Visual and semantic determinants of object recognition deficits in early Alzheimer's disease

The visual and semantic factors underlying recognition deficits in Alzheimer's disease were investigated. To control for object form we used triads of computer-generated blobs. Triads either shared visual features or were visually distinct. Triads were artificially labeled using semantically close or distinct, biological, or nonbiological names (e.g., lion, tiger, zebra or violin, guitar, banjo vs robin, donkey, frog or kite, wrench, carriage). On test trials blobs were presented alone and patients attempted to recall their "name." Visually close triads were misidentified significantly more often than visually distinct triads. Irrespective of whether biological or non-biological labels were used, triads labeled with semantically close names were misidentified significantly more often than semantically distinct triads. Hence, problematic object categories are those whose exemplars are visually and semantically close.