Midazolam amnesia and short-term/working memory processes

https://doi.org/10.1016/j.concog.2005.03.004Get rights and content

Abstract

We examined whether midazolam impairs short-term/working memory processes. We hypothesize that prior dissociations in midazolam’s effects on short-term/working memory tasks and episodic memory tasks arise because midazolam has a larger effect on episodic memory processes than on short-term/working memory processes. To examine these issues, .03 mg/kg of participant’s bodyweight of midazolam was administered in a double-blind placebo-controlled within-participant design. Performance on the digit span and category generation/recall tasks was examined. The results of Experiment 1 demonstrated that: (1) midazolam impaired performance on the digit span task; (2) midazolam did not impair performance on the category generation task; (3) midazolam impaired performance on the category recall task; and (4) midazolam’s effect on category recall was four times as large as its effect on digit span. The results of Experiment 2 demonstrated that midazolam did not impair digit span performance when the digit span task was administered at a later time. These results suggest that midazolam can impair short-term/working memory processes, but these effects are substantially smaller than midazolam’s effect on episodic memory processes. Moreover, they demonstrate that conscious awareness of materials during study is not sufficient to produce episodic memory.

Introduction

Benzodiazepines are well known for producing anterograde amnesia (Curran, 2000, Mewaldt et al., 1983). The benzodiazepine, midazolam, produces dense anterograde amnesia (Polster, McCarthy, O’Sullivan, Gray, & Park, 1993). This amnesia appears to reflect midazolam’s specific effect on encoding processes in episodic memory. To demonstrate this, Polster et al. (1993) administered midazolam prior to study and/or test on a recognition memory task—a standard test of episodic memory. Polster et al. found that midazolam only produced amnesia when injected prior to study. Moreover, there was no evidence that injecting midazolam prior to test reduced the amnesia produced by injecting midazolam prior to study. These results suggest that midazolam amnesia does not reflect impaired retrieval from long-term memory and that it is not a state-dependent effect. In the same vein, Veselis, Reinsel, Feschenko, and Wronski (1997) demonstrated that midazolam amnesia can occur independently of sedation, suggesting that midazolam amnesia is not due solely to general effects on arousal level or motivation. (See Mewaldt et al. (1983) for similar results with other benzodiazepines.)

In considering theoretical explanations of midazolam amnesia, an important question concerns whether administering midazolam impairs short-term/working memory processes (Miyake, 2001). Classical theories of episodic memory (e.g., Raaijmakers & Shiffrin, 1981) assume that encoding in episodic memory relies heavily on short-term/working memory processes. Thus, midazolam’s effect on encoding processes in episodic memory may arise because it impairs short-term/working memory processes.

Determining whether midazolam impairs short-term/working memory processes also provides evidence on the role of consciousness in episodic memory encoding. While it is clear that conscious experience of materials during study contributes to episodic memory performance (Jacoby, 1998), this does not imply that consciousness of materials is sufficient to produce episodic memory. If midazolam can impair episodic memory performance without impairing short-term/working memory processes, this would suggest that consciousness of materials during study is not sufficient to produce episodic memory.

At sufficiently high doses, benzodiazepines produce extreme sedation and unconsciousness (Andrade, Sapsford, Jeevarathum, Pickworth, & Jones, 1996). Under these conditions, short-term/working memory processes are necessarily impaired because participants do not attend to presented information. We focus here on the more interesting case in which the dose of midazolam is set so that the associated sedation is minimal. In this case, participants are responsive and can perform a variety of tasks in the midazolam condition, even though they demonstrate dense anterograde amnesia. For example, Hirshman, Fisher, Henthorn, Arndt, and Passannante (2003) demonstrated that participants administered a .03 mg/kg of bodyweight dose of midazolam demonstrated no impairment in a category generation task, even though their later recall of these category items was substantially impaired.

Current evidence on whether midazolam affects short-term/working memory processes is ambiguous. There is substantial prior evidence (Ghoneim and Mewaldt, 1975, Knopman, 1991) that benzodiazepines such as diazepam and lorazepam can impair performance in episodic memory tasks without impairing performance on short-term/working memory tasks. Consistent with these results, Hennessy, Kirkby, and Montgomery (1991) have demonstrated that midazolam can impair performance on tasks indexing episodic memory processes (e.g., delayed recall of verbal and visual stimuli) without impairing performance on tasks indexing short-term memory/working memory processes (e.g., immediate recall). Similarly, demonstrations that benzodiazepines affect the primacy, but not the recency, portion of the serial position curve (e.g., Mewaldt et al., 1983) are broadly consistent with the view that midazolam does not affect short-term/working memory processes.

Conversely, Rammsayer, Rodewald, and Groh (2000) have demonstrated that administering 11 mg of midazolam prior to study impairs performance on an immediate word recall test that is generally assumed to index short-term/working memory processes. This dose, however, is likely to have produced extreme sedation, impairing short-term/working memory by rendering participants unconscious.

Results from positron emission tomography (PET) studies (Veselis et al., 1997) have demonstrated that midazolam decreases regional cerebral blood flow in prefrontal cortex, an area repeatedly implicated in short-term/working memory processes (e.g., Tsukiura et al., 2001). However, this result is not conclusive in that areas of prefrontal cortex are also known to mediate performance in episodic memory tasks (e.g., Cabeza, Locantore, & Anderson, 2003). Thus, Veselis et al.’s results may reflect midazolam’s effects on regions of prefrontal cortex that mediate episodic memory processes exclusively.

Two contrasting interpretations of the preceding results are viable. One interpretation is that, excluding cases in which sedation is extreme, midazolam produces effects that are selective to episodic memory processes. A second interpretation is that midazolam can produce effects on short-term/working memory processes, as well as episodic memory processes, but the latter effects are substantially larger than the former effects. In this interpretation, the dissociation demonstrated by Hennessy et al. reflects midazolam’s larger effect on episodic memory processes. The failure to demonstrate an effect on immediate recall in their study reflects limitations in the experimental power to detect midazolam’s smaller effect on short-term/working memory processes.

The second interpretation predicts that, given an appropriately sensitive experiment, midazolam should impair performance on tasks indexing short-term/working memory processes. Further, midazolam’s effect on tasks indexing short-term/working memory processes should be smaller than its effect on tasks indexing episodic memory processes.

Experiment 1 explores these issues by examining the effect of midazolam on the digit span task (Conklin, Curtis, Katsanis, & Iacono, 2000) and the category generation/recall task (Hirshman et al., 2003). We used the digit span task because it is considered a prototypical index of short-term/working memory processes, with forward digit span assumed to index passive short-term memory processes and backward digit span assumed to index working, as well as short-term, memory processes (Conklin et al., 2000). We used the category generation/recall task because of the prior demonstration (Hirshman et al., 2003) that midazolam impairs participants’ category recall without affecting category generation in this paradigm.

If midazolam impairs digit span, we can compute ή2 (Hays, 1981), a standard measure of effect size reflecting the variance accounted for by the experimental variable relative to the overall variance of the data, to compare midazolam’s effects on digit span and category recall. This permits comparison of the size of midazolam’s effect on digit span and category recall even though these measures may have different variances.

Section snippets

Experiment 1

Experiment 1 examined the effect of an intravenous injection of .03 mg/kg of bodyweight of midazolam on performance on the digit span and category recall tasks. This dose matched those used in our previous studies (Hirshman et al., 2001, Hirshman et al., 2003) that demonstrated effects of midazolam on episodic memory tasks.

Methods

The methods of Experiment 2 replicated those of the digit span task in Experiment 1 with two exceptions. First, there were 36 participants in the experiment. Twelve participants were from The University of North Carolina at Chapel Hill community and 24 participants were from the Denver community. Second, the digit span task began 21 min after the injection of midazolam or saline.

Results and discussion

The mean number of correct trials as a function of Type of Drug (midazolam vs. saline) and Type of Task (forward digit

General discussion

Our experiments present two results. First, midazolam can impair performance on the digit span task hypothesized to index short-term/working memory processes. Second, midazolam’s effect on the digit span task is smaller than its effect on a category recall task hypothesized to index episodic memory processes. Together, these results are consistent with the hypothesis that midazolam can impair short-term/working memory processes even though these effects may not be as large as midazolam’s

Acknowledgments

The work presented here was supported by a grant from the National Science Foundation to the second author. The current work was conducted in compliance with the procedures of the Human Subjects review boards of the University of Colorado and the University of North Carolina.

References (21)

There are more references available in the full text version of this article.

Cited by (20)

  • Using arterial spin labeling perfusion MRI to explore how midazolam produces anterograde amnesia

    2012, Neuroscience Letters
    Citation Excerpt :

    GABA is the primary inhibitory neurotransmitter in the mammalian central nervous system and the GABAA receptors are expressed in cerebral cortex, hippocampus, basal ganglia, thalamus, cerebellum, and brainstem [31]. Midazolam, like benzodiazepines in general, promotes transient anterograde amnesia [2,6,11,15] and as such, provides a promising tool for studying human memory that finesses problems inherent in patient populations. Some of our previous work has suggested that midazolam-induced memory effects occur by inhibiting the formation of new associations [14,17,18].

View all citing articles on Scopus
View full text