Multi-level dynamics of task representation during learning

Poster Session A - Saturday, April 13, 2024, 2:30 – 4:30 pm EDT, Sheraton Hall ABC

Dr. Guochun Yang¹ (guochun-yang@uiowa.edu), DR. Jiefeng Jiang¹; ¹University of Iowa

While humans can rapidly learn new tasks, the underlying task representation is less known. We hypothesize that a task can be represented at different levels that may change with learning. For instance, cooking can be represented as a holistic event (event level), as an array of independent subtasks (subtask level), or as subtasks organized with meaningful orders (order level). We designed a delayed matching paradigm. On each trial, participants needed to memorize a five-feature stimulus and choose matching options of a cued feature after a delay. Five trials form a sequence, each having a fixed order of cued features. A good sequential memory can predict the feature to be cued and reduce memory load. We tested the prediction that learning intensity influences the representational level by training participants with one/three/eight sequences within three sessions of equal length. Behavioral analysis (n = 10) examined whether reaction time (RT) was influenced by the experience of learned sequence and orders as well as memory loads, each reflecting a representation level. Results showed an increase in order-level RT modulation during the first session, followed by a rapid decrease in the second session, persisting until the third session. The subtask-level modulation effect showed an opposite trend. EEG data revealed above-chance decoding of cued features before cue was presented, and the decoding accuracy showed a decreasing trend with time, potentially suggesting a decrease in event-level representation and an increase in subtask-level representation. In summary, our findings underscore the dynamic changes in task representation during learning.

Topic Area: EXECUTIVE PROCESSES: Goal maintenance & switching