Molyneux’s answer: Situated predictive processing

Abstract

Molyneux’s problem asks whether a person blind from birth, upon gaining sight, could immediately recognize and distinguish objects by sight alone that were previously known only by touch. Historical and contemporary empirical studies have explored this question with inconclusive results due to empirical limitations. More recently, Held and colleagues (2011) found that treated congenitally blind individuals cannot immediately recognize objects previously familiar through touch. Piller and colleagues (2023) further reported the absence of visual illusions in blind and recently visually-restored individuals. Nevertheless, cross-modal mappings gradually develop post-sight restoration. These findings suggest a reluctance of the mind to make cross-modal inferences, aligning with the predictive processing (PP) framework. PP posits that the mind generates top-down predictions about sensory stimuli, updating internal models through prediction errors when expectations are not met. With no prior visual experience, generative models in congenital blind individuals fail to produce accurate predictions. PP’s representational claims have been challenged by 4E cognitivists, who emphasize embodied, embedded, extended, and enactive aspects of cognition. This paper proposes a Situated Predictive Processing (SPP) framework that integrates PP with 4E cognition through the concept of situated mental representations, offering a new perspective on the Molyneux’s problem and emphasizing the role of experience and situatedness in the gradual development of visual-tactile mappings post-sight restoration.