Upon arrival at the laboratory, participants were endowed with $3 andtold that they would take part in an auction (Figure 1A). The auction procedurefollowed the procedure of a Becker-Degroot-Marschak (BDM) auction 20. Single pictures of food itemsappeared on the screen one at a time and participants placed their bid for eachindividual item by selecting a value on a visual analog scale at the bottom ofthe screen using a mouse. Participants were explicitly told that their beststrategy for the auction was to bid exactly what the item was worth to them tobuy from the experimenter at the end of the session. At the end of the sessiona single trial was selected at random and played out such that the computergenerated a counter bid, which was a random number between 0 and 3 in 25 centincrements. This number was compared to the participant’s bid on the randomlyselected trial and if the computer bid was higher than the participant’s, theparticipant could not buy that item. If, however, the computer bid was lowerthan the participant’s, then the latter was offered that item at the computer’sbid lower price. This auction provided us a measure of willingness-to-pay (WTP)for all 60 food items per participant.

We used WTP to rank order the foods foreach participant from most preferred (highest WTP) to least preferred (lowestWTP, Figure 2A). Items were split into high-value and low-value items. Itemswere then placed into one of two training conditions; Go items required abutton press during training and NoGo items required no response from theparticipant in adaptations of the cue-approach task 3. Item assignment to Go and NoGoconditions based on their rank order was counterbalanced across participants.