At the heart of the target article is the claim that there are distinct systems in the human brain for the conscious and unconscious control of behavior. Fleming and Michel (F&M) identify these systems with model-free versus model-based controllers, but fundamentally the same distinction has gained popularity under a variety of labels, including the System 1/System 2 and implicit/explicit distinctions (Newell & Shanks, Reference Newell and Shanks2023). Others, like F&M, have proposed evolutionary accounts of the origins of these distinct systems (Reber, Reference Reber1992; Sherry & Schacter, Reference Sherry and Schacter1987).

F&M present the high-level evidence base for this distinction in the Qualitative differences in conscious control section of their article. Within the limited space available for this commentary, I briefly and critically reevaluate all of the substantive evidence that F&M cite (ignoring animal studies, which conceptually cannot bear on this issue).

First, F&M claim that subliminal stimuli may evoke defensive or motivational responses. The evidence against this claim is compelling (see meta-analysis by Mertens & Engelhard, Reference Mertens and Engelhard2020). As an example, the highly cited demonstration of unconscious instrumental conditioning by Pessiglione et al. (Reference Pessiglione, Petrovic, Daunizeau, Palminteri, Dolan and Frith2008) was not replicated in a high-powered registered report by Skora et al. (Reference Skora, Livermore, Dienes, Seth and Scott2023), and the necessity of awareness extends to habits (Höhmann, Jocham, & Skora, Reference Höhmann, Jocham and Skorain press). Dozens of studies and several reviews support the claim that learning the motivational value of stimuli (evaluative conditioning) only proceeds in participants aware of the relevant contingencies (Moran et al., Reference Moran, Hughes, Hussey, Vadillo, Olson, Aust, Bading, Balas, Benedict, Corneille, Douglas, Ferguson, Fritzlen, Gast, Gawronski, Giménez-Fernández, Hanusz, Heycke, Högden and De Houwer2021; Stahl, Haaf, & Corneille, Reference Stahl, Haaf and Corneille2016; Van Dessel et al., Reference Van Dessel, Hughes, Perugini, Smith, Mao and De Houwer2025).

In addition to conventional fear conditioning studies, F&M cite the innovative work by Taschereau-Dumouchel et al. (Reference Taschereau-Dumouchel, Cortese, Chiba, Knotts, Kawato and Lau2018) on neural reinforcement and the extinction of physiological fear responses. This impressive research does not, however, provide strong support for Taschereau-Dumouchel et al.’s claim that “fear responses can be ‘reprogrammed’ outside of conscious awareness.” As Dienes and Seth (Reference Dienes and Seth2022) have noted, participants may have inferred “correlated hypotheses” (Dulany, Reference Dulany1961), which led Taschereau-Dumouchel et al. to misclassify them as unaware.

F&M offer evidence from the process-dissociation procedure (PDP; Jacoby, Reference Jacoby1991). They describe a situation in which participants are presented with a briefly presented prime word (e.g., “referee”) and are then required to complete a stem (e.g., “ref…”) with any word that comes to mind, except the prime word. According to F&M, “when the prime is subliminal, the frequency of word stems completed by the prime is higher than expected by chance – showing inflexible, habitual responding” of the sort that is independent of consciousness. Although they do not cite any studies showing this pattern, it has indeed been reported (Debner & Jacoby, Reference Debner and Jacoby1994). However, later studies either did not replicate the effect or showed that it was attributable to partial prime awareness (see Fisk & Haase, Reference Fisk and Haase2015). Fisk and Haase (Reference Fisk and Haase2015) wrote that “accumulating evidence… increasingly suggests that exclusion tasks lack validity for studying unconscious perception. Exclusion failure effects may represent weak conscious perception rather than unconscious perception.”

F&M proffer the existence of “blindsight” as evidence that individuals “can perform visual tasks in the absence of reported conscious visual experiences,” but there is plentiful evidence that blindsight is simply a manifestation of near-threshold vision combined with response conservatism (Phillips, Reference Phillips2021).

F&M cite the apparent uncoupling of action and perception in the dorsal and ventral visual streams. A number of dissociations in both normal participants and individuals with damage to the ventral visual stream (most famously, D.F.) have been reported that are consistent with unconscious visual guidance of actions. Here again, recent research has shown that many of these claimed dissociations are either unreplicable or can be explained without reference to the conscious/unconscious distinction. For instance, the claim that perception but not action is susceptible to visual illusions and Garner interference has not held up to scrutiny (e.g., Bhatia et al., Reference Bhatia, Osenberg, Janczyk and Franz2025; Schenk & McIntosh, Reference Schenk and McIntosh2010).

F&M characterize Pavlovian conditioning as even more primitive than habitual behavior, which makes it all the more striking that consciousness is a necessary prerequisite in humans for this form of learning (Lovibond & Shanks, Reference Lovibond and Shanks2002; Lovibond & Westbrook, Reference Lovibond and Westbrookin press).

F&M claim that “…the existence of masked priming effects on behavior… is a prima facie example of short-timescale unconscious influences on action (Vorberg et al., Reference Vorberg, Mattler, Heinecke, Schmidt and Schwarzbach2003).” Forty years after Holender’s (Reference Holender1986) famous critique of this field, published in this journal, we are no closer to having a body of robust, replicable masked priming effects. Haase and Fisk (Reference Haase and Fisk2015) found clear prime awareness in the metacontrast masking task employed in the one study cited by F&M as demonstrating unconscious priming (Vorberg et al., Reference Vorberg, Mattler, Heinecke, Schmidt and Schwarzbach2003). The inference that primes can affect behavior unconsciously has been shown to rest on a range of questionable assumptions (e.g., Meyen et al., Reference Meyen, Zerweck, Amado, von Luxburg and Franz2022; Sand & Nilsson, Reference Sand and Nilsson2016; Stein, van Gaal, & Fahrenfort, Reference Stein, van Gaal and Fahrenfort2024; Stockart et al., Reference Stockart, Schreiber, Amerio, Carmel, Cleeremans, Deouell, Dienes, Elosegi, Gayet, Goldstein, Halchin, Hesselmann, Kimchi, Lamy, Loued-Khenissi, Meyen, Micher, Pitts, Salomon and Mudrikin press).

Where does this brief review leave us? If the contemporary human brain is not well-described by the dual-system unconscious/conscious viewpoint, then F&M’s evolutionary story, whatever its other merits, is attempting to answer an imaginary problem. There is nothing to see here.

A final meta-science comment. Each side can point to supporting studies. We can engage in a tallying exercise to count the relative numbers of studies confirming or contradicting the claimed role of unconscious vision. Such an exercise would not be productive. The larger point is that this is a degenerating research program that absorbs enormous amounts of research effort but which has failed to yield clear, reproducible examples of the unconscious control of behavior. A telling sign of this is the age of most of the studies cited by F&M. The scientific literature doubles approximately every 9 years, so the fact that key evidence for F&M’s claim comes from studies published 15–35 years ago (e.g., Aglioti, Desouza, & Goodale, Reference Aglioti, Desouza and Goodale1995; Debner & Jacoby, Reference Debner and Jacoby1994; Goodale et al., Reference Goodale, Milner, Jakobson and Carey1991; Morris, Öhman, & Dolan, Reference Morris, Öhman and Dolan1998; Persaud & Cowey, Reference Persaud and Cowey2008; Pessiglione et al., Reference Pessiglione, Petrovic, Daunizeau, Palminteri, Dolan and Frith2008; Vorberg et al., Reference Vorberg, Mattler, Heinecke, Schmidt and Schwarzbach2003) is revealing. This is not a research program that is cumulatively building a compelling base of replicable evidence. It is time to accept that most meaningful human behaviors are grounded in conscious mental states.