Abstract
A cornerstone of cognitive science is that mental systems are limited: There is a maximum amount of information they can process or store, beyond which performance breaks down. Yet so far the study of such limits has focused on core systems like attention and memory— are there also limits on higher-level thinking, and can models of visual processing be adapted to account for these limits as well? Across 20 studies and over 5000 participants, we asked people to identify with imagined versions of themselves (e.g., a younger “you” vs. older “you”) and found that people can only personally identify with one imagined self at a time, a limit that occurred across decision-making, visual associative learning, and visual long-term memory. In the associative learning task, participants learned associations between imagined selves and simple shapes, then the shapes were rapidly presented with either the correct or incorrect labels and participants performed rapid match-mismatch categorizations. In the long-term memory task, participants were told that various images were assigned to each of the imagined selves, then completed a surprise recall test of these associations. In both tasks, participants exhibited standard self-reference effects when they learned associations or remembered items for only one imagined self. Yet they performed no better overall when they imagined two selves rather than just one, suggesting a limit on the self-reference effect. Despite this limit, we also discovered ways in which self-related processing is flexible. For instance, long-term memory of images associated with multiple imagined selves was better when those selves were conceptually distinct (r = 0.86, p = .006). In sum, we found that the singular self-reference effect is aptly named, since it did not increase as the number of imagined selves did. More broadly, the notion of cognitive limits may hold promise for carving thinking at its joints.