The emerging UK national research network in neurodegenerative diseases and dementias aims to promote large-scale community-based studies of therapeutic interventions, based in primary care. However, trials in primary care settings can be problematic, a common difficulty being the recruitment of a large enough sample. The article discusses recruitment issues in a multi-centre randomized controlled trial of differing educational approaches to improving dementia care in general practice. Sample size calculations based on community studies of prevalence may be misleading in intervention trials which may recruit practices with atypical demography. Recruitment rates for practitioners in this study were lower than expected. Professionals excluded themselves from the study mainly due to pressures of time and staff shortages, and we detected both ambivalent attitudes to primary care research and a perception that research into dementia care was not a high priority. Evaluation of the quality of care may be perceived as criticism of clinical practice, at a time when general practice is undergoing major administrative and contractual changes. Variations in Research Ethics Committee conditions for approval led to different methods of recruitment of patients and carers into the study, a factor which may have contributed to disparate levels of recruitment across study sites. Patient and carer levels of recruitment were lower than expected and were affected partly by carers' time pressures and other family commitments, but largely by problems in identifying patients and carers in the practices. The development of research potential in primary care is at an early stage and studies reliant on patient recruitment in general practice must allow for multiple obstacles to enrollment. This is particularly relevant for studies of dementia care, where the prevalence of dementia in a demographically average population is low and the incidence very low, compared with other disabilities. Professionals may give dementia low priority in allocating practice time for research projects, and strategies to address this problem are needed.