Key evidence & numbers
| Version | Information tracked | Scoring detail to preserve |
|---|---|---|
| Single visual n-back | One visual identity or location stream | Matches, false alarms and response timing |
| Single auditory n-back | One auditory stream | Audio presentation and response rules |
| Dual n-back | Two streams in parallel | Separate-stream and combined scoring rules |
N-back research includes many implementations, not one standard game. Owen and colleagues distinguish identity, location and content differences, while the training meta-analysis considers variations in training paradigms and outcomes. Those design differences make a bare n level an incomplete description of performance. Owen et al., 2005 Soveri et al., 2017
What additional demand does dual n-back introduce?
Monitoring two streams adds coordination requirements, but the precise difficulty depends on how streams are timed and scored. It is not valid to multiply a single-task score by two or subtract a fixed number of levels to estimate a dual score. Working-memory theory includes multiple representations and control processes rather than one task-independent number of slots. Baddeley, 2012 Cowan, 2001
This distinction also explains why a single digit-span score cannot settle which version someone will find easier. Span procedures impose different recall and processing requirements, and the meta-analysis comparing n-back with complex span rejects treating them as interchangeable. Conway et al., 2005 Redick et al., 2013
Is dual n-back scientifically better?
That question needs a defined outcome: better performance on the practised task, transfer to another memory task or improvement in broad reasoning are different endpoints. The training meta-analysis separates these outcomes. More demanding or more elaborate training is not by itself evidence of a larger general benefit. Soveri et al., 2017
What affects it?
What does the transfer evidence allow?
Related-task improvements can occur, particularly when the assessment resembles training. The broader working-memory review finds that claims about intelligence or other far-transfer outcomes weaken under appropriate control comparisons. These findings do not support promising that dual n-back makes a person generally smarter. Melby-Lervåg et al., 2016 Soveri et al., 2017
They also do not imply that learning the game is worthless. A person can become more skilled at maintaining the rule, attending to the stimuli and responding accurately. The scientifically defensible description stays with the demonstrated task or closely measured transfer outcome. Soveri et al., 2017
Why online levels do not form a common scale
Stimulus pace, target proportions, adaptation thresholds, lures and response windows can differ. A slower dual task can offer different demands from a faster single task. Without matched procedures and calibration, equal n labels do not define equal ability or representative age and sex norms. Owen et al., 2005 Redick et al., 2013
Reliability must also be measured for the particular score, especially a difference between versions. The reliability-paradox literature shows why robust experimental tasks can still yield noisy individual differences. A single level reached once should not become a fixed label for your working memory. Hedge et al., 2017
If you compare versions informally, keep their details in view and track correct responses and false alarms. Treat any preference or progress as task-specific. The evidence does not supply a universal best n level, a guaranteed training dose or a conversion from dual-task performance to IQ. Conway et al., 2005 Melby-Lervåg et al., 2016
Try the related free test
N-back test
Try the n-back test and follow the version described on the page. If you use a separate dual implementation, record it separately rather than treating the n labels as one common scale. Helibrain's task does not establish broad training benefits.
Common questions
Is dual 2-back the same as single 4-back?
No. The number specifies lag; adding another stream changes coordination demands rather than simply doubling the lag. Owen et al., 2005
Will dual n-back improve every memory task?
The reviews distinguish task learning, related transfer and far transfer. Broad improvement is not established by a higher training score. Soveri et al., 2017 Melby-Lervåg et al., 2016
Which version should I use to measure IQ?
Neither version supplies a direct validated IQ conversion from its n level. Different working-memory tasks are not interchangeable. Redick et al., 2013
Sources
- Owen et al. (2005). N‐back working memory paradigm: A meta‐analysis of normative functional neuroimaging studies.
- Redick et al. (2013). Complex span and n-back measures of working memory: A meta-analysis.
- Cowan (2001). The magical number 4 in short-term memory: A reconsideration of mental storage capacity.
- Baddeley (2012). Working Memory: Theories, Models, and Controversies.
- Conway et al. (2005). Working memory span tasks: A methodological review and user’s guide.
- Soveri et al. (2017). Working memory training revisited: A multi-level meta-analysis of n-back training studies.
- Melby-Lervåg et al. (2016). Working Memory Training Does Not Improve Performance on Measures of Intelligence or Other Measures of “Far Transfer”.
- Hedge et al. (2017). The reliability paradox: Why robust cognitive tasks do not produce reliable individual differences.
Explore your daily routines in Helibrain
The app has short tests of reaction time, working memory, self-control, reasoning and creativity. Retake them whenever you like, see your trend, and find habits that may help.
Start free in your browser
