Home Curriculum Vitae Publications Conference Reports Forthcoming Extramural Colloquia Expert Testimony Teaching Healthcare The Human Ecology of Memory Research Archive Publications and Reports Rants


Consciousness Requires More than Attention


John F. Kihlstrom

Yale University


Link to Polish translation by Valeria Aleksandrova.

The cognitive revolution made the study of consciousness respectable again, if only in the form of studies of attention and imagery. The legitimation of consciousness was not inevitable, however: one of the dirty secrets of cognitive psychology is that the study of cognition can get along perfectly well without any interest in consciousness at all. The philosopher Owen Flanagan (1991) has pointed out four reasons for this state of affairs:

Positivistic Reserve: Flanagan points out, first, that cognitive psychology inherited some of the methodological assumptions of the behaviorism which it replaced, in particular an emphasis on publicly observable behavior as the window onto the mind. Because consciousness is inherently private, and too metaphysical, it still seems somehow beyond the pale of a science of the mind.

Piecemeal Approach: Furthermore, even among those cognitive psychologists who understand that they are studying consciousness, there is a tacit assumption that an understanding of consciousness will emerge, in a bottom-up fashion, from studies of individual phenomena such as attention or imagery. Many cognitive psychologists have made their careers by studying the phenomena of consciousness, such as attention, episodic memory, and imagery, without ever referring to consciousness itself. The effect has been to marginalize consciousness, as a topic which, like pornography, is too embarrassing to discuss in polite company even if we might admit privately that it=s what we=re really interested in.

Conscious Inessentialism: To make things even worse, the doctrine of computational functionalism, which underlies so much contemporary modeling of cognitive processes and systems, assumes that we can produce a perfectly adequate description of human information processing solely in terms of the functional relations between stimulus inputs and response outputs. After all the effort to get past radical behaviorism, this throwback to the connectionism of Thorndike and the radical behaviorism of Skinner renders consciousness, once again, inessential to the study of the mind.

Epiphenomalist Suspicion: Finally, many of those computational functionalists who, however grudgingly, admit that consciousness is part of the human experience, nonetheless argue that consciousness is the endproduct of cognitive functioning, and plays no causal role in human experience, thought, and action. Over a hundred years ago, Huxley asserted that we were Aconscious automata@; and many modern psychologists seem to agree. For example, connectionist analyses of cognition state or imply that conscious awareness is the last thing that happens, after the network has settled into a steady state -- that is, after all of the interesting and important work is done. The thrust of this argument is that while we humans may happen to be conscious, nothing much hangs on this fact, and things wouldn=t be any different if we weren=t conscious at all.

All this sounds pretty bad, if you=re interested in consciousness, as indeed it is, but the upside is that conscious inessentialism and the epiphenomenalist suspicion, taken together, provide cognitive psychologists with ample motivation for exploring the psychological unconscious -- that is, the idea that conscious experience, thought, and action is influenced by percepts, memories, and other mental states which are inaccessible to phenomenal awareness and independent of voluntary control. So it=s one of the ironies of contemporary cognitive psychology that many of those who might have made a science of consciousness have instead gravitated, knowingly or not, toward a science of the mind which gives precedence to unconscious mental processes.

So far as modern psychology is concerned, the psychological unconscious began life as a kind of wastebasket: the repository for unattended inputs, or memories which were rendered unavailable by decay or displacement. Later, acceptance of the distinction between automatic and effortful processes led to the idea that unconscious mental processes were executed automatically, without drawing on attentional resources. The upshot has been the identification of the unconscious with the unattended, and the rise of the notion that unconscious processing is limited to perceptual and other lower-level, presemantic analyses. Hence, the notion that implicit memory is mediated by a perceptual representation system, and the idea that semantic processing is not possible without conscious identification of the stimulus.

The situation is exemplified by the following syllogism: Unconscious processes are held to be automatic, and automatic processes are defined as those that don=t consume attentional capacity; therefore, unconscious processes don=t consume attentional capacity. I want to argue that this conventional wisdom on the psychological unconscious is mistaken -- that automatic, unattended, and presemantic processing do not exhaust the psychological unconscious.

My first example comes from posthypnotic amnesia -- the difficulty that some subjects have, after the termination of hypnosis, in remembering the events and experiences which transpired while they were hypnotized. Consider, for example, an experiment from my laboratory in which subjects memorized a list of 15 unrelated words to a criterion of two perfect repetitions (Kihlstrom, 1980, Experiment 1). Then they received a suggestion for posthypnotic amnesia, worded as follows:

When you awaken from hypnosis you will not be able to remember the words that you learned, or that you learned any words, until I say to you, Anow you can remember everything@. You will not remember anything until then.

When these subjects were awakened and tested by the method of free recall, they remembered virtually none of the list items, while control subjects remembered them perfectly well. Later, however, they completed a free-association test in which half the cues targeted items from the study list as the most frequent normative response, while the other half targeted matched controls which had not been studied. The amnesic subjects generated lots of target items on this test. Moreover, comparing associations to critical and neutral targets, the amnesic subjects showed a clear priming effect -- a priming effect that was equivalent to that shown by control subjects who were nonamnesic. Following the free-association test, the subjects were given a second test of free recall, which yielded only slightly better results than the first one. Apparently, generating significant numbers of list items as free associates generally did not remind them of any of the words they had memorized earlier. However, when administered the pre-arranged reversibility cue, ANow you can remember everything!@, intended to cancel the amnesia suggestion, free recall returned to baseline levels.

In a conceptual replication, another group of subjects showed a dense posthypnotic amnesia for a list of categorized words, but showed significant priming when asked, while amnesic, to generate instances of target and control categories (Kihlstrom, 1980, Experiment 2). Last year at these meetings, Jennifer Dorfman and I reported on a further study of posthypnotic amnesia in which the tests of explicit and implicit memory were matched for their cue value (Dorfman & Kihlstrom, 1994). Again, we obtained a nice priming effect on the free-association task, despite a gross impairment in cued recall.

Preserved priming on free-association and category-generation tasks, in the face of impaired recall, is a familiar dissociation between explicit and implicit memory. But the case of posthypnotic amnesia is different, in at least two respects. First, the priming which is preserved is semantic priming, and relies on the formation during encoding, and preservation at retrieval, of a semantic link between cue and target. This priming reflects deep, semantic processing, of a sort that cannot be mediated by a perceptual representation system. Second, in contrast to the typical explicit-implicit dissociation, the items in question were deeply processed at the time of encoding. Recall that the list was not just presented for one study trial, but rather deliberately memorized over the course of several study-test cycles to a strict criterion of learning. Moreover, the encoding activity was Adeep@ enough to produce semantic priming on the free-association test. Third, the impairment in explicit memory is reversible: posthypnotic amnesia is the only case I know where implicit memories can be restored to explicit recollection. Again, this testifies to the degree to which the items were encoded at the outset Taken together, then, these properties of priming in posthypnotic amnesia reflect the unconscious influence of semantic representations formed as a result of extensive attentional activity at the time of encoding. The priming itself may be an automatic influence, but it is not the sort that is produced by automatic processes mediated by a perceptual or otherwise presemantic representation system.

A second example is provided by a related phenomenon, posthypnotic suggestion, which occurs when a subject responds, after hypnosis has been terminated, to a suggestion which was administered while he or she was hypnotized. For example, we might suggest that when the experimenter taps his or her pencil, the subject will move from one chair to another. Response to posthypnotic suggestions appears to have a quasi-compulsive quality to it, especially when -- as so often happens -- the subject is unaware that he or she is responding to the experimenter=s cue. It appears to be an automatic response to stimulation. But appearances are deceiving, as indicated by a series of studies performed for a doctoral dissertation in my laboratory by Irene Tobis (nee Hoyt).

In one these studies, highly hypnotizable subjects were presented with three three-digit strings on a computer screen, and then told to press one of three response keys to indicate in which string the digit 7 appeared (Hoyt, 1990, Experiment 2). After 20 trials, the subjects were hypnotized and given a posthypnotic suggestion to press a key to indicate where the digit 3 appeared (of course, these conditions were counterbalanced); this suggestion was then covered by a further suggestion for posthypnotic amnesia.

After termination of hypnosis, the subjects were seated in front of the computer again, and reminded to press keys in response to the digit 7. The purpose of this exercise, of course, was to put the subject in conflict between the nonhypnotic instruction to respond to 7s, and the posthypnotic suggestion to respond to 3s. The subjects then worked through four different types of trials (18 of each) in which both 7s and 3s, or 7s but not 3s, or 3s but not 7s, or neither 7s nor 3s, appeared.

All the subjects showed amnesia for the posthypnotic suggestion. That is, they came out of hypnosis unaware that they had been told anything about the number 3. In the no-conflict trials, where cues of only one or the other type were presented, the subjects showed a high rate of response, catching about 75% of the targets; in the conflict condition, however, response rate diminished significantly, to about 50% of the trials. In the conflict condition, response to the instruction was negatively correlated with response to the suggestion -- a reasonable (though not inevitable) tradeoff, given that only one response was required per trial; but interestingly, this tradeoff also occurred in the no-conflict conditions.

And, in fact, subjects= preferences in the no-conflict condition were highly correlated with their preferences in the no-conflict conditions. That is, a subject who preferred to respond to the suggestion when suggestion and instruction conflicted also preferred to respond to the suggestion when the suggestion and instruction did not conflict.

Thus, response to the posthypnotic suggestion comes at the expense of response to the nonhypnotic instruction, and vice-versa, even when there is no actual conflict between them. This could not occur, of course, if response to the posthypnotic suggestion consumed little or no attentional resource. Apparently it does, and attentional resources devoted to executing the posthypnotic suggestion are not available for the task of responding to the nonhypnotic instruction, and vice-versa. Even though the posthypnotic suggestion is executed outside of the subject=s awareness, it is not automatic in the technical sense of being attention-free.

The identification of the unconscious with the pre-attentive and the automatic is appealing on theoretical grounds, but if the phenomena of hypnosis are to be taken seriously, it is misleading. Deep, semantic processing can occur without retrospective awareness of what has been processed; and behavior executed outside of awareness can nonetheless consume attentional resources. In other words, consciousness can be from attention. As Hilgard (1977) has suggested, these phenomena, and others like them, observed in hypnosis and other states, indicate that consciousness can be divided, so that attentive, semantic processing can proceed outside phenomenal awareness. How that division occurs is not well understood. My own view is that consciousness requires that mental representations of perception and memory be connected to a mental representation of the self as the agent or patient of some event, or the stimulus or experiencer of some state represented by the percept or memory in question (Kihlstrom, 1980). Whatever the fate of that particular idea, it is apparent that consciousness requires more than attention.


Author Notes

Paper presented at the annual meeting of the Psychonomic Society, Los Angeles, November 1995. This research was supported by Grant #MH-35856 from the National Institute of Mental Health.



Dorfman, J., & Kihlstrom, J.F. (1994, November). Semantic priming in posthypnotic amnesia. Paper presented at the annual meeting of the Psychonomic Society, St. Louis.

Flanagan, O. (1994). The Science of the Mind. 2nd Ed. Cambridge, Ma.: MIT Press.

Hilgard, E.R. (1977). Divided consciousness: Multiple controls in human thought and action. New York: Wiley-Interscience.

Hoyt, I.P. (1990). Posthypnotic suggestion versus ordinary instruction: Compliance and attention. Unpublished doctoral dissertation, University of Wisconsin.

Kihlstrom, J.F. (1980). Posthypnotic amnesia for recently learned material: Interactions with "episodic" and "semantic" memory. Cognitive Psychology, 12, 227-251.

Kihlstrom, J.F. (1995). Consciousness and me-ness. In J. Cohen & J. Schooler (Eds.), Scientific approaches to the question of consciousness (pp. xxx-xxx). Hillsdale, N.J.: Erlbaum.


This page last revised 04/08/10 02:58:49 PM .