Object perception encodes mutually exclusive possibilities.

Abstract

Human adults represent and reason about possibilities: multiple, mutually exclusive, non-actual ways the world could be. It is debated whether this is a late-developing, human-specific capacity and whether it requires natural language. We propose that instead of a unitary high-level capacity, multiple cognitive capacities might independently represent possibilities in different ways and with different properties. Our case study is object perception, a capacity shared with infants and non-human animals. Using methods that tap specifically into visual processes, across four experiments (N = 160), we found possibility representations in vision tasks that do not involve high-level reasoning or language. In Experiments 1 and 2, we found evidence for the visual encoding of possible object identities: When there were two possible ways that features might map onto visual objects, both possible mappings were primed. In Experiment 3, we found that visual feature-binding is a necessary precondition for such priming, ruling out low-level associative and high-level reasoning-based alternatives. Finally, in Experiment 4, we discovered that the visual system cannot encode one visual object possibly being at two possible locations. This pattern helps to explain a host of previous findings from the comparative and developmental literature and suggests that the unit of analysis for possibility representations should be the cognitive capacity rather than the agent.