The goal-control model of the functional impairment in dementia posits two different underlying mechanisms: decay of task goals (reduced task accomplishment) and poor control over goal execution (high error rates). Here we present a case series in which we explore the effects of a performance-based, functional intervention on two participants. Outcomes were evaluated using the goal-control framework.

Two participants with dementia (CS: age 70, 14 years of education; EM: age 93, 18 years of education) completed neuropsychological tests (scored using age, education, and IQ-adjusted norms) and baseline testing with the Naturalistic Action Task (NAT; a validated performance-based task of everyday function including a Breakfast and Lunch task). The Virtual Kitchen (VK) was used to train, through repeated performance, either the Breakfast (CS) or Lunch (EM) tasks for 30 minutes (or 10 total repetitions) per day over 5 days. After VK training, participants performed the NAT Breakfast and Lunch tasks again to evaluate improvement on the trained and untrained tasks. Baseline and post-training NATs were scored for task accomplishment and errors by two coders observing video recordings. Z scores were derived by calculating accomplishment and error change scores for each participant relative to the mean and standard deviations of change scores from a cohort of 36 healthy controls (mean age: 73.3, SD: 6.44; mean education: 17.42, SD: 2.17).

Both participants exhibited similar cognitive profiles: high estimated IQ; low MMSE (total = 19 for both CS and EM; 1st percentile); anterograde amnesia, slowed processing speed and impaired executive function; average scores on tests of attention, language, and self-reported depression. Informant report of daily functioning (FAQ) suggested that EM (FAQ=28) exhibited greater functional impairment than CS (FAQ=9). Both participants completed all VK training sessions. Z scores of the change from pre- to post-training showed significant increases in task accomplishment on the trained task (trained condition change z scores: EM = +27.69; CS =+ 6.06), but significantly less improvement or worse task accomplishment on the untrained task (untrained condition change z scores: EM = +4.06; CS = -13.69). The training did not reduce errors, as error rates increased for both participants on the trained task.

The participants presented in this case study exhibited comparable cognitive profiles, including marked anterograde amnesia. Our results suggest that repeated training in a virtual context can improve specific aspects of functioning on real, life everyday tasks. Further, according to the goal-control framework, repeated practice reduces the decay of the task goal, as represented by greater task accomplishment, but does not improve executive control over the task execution. Important future directions are to determine if people with different cognitive profiles will demonstrate different benefits from VK training and to examine if virtual training of personally relevant, everyday tasks can promote independent living and improve quality of life.