Background: Many people participating in dementia research may lack capacity to give informed consent and the relationship between cognitive function and capacity remains unclear. Recent changes in the law reinforce the need for robust and reproducible methods of assessing capacity when recruiting people for research.Aims: To identify numbers of capacitous participants in a pragmatic randomised trial of dementia treatment; to assess characteristics associated with capacity; to describe a legally acceptable consent process for research.Methods: As part of a pragmatic randomised controlled (...) trial of Ginkgo biloba for mild-moderate dementia, we used a consenting algorithm that met the requirements of existing case law and the exigencies of the new Mental Capacity Act. We decided who had capacity to give informed consent for participation in the trial using this algorithm and sought predictors of capacity.Results: Most participants with mild-moderate dementia in this trial were unable to give informed consent according to the legal criteria. When adjusted for confounding, the Mini Mental State examination did not predict the presence of capacity.Conclusion: Cognitive testing alone is insufficient to assess the presence of capacity. Researchers and clinicians need to be aware of the challenging processes regarding capacity assessment. We outline a procedure which we believe meets the ethical and legal requirements. (shrink)