Decomposing the n-back task: An individual differences study using the reference-back paradigm

- Working memory (WM) updating is controlled by a gating mechanism.
- Gating and WM updating are difficult to differentiate empirically.
- The newly developed reference-back paradigm enables to measure them separately.
- We used the reference-back paradigm to study individual differences in n-back.
- The results showed that gating is a key source of individual differences in n-back.

Working memory (WM) has two major functions: Maintenance, which is the ability to shield information from being overwritten by irrelevant information, and updating, the ability to modify the maintained information when needed. These two conflicting demands are suggested to be controlled by a gating mechanism (for review see O'Reilly (2006)) which enables selective control over updating. Information is robustly maintained in WM when the gate is closed, while opening the gate enables updating. In the present study, we utilized the reference-back paradigm in order to examine their unique contribution to individual differences in n-back, presumably the most widely-used WM updating task. The reference-back is composed of two types of trials: reference trials which require both matching (i.e., a same/different judgment) and WM updating, and comparison trials which require matching and maintenance. Eighty-eight participants performed the following tasks: 2-choice RT, 1-back, 2-back and the reference-back task. A multiple regression approach was taken in order to explain individual differences in 1-back and 2-back. The reference-back task enabled separating the contribution of the matching decision (difference between mismatch and match), gate-opening (the switch cost in reference trials), gate closing (the switch cost in comparison trials) and WM updating (the difference between reference and comparison trials) to task performance. An intrusion component (WM based proactive interference) was also calculated from 2-back performance. The results indicate that RT in 1-back is mainly predicted by gate opening and by WM updating while 2-back is mainly predicted by gate closing and intrusion. These results confirmed that n-back is not merely an updating task, but also that controlling the contents of WM is the main source of individual differences in the task. The implications for understanding the n-back task and WM updating in general are discussed.