Do students learn by talking to each other?

by Stephanie Chasteen on May 30, 2014

Here is another re-post from an article I wrote on the i>clicker blog.


This month I’d like to highlight a study which I think is crucially important in cementing the value behind peer instruction. It’s not new work anymore, but it so elegantly answers a key question – “do students learn by talking to each other?” – that it’s worth revisiting.

Peer Instruction is a method of using clickers to help deepen students’ understanding of material by asking students a question to consider on their own, they vote, and then they talk to their neighbors. After this peer discussion, students vote a second time, and the instructor holds a whole class discussion about the results. This method was popularized by Eric Mazur, who showed that students converged to the right answer after discussing with their peers, as in the following graph of student responses, adapted from his book Peer Instruction:

Before Discussion After Discussion
Results chart before discussion Results chart after discussion

Students converge on answer “B” after discussion (which is the correct answer). But this begs the question, are they learning, or is the right answer just spreading around through discussion?

Some of my colleagues at the University of Colorado came up with an elegant way to test this. This paper was published in Science magazine in 2009 (Smith et al., Science, 323 (5910): 122, 2009;). The researchers reasoned, if students learn from discussion, then they should be able to apply that new understanding to better answer a similar question. They developed 16 pairs of questions – each pair of questions looked different to students (they had a different “cover story”) but actually tested the same idea. The authors showed these question pairs to two independent experts, who agreed that these “isomorphic questions” indeed tested the same content.

The study went like this: First, they showed the first question (Q1) to students, who thought about it and answered it on their own. Then, they asked students to turn to their neighbors to discuss the question, and they voted again. Finally, students were shown the isomorphic (Q2) question, and asked to answer it on their own. Thus, Q2 served as a ‘test’ question, to determine how much students learned from their discussions. Data was gathered from 350 students in this study. The study design is outlined below: 
StudyDesignOutlineSo what did they find? Students were better able to answer Q2 than they were able to answer Q1, showing that (most) students learned from discussion with their peers – see the graph below.

Chart results
So, 20% more students, on average, were able to answer Q2 correctly after having talked to their neighbors. Even more compelling, if we look at the students who got the first question wrong when answering on their own, 77% of those students got Q2 right (caveat; this was true only for those students who got Q1 right after discussion). So, students were able to apply new understanding, gained through discussion with peers, to answer a novel question correctly.

Another compelling part of the study shows that students can piece together the right answer even when nobody in the group originally has the right answer. It’s very interesting to look at student responses to the question by difficulty level (see graph below). You can see that, for easy and medium questions, about 10-15% more students get Q2 correct than Q1, showing modest gains in understanding. But results are rather surprising for the most difficult questions – where only about 20% of the students get the answer to Q1 right on their own, so very few students know the answer in advance. There is a proportionally huge gain for these questions, with 50% of students answering them correctly after discussion. Statistically speaking, on these difficult questions, it was impossible for every student group to have a student who had the correct answer to start with – but these groups were able to arrive at the correct answer. Students are able to construct their own knowledge through discussion.Chart2On end-of-semester surveys, students also indicated that they don’t need a student with the correct answer in order for their group to make sense of the question. One example quote: “Discussion is productive when people do not know the answers because you explore all the options and eliminate the ones you know can’t be correct.” 

So, the take-home message is:
1. Students learn through discussion with their peers, as they are able to synthesize and make sense of a 
question together
2. Peer instruction does not work primarily by transmitting the correct answer from student to student

Therefore, it’s important to include peer discussion of questions – and to use challenging questions – in your use of clickers.

In future posts, I’ll look at some more of the research behind the effective use of clickers.

– See more at:

{ 1 comment }

Ryan June 6, 2014 at 7:18 am

This can work without clickers as well. I’ve done mini-tests with students at the beginning of a set of lessons. They’re just 10 multiple-choice questions, and you can mark them quickly with a bubble sheet type program (I use GradeCam). The students do them separately first, then redo the quizzes as a group. They don’t know their marks until the first group quiz, and they’re only told which ones are wrong much later (when it’s going too long). I’ve seen pretty good results testing it this year, and I’m planning on doing it much more next year.

Comments on this entry are closed.

Previous post:

Next post: