The problem of long-term assessment of cognitive change in dementia is defined as that of avoiding floor and ceiling effects in testing. Unlike diagnosis, the establishment of norms is viewed as secondary. This perspective allows the formulation of strategic guidlines in devising test batteries for longitudinal studies of cognitive change in dementia. A number of tests complying with these guidelines are described, tapping a wide range of cognitive abilities such as semantic analysis of verbal and visual material, semantic memory, object recognition, visuoconstruction and serial planning.