To the Editor:—
Memory formation during anesthesia is a hotly debated topic. In carefully controlled studies of the phenomenon, patients are typically presented with a series of word stimuli for which memory is tested after recovery. Besides questions into recollection, other tests are suitable. One such test is a word stem completion task in which the first (three) letters of a word are presented, which the subject completes to a word. Completion to a word presented during anesthesia must be taken as evidence of memory if a word surfaces more often when people were presented with it compared with a group of people who were not presented with it. The task has been used in numerous studies of memory function in general and during anesthesia in particular.
In the November 2006 issue of Anesthesiology, Stonell et al.1 report an elegant use of the word stem completion test to address sex differences in memory formation during anesthesia. They used the task in combination with the process dissociation procedure (PDP), which allows a closer look of the memory formed. In particular, the PDP touches upon the question whether the memory arose unconsciously (“implicit memory”) or consciously (“explicit”). Rather than relying on different tests (e.g., a recall questionnaire and a word stem completion task), the PDP relies on a single task that is performed under two, only marginally different, instructions. The purpose of this setup is to contrast performance on the same task when the subject tries to retrieve and use information (inclusion instruction) versus memory retrieval without using the information (exclusion instruction). By contrasting the two conditions, by virtue of their similarity, estimates for implicit and explicit memory can be calculated using relatively simple mathematical equations.2 Using PDP, we reported evidence for implicit memory function during general anesthesia.3,4
The PDP relies on various assumptions, the least controversial of which is that tasks are comparable. Or, as the founding father of the procedure put it,2,5 subjects must use the same “response criterion” in the inclusion and exclusion conditions if one wants to use the calculations that render the PDP so popular. Subjects should have, or rather, be enabled to have, a similar tendency to use previously presented items in both conditions. If not, the parameters that represent the two bases of memory in the various equations do not represent the same concept and, therefore, cannot be mathematically extracted.
Some evidence for equal response criteria comes from the distractor hit rate, also referred to as the “base rate,” which represents the probability of responding with a study word without being previously presented with it. Some subjects, for example, will complete the stem COU__ to couch simply because it works, regardless of whether they heard the word during anesthesia. Base rates, in other words, establish chance performance and tell us something about response tendencies. By the same token, differences between base rates in the inclusion and exclusion condition indicate that subjects used different response criteria (which violates the PDP).
Stonell et al.1 observed significantly different base rates (Kate Leslie, M.D., written communication, November 2006) but nonetheless calculated PDP estimates based on which the authors suggested that both implicit and explicit memory function contributed significantly to the memory effect observed. This conclusion is hard to reconcile with the PDP model and clinical findings so far, where either implicit or explicit uses of memory usually account for observed effects: Hit rates are either boosted in both the inclusion and exclusion conditions (indicative of implicit memory) or they are in the inclusion condition only (indicating explicit memory). Because the rates across all conditions in the work of Stonell et al. varied substantially, it is hard to discern what was at play in this study. From the inclusion condition, it can be derived that reliable memory was formed, but we can only guess the type of learning that may be held accountable for this effect.
A likely reason for the different base rates in the study by Stonell et al.1 is the instructions given to the subjects upon stem completion testing. In our studies, we have observed comparable base rates and instruct patients in both conditions to use the word stem as an aid (cue) to recall words presented during anesthesia.3,6,–8 Memory retrieval is thus encouraged in both parts of the test. In the inclusion condition, subjects are then told to complete stems with the recalled word, whereas such words are not to be used for stem completion in the exclusion condition. In contrast, Stonell et al. instructed subjects in the exclusion condition to use words not heard during anesthesia. Although subtle, the distinction between their and our instructions is important to the PDP procedure and its calculations. Because Stonell et al. gave dissimilar instructions in the inclusion and exclusion conditions, encouraging memory in one but not in the other part of the test, the tasks are not directly comparable and the PDP assumption of equal criteria is violated. The dissimilarity may have caused subjects to complete fewer exclusion than inclusion stems and to use unusual words in the exclusion part of the test, as the authors noted in the discussion of their report, observations that are in line with the notion of different response criteria. The conclusion, therefore, that both implicit and explicit memory function are to be held accountable for the memory observed by Stonell et al. was inappropriately drawn, although the conclusion that memory was formed at Bispectral Index values between 50 to 55 is clearly correct.