The compelling demand for dementia research is supported by the large and demographically associated increase of dementia for which no causal treatment is available. However, during its progressive course, dementia destroys the capacity for self-determination of the ill person and thus an essential prerequisite for participation in research, i.e. a valid consent toward a research intervention. Accordingly, not only sufficient information about all issues that are relevant for decision but also a flawless assessment of the capacity to consent is important; however, currently this is not satisfactorily possible. This paper tries to find answers for related questions, such as “why?” and “how?”, of its assessment including that of a substitution of consent of potential research participants for whom consent is no longer possible. In a second section, the equally underdeveloped benefit-risk evaluation will be discussed with two specific research examples, a diagnostic and a therapeutic research intervention with patients with dementia.