Using participant hedonic ratings of food images to construct data driven food groupings.

Little is known regarding how individuals' hedonic ratings of a variety of foods interrelate and how hedonic ratings correspond to habitual dietary intake. Participant ratings of food appeal of 104 food images were collected while participants were in a fed state (n = 129). Self-reported frequency of intake of the food items, perceived hunger, body mass index (BMI), and dietary restraint were also assessed. Principal components analysis (PCA) was employed to analyze hedonic ratings of the foods, to identify component structures and to reduce the number of variables. The resulting component structures comprised 63 images loading on seven components including Energy-Dense Main Courses, Light Main Courses and Seafood as well as components more analogous to traditional food groups (e.g., Fruits, Grains, Desserts, Meats). However, vegetables were not represented in a unique, independent component. All components were positively correlated with reported intake of the food items (r's = .26-.52, p <.05), except for the Light Main Course component (r = .10). BMI showed a small positive relation with aggregated food appeal ratings (r = .19; p <.05), which was largely driven by the relations between BMI and appeal ratings for Energy-Dense Main Courses (r = .24; p <.01) and Desserts (r = .27; p <.01). Dietary restraint showed a small significant negative relation to Energy-Dense Main Courses (r = -.21; p <.05), and Meats (r = -.18; p <.05). The present investigation provides novel evidence regarding how individuals' hedonic ratings of foods aggregate into food components and how these component ratings relate to dietary intake. The notable absence of a vegetable component suggests that individuals' liking for vegetables is highly variable and, from an empirical standpoint, not related to how they respond hedonically to other food categories.