Professor Wahlund has made a convincing case that structural imaging is not useful as adjunct investigation for all cases of dementia in the primary care situation. In my opinion, no-one has ever suggested that it would or should be. Indeed, there would be insufficient resources to do so and it would be overkill, to say the least. Primary care is not the setting in which brain imaging should be performed and so I agree with Wahlund in this regard. However, following primary care evaluation of a patient with suspected dementia, a referral to a second line setting often occurs. This can happen for a number of reasons, including the need for treatment, and it is here that imaging should always be performed. The first approach then should always be a structural one, not only to exclude the odd case of a primary glioma or subdural hematoma, but to show evidence of medial temporal lobe atrophy, shown by the Swedish SBU to be of high diagnostic accuracy for AD (SBU, 2006). Additionally, vascular changes may be demonstrated, which may prompt the specific treatment of risk factors, or else focal atrophy suggestive of another specific neurodegenerative disorder may be seen (Scheltens et al., 2002; van der Flier and Scheltens, 2005). The recent surprising observation of a relatively high frequency of microbleeds in AD patients (18%) and VaD patients (65%), which may caution the clinician against the use of oral anticoagulation, again underlines the need to perform MRI in demented patients (Cordonnier et al., 2006). The only exception to this would be the sound judgment of the clinician that no nosological diagnosis is warranted in a specific case (usually because of high age, presence of severe dementia and/or extensive comorbidity). It is important to note that, in the case of medial temporal lobe atrophy, however it is measured, there is a tendency for reduced diagnostic utility (reduced specificity) at a higher age, making age specific norms a requirement (van de Pol et al., 2006).