The main objectives were to study relationships between the design of group living (GL) units and psychiatric symptoms in demented patients before, 6 months after, and 1 year after admission to GL units. The study population comprised 105 demented elderly (83 +/- 6 years), 37% with dementia of Alzheimer's type and 58% with vascular dementia. The patients were relocated by the municipal care planning team after clinical examination. An observational scale (the Organic Brain Syndrome scale) was used to assess confusional symptoms and disorientation. The physical environment was assessed by an architect using the Therapeutic Environment Screening Scale, which evaluates general design, space, lighting, noise, communication area, floor plan, and related factors. Less than 15% of the patients had no signs of dyspraxia, hallucinosis, dysphasia, or depression at admission, whereas 66% or more reported lack of vitality, aggressiveness, or restlessness. Fourteen out of 18 units had a corridor-like design (group A), one unit an L-shaped design (group B), and the others a square or H-shaped design (group C). Patients living in the B unit had less disorientation than the others at the 6-month follow-up. After 1 year, the patients in the A units had more dyspraxia, lack of vitality, and disorientation of identity. The communication areas in the units were negatively associated with "disorientation for recent memory" and "lack of vitality," adjusted for type of dementia (r = -0.13 to -0.16). The size of the activity area, indoor public rooms in square meters, was not correlated to confusional reactions and disorientation. In conclusion, a GL unit design that facilitates perception without reducing the communication area is to be preferred.