To the Editor:
I read with great interest the article by Hadzidiakos et al .1in the August issue of Anesthesiology. These investigators conducted a study of memory function under anesthesia using the process dissociation procedure (PDP), a method that my colleagues and I have used in the same context in the past.2–4In contrast to our studies, Hadzidiakos et al. report no evidence of memory function in terms of word stem completion test performance, a discrepancy for which the authors provide plausible explanations such as the depth of anesthesia and midazolam premedication. However, notwithstanding their null finding, one of the PDP models—the original—produced parameters suggesting the presence of controlled (explicit) and automatic (implicit) memory processes. By extending the model to include guessing parameters, the authors go on to show that the original model produces faulty estimates and that other published results using the original model are faulty. That is, Hadzidiakos et al. find no evidence of any memory processes in three of the four inspected studies when the extended measurement model is applied. They conclude that in these studies there was no contribution (i.e. , evidence) of memory at all and that past findings are spurious.
I take issue with this conclusion for several reasons. Foremost, a model that generates discrepant parameters depending on its assumptions or underlying structure should not invalidate the behavioral findings it attempts to model. When significant differences are found in patients' postoperative behavioral responses to old material presented under anesthesia versus new material not presented before, this difference is real and evidences memory for old material regardless of how the underlying process is labeled. Dismissing these behavioral observations ignores an overwhelming body of evidence in favor of implicit memory (“priming”) phenomena in the cognitive psychology and neurology literature and surely cannot have been the intent of Hadzidiakos et al . Although their critique of the original PDP model may be warranted in that the modeling heavily depends on (controversial) assumptions and proper test instructions,5the evidence for or against memory function under anesthesia is based on actual response data and not on models. In many of the anesthesia studies cited by Hadzidiakos et al. , and many more, response tendencies demonstrated memory for material presented under anesthesia, and the quest for understanding this phenomenon continues.6Therefore, it would be wrong to imply or believe that memory function under anesthesia is a spurious phenomenon. Second, the authors failed to include studies that used the extended PDP model and found evidence of automatic memory processes.2,7Although one study may not have properly implemented the PDP methodology and produced skewed estimates as a result,5another found robust evidence of implicit memory function under seemingly adequate levels of anesthesia based on patient response data and Bucher's PDP model.2It is not clear why this evidence was disregarded.
I commend the authors on undertaking their study and welcome their critical examination of a popular yet tricky methodological approach but regret their simplified argument and failure to distinguish between modeled and actual reality.
Emory University School of Medicine, Atlanta, Georgia. email@example.com