**4. Discussion**

The present study reports case PL518, an architect who lost his ability for visual imagery following a bilateral PCA stroke 35 months prior to this investigation. We compare his performance across a range of perceptual and cognitive tests and a visual imagery questionnaire with four other PCA stroke patients, an architect with a large right hemisphere lesion and three bilateral cases. PL518s profile on the perceptual and cognitive tests was similar to other cases with the exception that PL518 reported severe visual imagery problems following his stroke. Lesion profiles were also comparable with the exception that PL518 showed selective damage in the right lingual gyrus and left medial posterior fusiform gyrus. It is tempting to sugges<sup>t</sup> that these are both candidate regions for specific involvement in visual imagery.

However, Bogousslavsky and colleagues [60] described a man whose lingual gyrus was destroyed in both hemispheres, while only the middle third of the fusiform gyrus on the left side was a ffected. His visual imagery was intact for colors, faces (human and animal) and places (streets). The authors concluded that the fusiform gyrus and underlying white matter, rather than the lingual gyrus, was a principal structure for color integration, face recognition, visuo-verbal processing, and corresponding visual imagery. The fact that the current primary case, PL518, had selective damage to the left fusiform gyrus is also more in alignment with other research indicating that left hemisphere regions are more consistently implicated in generating mental imagery than corresponding right hemisphere regions [4,7,22,24,61–68].

A seeming counterexample comes from de Gelder and colleagues [69]. They described patient TN who had bilateral cortical blindness due to lesions in the primary visual cortices in both hemispheres. The lesion also reached some high-level visual ventral areas, including parts of the left posterior fusiform gyrus. Despite this damage, de Gelder and colleagues [69] argued that TN was able to generate visual mental imagery. However, judging from the lesion reconstruction (their Figure 2), the left medial posterior fusiform might have been at least partially spared in this patient. Also, the imagery tasks used involved a significant motor or action component, and correspondingly TN's functional activation pattern in the imagery conditions was primarily fronto-parietal.

Fitting with a role of the left fusiform gyrus in visual imagery, some developmental prosopagnosics appear to have functional abnormalities in this region [70–72] as well as reduced or absent mental imagery, not only for faces but also for objects and scenes [73]. Barton and Cherkasova [74] examined face imagery in prosopagnosics for featural imagery (questions regarding facial features, e.g., "Who has a wider mouth: Sophia Loren or Ingrid Bergman?") as well as facial configurations (questions on overall face shape or configuration, e.g., "Who has the more angular face: George Washington or Abraham Lincoln?"). In acquired prosopagnosics, they found that right-sided occipito-temporal lesions affected imagery for facial configuration but not for facial features, while bilateral lesions additionally impaired imagery for facial features [74]. This fits well with the left fusiform gyrus responding more to facial features while the right fusiform gyrus is more involved in configural processing [75]. It is possible that the generation of mental imagery heavily relies on the assembly of separately stored visual features or parts, and that this generation of multipart images specifically taxes left hemisphere regions [37,66]. This is consistent with PL518s description of the fragmented minimal visual imagery that he possibly still has (e.g., visualizing bits of elephants).

Compared to before his stroke, PL518 seems to make greater use of verbal strategies (e.g., recalling a list). If PL518 still has some mental imagery, it nonetheless mostly seems to be based on an altered strategy which could be described as motor, action-based, or spatial, such as mentally tracing a line around a shape or doing mental rotation by physically trying to move things inside the head. This is reminiscent of patient MX [4] who also reported the loss of the experience of visual imagery as well as an unusual or altered strategy when attempting a mental rotation task, where he needed to match individual blocks and angles perceptually when making his decision.

The two architects, PL518 and PM024, had similar functional deficits, including prosopagnosia, but described vastly di fferent visual imagery (minimal vs. very clear and lively). It is tempting to speculate, therefore, that the additional left hemisphere a ffection in PL518 contributes significantly to his disruption of imagery. In particular, the small patch in the medial left fusiform gyrus where PL518 has unique damage compared to all the four other patients presents as a good candidate for playing a critical part in the generation of visual mental imagery. While our findings indeed sugges<sup>t</sup> that this region is an important node in the cerebral network underlying visual imagery, other areas, including right hemisphere ventral occipito-temporal areas, left hemisphere areas further anterior in the temporal lobe (see e.g., [74]), more posterior areas in the left occipital lobe, and regions outside of the ventral visual stream are also likely to partake in at least some aspects of visual imagery. For example, while mental imagery generation might mainly depend on structures in the posterior left hemisphere, right parietal regions have been found to be important for spatial comparisons of the contents of visual

imagery [76], see also [77]. The right hemisphere could also have some ability to generate visual imagery for overall shape [66], and had we included sensitive measures of configural processing deficits in mental imagery in addition to the VVIQ, it is possible that subtle deficits in PM024 could have been discovered. It is also worth noting that the aphantasic architect PL518 had bilateral damage, while mental imagery generation could possibly be taken over by the right hemisphere in cases of unilateral left hemisphere disruption [76].

The most commonly used questionnaire to measure mental imagery is various versions of the VVIQ [39,50]. The VVIQ has good psychometric qualities and vividness correlates with some other behavioral and neural measures of visual imagery [53,78,79]. These questionnaires do have their limitations, though, as they rely on self-reporting and only measure overall vividness of visual mental imagery. Mental imagery is, however, of a multimodal nature [80] and includes for example smell, touch, sound and taste. Also, there are several di fferent aspects of visual imagery, and in order to capture this more completely, a measure would need to include items specifically for spatial imagery, as well as imagery for colors, objects, places, faces, and even subsets of these such as featural vs. configural face imagery. More fine-grained mental imagery questionnaires and additional behavioral measures that likely rely on mental imagery, such as the clock task [81–83], the taller/wider task [66,83], or mental letter construction [84], animal tails test [8], drawing objects from memory [85,86] and the binocular–rivalry technique [87,88], would provide further insights into whether mental imagery deficits are due to a loss of all imagery across modalities, specific loss of visual imagery, or specific loss of subcomponents of visual imagery. Such specific aspects of mental imagery were not directly assessed in the present study.

It is still debated whether imagery and perception may be dissociated, or whether they depend on common networks. In one sense, the current results support the former as some patients with heavy damage to ventral stream areas and associated problems with visual cognition nonetheless appear to have intact visual imagery. Our neuropsychological approach suggests that some ventral stream regions might not be necessary for visual imagery despite containing information on imagined objects [21,28,29,89,90]. On the other hand, the areas specifically associated with PL518s visual imagery loss are better known for their role in visual perception. A key di fference between imagery and perception could however lie in their di fferent network dynamics where imagery is dominated by top-down feedback [21,89,90]; this could even map onto di fferent cortical layers within the same region [91,92]. Even if a region serves both perception and imagery, is it still possible that distinct computations and separable subpopulations of neurons are involved.

It should finally be noted that individual di fferences in premorbid ability for imagery might play a role in the e ffects of stroke on these abilities. PL518 reported that his abilities for visual mental imagery had been above average before his stroke. These abilities had enabled him to visualize the spatial and visual attributes of buildings and rooms in rich details and contributed greatly to his achievements as an architect. This fits a general pattern noted by Farah [7] where many cases of acquired deficits in visual imagery involved people whose day-to-day activities had likely demanded visualization. As the normal variability in visual imagery from congenital aphantasia to hyperphantasia becomes better understood, this factor may perhaps help explain variability in the e ffect of brain injury on visual imagery.
