Objective We hypothesized path tortuosity (an index of casual locomotor variability) measured by a movement telesurveillance system would be suitable for assisted living facility residents clinically diagnosed with dementia. Background We examined the relationship of dementia to path tortuosity and to movement speed and path length variability, both of which increase in dementia. Methods Daytime movements of 25 elders (19 female; 14 with dementia; average age 80.6) were monitored for 30 days using radio transponders measuring location with a maximum accuracy of 20 cm. After 30 days, the Mini Mental State Exam (MMSE) and Revised Algase Wandering Scale-Community Version (RAWS-CV) were administered. Results Fractal Dimension (Fractal D), a measure of path tortuosity, correctly classified all but 2 residents with dementia; sensitivity 0.857, specificity 0.818 while the MMSE had 6 misclassifications, a sensitivity of 0.857 and a specificity of 0.727. Individual logistic regressions of dementia diagnosis on predictors MMSE and Fractal D were significant, but a logistic regression using both predictors found Fractal D marginally predictive of dementia (p=0.055) while the MMSE was not (p=0.168). Although significantly correlated with Fractal D, rate of travel and mean path distance were not predictive of dementia. Fractal D correlated negatively with overall MMSE (r= -0.44, n=25, p<0.05) but the relationship was mediated by MMSE Geographical Orientation items. Fractal D was unrelated to the RAWS-CV. Conclusions Telesurveillance-measured path tortuosity is greater in persons diagnosed with dementia. Persons with dementia have relatively more impaired spatial memory which is required for successful navigation. Application Automatic monitoring of direction, length and speed of unconstrained movements.