Method at a glance
- Modality: visual emotion induction with multi-label self-report
- Measures: Emotional granularity and felt intensity in response to standardized affective stimuli — how differentiated a person’s emotion labelling is when everyone is looking at the same pictures.
- Procedure: Normed affective images (IAPS) are presented twice each. On the first presentation the participant rates experienced valence and arousal (SAM sliders); on the second they rate how strongly they feel each of a fixed set of discrete emotions (e.g. amusement, happiness, satisfaction, sympathy, fear, anger, disgust, sadness). Granularity is the reverse-scored intraclass correlation across same-valence labels; intensity is the highest-rated label per trial, averaged.
- Also called: ED task, emotion differentiation paradigm
Strengths
- Holds the eliciting event constant across participants, so between-person differences in labelling cannot be attributed to having had different days — the control the DRM gives up.
- Runs in one session, online, with normed stimuli, so it scales.
- Separates the valence/arousal rating from the discrete-label rating by presenting each image twice, keeping core-affect ratings from anchoring the categorical ones.
Limitations
- Normed pictures reliably move valence and arousal but may not evoke specific emotions, which is precisely what a differentiation measure needs. This is the authors’ own diagnosis of their null result.
- Granularity is bounded by the label set. Eight labels (Ventura-Bort et al. 2021) appears to be too few; the studies that got it working used more labels or restricted to a single valence (Nook et al. 2018; Erbas et al. 2019; Israelashvili et al. 2019).
- One rating per image, so the repeated observations that the granularity index needs come from trials rather than from occasions — a different (and arguably weaker) source of variance than sampling a life.
- Distressing content: the Ventura-Bort implementation warned participants about mutilation and sexual imagery and let them opt out, which makes the ED subsample self-selected relative to the rest of the study (127 of 131 here, so a small effect in this case).
Connections
The laboratory counterpart to computing Emotional granularity from daily life: instead of waiting for a person’s own emotional episodes, present everybody the same ones. Adapted from Nook et al. (2018) and Israelashvili et al. (2019); the wiki holds one implementation, the online version in Ventura-Bort, Wendt & Weymar (2021).
Structurally it is Affective picture viewing with a second pass added. Pass one is the classic IAPS + SAM protocol that stands behind the wiki’s Dunn disagreement. Pass two — rate each of eight discrete emotions on the same image — is what converts it into a differentiation instrument.
The wiki’s one datum is a null, and it is informative
In 127 participants, no measure derived from this task was predicted by anything: not intensity, not arousal, not positive or negative granularity, by either factor of interoceptive/emotional self-report, with no interactions. The same participants’ DRM data produced four significant effects.
Two things follow, and the wiki should keep them apart.
The methodological reading (the authors’). The task under-specifies the emotion. Normed pictures modulate the core-affect dimensions reliably and the categorical layer unreliably, and eight labels cannot represent whatever state a mutilation photograph produces. Prior successful uses had more labels or one valence. This is a diagnosis of the implementation, and it comes with a design recommendation: when using standardized stimuli, a wider label set is required.
The substantive reading (the wiki’s, and speculative). If granularity is the resolution of a person’s own emotional life, an instrument that supplies both the event and the vocabulary may be measuring how well the supplied labels fit a supplied picture — a question about stimulus norms as much as about the person. The DRM/ED contrast in this one paper is consistent with that, and does not establish it: the two tasks also differ in number of observations, in valence coverage, and in recall vs. online rating.
Either way, the practical note for reading this wiki: a null on the ED task is not evidence of no relationship with granularity, and the paradigm has not yet demonstrated in the wiki’s sources that it can detect one.