Skip to main content
v2026.11,986 entries · CC-BY 4.0

Cogstate: Computerized Cognitive Test Battery for Research

What the Cogstate Brief Battery measures, how its scores are transformed, and what validity and test-retest studies report.

Written and maintained by CASRAI Editorial Board

Published

On this page

Cogstate is a neuroscience technology company and also the name researchers use for its computerized cognitive tests, most often the Cogstate Brief Battery (written CogState in older papers). Researchers use it to measure speed, attention, learning and working memory in a few short card and maze tasks. This page describes the instrument for people choosing or interpreting a research measure. It is not a diagnostic guide and does not interpret anyone’s personal results.

What is Cogstate? (definition)

Cogstate describes itself as a neuroscience solutions company whose offerings include digital cognitive assessment, endpoint data quality (rater services), remote assessment, healthcare cognitive screening, scientific consulting and research collaborations (Cogstate website, read 2026-11-08). Its own materials state that its tests have been used in drug efficacy and safety trials, including Alzheimer’s disease trials, and in studies of placebo response in major depressive disorder. The company’s headline figures (about 25 years of operation, 2.4 thousand academic and clinical trials, 2 million digital tests and 111 languages) are the company’s self-reported claims, not independently audited numbers.

The instrument is commercial and licensed through the company; access is arranged with Cogstate rather than downloaded as an open scale. For how a battery like this fits the wider field, see psychometrics. For a paper-and-pencil comparison instrument, see the RBANS; for a large study that uses cognitive endpoints, see ADNI.

Cogstate Brief Battery tasks and outcome measures

The tasks below follow the description in the 2015 validation study by Bangirana and colleagues, which administered the battery to children in Uganda (see Sources). The tasks are computer-based and use playing-card or maze stimuli, which keeps the language demand minimal.

Task Construct Outcome metric Better performance
Detection Psychomotor speed Mean log10 reaction time of correct responses Lower
Identification Visual attention Mean log10 reaction time of correct responses Lower
One Card Learning Visual learning and memory Accuracy: arcsine square root of the proportion correct Higher
One Back Memory Working memory Accuracy: arcsine square root of the proportion correct Higher
Groton Maze Chase Visuomotor speed Correct moves per second Higher
Groton Maze Learning Spatial working memory and learning Total errors Lower

Scoring: formulas and a worked example

Raw reaction times and proportions are skewed or bounded, so the battery’s outcomes are transformed before analysis. Scores are then often standardized against a reference group. In the Bangirana study, scores were converted to age-adjusted z-scores:

z = (x − Mage) / SDage, where x is the child’s transformed score, Mage is the mean transformed score of the control group at that age, and SDage is that group’s standard deviation. See z-score for the general concept.

Illustrative example with made-up numbers, not data from any study. A participant’s mean correct reaction time on Detection is 800 ms. Step 1: log10(800) = 2.9031. Step 2: suppose the reference group at that age has M = 2.9500 and SD = 0.1000. Step 3: z = (2.9031 − 2.9500) / 0.1000 = −0.469. Because lower reaction time is better, the negative sign means faster than the reference mean; many analysts multiply speed scores by −1 so that higher always means better.

For an accuracy task, 18 correct out of 20 gives a proportion of 0.90. The transform is arcsin(√0.90) = arcsin(0.9487) = 1.2490 radians, which is then standardized in the same way.

Interpreting z-scores from the battery

z-score (after sign alignment) General reading
About 0 Near the reference group’s mean
+1 One reference SD better than the mean
−1 One reference SD worse than the mean
Beyond ±2 Far from the reference mean; in a normal distribution about 5% of people fall outside this range

This table describes the z-score scale in general, not clinical cut-offs. A z-score is only as meaningful as the reference sample behind it, so the age range, language and setting of the control group matter.

Validity and reliability evidence

Validity

In the Bangirana study (n = 230 children, mean age 6.99, SD 1.67, range 5 to 13), the battery was compared with the KABC-II and the TOVA. One Card Learning correlated r = 0.53 and One Back r = 0.55 with TOVA d’; Detection correlated r = 0.38 with TOVA reaction time; Identification showed no significant correlations; factor analysis suggested three factors (accuracy, speed and maze). In a separate adult study, Maruff and colleagues (2009) reported, in the abstract, construct validity correlations with conventional measures of r = .49 to .83 in healthy adults, and lower performance in groups with mild head injury, schizophrenia and AIDS dementia complex than in controls (Cohen’s d from −0.60 to −1.80). The authors concluded that construct and criterion validity were acceptable. For the logic behind these comparisons, see convergent and discriminant validity.

Reliability

Eight-week test-retest correlations in the Bangirana control group were 0.32 (Detection), 0.43 (Identification), 0.50 (One Card Learning), 0.57 (One Back), 0.42 (Maze Chase) and 0.35 (Maze Learning). The authors described these as low to moderate and noted that the study was not designed primarily to establish validity and did not assess sensitivity or specificity. See test-retest reliability for how to read such coefficients.

Assumptions and common mistakes

  • Treating one study as a property of the instrument. The figures above come from specific populations (Ugandan school-age children; adult clinical and healthy groups). They do not transfer automatically to your sample.
  • Mixing score directions. Reaction time and error scores improve downward; accuracy and moves per second improve upward. Align the sign before combining tasks.
  • Analyzing untransformed scores. Use the transformed outcomes the battery defines, and report which transform you used.
  • Ignoring score distribution limits. Short tasks can show floor and ceiling effects in some groups; check the distribution before choosing an endpoint.
  • Using the z-score as a diagnosis. A research z-score is a position relative to a reference group, not a clinical judgment.

Using Cogstate in clinical research

Trial sponsors use computerized batteries because they standardize administration and capture reaction times that paper tests cannot. When considering it as an endpoint, the published reliability coefficients above are the main caution: moderate test-retest correlations limit how precisely an individual’s change can be detected. Decide the endpoint, reference data and licensing route in the protocol and statistical analysis plan before enrolment.

How to report it (APA 7th style)

Name the task, the outcome metric, the transformation and the reference group. Example wording: “Psychomotor speed was measured with the Cogstate Detection task (mean log10 reaction time of correct responses) and standardized against age-matched control means (z = (x − M)/SD).” Cite the validation paper for the version you used, for example:

Bangirana, P., Sikorskii, A., Giordani, B., Nakasujja, N., & Boivin, M. J. (2015). Validation of the CogState battery for rapid neurocognitive assessment in Ugandan school age children. Child and Adolescent Psychiatry and Mental Health, 9, Article 38.

How to calculate the z-score in Excel, R or Python

Excel: with the reaction time in milliseconds in A2, control mean in B2 and control SD in C2, enter =(LOG10(A2)-B2)/C2. For accuracy, use =(ASIN(SQRT(18/20))-B2)/C2.

R: z <- (log10(800) - 2.95) / 0.10 gives −0.469, and asin(sqrt(18/20)) gives 1.249.

Python: import math; z = (math.log10(800) - 2.95) / 0.10; acc = math.asin(math.sqrt(18/20)).

FAQ

What does Cogstate measure?

The Cogstate Brief Battery measures psychomotor speed, visual attention, visual learning and memory, working memory, visuomotor speed and spatial learning through six short card and maze tasks.

Is Cogstate free to use?

It is a commercial platform provided through the company; arrange access and terms with Cogstate directly.

Is Cogstate a diagnostic test?

This page treats it as a research measure. Validation studies report group comparisons and correlations, not diagnostic accuracy, and the Bangirana study states it did not assess sensitivity or specificity. Questions about personal cognitive health belong with a clinician.

How reliable is it over time?

In one 8-week study of children, test-retest correlations ranged from 0.32 to 0.57, which the authors called low to moderate.

Sources

  • Cogstate. Company website, https://www.cogstate.com/ (read 2026-11-08).
  • Bangirana P, et al. (2015). Child Adolesc Psychiatry Ment Health, 9:38. https://pmc.ncbi.nlm.nih.gov/articles/PMC4536703/ (read 2026-11-08).
  • Maruff P, et al. (2009). Validity of the CogState brief battery. Arch Clin Neuropsychol, 24(2):165-178. doi:10.1093/arclin/acp010 (abstract-level figures).

Follow CASRAI

Research-administration guidance, standards updates and independent tool reviews.

Ask CASRAI · free to try

Ask about Cogstate: Computerized Cognitive Test Battery for Research

2 free questions a day, no account.

An AI assistant for research: funder rules, compliance, methods and publishing. It links the sources behind its answers and labels answers that come from a web search.

Answers draw on CASRAI's guides and dictionary plus the federal and funder documents we index: Federal Register, Grants.gov, Regulations.gov and UKRI.

Works on this site and inside Claude, Cursor and the AI tools you already use.

Everything CASRAI publishes — this page, the dictionary, the guides and the news — stays open to read, with no account and no card.

Referenced across the research world

University of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logoUniversity of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logo
  • University of Cambridge logo
  • Columbia University logo
  • Crossref logo
  • University of Edinburgh logo
  • Harvard University logo
  • University of Oxford logo
  • Princeton University logo
  • Stanford School of Medicine logo
  • University College London logo
  • ORCID logo

View CASRAI adoption →