Investigating the Generation-First-Instruction-Later Method for its Effects on Learning and Transfer - A Proposal to Study Analogical Reasoning as the Generation Task

Project Number
OER 02/15 HJS

Project Duration
July 2015 - December 2017


The generation-first-instruction-later (G-I) method leverages on student-generated ideas for instruction and leads to students' initial failure after generation and subsequent success after instruction. The generation phase activates and differentiates students' prior knowledge and the compare and contrast in the instruction phase help students attend to their knowledge gaps and encode the new knowledge. While problem-solving is frequently studied as the generation task, this study seeks to investigate analogical reasoning (i.e., students generating analogous problems, in short, SGA, and students generating structural mappings between a given problem and the analogous problem, in short, AE) as the generation task. Involving SGA and AE as the generation tasks allows further investigation on the learning mechanisms involved in the G-I method in three ways. First, recent studies find that compared to the problem-solving task, the problem-posing task in the generation phase produces better effect on transfer. A fine-grained investigation on transfer effect is possible when SGA and AE are used as generation tasks. Second, SGA and AE both embody the compare-and-contrast mechanism. By extending the compare-and-contrast mechanism from the instruction phase to the generation phase, the study seeks to better understand the learning mechanisms involved in the G-I method and the effect on learning outcome. Third, between SGA and AE there is a trade-off on degrees of freedom in generation which allows further differentiation of the learning mechanisms involved. To achieve the research objectives, a two (activity sequence: G-I vs I-G) by two (generation task: SGA vs AE) factorial experimental design is proposed for the main study. 7th graders (n=320) will take part to learn two-variable algebra. They will be assigned using proportional stratified random sampling based on their analogical reasoning ability. The participants will work individually in the generation phase and receive instruction as a group in the instruction phase. All groups will be taught by the same instructor. The participants' math ability, prior algebra knowledge and analogical reasoning ability will be measured in pretest. The participants will self-report their mental efforts and attend a posttest after each phase (i.e., the generation phase and the instruction phase). They will also self-report their confidence level on each answer in the posttests. The last posttest involves transfer tasks to isomorphic problems, manipulated problems and extended problems. The artefacts (e.g., analogous problems and mappings of similarities and differences) generated by the participants in the generation phase will also be coded. The main data analysis involves two steps. First, to analyze the learning outcomes, MANCOVAs will be applied to compare the two conditional groups with math ability, prior algebra knowledge and analogical reasoning ability as covariates. Second, to analyze the learning processes, MANOVA will be conducted to compare the two conditional groups to check the multivariate effects of process measures, such as mental effort, the number of generation and the quality of generation. For within-subject differences, regression analysis will be applied to check to what extent the process measures can predict the learning outcome. Findings contribute to the understanding of the mechanisms involved in the G-I method.

Funding Source

Related Links
ReEd Vol 20 (2017): Solving Problems through Analogies

Related Projects