Benefits, Definitions, and Underpinnings

Crouch CH, Mazur E (2001). Peer instruction: ten years of experience and results. Am J Phys 69, 970–977. In this paper, student performance data on clicker questions after PI, pre-post concept assessments, and exams, was collected in multiple courses over several years. The authors compare students’ performance in PI courses to performance in traditional courses, improvements over time in PI courses, and also discuss the effects of different implementations of PI. Concept assessment performance improves dramatically from traditional courses to those using PI, as does problem solving. Other quantitative measures, such as a repeated exam questions from a traditional course, also showed that students in the PI course outperformed the traditional students with an effect size of 0.57. Performance on ConcepTests during a single course (Fall 1997) before and after PI showed that almost half of the correct answers were arrived at after PI, with students infrequently shifting to incorrect answers. The authors also show that improvements seen in PI courses are not instructor-dependent, nor were they level-dependent (improvements seen in both algebra and calculus-based physics). The authors also discussed using reading incentives in the form of “warm up questions” to help students come to class prepared, showed evidence that motivating students is challenging, and evidence that student attitudes can be mixed in response to this teaching style.

Vickrey T, Rosploch K, Rahmanian R, Pilarz M, Stains M (2015). Research-based implementation of peer instruction: a literature review. CBE Life Sci Educ 14, 1–11. This review is based on studies that examined specific effects of PI in college-level STEM courses. In physics, several studies indicate that PI typically results in learning gains of 30-70% of students’ potential gain on concept inventories; this effect is observed across institution type and instructors. Studies in geology, genetics, and calculus also showed PI-related gains, as did studies that examined problem-solving skills and student retention in courses implementing PI. The authors also reviewed student attitudes regarding PI, finding that students are generally positive, especially regarding the immediate feedback of PI. The authors also describe the results of empirical studies identifying evidence-based practices related to each of the steps of PI. In brief, they review evidence that challenging questions lead to greatest improvement; that individual thinking and commitment to an answer improves students’ learning experiences; that clickers and low-tech voting tools can effectively be used with PI; that peer discussion is a key element of PI but that improvements in student responses may overestimate student understanding; that instructor explanation of the purpose of peer discussion as well as answers to individual questions are key; that awarding participation points rather than points for correct answers leads to better peer discussions. They also review other implementation choices that need further research for clarification and highlight the need for research on the relationship between learning gains and individual student characteristics. Finally, the authors offer a flow chart describing evidence-based implementation of PI that can be a useful tool for instructors.

Smith MK, Wood WB, Adams WK, Wieman C, Knight JK, Guild N, Su TT (2009). Why peer discussion improves student performance on in-class concept questions. Science 323, 122–124. In PI, students answer conceptual multiple choice questions individually, discuss the questions with their neighbors, and then revote before the instructor explains the correct answer. Typically, the number of students providing a correct response increases after the peer discussion. These researchers investigated whether the peer discussion promotes understanding or whether students are persuaded to vote for correct answers by peers using sixteen paired clicker questions in a high enrollment genetics class. Students voted on the first question individually, discussed it with peers and revoted, and then voted on a second, similar question before instructor explanation. As expected, the percent of students answering the first question correctly increased after discussion. Importantly, the percent of students answering the second, similar question correctly was significantly higher than for Q1, indicating that students’ conceptual understanding increased. Peer discussion improved student understanding for questions of different difficulty levels, but the greatest benefit was observed for the most difficult questions. Statistical analysis suggested that the improvement extended to groups in which none of the students understood the concept when answering the first question. Further, students who answered Q1 incorrectly both before and after discussion demonstrated a better-than-chance probability of answering Q2 correctly, indicating that peer discussion had a delayed, unexpected benefit. Thus peer discussion is an essential element for deriving benefit from clicker questions.

Smith MK, Wood WB, Krauter K, Knight JK (2011). Combining peer discussion with instructor explanation increases student learning from in-class concept questions. CBE Life Sci Educ 10, 55–63. The authors asked whether peer discussion, instructor explanation, or a combination of both led to more conceptual change in genetics courses. Student responses to paired, isomorphic clicker questions were used to compare the conditions. For each question pair, students voted on the first question individually. In the first condition, students then discussed the question with peers and revoted prior to learning the correct answer with no explanation. In the second condition, students volunteered reasons for their answers and heard the instructor’s explanation. In the third condition, students engaged in peer discussion, revoted, volunteered reasons for their answers and heard the instructor’s explanation. In all three conditions, students then voted individually on a second, isomorphic question (Q2). Both peer discussion and instructor explanation significantly improved student performance on Q2, and the combination of peer discussion and instructor explanation produced larger learning gains than either alone. When the authors compared the effects of the treatments for weak, medium, and strong students, they found that strong students derived the greatest benefit from peer discussion while weak, nonmajor students benefited most from instructor explanation.

Cortright RN, Collins HL, DiCarlo SE (2005). Peer instruction enhanced meaningful learning: ability to solve novel problems. Adv Physiol Educ 29, 107–111. The authors used a crossover design in a small physiology course to compare performance on in-class quiz questions with and without PI. Two kinds of questions were given to students: content mastery questions, and novel problems, which required students to extend their learning to a new context. Students self-sorted into permanent groups of 3-4, and were assigned to a side of the room. On one side, students were allowed to discuss their answers to in-class questions before answering, and on the other side, students answered individually. After the first exam, the conditions were switched. In the final third of the course, students were presented with a single novel question in each class period, and switched from day to day whether they answered with or without peer discussion. When students answered questions with peer discussion, they averaged 59% correct, significantly higher than 44% correct when answering without peer discussion. In addition, the novel problem-solving task also yielded a significant difference, with students averaging 47% correct with peer discussion vs. 24% without. Students also completed a questionnaire about the peer instruction process, reporting that they understood the educational goals of PI, as well as the nature and value of the activities, and that it facilitated their learning of the material.

Brooks BJ, Koretsky MD (2011). The influence of group discussion on students’ responses and confidence during peer instruction. J Chem Educ 88, 1477–1484. This study related written student explanations for answers before and after group discussion to student performance and confidence. Additionally, the authors studied whether the display of the voting histogram for individual answers affected the answer choices of students after discussion. Two cohorts of students in a chemical thermodynamics course answered the same 5 question pairs, each on a different topic, using the typical PI cycle. After each vote, students were given time to write an explanation of their answer choice, and reported their confidence using a Likert scale of 1-5. In both cohorts, significantly more students changed from incorrect to correct than from correct to incorrect when the most common answer was correct, but not when the most common answer was incorrect. More students also changed their answer to match the consensus answer than from the consensus to another answer. This held true regardless of whether the consensus answer was correct, and whether or not the histogram of votes was shown. In looking at student explanations of their answers, explanation scores increased from individual to after-discussion for students who answered correctly both times, and for those who went from incorrect to correct. No significant correlation was found between confidence ratings and either correctness or consensuality. The authors conclude that group discussion during peer instruction helps students construct deeper explanations in all circumstances. Students who choose the correct but not consensus answer after discussion are less confident in their response than other students, but had the best explanations. The authors point out that instructor-led discussion may be critical for situations in which most people answer the question incorrectly.

Johnson DW, Johnson RT, Smith KA (2014). Cooperative learning: improving university instruction by basing practice on validated theory. J Excellence Coll Teach 25, 85–118. This review describes cooperative learning, its basis in social interdependence theory, and the conditions that produce cooperation: positive interdependence, individual accountability, promotive interaction, social skills, and group processing. The authors provide a meta-analysis of university studies comparing efficacy of cooperative, competitive, and individualistic learning, reporting moderate to large benefits of cooperative learning for student achievement and other measures. The authors also provide information about implementing this pedagogy, including descriptions and suggestions for formal and informal cooperative learning. They define formal cooperative learning as students working together toward common learning goals on common assignments, where the learning goals, assignments, needed skills, and size and structure of the group are defined by the instructor, and where the interaction lasts from a single class period to multiple weeks. They define informal cooperative learning as ad-hoc student groups working together toward a common learning goal for a shorter period (a few minutes to a class period). The authors conclude by describing ways in which cooperative learning can serve as the basis for other forms of active learning.

Springer L, Stanne ME, Donovan SS (1999). Effects of small-group learning in science, mathematics, engineering, and technology: a meta-analysis. Rev Educ Res 69, 21–51. This article describes three theoretical perspectives on the benefits group work can provide: a motivational perspective, an affective perspective, and a cognitive perspective, each of which can provide a rationale for instructional decisions surrounding group work. The authors performed a meta-analysis of 39 reports of small-group learning in postsecondary STEM courses from 1980 to c. 1999, finding a significant positive main effect of small-group learning on achievement, persistence, and attitudes. They examined the effects of small-group learning based on gender and racial or ethnic composition of the group, finding little evidence that gender composition altered outcomes but observing greater benefits for groups composed primarily or exclusively of African-American or Latinx students.

Chi MTH, Bassok M, Lewis MW, Reimann P, Glaser R (1989). Self-explanations: how students study and use examples in learning to solve problems. Cogn Sci 13, 145–182. This comprehensive article examines how students learn via self-explanation, and suggests that student problem-solving is due to their ability to understand and learn from text and from worked-examples. Students solved problems on the application of Newton’s laws of motion and talked about their process as they worked. Two groups (4 students per group) were defined after the study as “good” and “poor” problem solvers based on their problem-solving success. The article details student processes in solving isomorphic problems and transfer problems, and describes the kinds of explanations that students construct. For student’s explanations, the authors found that good problem-solvers generated a significantly higher number of lines of explanation per problem than poor problem-solvers. Good problem-solvers also spent more time on each example. In describing the importance of “explaining”, the authors found that good problem-solvers generated twice as many explanations as poor, and that these explanations included more physics content. In fact, the number of physics explanations a student gave was highly correlated with their subsequent success in solving an isomorphic problem. Good problem-solvers also invoked additional components of physics in generating self-explanations, while poor did not. The authors argue that these types of behaviors enable students to learn more deeply and effectively, suggesting that explanatory practices can benefit student learning.

Chi MTH, de Leeuw N, Chiu M-H, Lavancher C (1994). Eliciting self-explanations improves understanding. Cogn Sci 18, 439–477. Eighth grade students were prompted to either explain what they had read (self-explanation), or to read a passage on human circulation a second time, and were tested on their understanding with a set of pre- and post-test questions. All students made gains from the pre- to post-test, but the self-explanations group made significantly higher gains on all 4 types of pre-post questions, and particularly higher gains on questions requiring knowledge inference and synthesis. The authors characterized the nature of student explanations by coding the number inferences and their correctness. Students who used a high number of inferences had significantly more correct explanations, and higher (though non- significant) learning gains, than those who used a low number of inferences. Only those with high inference spontaneously drew diagrams to assist in answering the questions. Student incoming ability (as measured by the California Achievement Test) was not a predictor of learning gains or of self-explanations. The authors present additional information on how students integrate information, and how they make models as they engage attempt to understand the text. The authors conclude that generating explanations while attempting to learn helps students integrate new information with old knowledge, and improves overall learning.

Osborne J (2010). Arguing to learn in science: the role of collaborative, critical discourse. Science 328, 463–466. This review discusses the use of argumentation in developing reasoning skills in science students. Osborne presents research on how argumentation is not typically taught, and does not come naturally, how students with different viewpoints are more likely to engage in argumentation that is meaningful, and how students encouraged to argue construct deeper knowledge and use more biological terms and ideas in their discussions. In addition, he discusses findings supporting interactivity over individual constructivist activities, and the need to explicitly teach strategies for arguing using reasoning in order for students to make gains in these skills.

McDonnell L, Mullally M (2016). Research and teaching: teaching students how to check their work while solving problems in genetics. J Coll Sci Teach 46, 68–75. Students in this study were enrolled in a 2nd year genetics course at University of British Columbia, required for all biology majors. The authors used think aloud interviews with students and experts to develop a rubric on work checking as a measure of metacognition. Sixty students then participated in a “work checking intervention” after taking a quiz before the lesson, and another quiz two days later. The work checking intervention included explicit instruction, practice, feedback, and assessment. Only 35% of the students checked their work prior to the intervention (prequiz), while 98% checked their work on the postquiz. Most students (65%), checked their work superficially, but were still more likely to achieve a higher performance score than those who did not; those who checked thoroughly received the highest performance scores. When students received feedback that they had not answered a question correctly, only one quarter of students were able to raise their score upon additional reflection. Having students repeatedly reflect on their work and make decisions that are based on that reflection may result in more thorough monitoring, and potentially an improvement in their problem-solving skills.