Abstract
Humans can rapidly and easily extract categorical information from complex natural scenes (Thorpe, Fize et al. 1996). It is unclear whether rapid object recognition is limited to conscious perception. We investigated this by recording event-related potentials (ERPs) in both conscious and unconscious conditions, while using a continuous flash suppression (CFS) paradigm to suppress the target images during the experiment (Tsuchiya and Koch 2005). We equated the histograms of the images to minimize potential low-level confounds in our study. Image contrast was adaptively controlled to ensure 50% of the images were seen during CFS. After each trial, subjects were asked whether they saw any image. Subsequently, they decided/guessed the target category (animal or not). Trials were then separated into seen and unseen. In experiment 1, accuracy was 77% vs. 49% for seen and unseen, confirming subjects were truly unaware of "unseen" images. We compared the ERP waveforms in two time windows: 150-200ms and 250-300ms. The results showed animals induced higher amplitude than non-animals in seen trials; but smaller amplitudes than non-animal images in the unseen trials (150-200ms: F (1, 15) =31.8, p <0.001; 250-300ms: F (1, 15) = 12.8, p <0.001). As a control experiment, we replaced the generic non-animal images with vehicle images, now asking subjects to discriminate between animals and vehicles. Similar to experiment 1, animal images induced bigger amplitudes than vehicle images on seen trials, but smaller amplitudes in the unseen trials (150-200ms: p <0.001; 250-300ms: p <0.001). Under CFS conditions, unlike the unseen stimuli, the seen animal stimuli produced bigger amplitudes than non-animals, opposite to previous reports without masking (Thorpe, Fize et al. 1996). Our results indicate the brain responds in a special way to animal stimuli even in unawareness, and the rapid processing of animal images might differ between conscious and unconscious conditions.
Meeting abstract presented at VSS 2014