The analysis, published in Frontiers in Psychology, drew on 4,426 United States fifteen-year-olds who participated in PISA 2018. Seventeen ordinal questionnaire items represented five motivational attributes: learning goals, attitudes toward competition, motivation to master tasks, general self-efficacy, and fear of failure. Rather than collapsing these items into the official PISA index scores, the researchers built a theory-informed Q-matrix that linked each item to its focal construct, and, in three cases, to a second attribute when the item wording clearly expressed an additional motivational process. A sequential G-DINA model, a flexible diagnostic classification model designed for ordered responses, then estimated which attributes each student possessed.

The technical machinery matters here. In the sequential formulation, endorsing a higher response category is treated as progression through successive category transitions, and each transition can carry its own attribute specification. That means the model retains the full ordinal information in Likert-type responses instead of dichotomizing them, and it can test whether the strongest level of endorsement reflects the same motivational structure as a weaker one. The initial Q-matrix contained 54 category-transition specifications; 48 were retained unchanged, and six modifications across three items were supported by empirical diagnostics and expert content review. Three experts independently rated all 85 item-attribute pairs, agreeing on 96.5 percent of them, with Fleiss’ kappa of 0.83.

Fit was satisfactory in both a development sample of 3,098 students and an independent validation sample of 1,328 students, with standardized root mean square residuals of 0.031 and 0.034 respectively. Attribute reliability estimates ranged from 0.77 to 0.88. Notably, the theory-informed model outperformed a fully construct-aligned specification in which every item was linked only to its original PISA construct, suggesting that the multi-attribute relationships were doing real psychometric work rather than adding complexity for its own sake.

The prevalence results painted a picture of a broadly motivated population. Survey-weighted estimates indicated that learning goals had the highest prevalence at 0.77, followed by general self-efficacy at 0.72, motivation to master tasks at 0.66, attitudes toward competition at 0.56, and fear of failure at 0.52. All five attributes were more prevalent among students with higher core-stage reading performance, the PISA routing variable that reflects performance on automatically scored items during the common core stage of the adaptive reading assessment. The largest between-group difference appeared for self-efficacy, at 0.14, with learning goals and motivation to master tasks each showing differences of 0.12.

The joint configurations, however, delivered the study’s most striking findings. Eight configurations each characterized more than 3 percent of the population and together accounted for 65.9 percent of it. The most prevalent pattern, held by 14.4 percent of students, combined learning goals, motivation to master tasks, and self-efficacy without competition or fear of failure. The second, at 12.1 percent, added competition to that trio. Both were substantially more common in the higher-performance group. Meanwhile, the broadly low-endorsing configuration, in which students possessed none of the five attributes, was more than three times as prevalent in the low-performance group, at 10.8 percent versus 3.2 percent, showing the largest negative group difference of any pattern.

Fear of failure provided the clearest demonstration of why configurations matter. Its marginal prevalence differed only modestly between performance groups, at 0.52 versus 0.47, a difference of just 0.06. But the configurations containing fear told two very different stories. In the higher-performance group, fear more often co-occurred with learning goals, motivation to master tasks, and self-efficacy; the all-five-attributes configuration and the fear-plus-trio configuration together characterized roughly 19 percent of the population and were both more prevalent among stronger readers. In the low-performance group, fear appeared in isolation or combined only with competition, patterns that were both significantly more common among weaker readers. The same attribute, in other words, lived within radically different motivational worlds depending on the company it kept.

Competition showed a parallel split. The configuration combining competition with learning goals, effort-related persistence, and self-efficacy was substantially more prevalent in the higher-performance group, while the pattern of competition plus fear of failure, without any learning-, effort-, or efficacy-related attributes, was more prevalent in the low-performance group. This echoes prior research suggesting that competitive orientations can be associated with both greater engagement and greater anxiety, and it cautions against treating the PISA competition items as a direct measure of performance-approach goals. The authors also note that the absence of fear in the broadly low-endorsing configuration should not be read as favorable, since low evaluative concern can sometimes accompany withdrawal from achievement contexts.

The researchers are careful about what these cross-sectional, self-report data can and cannot show. The core-stage performance variable is not a diagnosis of reading difficulty or a final proficiency level, and the observed differences cannot establish whether motivation shaped performance or prior performance shaped motivation. Longitudinal meta-analytic evidence, including Vu and colleagues’ 2024 finding that prior achievement generally predicts subsequent motivation more strongly than the reverse, suggests both directions are plausible. The configurations are probabilistic classifications, not fixed student types, and the authors warn against using them as stand-alone screening tools without triangulating with direct performance measures, classroom observations, and well-being indicators.

Even with those limits, the study’s implications for assessment are hard to ignore. Students with low core-stage reading performance were anything but uniform: some showed broadly low endorsement, some showed isolated fear of failure, some combined competition with fear, and others endorsed learning goals, persistence, and confidence in managing challenges. A single scale score would have flattened that heterogeneity into a modest mean difference. What the diagnostic classification approach reveals instead is that understanding adolescent achievement motivation may require asking not just whether a characteristic is present, but which characteristics it travels with, a question that could reshape how educators interpret motivational data and design support for struggling readers.

Subject of Research: Diagnostic classification analysis of achievement motivation configurations and core-stage reading performance among United States adolescents in PISA 2018

Article Title: Achievement motivation configurations and core-stage reading performance among United States adolescents: a diagnostic classification analysis

Article References: Hwang, J., & Kim, D. (2026). Achievement motivation configurations and core-stage reading performance among United States adolescents: a diagnostic classification analysis. Frontiers in Psychology, 17, Article 1963203. https://doi.org/10.3389/fpsyg.2026.1963203

Image Credits: AI Generated

DOI: 10.3389/fpsyg.2026.1963203

Keywords: Achievement, motivation, configurations, core-stage, reading, performance, United, States, adolescents, diagnostic, classification, analysis