Objective: This study examines the performance of AD patients on the WAIS-III Block Design subtest to analyze the nature of constructional deficits. Method: The sample is 15 patients with AD (5 men and 10 women, mean age = 73.6; mean MMSE = 21.8) and 15 healthy controls (5 men and 10 women, mean age = 74.1; mean MMSE = 29.1). All participants were administered the Estonian version of WAIS-III Block Design subtest. In addition to standard scoring we also noted qualitative observations: number of moves within time limit, proportion of blocks correctly placed on each design, single block errors (divided to placement errors and rotation errors), broken configuration errors (when parameters of the model were not accounted) and design rotations (when entire design was rotated more than 30°). Results: AD patients (M = 7.4, SD = 3.9) performed significantly worse than control group (M = 11.5, SD = 3.1), U = 176.0, p < .008. The control group made more moves than patients (U = 158.0; p < .02). They also had proportionally more correctly placed blocks, U = 166.0, p < .026. Most of the single block errors in control group were placement errors while AD patients made both, placement and rotation errors. Patients also tended to rotate the whole design but this was rare. Broken configuration errors were also rare but they were observed in both groups. Conclusion(s): Lower performance of AD patients in Block Design can be partly explained by slower mental computation in deciding about the block placement, and partly by spatial deficits in rotating blocks when looking for correct angles with the design.