Oğul Can Yurdakul¹ (), Xueyi Huang², Angela Shen¹, Emil Olsson¹, Ali Ekhlasi¹, Michaela Klímová², Jorge Morales², Megan A. K. Peters^1,3,4; ¹University of California Irvine, ²Northeastern University, ³Canadian Institute for Advanced Research, ⁴University College London

Mental imagery research has long treated vividness as a central but poorly specified construct. However, questions remain: Is ‘vividness’ objectively definable? Does it reflect a single latent dimension or multiple interacting dimensions? How does it relate to perceptual reality monitoring? These questions have theoretical implications for higher-order theories of consciousness – which disagree about whether vividness constitutes a content-invariant reality-monitoring signal or a separable higher-order dimension – and methodological implications for developing scalable, data-driven approaches for quantitatively characterizing vividness across individuals, contents, and perception versus imagination. Here, we introduce a novel computational approach for recovering the multidimensional structure underlying vividness judgments using data collected as pilot for a large, preregistered behavioral experiment (full experiment reported by Huang et al., this VSS). Participants (n = 88 in the pilot) generated mental images and rated vividness, then reproduced their imagery experience on abstract Voronoi images across three dimensions: sharpness, opacity, and color saturation. Rather than pre-specifying the form of the mapping to vividness, we use Gaussian process regression to flexibly model joint relationships between vividness and these dimensions while also capturing uncertainty in the fitted model arising from uneven sampling. Our approach estimates functions linking vividness to these dimensions to characterize nonlinear, heterogeneous mappings between reconstruction dimensions and vividness, and individual differences in how these dimensions contribute to vividness reports. Crucially, our approach avoids parametric assumptions of linear models and enables recovery of potentially nonlinear manifolds linking these interacting dimensions. Preliminary results reveal peaks in the joint space which suggest nonmonotonicity and individual differences in how these features define vividness.Our framework offers a generalizable, theory-flexible approach for quantifying the latent structure of imagery vividness and its relationship to perceptual dimensions, facilitating future psychophysical and neuroimaging experiments characterizing imagery’s relationship to perception and the neurocomputational processes underlying reality monitoring.

Acknowledgements: This work was supported by the Templeton World Charity Foundation, Inc. (funder DOI 501100011730) under the grant https://doi.org/10.54224/22032 (to JM & MAKP) and in part by the Canadian Institute for Advanced Research (CIFAR, Fellowship in Brain, Mind, & Consciousness; to MAKP).