Everyone wants to be better at focusing, remembering, and problem-solving, right? The self-help market recognizes and exploits this and is teeming with brain games that promise a sharper mind. Many of these games are based on tasks that cognitive psychologists use to study working memory, the system involved in short-term processing and storage of information. Whether or not training on these games (or cognitive tasks) leads to improved performance on other memory and problem-solving tasks – called transfer in the literature – remains controversial. Although people do improve on working-memory tasks through training, evidence in support of far transfer – which would suggest their working-memory capacity has actually increased – is limited.

A big issue for working-memory training programs is that they lack theoretical grounding. That is, there is no basis for why training procedures should improve memory. In their recent article published in the Psychonomic Society journal Memory & Cognition, Weng Tink Chooi and Robert Logie (pictured below) aimed to reveal the cognitive processes that occur during working-memory training.

To do so, the authors focused on homophones, sets of words that have the same pronunciation but different meanings (e.g., blue and blew). Homophones not only make for punny humor, but they can also offer insight into the strategies that people use when performing a specific working-memory task: the adaptive N-back.

The adaptive N-back involves studying single words for two seconds each and then responding to whether the current word matches the word that was a given number ‘N’ back from the current word. The task “adapts” to the ongoing performance by becoming harder or easier (i.e., the value of ‘N’ increases or decreases). Chooi and Logie used eight sets of homophone words (e.g., blue-blew; vain-vein) as stimuli.

Across 5 weeks, university students in Malaysia completed 15-20 thirty-minute training sessions on the N-back. The authors investigated how error rates changed during training on the task.

Before diving into the results, let’s consider the types of errors that the students could make on this task, given they are important for understanding the strategies students may use during training. Examples are explained below (and in the figure below).

1. Target missed. Error due to missing the target word when it was in the correct position. For example, if a participant responded “no match” to the word blew in panel a.
2. Foil false alarm. Error caused by identifying a word when it did not match the target word in the N-back position. For example, if a participant responded “match” to the word maid in panel a.
3. Lure word false alarm. Error due to false homophones of the target words in n-back position. For example, if a participant responded “match” to the word sole in panel b, given the word 3-back was the homophone soul.
4. Lure position false alarm. Error caused by a word identical to the target word but at the wrong position – either one before or one after ‘N’. For example, if a participant responded “match” to the word blue in panel c, given the target word was at 2-back, not 3-back.

What did they find? Overall, performance improved with training – that is, students made fewer errors across time. However, the type of errors made provided insight into the processes that students were engaging in during the task. Across training, the authors found an increase in the proportion of errors from false alarms due to lure phonologically similar words. Although errors decreased with training, when students did make an error it was likely to be from a homophone (e.g., responding to blue instead of blew).

This finding suggests that the students have been engaging in a phonological rehearsal strategy during the N-back, which was generally a good strategy given performance improved. Repeating prior words in your mind can be helpful for reducing errors on the N-back. That is, until a pesky homophone comes along!

Given the students relied on a sub-vocal rehearsal strategy, the authors wondered whether training on the N-back would improve the participants’ ability to remember phonologically encoded sequences in order after training.

To investigate this idea, the authors compared performance on serial-order recall tasks before and after N-back training. The serial-order recall tasks involved recalling stimuli (letters or numbers) in the order they were presented. Performance on the recall tasks was then compared to another group of students who completed an adaptive Operation Span (OSPAN) training task. The OSPAN task involved alternating between processing arithmetic equations and remembering words to be recalled later. Thus, it did not readily provide opportunities for students to engage in sub-vocal rehearsal.

The authors predicted that N-back training would improve performance on the serial-order recall tasks whereas OSPAN training would not. Surprisingly, neither training group showed improvements on the serial-order recall tasks – reinforcing just how hard it can be to observe transfer for some working-memory tasks. However, the authors did replicate the distribution of errors from their first experiment. Once again, they observed an increase in the proportion of errors from false alarms due to homophones.

In a third experiment, the authors replicated the error patterns with a British sample and asked students to report on their strategy use throughout training. Consistent with the idea that students relied on phonological strategies for the N-back, many students said they used sub-vocal rehearsal, and the use of this strategy (along with other phonological strategies) increased across training sessions. However, the authors caution that these self-reported strategies did not predict N-back performance.

To sum up, people rely on a sub-vocal rehearsal strategy during the N-back, which improves their overall task performance but makes them fall prey to homophones. Though no transfer effects were found, their study is an important contribution toward understanding when and why transfer effects do and do not occur during working-memory training.

Featured Psychonomic Society article:

Chooi, WT., & Logie, R. (2020). Changes in error patterns during N-back training indicate reliance on subvocal rehearsal. Memory & Cognition, 48, 1484–1503. https://doi.org/10.3758/s13421-020-01066-w