What Is the N-Back Working Memory Test?
The N-Back Test is a browser-based working-memory challenge in which one of nine squares lights up at a time. Your job is to decide whether the current square is in the same location as the square N positions earlier. For example, in 2-Back mode, compare the present location with the location shown two items ago. If they match, press Match; otherwise, let the trial pass without a response.
The task exercises short-term information tracking and continuous updating. You must keep a moving record of recent positions rather than memorize one fixed list at the beginning. With every new square, the reference position changes. That makes the test different from digit recall, where a sequence is displayed and then reproduced after it ends.
ReflexBenchmark offers three separate difficulty modes: 1-Back, 2-Back and 3-Back. Each mode uses a 3×3 grid and 30 scored trials after its initial warm-up squares. Results describe your performance on this particular timed browser task. They are not clinical diagnoses, intelligence scores or evidence that practice will improve unrelated cognitive abilities.
How to Play N-Back Step by Step
- Select your level. Start with 1-Back if you are new, or choose 2-Back or 3-Back for a longer comparison window. The selected level is displayed above the grid.
- Press Start Test. Watch the square positions carefully. The first N squares are unscored warm-up stimuli and establish the initial comparison history.
- Watch each new square. It lights up for approximately 850 milliseconds; the response window lasts 1,700 milliseconds from its appearance.
- Respond only to a match. Tap Match, press M or press Space when the location equals the square N positions earlier. Do nothing when it differs.
- Finish all 30 scored trials. The display shows progress and registered match hits. An ignored match becomes a miss; pressing on a non-match becomes a false alarm.
- Review your result. Compare balanced accuracy, raw accuracy, hits, correct rejections, response time and the five six-trial graph blocks. Try the same level again to make a meaningful comparison.
1-Back vs 2-Back vs 3-Back
The number before “Back” tells you how far into the recent sequence you must compare. It does not change the total number of scored decisions or the target/non-target balance. You can switch difficulty before a run, but a run already in progress must be completed or interrupted before switching.
| Mode | Compare with | Warm-up stimuli | Scored decisions |
|---|---|---|---|
| 1-Back | Immediately previous square | 1 | 30 |
| 2-Back | Square two steps earlier | 2 | 30 |
| 3-Back | Square three steps earlier | 3 | 30 |
In any level, 10 of the 30 scored trials are true matches and 20 are non-matches. The order is shuffled for each run. Matching locations are created according to the selected N, so the game does not rely on a simple repeated visual pattern that works the same way at every difficulty.
How N-Back Scoring Works
The main displayed percentage is balanced match-detection accuracy, not just the percentage of all trials you answered correctly. This matters because two-thirds of the scored trials are non-matches. A player who never presses Match can reject many non-matches correctly, but has detected none of the real matches.
To make both kinds of decisions count equally, the game calculates a hit rate from 10 match trials and a correct rejection rate from 20 non-match trials. The balanced score is the average of these two rates, rounded to a whole percentage using the game's rounding rule.
For example, getting 7 of 10 matches correct and correctly ignoring 16 of 20 non-matches gives a 70% hit rate and an 80% rejection rate, for a balanced score of 75%. Raw accuracy in that example is 23 of 30, or approximately 77%. These are different measurements of the same run.
What Each Result Metric Means
| Result | How to interpret it |
|---|---|
| Hits | Correct Match presses on the 10 actual match trials. |
| Missed matches | Actual matches where you did not press Match before the response window ended. |
| Correct rejections | Non-match trials correctly left without pressing Match, out of 20. |
| False alarms | Non-match trials on which Match was pressed. |
| Median match response | The middle response time among successful match presses; shows “No hits” if none occurred. |
| Overall correct | Hits plus correct rejections, out of 30 scored trials. |
Pressing Match multiple times does not earn extra credit: the interface accepts at most one response per trial. The button and M/Space shortcut are alternatives for the same action. Silence is an intentional response on a non-match, not a failure to participate.
Understanding Your Five-Block Performance Graph
After the 30 scored trials, the test divides your results into five consecutive blocks of six trials. Each bar shows how many of those six decisions were correct, counting both detected matches and correctly ignored non-matches. A bar can therefore range from 0 to 6. These are measured outcomes from the run, not a decorative progress pattern.
The complete trial history distinguishes MATCH from OTHER (non-match) events and records whether a press happened. When a match is pressed, the history also shows its measured response time. An omitted press is displayed differently from a registered key or touch response. Use the history to look for stretches of missed matches, false alarms or steadier performance as the run continues.
The timing values include your browser, display and input latency. They are best compared using the same device, browser conditions and N-Back level. A different refresh rate, keyboard or amount of background activity can change response timing without any meaningful change in memory ability.
Practical Tips for Better N-Back Accuracy
- Keep your eyes near the grid center. Look for where the illuminated cell appears without moving attention to the Match button after every square.
- Use a compact mental rhythm. In 1-Back, hold the previous location. In 2-Back and 3-Back, update a small rolling sequence after each new stimulus.
- Favor accuracy over frantic pressing. A false alarm lowers your correct-rejection rate. Only respond when the comparison is clear.
- Practice one level consistently. Comparing three 2-Back attempts is more informative than treating a 1-Back score and a 3-Back score as equivalent.
- Review both mistake types. Many misses may mean you lost the recent position sequence; frequent false alarms may mean you pressed on visual familiarity rather than the exact N-step position.
- Take breaks between attempts. Rushing repeatedly can make sustained attention harder. Short, focused sessions make the results easier to interpret.
Common Mistakes and How to Avoid Them
Comparing with the last square in every mode: That rule works only in 1-Back. In 2-Back you must skip one position; in 3-Back you must compare three positions earlier. Recheck the chosen level before starting.
Pressing for any repeated location: A square may have appeared earlier but still not be an N-Back match. The game checks one exact historical position, not whether a location has ever been shown during the session.
Pressing during warm-up: The first N squares supply memory context but are not scored. Wait until the display identifies scored trial 1 before judging matches.
Always ignoring or always pressing Match: Either shortcut fails to distinguish the two types of trial. With the fixed 10-match/20-non-match balance, either constant-response strategy earns only 50% balanced accuracy.
Switching browser tabs: Hiding the tab interrupts an active test. The run is not submitted as a completed score, and you must restart. Leave the page visible while playing.
What This N-Back Test Can and Cannot Tell You
Higher scores indicate stronger performance on this specific set of spatial comparisons under its chosen difficulty and timing. The result can help you practice the task, observe changes across comparable attempts and identify whether missed targets or false alarms are a bigger source of errors. It cannot diagnose a memory disorder, determine IQ, measure everyday functioning or prove general cognitive improvement.
Working-memory research uses many variations of N-Back tasks, including different stimuli, pacing rules and scoring methods. A result here should not be assumed interchangeable with a research laboratory's protocol. Your score is a browser-game outcome, and any attempt-based rank depends on enough previously recorded comparable attempts at the same level.
For different memory challenges, try the Digit Span Test for immediate numeric sequence recall, the Corsi Block Test for spatial order recall or the Visual Memory Test for grid-pattern recall. Each task uses different rules, so compare progress within a test rather than comparing raw scores across them.