A sample of 144 participants underwent the Trier Social Stress Test (TSST), a psychosocial stress manipulation involving a mock interview and a mental arithmetic task, or a matched control procedure. Physiological stress was estimated via a collection of 7 saliva samples over the course of the experiment analysed for cortisol and alpha-amylase, as well as via the mean heart-rate measured before and during the experimental manipulation. Subjective stress was assessed via the Positive and Negative Affect Scale as well as four Visual Analogue Scales at 6 points over the time course of the experiment. Participants solved an incentive-compatible food-choice task before, immediately after and in the aftermath of the experimental manipulation. In each trial of the food-choice task, participants had to choose one out of a set of two to seven snack bundles. Each snack bundle consisted of specific amounts of a sweet or salty snack and a fruit or vegetable. The snacks for both categories were selected to be similarly attractive according to the previously provided online ratings of the participants. The design of the food-choice task allows for the calculation of revealed preference consistency indices. The dataset further contains several self-report questionnaires administered to the participants before the experimental session, including the Trier Inventory of Chronic Stress.