Abstract
Prior expectations influence how we recognize objects. As suggested by recent evidence, this may be done by altering internal representations. However, how expectations of complex everyday objects affect representations remains largely unknown. Such objects are composed of multiple features that may be affected differently. For example, more generic low-spatial-frequency features could be represented when there are no specific expectations about the incoming object; when there is an expectation, subjects might focus on more specific high-spatial-frequency features to try to confirm their expectation. In the present study, subjects had to perform a 4AFC object categorization task. In the expectation condition, an object name was shown prior to the object image and indicated the most likely object to appear next (with 50% validity); in the no-expectation condition, a random string of letters appeared prior to the image. We randomly sampled spatial frequencies (SFs) across 400 ms on each trial. After reverse correlating accuracy with SFs shown at each moment for each condition, we observed that low SFs (~1-25 cycles/image) throughout recognition were significantly more used to categorize objects when there were no expectations than when there were valid expectations (p < .05), indicating that subjects focus on coarser features when they have no specific expectation. We further observed that there was significant variance in the use of high SFs (~35 cycles/image) late during recognition across object expectations (p < .05), indicating that subjects alter their representation in specific ways depending on their specific prior expectation. In summary, subjects focus on generic coarse features when they have no expectation, and they use fine features differently depending on the specific expectation. These results reveal the mechanisms underlying the effects of expectations on the recognition of real-world complex objects.