Stephen M. Fleming and Matthias Michel’s proposal regarding the function of visual consciousness is predicated on the claim that it is “surprisingly slow.” Accumulating findings over the last 25 years are radically challenging the view that visual perception is arranged hierarchically and thus slow (Duncan & Barrett, Reference Duncan and Barrett2007; Gilbert & Sigman, Reference Gilbert and Sigman2007; Pessoa & Adolphs, Reference Pessoa and Adolphs2010). The emerging view is of a visual brain composed of multiple parallel processing pathways to provide fast but coarse visual processing, which was critical in the ancestral environment of evolutionary adaptiveness. The hundred-fold expansion of visual range associated with the sea-to-land transition meant far broader visual exposure to survival threats that could strike fast and without warning. The postdictive psychological effects that the authors rely on to substantiate the claim that visual perception is slow lack ecological validity; they are not relevant to the evolutionary context. This undercuts their proposal that visual consciousness evolved to support reality monitoring in the broader visual horizons associated with the sea-to-land transition.

The speed of visual processing can be accurately assessed by measuring and comparing the response latencies of brain regions. Pessoa and Adolphs (Reference Pessoa and Adolphs2010) gathered this data in the macaque cerebral cortex and showed that visual response latencies are surprisingly short, indicating very fast cortical processing. In both the primary visual area V1 and the medial temporal area V5 for motion perception, the earliest response latencies are <40 ms and mean latencies are 67–72 ms. Even in the inferotemporal cortex, the so-called “late” visual cortex for object recognition, latencies are as short as 60–85 ms (Lamme & Roelfsema, Reference Lamme and Roelfsema2000), and in the frontal eye field regions as short as 40–70 ms (Schmolesky et al., Reference Schmolesky, Wang, Hanes, Thompson, Leutgeb, Schall and Leventhal1998). Strikingly, these latencies overlap with those of area V1 and with the subcortical pulvinar (60–80-ms; Ouellette & Casanova, Reference Ouellette and Casanova2006).

Computational models based on a hierarchical model of vision do not fit these surprisingly short latencies, especially those of higher-order visual regions (Capalbo, Postma, & Goebel, Reference Capalbo, Postma and Goebel2008). They are better explained by the existence of multiple visual pathways, including several “short-cuts” in the visual ventral stream to generate faster transmission to the inferotemporal cortex. For example, the lateral geniculate nucleus (LGN), which transmits visual signals from the retina to primary area V1, also has direct connections to the secondary area V2 and to the medial temporal area V5 for motion perception. Other shortcuts link V1 to later visual and inferior temporal areas in the ventral stream (Nakamura et al., Reference Nakamura, Gattass, Desimone and Ungerleider1993; Rempel-Clower & Barbas, Reference Rempel-Clower and Barbas2000). Longer-range shortcuts connect regions in the visual ventral stream with the orbitofrontal cortex, which is critical to evaluating the biological significance of stimuli (Rempel-Clower & Barbas, Reference Rempel-Clower and Barbas2000).

Pessoa and Adolphs (Reference Pessoa and Adolphs2010) mapped these visual pathways to demonstrate how the initial processing of visual signals occurs along multiple parallel channels simultaneously in a rapid, feedforward “cortical sweep” that allows higher-order visual regions to respond with remarkably short latencies. The disadvantage of a parallel distributed architecture is that initially, it may provide relatively coarse information about a stimulus. It would be evolutionarily adaptive, however, to prioritize coarse information processing to provide faster perception of global properties of a stimulus. Studies consistently find that humans see “the forest before the trees,” perceiving the global properties of a stimulus significantly earlier than its local features (Kimchi, Reference Kimchi1992). Behavioral and electrophysiological studies suggest perception of global stimulus properties is remarkably fast (e.g., Li et al., Reference Li, VanRullen, Koch and Perona2002; Thorpe, Fize, & Marlot, Reference Thorpe, Fize and Marlot1996). Greene and Oliva (2009) found that just a 19–67 ms flash of a scene was needed for observers to attain 75%-correct classification of its global property (e.g., “natural scene”) and basic level categorization (e.g., “forest”).

These perceptual distinctions should be kept in mind when considering the postdictive effects reported by the authors, indicating conscious perception between 200–400 ms. These effects are generated by rather unusual stimuli, like the Poggendorff illusion and Vernier (see article Figure 2). It is not surprising that the perceptual phenomena produced by these complex stimuli take time; they naturally lead to the conclusion that visual consciousness is “relatively sluggish.” Similarly, in the example of metacontrast masking (p. 7), the target stimulus is often a small dot or something similar, and the masking stimulus is an annulus surrounding the dot. Thus, the annulus mask can be presented for 70 ms after the dot and render the dot invisible. It is not clear, however, how these stimuli are relevant to the authors’ proposal regarding the function of visual consciousness in the larger visual horizons associated with the sea-to-land transition – and thus broader visual exposure to a range of survival threats.

In visual awareness studies of evolutionary-relevant stimuli, postdictive effects occur much earlier, between 20–40 ms. The target stimuli are threats like snakes or angry faces, and the masks are composed of scrambled fragments of these images and thus many of the same lower-order visual features. Thus, the mask camouflages the threatening image, as might occur when predators or enemies are approaching. Nonetheless, most participants become visually aware of masked angry faces presented for longer than ∼20 ms (Dimberg, Thunberg, & Elmehed, Reference Dimberg, Thunberg and Elmehed2000; Milders, Sahraie, & Logan, Reference Milders, Sahraie and Logan2008), and of masked snakes or spiders presented for longer than ∼40 ms (Öhman & Soares, Reference Öhman and Soares1994; Siegel & Weinberger, Reference Siegel and Weinberger2012).

These considerations reveal that the authors regard visual processing in relative isolation, decoupled from evaluation of the biological significance of stimuli. In their opening example of an apple being unexpectedly thrown at you, it takes considerable time for the brain to produce a conscious perception of the apple. It takes considerably less time to produce conscious perceptions akin to “red thing moving directly at me,” the classic phenomenon Gibson (Reference Gibson1966) termed “direct perception.” In a visual brain equipped with multiple parallel processing channels, each feature (color, contours, motion direction) can be processed simultaneously by a functionally specialized system, rapidly generating the coarse perceptions needed to identify survival threats in broader terrestrial visual horizons.