Selection-for-prediction: rethinking early cognitive development.

For decades, core knowledge accounts have proposed that early cognition is organized by innate, domain-specific conceptual or concept-like representations of objects, physical events, number, and agents. I propose a different starting point: infants' behaviors may arise from domain-general attentional priors-evolutionarily shaped selection biases-interacting with perceptual mechanisms that organize information and support perceptual prediction, rather than from abstract, postperceptual representations. Attentional priors favor ecologically consequential information, including continuity, contact, and synchrony. When such information is disrupted, perceptual prediction error-rather than conceptual contradiction-may produce violation-of-expectation effects. Grounded in constraints on visual input and neural processing, this framework accounts for early successes and systematic failures, suggesting that development begins with perceptual-attentional mechanisms that shape selection, organization, prediction, and learning.