in parentheses indicate that the two groups in that pair attributed similar intensity ratings to the emotion word. All correlations in boldface are significant at p < .001.

at BOSTON COLLEGE on November 21, 2013jls.sagepub.comDownloaded from

266 Journal of Language and Social Psychology 32(3)

interest versus ihtimam were for faces predicted by Ekman (1980) to express fear, embarrassment, surprise, and hesitancy.

One pair—contempt and yhad—showed a major problem of translation not attribut-able to culture. The two groups responding in English used contempt similarly (r = .79) but not similar to the way the group responding in Arabic used its translation, yhad: r = .40 and .47, respectively. Observers’ ratings for specific faces suggested that contempt was used for faces expressing low-arousal hostility, discomfort, or annoy-ance, whereas the yhad was limited to faces expressing a more aroused anger.

The remaining three words—perplexity, hesitancy, and embarrassment—failed equivalence of profiles for all possible comparisons. Perplexity was used with only modest agreement across samples: mean r = .48. Modest correlations occurred not just for the translation of perplexity but even for the two English-speaking samples using the same word. Modest correlations for perplexity may be due to low reliabil-ity (Table 1, columns 5 and 6) for the two Palestinian samples with this word. Perhaps the concept is not readily accessible in Palestinian culture. Similarly, the words hesitancy and embarrassment were used with modest agreement across sam-ples: mean r = .39. Reliability for the English words was high (Table 1) and therefore cannot explain the modest correlations for the two English-speaking groups. Palestinians-in-English used embarrassment to refer to positive states, whereas Americans did not. The differences between Americans’ and English-speaking Palestinians’ use of hesitancy were unclear.

A similar pattern of results emerged when a profile was created with mean intensity ratings substituted for mean endorsement; Table 1 gives these correlations in parenthe-ses. Those words that were endorsed most often were also rated as most intense: r = .97, .99, .99, all ps < .001, in the three samples. Six words (happiness, anger, disgust, relaxation, sadness, and surprise) met the .80 criterion (mean r = .91, all ps < .001), two words (interest and fear) came close but failed (mean r = .67, all ps < .05), and four words (perplexity, hesitancy, contempt, and embarrassment) remained problem-atic for all comparisons (mean r = .39).

Equivalence of BreadthOur first operational definition of a word’s breadth was its mean endorsement across the 22 faces. Figure 1 illustrates how a word and its Arabic translation can differ in breadth; it shows the percentage of observers from each sample who applied the word sadness or its Arabic translation to each of the 22 faces. Figure 2 shows the breadth of each word seen in each of the three samples. Every word except happiness and its translation showed a difference in breadth. Overall, most English emotion words—whether used by Americans or by Palestinians—were applied more broadly than their Arabic counterparts.

Happiness illustrates a translation that passes the breadth test. All three groups applied the word to a similar percentage of facial expressions: 28% both by Americans

at BOSTON COLLEGE on November 21, 2013jls.sagepub.comDownloaded from

Kayyal and Russell 267

and Palestinians-in-English and 24% by Palestinians-in-Arabic (independent samples t statistics were all <1.00, all ps > .40).

Five emotion words—sadness, anger, fear, surprise, and contempt—showed a lan-guage difference without a culture difference: the English word was broader than its Arabic translation. In each case, Palestinians-in-English did not differ significantly from Americans (t tests; all ps > .30) but showed significantly greater breadth than did Palestinians-in-Arabic (t

100 = 2.14 for happiness, 4.40 for sadness, 5.21 for anger, 6.02

for fear, 3.03 for surprise, and 3.58 for contempt; all ps < .01). Palestinians-in-Arabic were least likely to apply sadness for all 22 faces, anger for 19, fear for 21, contempt for 17, and surprise for 14 facial expressions.

Two emotion words—interest and disgust —were reasonable cases of culture, but not language, differences in breadth. Americans applied these words to a significantly greater percentage of facial expressions than did either group of Palestinians, who did not differ from each other. The difference for interest occurred for all 22 faces, but for disgust for 12 faces.

Three words—relaxation, perplexity, and hesitancy—combined the two patterns just reported: Breadth was greater for English than Arabic language and for American than Palestinian culture. The three groups differed significantly from each other: Americans versus Palestinians-in-English, t

118 = 5.02, 2.81, 2.56, respectively;

Americans versus Palestinians-in-Arabic, t100

= 7.85, 5.38, and 8.30, respectively; Palestinians-in-English versus Palestinians-in-Arabic, t

100 = 3.75, 2.38, and 5.06,

respectively (all ps < .04). The percentage of faces to which relaxation, perplexity, and

Figure 2. Mean percentage of facial expressions to which each emotion word was applied by each of the three sample groups (irrespective of the specific facial expression to which each word was applied).Note. The dotted box indicates the five words for which Americans and Palestinians-in-English applied that word to an equal percentage of facial expressions but to a significantly greater percentage of facial expressions than did Palestinians-in-Arabic.

at BOSTON COLLEGE on November 21, 2013jls.sagepub.comDownloaded from

268 Journal of Language and Social Psychology 32(3)

hesitancy were applied was highest for Americans (37%, 53%, and 47%, respectively), followed by Palestinians-in-Arabic (23%, 41%, and 36%, respectively), followed by Palestinians-in-English (13%, 29%, and 14%, respectively).

Finally, embarrassment produced a pattern difficult to explain. Palestinians-in-English applied embarrassment to a significantly greater percentage of facial expres-sions (27%) than did either Americans (18%; t

118 = 2.50, p < .02) or Palestinians-in-Arabic

(16%; t100

= 3.15, p < .01), who did not differ from each other (t100

= 0.55, p = .59).A similar pattern of results emerged when the mean intensity ratings for each facial

expression were substituted for mean endorsement, although the pattern was less robust.

Multidimensional ScalingDespite differences found so far, there was also similarity. To explore that similarity, we used individual differences scaling (INDSCAL; Carroll & Chang, 1970), which is a nonmetric, weighted multidimensional scaling technique that provides a geometric representation of the similarities among stimuli—in this case, the 12 emotion words. In this study, the similarity between two words was defined as the correlation between their profiles of endorsement of the 22 facial expressions. Each sample was treated as a separate “subject,” and INDSCAL provides “weights” that assess how subjects dif-fer in their emphasis on each dimension of the space.

Number of dimensions. Stress, canonical correlations, variance accounted for, and interpretability all suggested a two-dimensional solution. Stress was substantially reduced by adding a second dimension (from .28 to .19) but not by a third (.14). All three samples shared two common dimensions, as indicated by high and significant first (r = .93-.97, all ps < .001) and second (r = .80-.93, all ps < .05) canonical correla-tions. The third canonical correlations were nonsignificant for two of the three com-parisons (Americans vs. Palestinians-in-English: r = .22, p = .53; Americans vs. Palestinians-in-Arabic: r = .02, p = .95; Palestinians-in-English vs. Palestinians-in-Arabic: r = .68, p =.03). Improvement in variance accounted for from a two- to a three-dimensional solution was relatively small (from .80 to .83). The third dimension was not readily interpretable. The two-dimensional solution is shown in Figure 3.

Interpretation. The first dimension could be interpreted as valence, contrasting unpleasant versus pleasant feelings, and accounted for 65.5% of the variance. The second dimension could be interpreted as contrasting response to novel versus familiar events. High values went to responses to unfamiliar situations (e.g., perplexity, inter-est, and surprise), low values to responses to familiar situations with established or expected reactions (e.g., happiness, anger, and sadness). This dimension accounted for 14.9% of the variance.

Sample differences. For each subject (sample), INDSCAL provides a weirdness score ranging from 0 to 1, with 1 indicating that the subject fits the space poorly. Americans, Palestinians-in-English, and Palestinians-in-Arabic had weirdness scores of .28, .02, and .39, respectively. All fit the space well, albeit with some differences.

at BOSTON COLLEGE on November 21, 2013jls.sagepub.comDownloaded from

Kayyal and Russell 269

INDSCAL also provides weights that indicate the relative importance of each dimension to each sample; weights range from 0 to 1, with 0 indicating no importance and 1 indicating high importance. Americans, Palestinians-in-English, and Palestinians-in-Arabic had weights of .76, .82, and .85, respectively, for Dimension 1 and weights of .53, .35, and .20, respectively, for Dimension 2. Thus, Americans emphasized both dimensions more equally than did either group of Palestinians, whereas the Palestinians-in-Arabic gave more than 4 times the emphasis to the first dimension than they did to the second.

Discussion and ConclusionEvidence continues to mount that when we translate a word, we risk changing its meaning. In the current study, only one word, happiness, had an Arabic translation, farah, that passed unequivocally both of the two criteria examined here: correlation of their profiles and equivalence of their breadth when used to describe the emotions

Figure 3. Emotion words in a two-dimensional space created by INDSCALNote. INDSCAL = individual differences scaling. Dimension 1 was interpreted as valence, contrasting unpleasant versus pleasant feelings. Dimension 2 was interpreted as contrasting response to novel versus familiar events.

at BOSTON COLLEGE on November 21, 2013jls.sagepub.comDownloaded from

270 Journal of Language and Social Psychology 32(3)

conveyed by faces. All other words, even though they passed the back translation test, lacked an exact translation by one or both of these criteria.

The largest differences were seen when the samples differed in both culture and language, but some differences were also seen when samples differed in culture but used the same language. Of the seven words that failed the equivalence-of-profiles test, four represented translation-only differences, none represented culture-only differ-ences, and three showed no clear pattern. For the equivalence-of-breadth tests, five words clearly represented language-only differences, two represented culture-only dif-ferences, three showed both patterns, and one showed no clear pattern. The two criteria examined here—equivalence of profiles and equivalence of breadth—are necessary but not sufficient. Thus, future research might benefit from examining further criteria.

Although differences were found between English emotion words and their Arabic translations, there was also similarity. This similarity was seen in the translation–back translation results and in the correlations reported in Table 1, which were positive for 11 of the 12 words and often exceedingly high. Multidimensional scaling suggested that similarity occurred in the valence and novelty coded by the words.

Acknowledgments

We thank Howie Giles and the anonymous reviewers for their helpful comments. We also thank Jeehye Choi, Amanda Greenwood, Kerrie Pieloch, Kristen Hewitt, Claudia Hernandez, Matthew Williams, Ciara James, Nicole Trauffer, and Marissa DiGirolamo for their help in data entry and manuscript preparation.

Declaration of Conflicting Interests

The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author disclosed receipt of the following financial support for the research, authorship, and/or publication of this article:

This research was funded by a grant from the National Science Foundation (5001525).

References

Carroll, J. D., & Chang, J. (1970). Analysis of individual differences in multidimensional scaling via an n-way generalization of “Eckart-Young” decomposition. Psychometrica, 35, 283-319.

Ekman, P. (1973). Cross-cultural studies of facial expression. In P. Ekman (Ed.), Darwin and facial expression: A century of research in review (pp. 169-222). New York, NY: Academic Press.

Ekman, P. (1980). The face of man: Expressions of universal emotions in a New Guinea village. New York, NY: Garland STPM Press.

Hurtado de Mendoza, A., Fernández-Dols, J. M., Parrott, W. G., & Carrera, P. (2010). Emotion terms, category structure, and the problem of translation: The case of shame and vergüenza. Cognition & Emotion, 24, 661-680.

at BOSTON COLLEGE on November 21, 2013jls.sagepub.comDownloaded from

Kayyal and Russell 271

Kayyal, M. H., & Russell, J. A. (2012). Americans and Palestinians judge spontaneous facial expressions of emotion. Manuscript submitted for publication.

Russell, J. A. (1991). Culture and the categorization of emotions. Psychological Bulletin, 110, 426-450.

Russell, J. A., & Bullock, M. (1986). Fuzzy concepts and the perception of emotion in facial expressions. Social Cognition, 4, 309-341.

Russell, J. A., & Sato, K. (1995). Comparing emotion words between languages. Journal of Cross-Cultural Psychology, 26, 384-391.

Wierzbicka, A. (2009). Language and metalanguage: Key issues in emotion research. Emotion Review, 1, 3-14.

Author Biographies

Mary H. Kayyal is currently a doctoral student at Boston College, working under the supervi-sion of Dr. James A. Russell. Her research focuses on the nature and development of emotion categories across cultures and languages. Recent first-authored publications include “What Made Sahar Scared: Imaginary and Realistic Causes in Palestinian and American Children’s Concept of Fear” in the Journal of Cognition and Culture, in press.

James A. Russell (PhD, University of California, Los Angeles) is a professor of psychology at Boston College. He studies various aspects of emotion and has recently proposed a conceptual framework, called psychological construction, to integrate these disparate aspects. He, along with Jaak Panksepp, was the topic of a recent book titled Categorical Versus Dimensional Models of Affect: A Seminar on the Theories of Panksepp and Russell.

at BOSTON COLLEGE on November 21, 2013jls.sagepub.comDownloaded from