Computerized Adaptive Testing (CAT)
Also known as: Adaptive Testing, Tailored Testing, Item-Adaptive Testing, Bilgisayar Destekli Uyarlanabilir Test
Computerized Adaptive Testing (CAT) is an individualized assessment methodology in which a computer algorithm selects successive test items based on a running estimate of each examinee's latent ability. Grounded in Item Response Theory, CAT dynamically tailors the item sequence so that each question is optimally informative given the current ability estimate. The framework was systematized and popularized by Howard Wainer and colleagues through the foundational primer first published in 1990 and expanded in the 2000 second edition.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
CAT is appropriate when individual-level precision matters more than group-level comparisons, and when an item bank of 100 or more calibrated items is available. It requires items pre-calibrated under a suitable IRT model (Rasch, 2PL, or 3PL), a reliable computer delivery system, and acceptable exposure-control constraints. Avoid CAT when retesting is common without item-bank refresh, when item security cannot be guaranteed, or when all examinees must receive identical items for legal comparability. Alternatives include multistage adaptive testing (MST) or linear on-the-fly testing (LOFT).
Strengths & limitations
- Achieves equivalent measurement precision with 40–60% fewer items than conventional fixed-form tests, reducing examinee fatigue.
- Provides real-time ability estimates and standard errors, enabling immediate score reporting.
- Adapts to each examinee individually, making the assessment experience more engaging and appropriate across ability levels.
- Supports flexible testing windows since each examinee receives a unique item set drawn from a large calibrated bank.
- Requires a large, pre-calibrated item bank (typically 100+ items per dimension) and significant upfront psychometric investment.
- Item exposure and content-balancing constraints add algorithmic complexity and can reduce measurement efficiency.
- Adaptive selection complicates score comparability when strict item-level fairness or legal defensibility requires identical items for all examinees.
- Ability estimation is less stable at the extremes of the ability distribution, where item bank coverage is typically thin.
Frequently asked
How many items does a CAT typically require compared to a fixed-form test?
Empirical studies consistently show that CAT achieves the same measurement precision as a full fixed-form test using approximately 40–60% of the items, depending on item bank size, stopping criterion, and ability distribution. For a 200-item fixed form, an equivalent CAT often converges in 50–80 items for examinees near the center of the ability distribution.
Can CAT scores be compared across administrations when different items are given?
Yes, provided items are calibrated on a common IRT scale. Because CAT scores are expressed as estimated latent trait values (theta) on the same underlying metric, comparisons across administrations and across examinees are valid. This is a core advantage of IRT-based adaptive testing: the score meaning is item-invariant as long as the measurement model fits.
What happens if an examinee performs at the extreme ends of the ability scale?
At the extremes, the item bank typically contains fewer well-targeted items, so the standard error decreases more slowly and the stopping criterion may not be met until the maximum item limit is reached. Scores near the floor or ceiling are less precise, and it is good practice to flag such cases and interpret them with appropriate caution.
Sources
- Wainer, H. (2000). Computerized Adaptive Testing: A Primer (2nd ed.). Lawrence Erlbaum Associates. ISBN: 978-0-8058-3511-3
How to cite this page
ScholarGate. (2026, June 2). Computerized Adaptive Testing (CAT). ScholarGate. https://scholargate.app/en/psychometrics/computerized-adaptive-testing
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Rasch ModelPsychometrics↔ compare