In a recent review paper1, I outlined a model of visual mental imagery proposing a reverse visual hierarchy starting from prefrontal areas back to sensory areas.

I would like to thank Paolo Bartolomeo, Dounia Hajhajate, Jianghao Liu and Alfredo Spagna for their correspondence on our Review (The human imagination: the cognitive neuroscience of visual mental imagery. Nat. Rev. Neurosci. 20, 624–634 (2019))1, which raises some important issues (Assessing the causal role of early visual areas in visual mental imagery. Nat. Rev. Neurosci. https://doi.org/10.1038/10.1038/s41583-020-0348-5 (2020))2.

Neuropsychological work reports that individuals with visual cortex damage can show some imagery-like processes without perception3,4, while damage to the temporal lobe can correspond to a lack of imagery/memory abilities5,6. Further, individuals with aphantasia (no imagery vividness or sensory imagery)7,8 show normal sensory perception (although this is yet to be tested in detail), suggesting a general dissociation, and has been linked to areas outside of early visual cortex9. Based on these reports Bartolomeo et. al., suggest a revision to the reverse visual hierarchy model I described, without a causal role of early visual cortex. Here I outline three reasons why this is not required.

Much of the neuropsychological work on imagery is actually correlational, often documented brain injury is correlated with subjective descriptions of imagery loss. Hence, we do not know an individual’s imagery abilities prior to the injury. For it to be causal, visual imagery would need to be accurately assessed before and after pre-defined or ‘planned’ and controlled damage or interruption to a particular brain region (or with a control site), in a similar manner to TMS work in humans or animal lesion work.

Perhaps more of a problem for imagery research is the variety of different ways it is measured1. Most typically in neuropsychological work discussed by Bartolomeo et al., subjective reports are used, perhaps most problematic are questions regarding the physical structure of objects as in4,10, which use questions like “Do tractors have two large wheels at the back or front?” Such knowledge questions can easily be answered without imagery. In contrast to this, much of the behavioural and fMRI work use objective task-based measures11. This is important because research suggests that these different measures do not rely on the same brain areas1,12. This measurement problem is well illustrated by the fact that many aphantasics always believed they had mental imagery, not fully understanding the sensory nature of imagery in others until it was graphically described.

Interestingly, new data indeed suggest a causative link between cortical excitability in early visual cortex and measures of imagery strength. Recent work has shown that the excitability of early visual cortex predicts and can modulate imagery strength in a ‘causal’ manner via brain stimulation13.

One way to reconcile these different findings is that early visual cortex has a causative role in the high definition precise features of visual imagery, but not lower fidelity imagery and not necessarily reports of imagery vividness or object descriptions. With such a model damage to primary visual cortex would result in a loss of the high-fidelity precise dimensions of imagery, if an individual did indeed have this to begin with, but not a loss in high-level imagery of places, faces, spatial imagery or imagery vividness. To put this another way, visual imagery, like visual perception, is not a unitary process. Different features, colour, form and motion, and their levels of precision are processed across different brain areas and most likely use a range of mechanisms. Further complicating this hierarchical spread is the extreme range of individual differences that naturally exist with imagery, together making cross-methodological comparisons and meta-analyses difficult.