In the target article, Firestone & Scholl (F&S) make the bold claim that higher-level cognition does not affect perception. They acknowledge that there have been a very large number of experiments that would seem to contradict them but argue that all of that previous work fell foul of at least one of six potential confounds. They challenged the academic community to find evidence for the effects of top-down cognition on perception that does not suffer from any of these potential confounds.

Hallucinations are one example that clearly meets the challenge. In psychotic disorders such as schizophrenia, hallucinations are most commonly reported in the auditory domain (e.g., hearing voices). However, pure visual hallucinations are also possible and are associated with hyperconnectivity between the amygdala and the visual cortex (Ford et al. Reference Ford, Palzes, Roach, Potkin, van Erp, Turner, Mueller, Calhoun, Voyvodic, Belger, Bustillo, Vaidya, Preda and McEwen2015), underscoring the role of top-down feedback in their formation. There is also a range of natural and artificial compounds known to cause visual hallucinations when consumed. These drugs commonly induce vivid, geometric, kaleidoscope-type patterns behind closed eyes. Depending on the specific drug and dose, they can also lead to pure hallucinations involving creatures and scenes (Vollenweider Reference Vollenweider2001). Such drug-induced hallucinations are further demonstrations of top-down effects on perception. They are impossible to explain or conceptualise in terms of changes occurring within the bottom-up flow of external sensory information through the visual hierarchy.

Although it could be argued that we should consider such hallucinations an anomaly, independent of normal cognitive processes, other forms of hallucination are harder to dismiss. About 10%–30% of individuals who suffer from a severe visual impairment, such as glaucoma, experience Charles Bonnet syndrome (CBS; Schultz et al. Reference Schultz, Needham, Taylor, Shindell and Melzack1996; Vukicevic & Fitzmaurice Reference Vukicevic and Fitzmaurice2008). These individuals typically have no other neurological or psychiatric conditions, yet they frequently experience hallucinations. It is not that CBS sufferers merely think or sense the hallucinations; rather, the hallucinations appear so realistic that the affected people sometimes mistake them for reality (Schultz et al. Reference Schultz, Needham, Taylor, Shindell and Melzack1996). In addition, the hallucinations will often interact with the participant's visual perception of the external world. For example, a hallucinated figure may obscure part of the visual scene that would otherwise be visible to the observer, thereby preventing the observer from seeing it – a clear example of top-down cognition influencing stimulus-based visual perception.

Mental imagery (i.e., visualisation) is a further example of top-down cognition obviously affecting perception. For some people, their mental images are exceptionally vivid, almost as vivid as visual perception (Pearson et al. Reference Pearson, Rademaker and Tong2011). Like hallucinations, these mental representations are pictorial: People actually see the mental images, as opposed to merely being aware of them (Kosslyn et al. Reference Kosslyn, Behrmann and Jeannerod1995).

Both hallucinations and mental imagery activate the visual cortex in a way very similar to that of normal visual perception, which helps explain why people describe hallucinations and mental imagery as true perceptual experiences. Whereas hallucinations typically activate higher cortical areas (ffytche et al. Reference ffytche, Howard, Brammer, David, Woodruff and Williams1998; Vollenweider Reference Vollenweider2001), mental imagery is able to additionally activate the primary visual cortex (Albers et al. Reference Albers, Kok, Toni, Dijkerman and de Lange2013; Slotnick et al. Reference Slotnick, Thompson and Kosslyn2005; Stokes et al. Reference Stokes, Thompson, Cusack and Duncan2009). The BOLD (blood-oxygen-level dependent) fMRI activation caused by mental imagery is so similar to that generated by observed visual stimuli that a model based on tuning to low-level visual features (e.g., spatial frequency and orientation) that was trained on the BOLD fMRI activity generated when participants observed real images was able to determine which of those images the participants were subsequently visualising solely from the BOLD fMRI activity generated during these visualisations (Naselaris et al. Reference Naselaris, Olman, Stansbury, Ugurbil and Gallant2015). Further, this low-level coupling between mental imagery and visual perception is unlikely to be epiphenomenal; magnetic pulses delivered to the primary visual cortex disrupted both mental imagery and visual perception to a similar extent (Blasdel & Salama Reference Blasdel and Salama1986).

Hallucinations and mental imagery demonstrate that top-down cognitive and emotional processes can affect perception in a manner that avoids all six of the potential confounds F&S raise. There is no doubt that perception can be altered depending on which locations, features, or objects the observer attends to in the external world (Collins & Olson Reference Collins and Olson2014; Vetter & Newen Reference Vetter and Newen2014). The fact that hallucinations and visual imagery are not driven by bottom-up stimuli, and that they are often unconstrained to specific locations in visual space, makes it difficult to conceptualize how such phenomena could be explained by such peripheral attention effects (confound 5) or indeed by low-level differences in the visual input (confound 4). Furthermore, because hallucinations and mental imagery are clearly perceptual, they also avoid confound 2 (perception vs. judgment) and confound 6 (memory and recognition). Additionally, they avoid the El Greco fallacy (confound 1) and cannot be attributed to demand and response bias (confound 3). As such, hallucinations and visual imagery avoid all of F&S's potential confounds.

In conclusion, it is clear that top-down processes can affect perception in a variety of ways. In clinical or drug-induced psychosis, a person's visual experience can be generated independent of bottom-up stimulation. In the case of mental imagery, percepts can be generated by directing top-down attention to internal mental states and representations. Finally, although CBS hallucinations appear to be beyond voluntary cognitive control (Schultz & Melzack Reference Schultz and Melzack1991), they are clearly also caused by top-down processes. Moreover, such hallucinations can interact with visual perception, thereby providing a clear demonstration of a top-down effect influencing stimulus-based visual perception. These examples show that top-down cognitive processes are not only able to penetrate visual perceptions, but also they can cause them.