Multiple cognitive deficits have been recognized in many neurological disorders but the specificity of the findings and the relationship to the underlying neuropathology remain obscure. Definitions of dementia have been proposed based on symptom profiles of the cognitive disorder and qualitative differences have been claimed between dementias of different aetiology. Some conditions have been claimed to show patterns of cognitive deficit that are distinguished from dementia and related to specific neuropathology or psychological processes, e.g. frontal lobe deficits in Parkinson's disease. Sometimes, a relationship has been established between certain cognitive deficits and particular neurochemical deficits which has led to the notion of specific drug treatment, e.g. cholinergic deficits and memory failure in Alzheimer's disease. However, these conclusions are often potentially flawed by methodological inadequacies. This critique presents some methodological issues relevant to the study of brain-behaviour and drug-behaviour relationships in syndromes of multiple cognitive deficit, using Parkinson's disease as the model. The following recommendations are made: rigid diagnostic criteria; representative patient groups; avoidance of arbitrary quantitative criteria to limit definitions of dementia; matching of groups for overall level of cognitive impairment in the search for qualitative cognitive differences related to neuropathology or effects of particular drugs; the use of suitable controls in patient groups, neuropsychological tests and treatment regimes; the use of specific quantitative tests of cognition, affect and motor disability; and longitudinal, compared with cross-sectional, study design.