In many cognitive psychology experiments, researchers ask their experimental subjects to make retrospective confidence ratings, judgments that estimate the likelihood that a response provided earlier was correct. Despite the ubiquity of these ratings in the laboratory, there is little scientific consensus on the ideal way to implement confidence ratings in research. This case report describes different ways that cognitive psychologists can collect (and have collected) confidence ratings in their experiments. It illustrates a number of theoretical considerations relevant to recording and analyzing confidence data and then describes the results of two recent experiments (works of Roediger and DeSoto in 2013) to show how the methods used to collect confidence may affect the data that are obtained. The implication is that collecting confidence data in a research project is more complicated than it appears.