Objective: To assess the psychometric properties of the Power of Food Scale (PFS) in diverse populations of obese and non-obese individuals. Methods: Data were obtained from adults in a clinical trial for a weight management drug (n=1,739; mean body mass index [BMI] [SD]=38.6 [6.7]) and a web-based survey (n=1,275; overweight and obese [BMI 27-76 kg/m2] and non-obese [BMI 18-27 kg/m2]). Exploratory and confirmatory factor analyses were employed to discover the structure of PFS using the clinical data. The model developed was then tested using data from the web-based survey. The relationship between PFS domains and BMI was also examined. Logistic regression was used in the web-based survey to evaluate the association between obese status and PFS scores. Results: Psychometric assessment of data from the clinical study indicated that the scale was best represented by a 3-factor, 2nd-order model - three domains and a composite domain (average of the three domains) - which was confirmed within the web-based survey (Bentler's Comparative Fit Index: 0.92 and 0.91, respectively). Cronbach's alpha for both data sets were high, ranging from 0.81-0.94 (three domains and a composite domain score). The relationships between BMI and each domain were subtle and approximately linear. An increase of one point in a PFS domain score increased the odds of being obese by 1.6-2.4 times (depending on the domain; domain scores range from 1 to 5). Conclusion: The structure of PFS is represented by a 3 factor, 2nd-order model with three domains (Food Available, Food Present, and Food Tasted) and a composite of them. This structure has high internal consistency and reliability, relates to BMI, and distinguishes between obese and non-obese subjects. The data indicate that the PFS can be used to evaluate the effects of treatment on patient perception of the power of food in trials of obese patients.