Working Memory Games: What the Training Research Shows

Updated 2026-10-25

Written by the Optics Trainer team from published research, with every source linked below.

Working memory games ask you to hold a few things in mind for a few seconds and use them: a string of digits, a path across a board, a running total. Most are built on tasks psychologists have used for decades, such as digit span, the Corsi block task and the n-back. This guide explains what working memory is, how those tasks work, and what working memory training studies in children, teens and adults found, including the research on ADHD. It ends with four Optics Trainer memory games and the lab task each one is modeled on.

What is working memory?

Working memory is the small mental space where you keep information active while you use it, like holding a phone number long enough to dial it, or keeping track of a recipe step while you measure the next ingredient.

A widely cited model of working memory comes from Alan Baddeley and Graham Hitch, who proposed it in 1974. It had three parts: a store for sounds and words, a store for visual and spatial information, and a control system that directs attention between them. In 2000, Baddeley added a fourth part, the episodic buffer. He described it as a limited-capacity store that binds information from the other parts and from long-term memory into a single episode.

One idea runs through all of this research: working memory holds only a little at a time. That limit is what span tasks are designed to find.

Classic working memory tasks: digit span, Corsi block and n-back

Most working memory exercises and games are versions of a handful of lab tasks. Each one changes what you have to hold and what you do with it.

Span tasks go back a long way. Joseph Jacobs published span experiments in 1887. The Corsi block task comes from a 1972 doctoral thesis by Philip Corsi and a 1971 paper by Brenda Milner, and a 2000 standardization study described it as widely used for measuring visual memory span. The n-back task is usually traced to a 1958 paper by Wayne Kirchner.

Common working memory tasks
Task	What you do	What it mainly loads
Digit span	Hear or see a list of digits, then repeat it, forward or backward. Lists grow until you make errors.	Verbal short-term and working memory
Corsi block (spatial span)	Watch blocks being touched one at a time, then touch them in the same order. Sequences grow.	Visuospatial short-term and working memory
Location or pattern recall	Study where items are on a grid, then recall the positions.	Visuospatial memory
N-back	Watch a stream of items and respond when the current one matches the one shown n steps earlier (for example 2-back).	Continuous updating of what is held in mind
Memory updating	Keep a running count or set in mind as items are added and removed.	Updating and keeping track

How psychologists use span tasks, and the limits of online versions

In research and clinics, span tasks are given in a fixed way and scored against norms, so one person's result can be compared with others of the same age. Many standard test batteries include a digit span and a spatial span task. A 2000 study by Kessels and colleagues set out a standard way to give and score the Corsi task, collected norms from 70 healthy adults and 70 people with brain lesions, and reported that it could be used to assess visuospatial short-term memory after brain injury.

A 1998 review of 25 years of Corsi studies found considerable inconsistency in how the task was given and scored, and even in the blocks used. Online working memory tests and games add more differences: screen size, timing, distractions and practice from playing every day. A free online working memory test or game score is not a clinical measure and is not comparable to a score from a standardized test given by a psychologist. If you are worried about memory or attention, ask your doctor or a psychologist.

What working memory training research found

Working memory training programs, apps and games ask people to practice tasks like these for weeks, usually with levels that get harder as they improve. Researchers then ask two questions. Do people get better at memory tasks (near transfer)? And do the gains carry over to unrelated skills such as reasoning, reading, maths or daily life (far transfer)?

Several large meta-analyses have pooled these studies:

Melby-Lervåg and Hulme concluded in 2013 that these programs "appear to produce short-term, specific training effects that do not generalize." Their 2016 review also found that bigger working memory gains did not go with bigger far-transfer gains.
In 2019, Sala and Gobet summarized several meta-analyses across training types and reported a minimal effect on general thinking skills, with differences between studies explained by design quality and statistical artefacts.
In their 2020 paper, Sala and Gobet concluded that "there is no reason to keep investing resources" in working memory training research with typically developing children.
Broader reviews report a similar pattern. A 2016 review led by Daniel Simons found extensive evidence that people improve on trained tasks and little evidence for everyday thinking. In the ACTIVE trial in older adults, the memory-strategy group (711 people, 10 sessions) improved on memory tests at 2 years, but at 10 years only the reasoning and speed groups still showed gains on their trained skills. Our guide to visual brain training covers both.

Meta-analyses of working memory training
Review	Studies and people	Memory tasks (near transfer)	Other skills (far transfer), as reported
Melby-Lervåg & Hulme 2013	23 studies, 30 comparisons; children and adults, typical and clinical groups	Reliable short-term gains; verbal working memory gains not sustained at follow-up	No convincing evidence of transfer to verbal or nonverbal ability, inhibition, word decoding or arithmetic
Melby-Lervåg, Redick & Hulme 2016	87 publications, 145 comparisons	Reliable gains right after training on verbal and visuospatial working memory	No convincing evidence vs active control groups for reasoning, verbal ability, decoding, reading comprehension or arithmetic
Sala & Gobet 2017	Typically developing children aged 3 to 16	Effect size g = 0.46 on related skills	g = 0.12; effects were smaller in better-designed studies
Sala & Gobet 2020	41 studies, 2,375 typically developing children	Small to medium, larger when the test resembled the training task	Intelligence, maths and language essentially unaffected, especially vs active controls

N-back training and fluid intelligence: Au 2015 and its critics

The n-back became famous in 2008, when Jaeggi and colleagues reported in PNAS that young adults who trained on a demanding dual n-back task gained on a fluid intelligence test, with more training linked to larger gains.

Other teams tested the idea. In a 2013 randomized study, Redick and colleagues gave young adults 20 sessions of adaptive dual n-back, 20 sessions of a visual search game as an active placebo, or no training. Both training groups got better at their own task, but the authors found no positive transfer to any of the intelligence and ability tests.

In 2015, Au and colleagues pooled 20 n-back studies of healthy adults aged 18 to 50 and reported a small but significant effect on fluid intelligence (g = 0.24). Melby-Lervåg and Hulme challenged the analysis, saying it did not account for differences between groups at baseline or separate active from passive control groups. In their re-analysis the effect was g = 0.13 and was no longer significant once studies without active controls were left out. Au and colleagues replied that coding differences explained the gap, that control-group type did not change the effect, and that their original conclusion stood. Neither team tested Optics Trainer.

Working memory training for ADHD and Cogmed

Several trials in this section used Cogmed, a program developed by Torkel Klingberg and Helena Westerberg at the Karolinska Institutet. According to Karolinska, the company was started in 2001 and bought by Pearson in 2010.

In a 2005 randomized, double-blind trial by Klingberg's group, 53 children aged 7 to 12 with ADHD, not on stimulant medication, were enrolled, and 44 completed more than 20 days of training. The training group improved on an untrained visuospatial memory task after training and 3 months later, and parents rated fewer symptoms.

Later reviews pooled many trials. Rapport and colleagues (2013, 25 studies) found that training short-term memory improved short-term memory (d = 0.63), while effects on academic skills and on behavior rated by blinded observers were not significant. Raters who knew which children were trained reported larger benefits than blinded raters. Cortese and colleagues (2015, 16 randomized trials, 759 children and teens) found gains on working memory tests, but ADHD symptom effects shrank when raters were probably blinded, academic effects were not significant, and working memory training on its own had no effect on ADHD symptoms.

Two later Cogmed trials by other research groups followed the same pattern. In a 2014 trial with 85 children aged 7 to 11 and a closely matched placebo version, Cogmed improved simple working memory storage but no other outcome, and the authors said it "should not be considered a viable treatment" for ADHD. A 2016 trial in 452 first graders with low working memory (not an ADHD sample) found a visuospatial memory gain at 6 and 12 months that was gone at 2 years, no other benefits, and lower maths scores at 2 years in the trained group.

For ADHD care, talk with your child's doctor or psychologist.

Our working memory games and the tasks they are modeled on

This section covers four Optics Trainer memory games. Each is modeled on a lab task, with game changes such as levels that adapt, timing and visual design. None was part of the studies above, and none is a test or measure of working memory. All four can also be played in a binocular mode, where each eye sees part of the display, with red-blue glasses on a phone, tablet or computer, or built in on Meta Quest. Optics Trainer does not have an n-back game.

Optics Trainer memory games and their lab models
Game	What you do	Modeled on	What we kept and changed	Binocular mode
Symbol Order	Watch a sequence of directional symbols flash on pads, then repeat it; the sequence grows	Corsi block / Simon-style sequence span	Keeps the growing sequence; uses symbols on pads instead of blocks	Yes
Number Flash	A number flashes briefly; type it from memory; numbers can get longer	Digit span with brief exposure	Keeps recall in order; adds a short display time, so speed of taking it in matters too	Yes
Memory Grid	Study the whole board, then uncover every matching pair	Pairs ("Concentration") spatial recall	Recall of where items were, rather than a strict sequence	Yes
Working Memory	Numbered tiles appear, then hide; tap them from lowest to highest	Numbered-location spatial span	Combines where each tile was with its number and order	Yes

Where Optics Trainer fits

Optics Trainer combines vision and brain training games on iPhone, iPad, Android, web browsers and Meta Quest. Its memory games are a way to practice and challenge yourself with tasks modeled on digit span, Corsi-style sequences and spatial recall, for kids, teens and adults who enjoy them.

Optics Trainer was not tested in any of the studies described here. We do not claim it changes memory, attention, IQ, school or work performance, or ADHD or dementia. It is not a test, a measure or a treatment, and it is not FDA cleared. Optics Trainer is not a substitute for professional diagnosis or prescribed care. Consult your eye care professional for guidance about your specific condition.

The phone and tablet apps start with a 7-day free trial. After that, mobile and web access is $14.99 a month or $149.99 a year. On the web, a free account includes 3 plays a day.

Download Optics Trainer

Get it on iOS Get it on Android Get it on Meta Quest Play in your browser

Frequently asked questions

What are working memory games?

They are games that ask you to hold information in mind for a few seconds and use it, such as repeating a sequence, recalling where tiles were, or keeping a running count. Most are based on lab tasks like digit span, the Corsi block task and the n-back.

Does working memory training work?

Meta-analyses report that people get better at working memory tasks after training. Melby-Lervåg and colleagues (2016) and Sala and Gobet (2020) found no convincing transfer to intelligence, reading or maths when trained groups were compared with active control groups.

Are working memory games good for kids?

A 2020 meta-analysis of 41 studies in typically developing children found small to medium gains on memory tasks and essentially no effect on intelligence, maths or language. Games can be a fun challenge, but those are the reported results.

Does working memory training help ADHD?

Meta-analyses by Rapport (2013) and Cortese (2015) found gains on working memory tests, but effects on ADHD symptoms were smaller or absent with blinded raters, and academic effects were not significant. Talk with your child's doctor about ADHD care.

Is dual n-back training worth it?

Researchers disagree. A 2015 meta-analysis reported a small effect of n-back training on fluid intelligence; critics re-analyzed it and found the effect was not significant against active controls, and the original authors disputed that re-analysis.

Are there free working memory games online?

Yes, many free online versions of digit span and Corsi tasks exist. Their scores are not standardized clinical measures. Optics Trainer's free web account includes 3 plays a day, and the phone and tablet apps start with a 7-day free trial.

What is the Corsi block test?

It is a visuospatial span task: you watch blocks tapped one at a time and repeat the order, with sequences getting longer. It comes from Philip Corsi's 1972 thesis and is widely used to assess visual memory span.

Sources

Related guides

Try Optics Trainer's memory games

Symbol Order, Number Flash, Memory Grid and Working Memory are modeled on classic span and recall tasks. Play them on phone, tablet, web and Meta Quest. The phone and tablet apps start with a 7-day free trial.

Start free