Background: Complementary therapies have become more commonly used over the last decade and have been applied to a range of health problems, including dementia. Of these, aroma therapy is reported to be the most widely used in the British National Health Service (Lundie 1994) and might be of use for people with dementia for whom verbal interaction may be difficult and conventional medicine of only marginal benefit. Aroma therapy has been used for people with dementia to reduce disturbed behaviour (e.g. Brooker 1997), promote sleep (e.g. Wolfe 1996), and stimulate motivational behaviour (e.g. MacMahon 1998). Objectives: To assess the efficacy of aroma therapy as an intervention for people with dementia. Search strategy: The Cochrane Dementia and Cognitive Improvement Group's Specialized Register was searched on 29 October 2002 to find all relevant trials using the terms: aroma therap*, "aroma therap*", "complementary therap*", "alternative therap*" and "essential oil". The CDCIG Register contains records from all major health care databases and is updated regularly. Additionally, relevant journals were hand searched, and 'experts' in the field of complementary therapies and dementia contacted. Selection criteria: All relevant randomized controlled trials (RCTs) were considered. A minimum length of trial and requirements for a follow-up were not included, and participants in included studies had a diagnosis of dementia of any type and severity. The review considered all trials using fragrance from plants defined as aroma therapy as an intervention with people with dementia. Several outcomes were considered in this review, including cognitive function, quality of life, and relaxation. Data collection and analysis: The titles and abstracts extracted by the searches were screened for their eligibility for potential inclusion in the review, which revealed 2 RCTs of aroma therapy for dementia. Neither of these had published results in a form that we could use. However, individual patient data from one trial were obtained (Ballard 2002) and additional analyses performed. Analysis of co-variance was used for all outcomes, using a random effects model. Main results: The additional analyses conducted revealed a statistically significant treatment effect in favour of the aroma therapy intervention on measures of agitation and neuropsychiatric symptoms. Reviewer's conclusions: Aroma therapy showed benefit for people with dementia in the only trial that contributed data to this review, but there were several methodological difficulties with this study. More well designed large-scale RCTs are needed before conclusions can be drawn on the effectiveness of aroma therapy. Additionally, several issues need to be addressed, such as whether different aroma therapy interventions are comparable and the possibility that outcomes may vary for different types of dementia.