Abstract
Prior expectations have been shown to affect object recognition. However, it is unknown whether the expectation of a specific object modulates how information is sampled across time during object recognition. Coarse information (low spatial frequencies, SFs) is typically sampled before more detailed information (high SFs). Some authors have suggested that low SFs processed early activate expectations about the object’s identity and that high SFs processed afterwards allow the confirmation and refinement of this hypothesis; an existent expectation could therefore reduce the need for confirmatory high SF information. In this study, we verified whether expectations influenced how SFs are used across time to recognize objects. On each trial, one object was randomly chosen among eighty, and all its SFs were randomly sampled across 333 ms. In half the trials (expectation condition), an object name was shown before the object; in the other half (no-expectation condition), it was shown after. Subjects had to indicate whether the name matched the object; it did so on 50% of trials. We first observed, after reverse correlating accuracy with SFs shown at each moment, that the early use of low SFs (1–30 cycles/image) was increased in the expectation condition. We then found that the late use of high SFs (~35 cycles/image), although not visible in the average results, was correlated with general recognition ability in the no-expectation condition, more so than in the expectation condition. Finally, we found that the early use of high SFs (~35 cycles/image) was affected differently by different expectations (i.e. different object names). Together, these results reveal how the processing of sensory information sampled across time during object recognition is modulated by expectations and they support the hypothesis that low and high SFs are affected differently by this modulation.