Bartlett's theory that remembering is a ‘reconstructive’ process is based largely upon the ways in which subjects change and distort prose passages when reproducing them from memory. If such changes and distortions are to serve as the foundation for a theory of remembering it is clearly desirable to be quite certain that the persons who make them really are trying to remember, and are not deliberately inventing material to fill in gaps in their memories. Three experiments were carried out in an attempt to shed light on this question. In the first experiment it was ascertained that strict instructions to avoid errors markedly lessened the number of errors which subjects make in reproducing prose passages. In the second it was found that subjects are fairly adept at picking out their own errors. In the third it was shown that conscientious subjects make notably fewer errors than less conscientious ones. It is concluded that for various reasons subjects in Bartlett‐type experiments often do more or less consciously invent material to bridge gaps in their memories, and that Bartlett's claims that the occurrence and nature of such changes and distortions shed light on the mechanisms of remembering must be regarded with caution.