Emotion differentiation dissected: between-category, within-category, and integral emotion differentiation, and their relation to well-being

ABSTRACT

Emotion differentiation, the ability to describe and label our own emotions in a differentiated and specific manner, has been repeatedly associated with well-being. However, it is unclear exactly what type of differentiation is most strongly related to well-being: the ability to make fine-grained distinctions between emotions that are relatively closely related (e.g. anger and irritation), the ability to make larger distinctions between very distinct emotions (e.g. anger and sadness), or the combination of both. To determine which type of differentiation is most predictive of well-being, we performed a comprehensive meta-analysis across six datasets. We examined the correlations between these three types of differentiation and several indicators of well-being (depression, emotional clarity, and self-esteem). Results showed that individuals differentiated most between very distinct emotions and least between more related emotions, and that an index computed across emotions from both the same and different emotion categories was most strongly associated with well-being indicators.

KEYWORDS:

• Emotion differentiation
• emotional granularity
• emotion categories
• emotion knowledge
• well-being

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

Peter Kuppens http://orcid.org/0000-0002-2363-2356

Notes

1. ED measures derived from this task have also been shown to correlate with ED measures derived from experience sampling methods. Specifically, dataset 6 (see below) included both the ED lab task and a week of experience sampling. Overall ICCs were computed for both the lab task data and the ESM data. Pearson’s correlation between the two indices was .22 (p < .01), meaning that both methods capture at least partially the same phenomenon.

2. To check whether the person or the couple should be the unit of analysis, we assessed the degree of nonindependence, i.e., the degree of similarity between the two members of the dyad on the variables of interest. In line with Kenny, Kashy, and Cook (2006), we assessed how strongly males and females correlated on the ED indices, on mean NA, and on the CES-D. Pearson correlations showed that there was a high level of independence within couples with regard to the ED indices and the CES-D (all p-values > .46) and a negative correlation for mean NA (r = −.27, p = .03). As such, in this dataset, the person will be treated as the unit of analysis.

3. Negative ICCs can be the result of negative correlations between the items (which is probably what happens with very high differentiators), but also of random error, and it is not possible to distinguish between the two. Moreover, a negative ICC is theoretically impossible to interpret (Giraudeau, 1996). Therefore, all negative ICCs were excluded from further analyses. This resulted in a large number of missing cases in some datasets. We examined whether the individuals with a missing ED index differed from individuals with ICC in terms of the well-being indicators. Independent t-tests showed that for between- and within-category differentiation, the no-ICC group had a significantly higher mean for most of the well-being indicators than the ICC group. For instance, for the between-category index which had the highest number of missings (as can be inferred from the lowest N in Table 3), across the whole sample, the no-ICC individuals rated significantly higher on the CES-D (t(496) = 5.37, p < .001), on mean negative affect (t(895) = 6.61, p < .001), and on the alexithymia-DIF scale (t(359 = 5.43, p < .001) than the group of individuals for which a reliable ICC could be derived. While it may seem problematic that the no-ICC group seems to have significantly better values on the well-being variables, this can in fact be expected. As discussed above, the negative values are likely a sign of higher levels of differentiation, which is in turn related to better well-being. Therefore, these findings are in line with our hypotheses.

Additional information

Funding

This work was supported by a postdoctoral fellowship from the KU Leuven and from the Flemish Fund for Scientific Research (FWO) [Fonds Wetenschappelijk Onderzoek] to the first author, the KU Leuven Research Council (grant GOA/15/003) and the Interuniversity Attraction Poles program financed by the Belgian government (IAP/P7/06).