Peer instruction is an active-learning pedagogy in which students answer short, conceptually based questions that are interspersed during instruction. A key element is the group discussion that occurs among students between their initial and final answers. This study analyzes student responses during a modified form of peer instruction in two chemical thermodynamics classes. It seeks to understand the changes in student thinking that result from group discussion. In addition to selecting multiple-choice answers, students are asked to write a short explanation of their choice and rate their confidence in their answer. The written explanations have been coded based on the depth and accuracy of the concepts applied. Two cohorts are studied. For Cohort A, initial results of the entire class are shown to the entire class before group discussion, whereas for Cohort B, they are not. After group discussion, the majority of students select the consensus answer, whether it is right or wrong, and whether they see intermediate results. However, display of intermediate results substantially impacts student confidence. A statistically significant number of students who initially had correct answers and did not change had a higher value code assigned to their explanations after group discussion. (Contains 3 tables and 5 figures.)