Working Memory Training and Transfer: What 2020s Research Actually Shows

Written by TOMOY · Published: April 9, 2026 · Updated: April 9, 2026

1. The Transfer Question

The central question in working memory training research is not whether training improves performance on the trained task — it does. The question is whether those gains transfer: to related cognitive tasks (near transfer), to everyday functioning (far transfer), or to real-world outcomes.

This distinction matters enormously, and the evidence for each type differs substantially. A training program that reliably improves your score on a specific task may or may not do anything for reading comprehension, attention at work, or academic performance. Understanding where the evidence is strong — and where it is genuinely uncertain — is the foundation for any honest conversation about what WM training is actually for.

2. What Near Transfer Shows

Near transfer refers to improvements on tasks that are similar but not identical to the trained task — for example, a different version of digit span, or a related working memory updating task. The research here is relatively consistent.

A 2026 meta-analysis published in npj Digital Medicine found evidence of near transfer with an effect size of g = 0.503 across included studies. This is a moderate, detectable effect. It suggests that WM training does improve related cognitive performance — not just the specific task being practiced.

What this means in practice: if you train consistently on backward digit span, you can reasonably expect improvement not only on that specific task but also on structurally similar tasks involving the phonological loop and central executive updating. The effect does not stay entirely within the narrow boundaries of the trained variant.

3. What Far Transfer Shows

Far transfer — gains that extend to functionally different cognitive domains such as reading comprehension, mathematical reasoning, or general fluid intelligence — is where the picture becomes more complicated.

The same body of research consistently finds that far transfer effects are smaller, less reliable, and more sensitive to study design than near transfer. A meta-analysis in SAGE Open (2025) reported a mean effect size of g = 0.23 for far transfer across studies, noting high variability across samples and methodologies.

This does not mean far transfer is impossible — but it means it is not guaranteed, and it cannot be assumed to follow automatically from near transfer gains. The variability in findings likely reflects real differences in training protocols, participant populations, outcome measures, and how transfer is operationally defined across studies. Whether the average effect is practically meaningful remains an open empirical question.

4. The Adaptive Difficulty Factor

One consistent finding across the transfer literature is that adaptive training protocols — where difficulty adjusts dynamically to the user's performance — tend to produce larger effects than non-adaptive designs.

Tzuriel and Schwartz (2024, British Journal of Educational Psychology) found that adaptive WM training in educational contexts showed more reliable transfer when embedded within structured support and feedback. The mechanism appears to be that adaptive designs maintain the learner at the threshold of their current capacity — the zone where cognitive demand is highest and memory consolidation is most active. When difficulty is fixed, the system either under-challenges experienced learners or overwhelms beginners, reducing the training signal in both cases.

This finding has practical implications: the specific structure of how you train matters, not only how much time you invest. Progressive, difficulty-adjusted practice is not merely a design preference — it is one of the more robust predictors of transfer in the available evidence.

5. What This Research Means for Practice

The research does not justify skepticism about WM training, nor does it justify strong claims about broad cognitive enhancement. What it supports is a more precise framing:

• WM training reliably improves performance on near-transfer tasks — this finding is consistent and moderate in magnitude.
• Far transfer is possible but variable — effect sizes are smaller, and the conditions under which it occurs are not fully understood.
• Training conditions matter: adaptive difficulty, sustained practice, and ecological validity appear to be meaningful moderators of the outcome.

For people using training apps, the honest expectation is that consistent practice will improve your performance on tasks like the ones you practice, and may support related cognitive functions. Whether those gains extend further — to fluid intelligence, academic performance, or everyday attention — depends on factors the research continues to investigate.

This question is not closed. It remains one of the most actively studied areas in applied cognitive psychology, and the answer will likely be more nuanced than either "training works" or "training doesn't work." The field is converging toward identifying the conditions, populations, and training structures that make transfer more or less likely — and that is a productive place for the science to be.