Computer tests that adjust to the person taking them may lift motivation and raise anxiety at the same time, according to a review published on October 19, 2026, in the journal npj Science of Learning. The umbrella review of meta-analyses on computerised adaptive testing also linked the format with slightly lower test scores.
The authors, Yishen Song, Hanya Li and Qinhua Zheng of the Research Center of Distance Education at Beijing Normal University in China, note that computerized adaptive testing (CAT) has been widely applied in large-scale assessments worldwide. Their stated aim was to bring the existing evidence together into a comprehensive action guide.
What the researchers did
An umbrella review works one level above ordinary research. Instead of collecting individual studies, it collects meta-analyses, which are studies that already combine the results of many earlier studies. This review brought together 10 meta-analyses reported in six papers. The samples behind those meta-analyses ranged from 2,830 to 32,735 people, with a median of 16,842.
The team searched four databases: PubMed, ProQuest PsycINFO, Scopus and Web of Science. The review was registered in advance in the PROSPERO register under the number CRD420251207547. The researchers combined the reported effect sizes, ran a fresh meta-analysis on them and rated the quality of the included work with a Joanna Briggs Institute (JBI) appraisal tool for systematic reviews and research syntheses.
What they found
The results point in two directions at once. On the positive side, CAT was linked with higher motivation among test-takers, with an effect size (written as d) of 0.08, and with more positive emotions (d = 0.13). On the negative side, it was associated with more anxiety (d = 0.08) and with lower test scores, where the median effect size was −0.08.
All four numbers are below 0.15. The authors sum up their main finding as the simultaneous presence of increased motivation and increased anxiety when CAT is used. They also point to a richer testing experience that the spread of technology has made possible.
The paper is an early version. The journal says it is sharing the article early to give faster access to peer-reviewed, accepted research, and that this version may be edited before a final Version of Record replaces it. The paper was received on January 15, 2026, and accepted on October 2. The work was funded by the National Natural Science Foundation of China, and the authors declare no competing interests.
How adaptive tests work in practice
Adaptive designs already shape some well-known tests. The GRE General Test, which its maker ETS says is accepted by thousands of graduate, business and law schools worldwide, uses what ETS calls section-level adaptation. According to the GRE test structure page, the Verbal Reasoning and Quantitative Reasoning measures each have two sections. The first section of each measure is of average difficulty. The difficulty of the second section depends on how well the test-taker did on the first.
ETS gives a simple example: a candidate who does very well on the first Quantitative Reasoning section will get a second section at a higher level of difficulty. Within a section, test-takers can move forward and backward, use “Mark” and “Review” features to skip questions and come back to them, and change their answers.
ETS’s guide to understanding GRE scores explains what happens next. The raw score is the number of questions answered correctly, and a process called equating converts it into a scaled score. Equating accounts for the differences in difficulty introduced by the adaptive design, so a given scaled score reflects the same level of performance whichever second section a candidate received.
Why it matters
Many students meet adaptive tests when they apply for further study, and the review suggests the format itself is linked to how test-takers feel during a test. ETS’s explanation is useful context here. On the GRE, a harder second section follows a strong first section, and the scoring takes the difficulty of each section into account. ETS also offers free POWERPREP Online practice tests so candidates can try the test design features before test day.
For teachers, schools and test designers, the findings are a starting point rather than a verdict. The review rests on 10 meta-analyses, every effect size it reports is below 0.15, and the published version may still change. The authors themselves call for more high-quality meta-analyses that look at CAT in a broader range of settings and with more varied outcomes.
What to watch
The next step is the final Version of Record of the paper. Beyond that, the authors want new research to cover a wider range of application scenarios and outcome variables, which would show whether the mix of higher motivation and higher anxiety holds across different kinds of tests and test-takers.


