This study examined the test-retest reliability and criterion validity of light (LPA), moderate (MPA), vigorous (VPA), and moderate-to-vigorous (MVPA) physical activity survey items in a subset of participants from a large prospective cohort. Participants included 423 women and 290 men aged 31–72 years in the Cancer Prevention Study-3 (CPS-3). Information on physical activity (PA) was collected using two different surveys: one survey which captures all activity performed during a typical 24-hour period in broad categories (24-hour survey), and a more detailed survey focused primarily on leisure-time PA (LTPA survey). One-year reliability was assessed by computing Spearman correlation coefficients between responses from pre- and post-study periods for both surveys. Validity was assessed by comparing survey-estimated PA with accelerometry, seven-day diaries, and a latent variable representing ‘true’ PA estimated through the method of triads. Reliability was considered acceptable for most items on the LTPA survey (range ρ = 0.45–0.92) and the 24-hour survey (range ρ = 0.37–0.61). LPA validity coefficients were higher for the 24-hour survey, while MPA, VPA, and MVPA coefficients were higher for the LTPA survey. Study results suggest that both CPS-3 PA surveys are suitable for ranking or classifying participants in our population according to overall PA category or intensity-specific activity level.