Abstract
When representing high-level stimuli such as faces or animals, we tend to emphasize their most salient features — such as a face's prominent cheekbones or a bird's pointed beak. Such ‘mental caricaturing' leaves traces in visual memory, which exaggerates these distinctive qualities when recalling and recognizing the stimulus at another time. How broadly does this phenomenon extend? In particular, might it also apply to more basic units of visual processing? Here, six experiments (N=700 Adults) explored how visual working memory exaggerates the features of even simple geometric shapes, automatically and without any task-related demand to do so. We generated a library of novel shapes, and gradually smoothed their contours — essentially altering the amount of information borne by the visible boundaries. In Experiment 1, participants saw a novel shape at the beginning of each trial; after a brief delay, they adjusted a copy of that shape to match the one they had just seen. Surprisingly, participants consistently reconstructed the shapes in “exaggerated” form — amplifying the curvature of the shapes' contours, enlarging their salient parts, and so on. Experiment 2 generalized this bias to new parameters, showing that the results were not specific to particular ranges of complexity values. Experiment 3 replicated the results with two-alternative forced choice; Experiment 4 ruled out response biases by comparing an online perception condition to a memory condition; and Experiment 5 ruled out strategic responding based on intuitive expectations about the reliability of memory. Finally, Experiment 6 amplified the effects in a serial reproduction experiment ending with a dramatically altered shape. Together, this work demonstrates a new memory distortion, whereby even the most basic units of visual processing are remembered as caricatured versions of themselves.