Using analogies to improve elementary school students' inferential reasoning about scientific concepts

School Science and Mathematics, Mar 2001 by Yanowitz, Karen L

Procedure. Texts were read aloud to individual participants, to control the number of presentations. Students were randomly assigned to conditions. The analogical texts were read once, and the nonanalogical texts were read twice. Therefore, all students heard the structural relations twice. Those who received the analogical texts heard the structure once in the context of the science domain and once in the context of the source domain. Students who received the nonanalogical texts heard two repetitions of the structure in the context of the science domain. All questions were read aloud, and participants' answers were recorded for later transcription. After the texts were read, participants were asked to describe what they could recall and the inference question was asked. Order of presentation of the texts was counterbalanced.

Results and Discussion

Preliminary analyses revealed no differences in performance for recall or inference questions as a function of text presentation order or gender.

Performance on recall. As in Experiment 1, recall was scored for the presence or absence of the main idea ofthe science concept and then summed to obtain a total recall score. Therefore, scores could range from 0 (no correct recall for any stories) to 5 (correct recall for all five stories). An undergraduate assistant and the author independently scored all participants' recall responses. Percent agreement on recall for each story ranged from 92% to 100% for each condition for each grade, and disagreements were resolved through discussion.

As seen in Table 2, participants showed a fairly high level of recall of the main ideas of the texts. A 2 (text type; analogy, nonanalogy) by 2 (grade; third, fifth) ANOVA was calculated to determine the effects of type of text on recall, and a marginal difference was obtained. Students who received nonanalogical texts showed slightly higher recall of the main ideas (M= 4.3, SD = 0.9) than those who received analogical texts (M = 3.8, SD = 1.3), F(l,69) = 3.4, p = .07). No significant difference was seen in recall as a function of age, and no significant interaction between grade and text type was obtained.

Performance on inference questions. As in Experiment 1, responses were scored as correct if they described a consequence that might occur as a result of the changed situation. For instance, the inference question for the enzyme concept asked, "What would happen if the shape of the enzyme changed?" Students who said, "It wouldn't fit inside the protein," or "It wouldn't be able to join up," were given credit for this question. Scores for each question were summed to obtain a total inference score and could range from 0 (no correct answers) to 5 (correct answers for all questions). The author and an undergraduate assistant independently scored all students' responses. Percent agreement on answers to each inference question ranged from 89% to 100% for each condition, and disagreements were resolved through discussion.

Table 2 also presents inference scores at each grade as a function of text type. A 2 (text type; analogy, nonanalogy) x 2 (grade; third, fifth) ANOVA revealed that students answered significantly more inference questions correctly if they had received analogical texts (M = 3.4, SD = 1.4) than if they had received nonanalogical texts (M= 2.5, SD = 1.1), F(l, 69) = 7.8, p