In the context of an upper-level psychology course, even when students were given an opportunity to refer to text containing the answers and change their exam responses in order to improve their exam scores, their performance on these questions improved slightly or not at all. Four experiments evaluated competing explanations for the students' failure to correct their answers. Experiments 1-3 ruled out ceiling effects, cognitive bias from a previous response and item selection effects, respectively, as explanations of the result. Experiment 4 showed that no more than 41% of the students comprehended the paragraphs well enough to find the answer. Furthermore, even this 41% of the students did not put sufficient effort into finding the answer, regardless of the impact on their grade, when they were not coerced to do so. (Contains 2 tables, 1 figure and 1 note.)