The first half of the target article is mainly devoted to showing that conscious perception is slow. This conclusion is a linchpin of the article since the basis of the sensory horizon argument is that consciousness emerged only when this slow processing would benefit the creatures who had it. This slow formation of a conscious percept is supposed to allow the formation of a higher-order pointer, supporting the target article’s theory of consciousness.

The evidence for the slowness of conscious perception depends on the fact that conscious perception can be influenced by events that occur up to 450 ms after the perception is initially processed. The authors’ reasoning is encapsulated here (p. 4):

This reasoning is problematic because it neglects the distinction between the formation of a conscious percept and the conscious percept’s vulcanization. By vulcanization, I mean the kind of consolidation that makes the percept unchangeable. As Charles Goodyear profitably showed in the mid-19^th Century, the formation of a tire can be followed later by a process of vulcanization in which the fully formed tire is heated together with sulfur under pressure, the vulcanized tire being much harder than the originally already-formed tire. The point of this analogy from tires to perception is that a conscious percept could be formed in under 200 ms (as indicated by some experimental paradigms), even if it is malleable by later events a couple of hundred ms later – before the conscious percept is vulcanized. (See (Vogel, Woodman, & Luck, Reference Vogel, Woodman and Luck2006) for mention of this distinction.)

The distinction between formation and vulcanization is a conceptual distinction, and it reveals a conceptual gap in the argument, but does the conceptual gap correspond to an empirical gap? In particular, is there an actual significant temporal gap between formation and vulcanization for conscious perception? Or do formation and vulcanization happen in fact at the same time? As I mentioned, some experimental paradigms, for example, some experiments by Victor Lamme and Michael Pitts (Supèr, Spekreijse, & Lamme, Reference Supèr, Spekreijse and Lamme2001, Sligte, Scholte, & Lamme, Reference Sligte, Scholte and Lamme2009, Pitts, Martinez, & Hillyard, Reference Pitts, Martinez and Hillyard2011), can be interpreted (controversially) as showing faster processing of conscious percepts. For example, there is some evidence that visual awareness negativity, which emerges at 200 ms reflects consciousness and is not just a predictor of later consciousness. It may be that these experiments are geared to formation rather than vulcanization.

But does the vulcanized conscious percept overwrite the originally formed conscious percept (as might be suggested by the phi phenomenon)? This question is as old as the hills, but whatever the answer is, we certainly know that perceptual representations are influenced by perceptual categories, and when perceptual representations are used in working memory – as they must be in order for there to be responses from participants in experiments like those that Fleming and Michel cite – the categorical influence gets stronger (Bae, & Luck Reference Bae and Luck2019, Kwak, & Curtis Reference Kwak and Curtis2022, Zhou, Lorist, & Mathôt, Reference Zhou, Lorist and Mathôt2022, Xu Reference Xu2023, Duan, & Curtis Reference Duan and Curtis2024, Printzlau, Bourganos et al. Reference Printzlau, Bourganos, Fukuda and Mack2025).

Duan and Curtis (Reference Duan and Curtis2024) showed that when perceptual representations of oriented grids are used in working memory, they are reformatted. They noted that a certain kind of bias affected perceptual decoding in V1 but not working memory decoding of the same stimuli. Zhou et al. (Reference Zhou, Lorist and Mathôt2022) showed subjects a number of colored disks at different locations, followed by a blank screen with a fixation point for a period between 100 ms and 2 seconds. Then, an indicator indicated the location of one disk, followed by a slider for continuous color report. See Figure 1. The upshot was that for delays over a second, working memory representations drift more toward prototype categories as working memory load increases. In addition, the best models of both perception and working memory involved a balance of continuous and discrete categorical coding.

Figure 1.

Subjects are shown in 1–4 colored disks, then the disks are replaced by a fixation point for up to 2 seconds, then an indicator points to the location of one of the disks. Subjects are asked to move a slider to indicate the color of the disk at that location. From (Zhou, Lorist, & Mathôt, Reference Zhou, Lorist and Mathôt2022) under a Creative Commons CC-BY license.

Furthermore, there are many different paradigms in which a saccadic response occurs hundreds of milliseconds before a button press (de Lissa, Sokhn et al. Reference de Lissa, Sokhn, Lasrado, Tanaka, Watanabe and Caldara2021), and saccades are known to be less influenced by past perceptions than perceptions, as indicated by perceptual report (Lisi & Cavanagh Reference Lisi and Cavanagh2015).

The upshot of all this is that perception is an ongoing temporal process in which perceptual representations evolve in a task-specific way as they are integrated into working memory. We may not know quite when the conscious perception is formed and what exactly the process of vulcanization consists of, or which stages are conscious, but there is plenty of space between the formation of the percept and the further processing that allows for the malleability that Fleming and Michel cite. Conscious perception may be fast even if malleable over time.

I started this response by noting that the basis of the sensory horizon argument is that consciousness emerged only when this slow processing would benefit the creatures who had it. But in the light of the distinction between formation and vulcanization, one might wonder whether the argument could go through if it were based not on conscious perception being slow but on the vulcanization of conscious perception being slow. Perhaps, there was no need for long vulcanization periods for creatures with short sensory horizons; so it was the emergence of long vulcanization periods that brought both fast and malleable conscious perception. Maybe so, but this is still bad news for the higher-order theories advocated by Fleming and Michel, if that fast formation is too fast for a higher-order pointer to form.