We are very interested in the letter from Karnath and colleagues and welcome the opportunity to continue this stimulating debate about the anatomy of neglect. Although the arguments advanced by Karnath and his colleagues are interesting, we disagree with them. We hope they will reconsider their interpretation of our findings (Mort et al., 2003), as well as those of Binder et al. (1992), in the light of our responses. In addition, we draw attention to the fact that the limited new data provided in their letter does not actually support well the controversial claim that neglect is associated with superior temporal gyrus (STG) damage (Karnath et al., 2001).
Karnath and colleagues state that ‘about one‐third of the sample studied by Mort et al. (2003) was normal or close to normal in cancellation’. To be precise, this was four out of 14 patients (M5, M7, M12 and M14) with infarction in the territory of the right middle cerebral artery (MCA) who manifested neglect. Contrary to what is implied, these patients were not included merely because of their rightward bias on line bisection but because they showed neglect in their clinical assessment, including observation of their daily behaviour. In addition, two of them (M5 and M7) showed robust neglect on cancellation tasks when they were admitted.
The suggestion is made that patients showing abnormal line bisection are naturally more likely to have posterior brain lesions, and that the presence of four patients in our group who performed relatively well on the cancellation task biased our region of maximal overlap towards parietal areas. We can confidently state that this was not the case. When those four patients are excluded from the MCA neglect group, the inferior parietal lobe (IPL) continues to be identified as the critical region. This applies both when this maximal lesion overlap is assessed in this group alone and, much more meaningfully, when it is directly contrasted with the non‐neglecting MCA controls (Fig. 1).
Likewise, exclusion of the same four patients makes no difference to the finding of commonality of angular gyrus involvement evident in the analysis of individual non‐normalized brains (Table 2 of our paper). This confirms that the IPL is the critical region associated with neglect in MCA patients, and further shows that our conclusion is not confounded by the presence of abnormal line bisection.
While Binder et al. (1992) did report a group of patients who showed neglect on cancellation but not line bisection, and who had lesions that included more anterior (subcortical and frontal) structures, there was no association between abnormal line bisection alone and damage to posterior structures. In all but two of 11 patients, the combination of line bisection and cancellation abnormalities was actually associated with parietal involvement. Thus, neglect patients with parietal involvement tended to be impaired on both line bisection and cancellation, whereas neglect patients with frontal involvement were impaired on cancellation.
To support their contention, Karnath and colleagues provide new data from 15 patients with hemispatial neglect (documented using cancellation and copying tasks). Very little information is provided in their letter about the methods of image acquisition and lesion demarcation. In particular, there is nothing to allay the concerns we raised about their original study (Karnath et al., 2001), including the quality of spatial resolution within the z plane. If, again, clinical imaging was used, poor z‐plane resolution could potentially confuse judgements about the involvement of IPL, STG and temporoparietal junction (TPJ) areas.
The lesion analysis performed by Karnath and colleagues does share with one of our analysis protocols the technique of native brain and lesion spatial normalization (Mort et al., 2003). However, it is important to note that we used 1 mm × 1 mm × 1 mm image acquisitions to perform normalization, and, as we have noted, the resolution of acquisitions is unfortunately not explicitly stated in the letter of Karnath and colleagues. In our study, we also performed a second analysis which did not rely on any normalization, and found that the IPL was the critical region using this technique as well. Thus, we stress that our conclusions do not rely solely on the normalization technique.
In the new normalized data set provided by Karnath and colleagues, there is an anterior region of maximal lesion overlap. Importantly, their figure clearly shows the area of maximal involvement lies deep in the inferior frontal lobe, not the STG. There is lesser overlap in the insula and, to some extent, the STG. However, we too found that the STG was damaged in half our MCA neglect patients (although it was not the critical site associated with the presence of neglect), presumably because of common vasculature involvement.
The inferior frontal lobe has been implicated in neglect by several previous studies of patients with focal lesions (Damasio et al., 1980; Husain and Kennard, 1996; Walker et al., 1998) and most researchers in the field would not find such an association totally surprising. However, in our analysis, frontal damage was neither necessary nor sufficient to produce neglect in patients with large lesions (Mort et al., 2003). It is important to note, though, that Karnath and colleagues’ new overlap map derives solely from patients with neglect. There is no contrast with non‐neglecting patients. This substantially weakens their argument. For conclusions about the anatomy of neglect, it is essential that anatomical comparison is made with a matched control group, as was performed in our study.
In summary, we conclude that the points made by Karnath and colleagues about line bisection abnormalities are misleading. The major conclusion from our findings is that the critical lesion region associated with neglect in MCA infarction lies in the IPL. The new data provided by Karnath and colleagues are too incomplete, especially in terms of information about the resolution of the imaging source data and absence of a contrast with non‐neglect controls, so that at present their worth cannot be properly assessed. We look forward to analysing a full version of this study in the future, and continuing this debate.
We thank Chris Rorden for his help with the original study and to the Wellcome Trust for its support.