Perceptual and Aesthetic Sensitivity to Systematically Manipulated Image Composition
Date:
Image Composition is said to influence human preference, yet it has rarely been operationalized and tested empirically in rich images. In this study, we focus on three arttheoretical principles often mentioned to underlie composition: Balance, Emphasis, and Repetition. In three online experiments, we directly manipulated and validated them separately, and tested their effects on aesthetic preference. Using a mix of generative AI and manual editing, we generated 180 images consisting of 60 motifs (realism and fantasy), with three intensity levels of each Composition principle (less,base, more). In addition to participants’ rankings, we recorded demographic backgrounds, art interest and knowledge, and the qualitative reasoning behind their choices. To verify our stimuli, our first Experiment (N=37) had participants rank image triplets from “least” to “most” for each compositional principle. Results from Plackett–Luce (PL) models showed that the manipulations performed as expected in nearly every instance, with only three exceptions. In Experiment 2 (N=152) we asked participants to rank image triplets with the same motif by their compositional quality or personal preference. While both criteria followed the expected patterns of “more” being ranked above the baseline condition and “less” being ranked below it (all PL), composition rankings showed a significantly stronger effect than preference rankings in the “more” condition (Z = 2.05, p = 0.041, Wald-type -test). To further test these effects, Experiment 3 (N=348) utilized a 2AFC design where participants chose between image pairs (across motifs) based on composition or preference. Analysis of item-level worth estimates (PL) indicated that the experimental manipulation successfully shifted rankings of composition, yet failed to produce a consistent effect on participant preference. Across three experiments, our manipulations consistently influenced participants’ rankings and choices regarding image composition but had a much weaker or negligible impact on participants’ personal preferences.
Acknowledgements: This work is funded by an ERC Advanced Grant (No. 101053925, GRAPPA) awarded to JW.
Recommended citation: Koßmann, L., Meulemans, M., Van Geert, E., De Winter, S., Willems, M. & Wagemans, J. (2026). Perceptual and Aesthetic Sensitivity to Systematically Manipulated Image Composition [Poster]. 48th European Conference on Visual Perception (ECVP), Bournemouth, UK.