We investigated whether previously reported differences between Alzheimer's disease and dementia with Lewy bodies (DLB) in resting occipital activity lead to activation differences within functionally specialized visual cortical areas and deactivation differences in the default network. Patients with Alzheimer's disease (n = 10; 5 male), DLB (n = 9; 4 male) and controls (n = 13; 5 male) performed three functional MRI (fMRI) scanning experiments involving visual colour, face or motion stimuli. Reaction time or accuracy in DLB and Alzheimer's disease differed significantly from controls but not between patient groups, with the exception of accuracy in the face task (DLB < Alzheimer's disease; P = 0.038). The most significant fMRI activations in the pooled data set were in left V4alpha for the colour task (Talairach coordinate: -30, -52, -24; P = 0.002 corrected), the right fusiform face area (FFA) for the face task (34, -48, -22; P = 0.005 corrected) and right intra-parietal sulcus (30, -66, 42; P = 0.003 corrected) for the motion task, with additional activity in right V5 (48, -64, 0; P = 0.015 corrected). Each task was associated with decreases in activity within the default network with prominent deactivation foci bilaterally in the posterior cingulate gyrus (+/-8, -48, 26; left P < 0.001; right P < 0.001 corrected) and medial frontal cortex (+/-18, 42, 32; left P < 0.001; right P < 0.001 corrected). Comparing patterns of task-related activity across groups, DLB patients showed more activation than Alzheimer patients within the superior temporal sulcus (STS) for the motion task (right STS: 44, 0, -20; P = 0.004 corrected; left STS: -40, -4, -26; P = 0.07 corrected). This difference could not be attributed to task performance, cognitive score or age [analysis of covariance (ANCOVA)F (2, 18) = 8.44, P = 0.003]. Within regions of interest, group activation differences were found for the face task (Alzheimer's disease > DLB P = 0.05; Alzheimer's disease > controls P = 0.14) and the motion task (DLB < Alzheimer's disease P = 0.031 and DLB < control P = 0.048). However, these differences could be explained by behavioural performance, failing to reach significance in the ANCOVA analysis. In the default network, group deactivation differences between controls and both patient groups were found for the colour and motion task (colour: control < Alzheimer's disease P = 0.02; control < DLB P = 0.019; motion: control < Alzheimer's disease P = 0.118; control < DLB P = 0.118) but could be accounted for by behavioural performance. The results suggest that cognitive fMRI can be used to detect both performance-dependent and performance-independent differences between Alzheimer's disease and DLB, reflecting the distribution of functional pathology in the two conditions.