Electrophysiological Correlates of Processing Unattended Objects in Visual Cognition

Research is divided as to what degree visually unattended objects are processed (Lachter et al., 2008; Carrasco, 2011). The hybrid model of object recognition (Hummel, 2001) predicts that familiar objects are automatically recognised without attention. However under perceptual load theory (Lavie, 1995), when objects are rendered unattended due to exhausted attentional resources, they are not processed.
The present work examined the visual processing of images of everyday objects in a short-lag repetition-priming paradigm. In Experiments 1-3 attention was cued to the location of one of two objects in the first (prime) display, with the unattended sometimes repeated in the second (probe) display. ERP repetition effects were observed which were insensitive to changes in scale (Experiment 1) but sensitive to slight scrambling of the
image (Experiment 2). Increasing perceptual load did not modulate these view-specific repetition effects (Experiment 3), consistent with the predictions of automatic holistic processing. In Experiments 4-7 a letter search task was used to render the flanking object image unattended under high load. In Experiment 5 distractor processing was observed in ERP even under high load. In Experiments 4, 6 and 7 a pattern of view sensitive/insensitive and load sensitive/insensitive repetition effects on RT (Experiment 4) and ERP amplitude (Experiments 6, 7) were observed that were difficult to interpret
under either the hybrid model or perceptual load theory, but may reflect fast view-based and slow view-independent processing of objects.
Overall, the properties of the view-sensitive repetition effects were generally consistent with those associated with the automatic/pre-attentive processing of the holistic route of the hybrid model. However, differences between the processing of objects rendered unattended via a spatial cue or perceptual load indicate that the bottom-up driven hybrid model and perceptual load theory may benefit from the consideration of the interaction of top-down biasing of processing (Tsotsos et al., 2008).