Mapping mental representations of fear-relevant stimuli through similarity modeling.

Anxiety disorders involve maladaptive fear processing, yet the underlying mental representations of fear-relevant stimuli remain poorly understood. Traditional approaches have typically relied on predetermined psychological dimensions, limiting the characterization of underlying representations and consequently constraining the mechanism-based refinement of therapies. Focusing on arachnophobia as a model condition, we used a similarity-based approach to reveal mental representations of spider-related images. From 705 individuals, we analyzed 49,141 pairwise similarity judgments among 314 spider-related images, constituting the largest dataset of its kind. Multidimensional Scaling identified a 5-dimensional representational space, with four dimensions strongly predicting fear ratings and capturing complementary information beyond fear. These findings indicate that perceptual representations are structured along specific dimensions with complex information that cannot be reduced to conventional predetermined dimensions, such as fear and disgust. Our approach provides a scalable and flexible framework for studying the representational structure of fear-relevant stimuli, offering potential applications for refining exposure therapy and anxiety treatment strategies.