Abstract
We investigated whether semantic expectations initiated by words influence object detection and, if so, whether effects are mediated by feedback from high to low levels or by high level activation involved in object detection. Participants viewed briefly presented displays divided into two equal-area regions by a central border; a familiar object (upright/inverted) was suggested on one side. Displays were masked to shorten reentrant processes. Participants reported which side they perceived as figure. Before each display, experimental group participants viewed a word denoting either the familiar object at a basic level (BL) or an unrelated (Unr) object in a different category. Control group participants viewed test displays only. We assessed object detection via speed/accuracy of reports that the object lay on the familiar configuration side. Prime effects were indexed by differences in these DVs between experimental and control groups. If semantic activation influences object detection via feedback to lower feature levels, BL primes should increase, and Unr primes should decrease, object detection accuracy. In Study 1 (90-ms displays) BL primes increased accuracy and reduced RTs for upright displays only, ps<0.04. No effects of Unr primes were observed. Hence, feedback to lower levels does not mediate semantic expectation effects. Results suggest the BL prime boosts activity in the neural population representing the familiar object, a population response that takes longer for inverted than upright objects. Study 2 (100-ms displays) allowed more time for display-generated semantic activation to build and to potentially conflict with activity initiated by the Unr prime. Following BL primes, accuracy increased for both orientations, ps< 0.01. Following Unr primes, detection RTs were longer for upright displays, p< 0.03, yet accuracy was unaffected, suggesting that RTs were longer because semantic conflict had to be resolved before object detection. These results implicate high level semantic activation in object detection.