Scene-object semantic incongruity across stages of processing: From detection to identification and episodic encoding

Abstract

Visual processes are assumed to be affected by scene-object semantics throughout the stream of processing, from the earliest processes of conscious object detection to the later stages of object identification and memory encoding. However, very few studies have jointly explored these processes in a unified setting. In this study, we build upon a change detection task to assess the influence of semantic congruity between scenes and objects across three processing stages, as indexed through measures of conscious detection, object identification, and delayed recognition. Across four experiments, we show that semantically incongruent targets are easier to detect than their congruent counterparts, but that the latter are better identified and recognized in a surprise memory test. In addition, we used eye-tracking measures, in conjunction with these three behavioral indexes, to further understand the locus of the advantage observed in each case. The results indicate that (i) competition with other congruent objects modulates the effects of congruity on target detection, but it does not affect identification nor recognition memory, (ii) the detection cost of scene-congruent targets is mediated by earlier fixations on incongruent targets, (iii) neither fixation times, dwell times, nor pupil dilatation are related to the effects obtained in identification and recognition; and (iv) even though congruent targets are both better identified and remembered, the recognition benefit does not depend on the identification demands. The transversal approach taken in this study represents a challenging but exciting perspective that holds the potential to build bridges over the seemingly different but related fields of conscious detection, semantic identification, and episodic memory.