Supplementary data for the preprint paper 'Comparing Likert and forced-choice formats in affective image evaluation.'

Abstract: Likert ratings and forced-choice judgments measure affective responses but they rely on different judgment processes and may express person-level response tendencies differently. We compared both formats on the same 74 images (valence or emotional intensity) across four online samples (N = 3,119). To summarize shared variance across correlated individual difference measures, we derived a positive self-evaluation (PSE) composite from personality, loneliness, social anxiety, self-esteem, and life satisfaction. Likert data were analyzed with linear mixed models; forced-choice decisions with Bradley-Terry generalized linear mixed models. Across samples, aggregate image-level scores were highly stable (split-sample r ≥ .998 for valence; r ≥ .989 for intensity), and Likert and forced-choice scores were strongly correlated (r = .979 for valence; r = .900 for intensity). However, forced-choice produced a clearer separation between social and non-social images. Associations with PSE also varied by format: in Likert ratings, higher PSE was accompanied by generally higher valence and intensity ratings for positive and neutral images, whereas in forced-choice, higher PSE was expressed primarily as a stronger relative preference for social (vs. non-social) positive content. These findings suggest that Likert and forced-choice provide similar and stable stimulus norms but differ in the category distinctions they emphasize and in how individual differences manifest.