Commentary
Learning under anesthesia: Checking the light in the fridge? Commentary on Deeprose and Andrade (2006)

https://doi.org/10.1016/j.concog.2005.11.001Get rights and content

Abstract

Research on learning under anesthesia has focused on showing that learning is possible in the absence of awareness. However, a simple dissociation between learning and awareness is conclusive only under strong additional assumptions, and the actual state of consciousness of an anesthetized person is difficult to determine. Instead of trying to establish complete unconsciousness, one might employ gradual anesthesia to experimentally vary the level of consciousness in a controlled fashion, checking whether cognitive processes exist that can change in opposite direction to measures of awareness.

Introduction

Anesthesia is of special interest because it complements traditional research in an interesting way: in standard paradigms on unconscious perception, one commonly investigates awareness by presenting degraded stimuli to an unchanged cognitive system; under anesthesia, one presents an altered cognitive system with intact stimuli. Still, the employment of anesthesia for demonstrating unconscious cognition poses the same fundamental problems as traditional paradigms. To demonstrate that a critical stimulus was processed unconsciously, one has to produce some dissociation between a direct measure (D) of awareness, and an indirect measure (I) of processing per se (Reingold and Merikle, 1988, Schmidt and Vorberg, in press, Shanks and St. John, 1994). For example, I might be any behavioral effect signaling that the critical stimulus has been processed (like a nonzero priming or learning effect), while D might be a behavioral or physiological measure that is supposed to signal the observer’s awareness of the critical stimulus. In this comment, I will focus on how variations in consciousness during anesthesia may be used to study dissociations between D and I.

The traditional criterion for unconscious perception has required D to equal zero, indicating the complete absence of any conscious processing. At the same time, it required I to be nonzero, indicating that the critical stimulus was nevertheless processed. Historically, this zero-awareness criterion has run into difficulties (Holender, 1986). The fundamental problem, it turned out, is that the zero-awareness criterion only works if we can draw a valid conclusion from zero performance in the direct measure to zero awareness in the observer (Reingold & Merikle, 1988).

Recently, Dirk Vorberg and I (Schmidt & Vorberg, in press) have critically examined the scopes and assumptions required by the zero-awareness criterion as well as alternative approaches. In our analysis, we assume that direct as well as indirect measures may depend on conscious as well as unconscious information about the critical stimulus. The dependency is supposed to be weakly monotonic, i.e., in the long run none of the measures will decrease in response to increasing conscious or unconscious information—a weak assumption that must be conceded for virtually any measurement situation. We make no assumption about the relationship of conscious and unconscious information, which are free to interact in arbitrary ways, as in mutual inhibition (Snodgrass, Bernat, & Shevrin, 2004). Establishing unconscious cognition then consists in refuting a Null Model which states that the influence of unconscious information is zero, so that both measures are driven by a single source of conscious information.

The traditional way of doing this is the zero-awareness criterion, which produces a simple dissociation of direct and indirect measures: zero D in the presence of nonzero I. In laboratory research, this type of dissociation is notoriously difficult to establish. Moreover, even if D is numerically zero, a simple dissociation can be regarded as evidence for unconscious processing only under an additional assumption: D must be an exhaustive, i.e., strongly instead of weakly monotonic measure of conscious information (Reingold and Merikle, 1988, Schmidt and Vorberg, in press, for a more general proof). This means that no change in awareness, however small, must escape detection by D; only then can we infer the absence of awareness from zero values of the direct measure. The exhaustiveness assumption is a strong requirement that is likely to be violated, and we cannot check whether a given measure conforms to this assumption, or whether such a measure even exists. If the exhaustiveness assumption is wrong, it can be argued that it was conscious processing alone that influenced both D and I, but that I was sensitive enough to detect it while D was not. Evidence for unconscious perception gathered from simple dissociations has left many researchers unconvinced, some of them maintaining that unconscious perception does not even exist (Holender, 1986, Holender and Duscherer, 2004).

It seems to me that such problems are aggravated rather than solved when unawareness is established by surgical anesthesia. Even though there is now convincing evidence that memory formation can take place under anesthesia (see the metaanalysis by Merikle & Daneman, 1996), there remains the obvious problem with the direct measure, i.e., how to detect whether the patient is conscious or not. Traditional clinical criteria deal with physiological signs such as pulse and respiration patterns, or the loss of certain reflexes, that do not address consciousness directly. Reactions to external stimuli are generally taken as indicating that the patient is aware of the stimuli and that narcosis should be deepened, but many reflexes and stimulus-triggered responses are independent of awareness and volition. EEG-derived measures like evoked potentials or the bispectral index (BIS) suffer from similar problems of interpretation: After all, BIS is regarded as a measure of consciousness only because it correlates with more direct behavioral measures, which may themselves be imperfectly correlated with awareness. (And even so, the range of BIS values deemed admissible during surgery seems arbitrary.)

Conversely, and more troublesome for surgical patients, there may be awareness despite the absence of external indicators. Suppose that we successfully demonstrate learning without any sign that the patient was conscious during anesthesia. This is a case of a simple dissociation, which requires a direct measure exhaustive for conscious information. We have seen that this assumption is problematic even if awareness is measured in awake, motorically unimpaired individuals. But under surgical anesthesia, the patients’ capacity to signal their awareness is likely to be compromised, because anesthesia serves additional functions of analgesia and paralytic muscle relaxation (Bonebakker, Jelicic, Passchier, & Bonke, 1996). Even if one arm is exempted from anesthesia, as in the “isolated forearm technique,” it is quite possible that anesthesia also affects processes of motor planning and cognitive control that are necessary for carrying out the intended movement. To me, monitoring a person disabled in such a way for signs of awareness seems similar to the notorious problem of checking whether the light in the fridge is really out when the door is shut.

An interesting alternative is to let awareness vary over experimental conditions. It may then be possible to establish a double dissociation (Schmidt & Vorberg, in press), which consists in finding an experimental manipulation that changes D and I in opposite directions. In particular, any pair of experimental conditions that leads to opposite ordering of direct and indirect measures gives evidence of a double dissociation, even when the conditions differ in more than one independent variable. It can be shown that double dissociations work under much milder assumptions than simple dissociations; most importantly, they do not require the restrictive exhaustiveness assumption. Examples of double dissociations come from response priming experiments where participants perform speeded keypress responses to the direction of an arrow-shaped mask that is preceded by an arrow-shaped prime. As the prime-mask interval increases, priming effects also increase, such that primes pointing into the same direction as the mask shorten response times, while primes pointing into the opposite direction prolong them. At the same time, prime identification performance can decline with prime-mask interval under suitable stimulus conditions, giving rise to opposite time-courses of the prime’s motor effects and its accessibility for conscious report (Mattler, 2003, Vorberg et al., 2003, Merikle and Joordens, 1997a, Merikle and Joordens, 1997b).

Double dissociations refute any explanation in terms of only a single type of information processing, i.e., any Null Model claiming that both direct and indirect measures are monotonically controlled by a single source of conscious information (Schmidt & Vorberg, in press). A surprising feature of double dissociations is that in order to avoid the exhaustiveness assumption, they require D to be nonconstant. This implies that double dissociations cannot be obtained in the complete absence of awareness; they require variation of awareness over a range of experimental conditions.

Instead of trying to establish complete unconsciousness, as in surgical anesthesia, one might therefore employ gradual anesthesia to experimentally vary the level of consciousness in a controlled fashion in volunteering participants. There have indeed been studies of this type investigating positive correlations between awareness and memory (Andrade, 1996, Chortkoff et al., 1993, Dwyer et al., 1992), but to my knowledge nobody has attempted checking whether some cognitive processes can change in opposite direction to measures of awareness. For learning effects reasonably independent of the participant’s state of awareness, demonstrating double dissociations under anesthesia might well be possible. For example, variables like the exposure time to the learning material, the way how learning is distributed over time, stimulus familiarity, and many other stimulus manipulations, might be quite independent of awareness if predominantly reflecting automatic, involuntary processes. If such variables are employed in tandem with different doses of anesthetic (e.g., by pairing increasing levels of stimulus familiarity with increasing doses of anesthetic), they might conceivably be able to enhance learning even under conditions of receding awareness, refuting a Null Model that claims that awareness and learning are both monotonically driven by a single source of (conscious) information. Given such double dissociations, it would no longer be necessary to rely on complete unconsciousness in the patient. Instead of trying to check the light in the fridge with the door shut, we could see what the light is doing as the door is slowly closing.

References (15)

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

Cited by (1)

Commentary on Deeprose, C., Andrade, J. (2006). Is priming during anesthesia unconscious? Consciousness and Cognition, 15, 1–23. Part of this work was supported by the German Science Foundation (Grant Schm1671/1-1 to Thomas Schmidt). Many thanks to Dirk Vorberg, who co-authored the manuscript on which this commentary is based.

View full text