**3. Results**

### *3.1. Visual Imagery*

In the clinical interview, PL518 reported an almost complete absence of visual imagery following his stroke. This was in stark contrast to his (in his own opinion) above average ability for visual imagery before his stroke that he had relied upon in his work as an architect. He said: "Before, my visualization abilities were pretty impressive. At my work, I could visualize and remember things that most people had not thought about. I would be sitting there and I would say, well, you can't do X, Y and Z, because you've go<sup>t</sup> this happening here and there. Now I have to look at the drawing and work my way through it." During the interview, he also described how it had felt to do a mental rotation task: "I cannot do it as quickly or the same way as I would have done before my stroke. Before, bang, I would just know the answer. Now it is a much more conscious process. It's almost as though I physically am trying to move things inside my head." He was then asked whether his di fficulty with mental rotation a ffected his ability to work as an architect, to which he responded: "Well I just do everything on the computer. That is one of the advantages of us using computers for these sorts of thing nowadays. You can see the stu ff happen." He also described how he is just about able to imagine very simple shapes, but this is done using something akin to motor or spatial imagery and he struggles to imagine more than one shape at a time: "If I tried to visualize shapes like a square, pyramid or sphere lined up next to each other, and I try and focus with a kind of spotlight on the corner of one shape, I can mentally trace a line around the shape. But as soon as I focus on one shape, the others disappear." When asked if he could imagine an elephant, he seemed to mostly think of the abstract concept of an elephant: "I can think of elephants, iconic elephants like Babar or Elmer, but I can only visualize bits of them. It's almost painful." When asked to describe the place he stayed during his last holiday and its surroundings he provided few very vague details about a couple of the bars from the

street they had lived on, and he apparently did not visually imagine himself there: "I am recalling almost like a list. I do the same when going somewhere. I have to remember a list".

PL518s copy and retention of the Rey figure are shown in Figure 1. The drawings were scored for accuracy according to the Taylor's (1969) method described in Spreen and Straus (1991). 35/36 points were given for the copy and 18/36 for the three-minute recall. While these scores are within the normal range, one could have expected patient PL518, with his background as an architect, to have adopted a more structured approach to drawing the figure in the recall condition. This drawing not only lacks many details but also some of the core elements. Also, some of the included elements are placed incorrectly.

**Figure 1.** PL518s performance on the Rey Complex Figure Test. (**A**): Copy. (**B**): three-minute recall.

PL518s complaints regarding his visual imagery were also clearly reflected in his responses on the VVIQ-modified where he scored 18 (i.e., a mean score of 1.13 per item), corresponding to minimal imagery [39]. None of the other patients reported any changes in the nature or vividness of their visual imagery following their strokes–neither in the VVIQ-modified nor the general questions; they all responded yes to the first three general questions about being able to see images in their minds, and no when asked if their visual imagery had changed following their stroke. Their respective scores on the VVIQ-modified were: PL502: 49 (mean: 3.06); PL545: 53 (mean: 3.31); PM006: 72 (mean: 4.5); PM024: 76 (mean: 4.75). See Appendix A Table A1 for the patients' responses to the individual questions.

### *3.2. Neuropsychological and Experimental Tests*

For the accuracy measures, PL518 is clearly impaired with faces, and shows a deficit (performing more than two standard deviations (SDs) from the control mean) on most individual face measures. He performs within the low–normal range on the object tests and is clearly on level with controls in the tests with word stimuli. For the RT measures, PL518 shows a deficit on most face measures (note that his RTs in the surprise recognition test may not be a good indicator of severity, as his accuracy in this test was very low). He responds with latencies within the normal range on the object tests but shows elevated RTs in the lexical decision and word reading tests.

Comparing the neuropsychological profile of PL518 to the other included patients, we find that one or more of them show deficits on the same tests/measures and in the cognitive domain(s) as PL518 (see Figure 2 for an illustration of their cognitive profiles on the selected tests, and Table S1 for an overview of test results). A comparison of the neuropsychological profile of the two architects (PL518 and PM024) shows that PM024 (with no aphantasia) shows the same pattern of performance as PL518 on most tests, including measures of face recognition, object recognition and word reading. Indeed, there is no measure on which PL518 shows a clear deficit, where PM024 is clearly within the normal range (see Figure 2). The key difference between the two patients, then, is in the measure of their visual imagery. Comparing PL518 to the three other bilateral patients (Figure 2 and Figure S1), again there is

no domain where PL518 is clearly impaired where the other patients are consistently within the normal range. In comparison to the three bilateral patients too, then, the key difference is in visual imagery.

**Figure 2.** Radar plots showing the results of PL518 (in red) and the other patients on all the included measures of object, word, and face recognition. Numbers denote z-scores based on the control means and SDs for the respective tests. Impaired performance (<sup>&</sup>gt;−2 SDs from the control mean) is marked by the dotted grey line, and scores closer to the center are more impaired (represents lower accuracy and slower RTs). Left panel (**A**,**C**) shows accuracy, right panel (**B**,**D**) shows RTs. Upper panel (**A**,**B**) shows PL518 vs. PM024 (architect with right hemisphere lesion). Lower panel (**C**,**D**) shows PL518 vs. the other bilateral patients. See individual radar plots comparing PL518 individually to bilateral patients in Figure S1.

### *3.3. Lesion Localisation*

PL518s lesion is most extensive on the right side, including damage to the occipital pole, the lingual gyrus, the whole fusiform gyrus and extending anteriorly to the parahippocampal region. On the left side, the lesion affects only the medial fusiform gyrus and lingual gyrus, while the left occipital pole, and lateral portions of the fusiform gyrus are spared. See Figure 3 and Table 2 for comparisons of lesion localization for PL518 and the other patients.

First, comparing the lesions of PL518 to the architect without aphantasia (PM024) shows that PL518 has selective left hemisphere posterior medial fusiform damage extending medially and anteriorly along the collateral sulcus, and selective right hemisphere damage to the superior medial lingual gyrus. Second, comparing the lesion of PL518 to the three patients with bilateral strokes but no aphantasia shows that PL518 has selective damage in the right fusiform gyrus and a portion of the right lingual gyrus, and additional smaller areas of selective damage in PL518 are found in the left fusiform gyrus. Combined, these comparisons reveal only small areas of selective damage in PL518 in the right lingual gyrus and left posterior medial fusiform gyrus.

**Figure 3.** (**A**) shows lesion overlap between PL518 and the other bilateral patients. (**B**) shows lesion overlap between PL518 and PM024. Left hemisphere depicted on the left. The lesions are in MNI space overlaid on the MNI152 template. See Figure S2 for individual comparisons of PL518 and the bilateral patients.

**Table 2.** Comparison of regions of interest within the occipital and temporal lobes affected in PL518 compared to other patients. The fusiform gyrus (FG) was segmented into four regions (FG1-4: corresponding to posterior medial, posterior lateral, anterior medial and anterior lateral, respectively) according to Lorenz and colleagues [57]. The occipital pole and the lingual gyrus were defined using a conventional atlas [58], and the parahippocampal region was identified using the images from Bouyeure and colleagues [59]. An x indicates that at least 10% of the corresponding region of interest was affected by a patient's stroke.

