Clinicians and researchers frequently quantify impairments and functional ability to monitor patient's symptoms and progress. For some patients, such as those with Parkinson's disease (PD), symptoms can fluctuate from day to day, making reliable measurement difficult. Multiple measures then may be required to obtain reliable data. Decisions must be made, balancing the optimum measurement schedule to obtain "good reliability" against burden to the patient. This investigation demonstrates the use of Generalizability Theory in determining the testing schedule when designing an experiment involving patients with known fluctuations of symptoms. In this investigation we use "Functional Axial Rotation" (FAR), a measure of spinal flexibility, to illustrate the use of Generalizability Theory for designing an experiment using participants who have PD. Measurements of FAR were taken on 13 participants, aged 60 or older, who were in early and mid-stages of PD. Three measurements were obtained on each of two consecutive days, and repeated on two consecutive days a week later, giving a total of 12 measures of FAR for each individual. Four sources of variation (subject, week, day and trial) were employed to estimate the reliability of FAR under several designs. Assuming different schedules of measures across weeks, days and trials, the estimated reliability of FAR for four measurements is in the range of 0.75 to 0.83, and for eight measurements in the range of 0.82 to 0.86. We discuss the use of this type of analysis in the determination of the optimum measurement design for experiments involving subjects with known fluctuations.