The evaluative space grid: A single-item measure of positivity and negativity

Abstract

The authors introduce the evaluative space grid (ESG), a two-dimensional grid that provides a single-item measure of positivity and negativity. In Study 1, ESG ratings of gamble outcomes were highly correlated with those obtained from conventional, less-efficient, unipolar measures, thus providing evidence for the grid's convergent validity. In Study 2, participants rated their moment-by-moment evaluative reactions to gamble outcomes with the grid every 100 ms; results replicated earlier findings that some outcomes elicit only positivity or negativity whereas others simultaneously elicit positivity and negativity. In Studies 3 and 4, the difference between the grid's positive and negative ratings of several types of stimuli and bipolar valence ratings were highly correlated, thus demonstrating the grid's generalisability and predictive validity. Study 4 also showed that ESG ratings predicted facial electromyographic activity, particularly in tasks involving strongly affective stimuli. Taken together, results indicate that the grid provides efficient, valid indices of positivity and negativity.

Acknowledgements

This research was supported by NIMH Grant P50 MH52384–01A1.

The authors thank Barbara Mellers for making available the data from Study 2 and Stephen Crites, Jonathan Levav, and Darcy Reich for helpful comments on previous versions.

Notes

¹Unless otherwise noted, all reported effects are significant at p<.05, two-tailed, and all effects involving independent variables with 3 or more levels were evaluated with the Huyn–Feldt correction. The ANOVA also revealed a main effect of response format. Post hoc tests indicated that grid ratings (M=1.63, SD=0.31) were higher than dichotomous-then-unipolar ratings (M=1.23, SD=0.15) but lower than simple unipolar ratings (M=1.70, SD=0.31).

²Participants also completed a gambles task similar to that of Studies 1 and 2. Results from the gambles task are excluded due to space constraints.

³An analogous ANOVA on the affect grid's arousal ratings yielded a main effect of valence category. Bonferroni-corrected post hoc comparisons revealed that negative, ambivalent, and positive attitude objects were rated more arousing than neutral attitude objects. No other contrasts were significant. The finding that negative and positive attitude objects were rated more arousing than neutral attitude objects replicates the curvilinear effect of valence on arousal (Lang, Greenwald, Bradley, & Hamm, 1993). The finding that ambivalent attitude objects were rated more arousing than neutral attitude objects is more novel, but is consistent with the hypothesis that ambivalence is uncomfortable (Cacioppo & Berntson, 1994).

⁴Larsen et al. (2003) also reported results from this study, but did not examine aspects of the data relevant to the results reported here (e.g., temporal stability of evaluative space grid ratings, task effects on EMG activity).