The effect of changing size and spatial frequency content of gray-scale object images in RSVP identification tasks

Purpose. Changing an object's size is associated with changes in its spatial frequency content while the opposite does not always hold. We investigated independently as well as conjointly the effects of size and spatial frequency (SF) content changes of images on high speed entry-level object recognition. Methods. In each of five experiments, 192 RSVP sequences, with 40 objects each (72 msec/item) were presented. Subjects reported the presence or absence of a prespecified verbal target. The relation between the target and distractors was varied in one of four ways : i) Homogeneous (HOM) vs. Heterogeneous (HET)-target and distractors appear at same size/spatial frequency or not and ii) only Size (Large or Small - 5:1) or only Spatial Frequency (10 vs. 2 cpd) or a combination of both. Results. 1. In all experiments a) performance was equally good (∼75%) in the HOM conditions despite the fact that there were large differences when the same images were presented as single trials. b) In the HET conditions performance improved (∼85%) for large size or low bandpassed targets and degraded (∼60%) with small size or high bandpassed targets. 2. This asymmetry in performance persisted even when masks were inserted after each image . 3. When size changes contradicted spatial frequency changes size changes determined performance, but detection of small targets in the HET conditions became more difficult. Conclusion. The results are consistent with the notion that the visual system sets up spatiotemporal "channels" dynamically which operate in restricted spatial frequency domains (e.g., Hughes, 1990;1986). The asymmetries highlight a bias towards processing low frequencies or large sizes when the content of the visual field changes rapidly. Our results provide evidence which suggests that these effects, well documented in low-level psychophysical studies, are preserved in higher level object recognition tasks.