Relational reasoning and scientific achievement : concurrent and longitudinal associations, underlying mechanisms, and causality

Abstract

Scientific achievement is crucial to academic success and societal development. Understanding its learning mechanisms can help improving pedagogy. Relational reasoning (RR), the domain-general ability of extracting meaningful patterns from informational streams, was proposed to be related to multiple domains of academic achievement, including science. Recent studies revealed associations between RR and performance in the scientific context (e.g., making medical diagnosis) but its roles in scientific reasoning and achievement remained unclear. The relation between scientific reasoning and achievement also remains to be elucidated. The present thesis aimed at addressing these issues by three studies, investigating the associations, underlying mechanisms, and potential causality between RR and science achievement. Study 1A aimed at identifying the concurrent associations of RR in scientific reasoning and achievement. A sample of 194 ninth graders were assessed on their RR, scientific reasoning, and scientific achievement. Scientific reasoning was proposed as the potential mediator. Linear hierarchical regression analyses and structural equation modelling were conducted to investigate the associations and mechanism. Findings revealed that after controlling the effect of cognitive abilities (verbal working memory, visuospatial working memory, and spatial skills), RR significantly predicted both scientific reasoning and achievement. Exploratory analyses revealed that among the four forms of RR, only analogy (similarity) and antithesis (opposition) predicted scientific reasoning, and only analogy predicted scientific achievement. Mediation analyses revealed that scientific reasoning mediated the relation between RR and scientific achievement. The results supported the concurrent associations of RR in both scientific reasoning and achievement. Study 1B elaborated on the RR-science relation via a longitudinal design. The study investigated whether RR measured at grade 5 could predict scientific achievement when students started to learn science formally at grade 7. A sample of 182 students were tracked from grade 4 to 7. Their RR, scientific reasoning, and achievement were measured at grade 5, 6, and 7 respectively. Results of structural equation modelling showed that students’ relational reasoning demonstrated an indirect effect to scientific achievement through scientific reasoning. A longitudinal relation of RR to scientific achievement was observed in Study 1B and its causality would be addressed in Study 2. Study 2 attempted to explore the potential causality of RR via two pre-registered online priming experiments. The first experiment recruited 146 English native speakers. Semantically distant analogy problems (furnace:coal::stomach:food) were used to prime participants’ analogical reasoning, while semantically near problems served as control. The second experiment recruited 173 participants. Two novel antithetical reasoning tasks were used to prime antithetical reasoning, while near analogy problems served as control. All participants completed respective RR and scientific reasoning measures. Results revealed that RR was significantly correlated with scientific reasoning, and priming efficiency correlated with both outcome measures. Contrary to the hypotheses, Bayesian ANCOVA analyses showed that the priming did not produce facilitative effect on RR performance. Priming efficiency was found to moderate the effect of priming in scientific reasoning. Future improvement on RR intervention were discussed. The findings shed light on the predictive role of RR on scientific reasoning and achievement, channeling momentum to further theoretical and pedagogical discussions.