Modeling Response Processes to Achieve an Appropriate Factor Structure for the Bafacalo Logical Reasoning Test

Responses to reasoning tests may reflect not only the ability the tests were designed to measure but also response processes distinct from that ability, which can obscure their factor structure when left unmodeled. The BAFACALO Logical Reasoning Test has been openly available since 2021, but the previous factorial study relied on specific factors derived from an exploratory analysis and on an estimator that treats dichotomous items as continuous, and its model fell just short of acceptable fit. As a syllogism-based test, it is open to response processes such as illicit conversion, the atmosphere effect, and acquiescence. This study reexamines its factorial validity by modeling these response processes, with data from 292 Brazilian public high school students. Five confirmatory factor models were compared: a unidimensional model, the bifactor model of the previous study, and three bifactor models with theory-driven specific factors for illicit conversion and the atmosphere effect, for acquiescence, and for both. Models were estimated with WLSMV and, for comparison, ML and MLMV. No model reached acceptable fit with ML or MLMV. With WLSMV, only the models including illicit conversion and the atmosphere effect were acceptable, and the model that also included acquiescence was selected. The general factor (logical reasoning) and the illicit conversion and atmosphere effect factor showed acceptable reliability, whereas the acquiescence factor did not. Modeling these response processes yielded an appropriate factor structure for the test, separating them from logical reasoning. These findings strengthen the evidence on the factor structure of this openly available test.