Flowing Toward Correct Contributions During Group Problem Solving: A Statistical Discourse Analysis

Abstract

Groups that created more correct ideas (correct contributions or CCs) might be more likely to solve a problem, and students' recent actions (micro-time context) might aid CC creation. 80 high school students worked in groups of 4 on an algebra problem. Groups with higher mathematics grades or more CCs were more likely to solve the problem. Dynamic multilevel analysis statistically identified watersheds (breakpoints) that divided each group's conversation into distinct time periods with many CCs versus few CCs, and modeled the groups' 2,951 conversation turns. Wrong contributions, correct evaluations of one another's ideas, justifications, and polite disagreements increased the likelihood of a CC. In contrast, questions, rude disagreements, and agreements reduced it. Justifications had the largest effects, whereas the effects of correct evaluations lasted 3 speaker turns. Some effects differed across groups or time periods. In groups that solved the problem, justifications were more likely to yield CCs, and questions were more likely to elicit explanations. Meanwhile, the effects of agreements and correct evaluations on CCs differed across time periods. Applied to practice, teachers can encourage students to evaluate others' ideas carefully and politely, express and justify their own ideas, and explain their answers to group members' questions.

View correction statement:

Erratum

Notes

^aEvaluation of the previous action (EPA): agreement [+], polite disagreement [–], rude disagreement [—-], ignore/new topic[*].

^bKnowledge content (KC): contribution [C], repetition [R], null academic content [N].

^cValidity: right [✓], wrong [X], null academic content [N].

^dJustification: justification [J], no justification [], null academic content [N].

^eForm of invitation to participate (IF): command [!], question [?], statement [_.].

^aVariables in full model. Classroom identification variables: Class_1, Class_2, Class_3. Group-level variables: correct solution, mean math grade, mean peer status, math grade variance, peer status variance, gender variance, race variance. Current speaker (0) variables: gender, race, math grade, peer status, correct evaluation, agree, politely disagree, rudely disagree, justify, question and command. Previous speakers' lag variables (i = 1..4): gender (−i), race (−i), math grade (−i), peer status (−i), correct evaluation (−i), agree (−i), politely disagree (−i), rudely disagree (−i), CC (−i), wrong contribution (−i), correct old idea (−i), justify (−i), question (−i), command (−i).

^aSeparate analyses for groups with each solution score showed substantial differences between groups that did and did not solve the problem correctly, and similar results across the latter unsuccessful groups. Thus, unsolved was coded as a binary variable (0 or 1) in the turn-level analysis to facilitate interpretation of the results.

* p < .05.

** p < .01.

* p < .05.

** p < .01.

*** p < .001.