Description

Study 2 will use a treasure task (employed in C.2.1.1), in which participants select amongst 3 keys to learn which feature of keys unlocks the treasure chest, revealing a trial unique object image (Fig. 8b). On each encoding trial participants view three multi-featural keys and a treasure chest. Each key varies on three dimensions: color (red, blue, yellow), shape of the handle (square, circle, triangle), and number of ridges (one, two, three). On each trial, all 9 dimensions of the keys are represented, but the dimensions represented on each key are randomized across trials. During hypothesis testing runs, participants are instructed to find the feature of a key that opens the treasure chest to reveal a trial-unique image. Critically, there is one target feature of a key that deterministically opens the treasure chest. The target feature of the key randomly changes after participants successfully open the chest 4 consecutive times. This design manipulates goal-oriented exploration by allowing participants to resolve uncertainty until they identify the target feature of the key. Uncertainty resolution is the greatest after the target features switches and incrementally reduces via individuals' active hypothesis testing. During control runs, participants will be instructed to select specific keys, and key selection and outcomes will be yolked to prior hypothesis testing runs. Participants will complete 3 runs of hypothesis testing runs and 3 control runs. Trial number for each run is determined by participant behavior, in that hypothesis testing runs end when participants complete 4 rule switches (~30 trials). Encoding is incidental in this task. Following encoding, participants will complete an immediate and delayed item memory test. On each trial, participants will view objects from the hypothesis testing and control conditions as well as new foil objects. Memory will be tested for half the stimuli immediately and at a 24-hour delay. These memory measures will be related to uncertainty resolution estimated by reinforcement learning, to determine how prior knowledge during information seeking has subsequent effects on memory.