Two equivalent special care nursing home units for elders with dementing illness were randomly designated as experimental and control units for an intervention called the “stimulation-retreat” model. This model introduced a set of staffing and program changes whose purpose was to diagnose, prescribe, and apply a package of care according to individual needs for additional stimulation or relief from stimulation (“retreat”). A total of 49 experimental and 48 control unit residents completed 12 months of care and were evaluated at baseline, 6 months, and 12 months. It was hypothesized that the intervention would not affect the basic disability (cognitive and activities of daily living functions), would improve negative behaviors and observed affects, and would have maximum impact in increasing positive behaviors and affects. Over time, most functions worsened, including negative attributes and affects. Lesser decline in positive affect and increases in external engagement, however, led to the conclusion that the intervention showed a marginally significant and selective effect on positive behaviors and affect.