Carlo Umiltà. The International Handbook of Psychology. Editor: Kurt Pawlik & Mark R Rosenzweig. Sage Publications. 2000.
Scientific psychology began in the nineteenth century as the study of those mental processes that are accompanied by conscious experience. When William James defined psychology as the science of mental life, he actually meant conscious mental life. For about half a century there had been investigations concerning psycho-physical laws, sensory processes, memory, intelligence, and hypnosis. Almost every investigator in these areas believed that consciousness was the key to studying mental events.
In the early part of the twentieth century, however, prominent scholars, like Ivan Pavlov and John Watson maintained that it was scientifically improper to deal with consciousness, because we can only observe physical behavior and brain events. As a consequence, human personal experience became almost a taboo subject in science. Behaviorism banned consciousness, and other mentalistic concepts, as a legitimate topic of investigation from mainstream experimental psychology. However, the study of consciousness survived in other areas of psychology. The contrast between conscious and nonconscious processes was at the core of psychoanalytic theories (Erdelyi, 1985), and of the study of dreaming (Foulkes, 1990) and hypnosis (Kihlstrom, 1985). Also, in social psychology research was to some extent concerned with consciousness (Uleman & Bargh, 1989). However, these studies did not directly address the issue of the relative contributions of conscious and non-conscious processes in cognition, which is of current concern in experimental psychology.
The turning point came when it became increasingly clear that a substantial amount of cognitive processing occurs independent of consciousness (for brief historical reviews, see Reber, 1992; Velmans, 1991). Earlier scholars, like Hermann Helmholtz (unconscious inferences) and Francis Galton in the nineteenth century, and the Wuerzburg school at the beginning of the twentieth century, had already realized that much cognitive processing takes place outside of consciousness. However, it was the discovery, in the late twentieth century, of a number of strikingly neuropsychological syndromes that made a really compelling case for the existence of nonconscious cognitive processes, and thus pointed to the need of consciousness as an explanatory construct.
By following the pioneering work of Clark Hull and Edward Thorndike in the first part of the twentieth century, the earliest studies with normal subjects touched on the issue of learning without consciousness or incidental learning (Reber, 1965). These studies, however, were plagued by problems of methodology and questions over the degree to which incidental learning was truly incidental and how much awareness there was on tasks that purportedly showed learning without consciousness. In the 1970s, papers on consciousness were published by respected investigators (Mandler, 1975; Posner, 1978; Shallice, 1972). They were instrumental in making consciousness a legitimate topic of investigation. Although severe problems of definition, theory, and methodology continued to plague the experimental study of consciousness, the attitude had changed; these difficulties were now seen as problems to be confronted and solved rather than as fatal flaws that question the legitimacy of research on consciousness.
Conscious versus Nonconscious Processing
Many studies with healthy persons investigated the processing of unattended information to collect evidence about nonconscious processing. In them, the term ‘attended’ is taken to mean ‘conscious’, whereas the term ‘unattended’ is taken to mean ‘unconscious’. The first relevant studies were those that went under the rubric of subliminal perception, in which the stimuli were presented using techniques designed to ensure that the material was not consciously processed.
Nonconscious Processing in Normal Subjects
Three main experimental paradigms were used: dichotic listening, parafoveal presentation, and visual masking. They involve looking for indirect evidence of nonconscious processing, that is for the influence of the outcome of nonconscious processing on the outcome of conscious processing.
In dichotic listening tasks, subjects are simultaneously presented with different messages to each ear, and are asked to attend to only one, ignoring the other. The task requires them to repeat back the message delivered to the attended ear. When questioned afterwards about the message delivered to the unattended ear, they are unable to report its meaning, or even the language in which it was spoken. However, they are able to report some of its physical characteristics, like whether it was speech or a tone.
In one of the earliest studies, Lewis (1970) found that the speed at which words delivered to the attended ear were repeated was influenced by semantically related words delivered to the unattended ear. Semantically unrelated words had no effect. In another study (Corteen & Wood, 1972), subjects were conditioned to expect a mild electric shock when they heard certain words. Subsequently, they participated in a dichotic listening experiment, in which, while they shadowed the attended message, the words previously associated with electric shocks were sometimes presented to the unattended ear. Although they claimed to have been unaware of the words in the unattended ear, the shock-associated words continued to produce changes in galvanic skin response. This occurred also with different words that were semantically related to the conditioned ones.
Such studies are cited as evidence that the processing of the unattended, nonconscious message includes the analysis of meaning. Their several replications (Velmans, 1991) indicate that the results are reliable. It is possible, however, that words in the unattended ear briefly attract attention and subjects become momentarily aware of the unattended words (Holender, 1986). If one cannot be certain that the message to the unattended ear has never become aware, the dichotic listening paradigm cannot provide a sound basis for the notion of nonconscious processing in the unattended ear.
The rationale is very similar to that of dichotic listening. Subjects are asked to focus attention on stimuli shown at fixation and to ignore stimuli presented outside the fovea. In a study (Underwood, 1976), a foveal ambiguous word, that is a word with several meanings, acted as the target, whereas a task-irrelevant word (the flanker) was presented in the parafoveal area. The subject could not report the flanker, which, however, biased the meaning of the target word. For example, the target word palm was interpreted as being related to plants or to body parts, depending on the meaning of the flanker (i.e., tree or hand).
The biasing influence of the meaning of unattended, parafoveal flankers on the reported meaning of foveal, attended targets is open to the same criticisms that can be leveled at dichotic listening studies (Holender, 1986). Because one cannot be absolutely certain that the subject never became aware of the flanker, it is not safe to rely on the subject voluntarily not paying attention to the to-be-ignored stimuli.
Under strict pattern-masking conditions, a stimulus cannot be reported because of data limitation, rather than because of a voluntary strategy on the subject’ part. If the masking procedure is really effective, the masked stimulus cannot be reported, no matter how hard the subject tries to report it. Therefore, the best available evidence of nonconscious processing originates from studies that used masking.
In a lexical decision task, subjects are asked to decide as soon as possible whether a letter string is a word or not. In ‘priming’ studies the target letter string is preceded by a prime, that is by a task-irrelevant word. Normally, both the prime and the target are presented supra-threshold and are both consciously processed. A prime such as bread facilitates the lexical decision for a semantically associated word, like butter, in comparison to a nonassociated word, like nurse. In ‘masked priming’ studies, the prime is masked and is not consciously perceived. It may, however, continue to facilitate the response to the following target. That is evidence of nonconscious processing of the masked prime. Marcel (1980) used two types of mask for the prime. When masked by a pattern, the prime was effective and produced facilitation, whereas with a random noise mask, the prime was not effective and did not produce facilitation. Marcel proposed that the pattern mask prevents access to consciousness by disrupting perceptual integration, but does not prevent nonconscious access to semantic knowledge. In contrast, the noise mask degrades the stimulus early in processing, and thus prevents access to semantic knowledge.
The effects of unattended stimuli were extremely sensitive to subtle differences in the experimental procedures. That produced heated discussions over the reliability of the results and their interpretation. The controversies were mostly concerned with the criteria that were used to determine consciousness in the subject, and the correct type of technique for determining thresholds (Styles, 1997).
The logic underlying the paradigms of dichotic listening, parafoveal presentation, and visual masking is that nonconscious information processing can be demonstrated by a dissociation between two indexes. One indicates unavailability of information to consciousness. The other indicates availability of information even though it is not available to consciousness.
Many investigators measured the first index solely in terms of subjective reports. But one should be very cautious in concluding that information was not available to consciousness merely because the subject denies having been aware of the stimulus. It is sounder to determine the absence of consciousness on the basis of tasks that measure discriminative capability. In particular, one should use present/absent tasks, which require subjects to distinguish between the presence or absence of a stimulus, or forced-choice tasks, which require discriminating among a set of known stimuli. Marcel’ (1980) experiments, and subsequent experiments in which his methodology was adopted, are especially convincing because they demonstrate the effects of the masked word, even when subjects cannot detect it, as evidenced by the fact that they cannot discriminate between its presence or absence.
Cheesman and Merikle (1985) proposed the distinction between a ‘subjective’ threshold and an ‘objective’ threshold. When the stimulus is below the subjective threshold, subjects claim they cannot see it, but may still be able to make a better than chance judgment about its presence/absence. The objective threshold is the point in perceptual processing below which subjects are at chance level at deciding whether the stimulus is present or absent. Holender (1986) maintained that it is the objective threshold that must be used to assess nonconscious processing. In contrast, Cheesman and Merikle maintained that the subjective threshold may be used, on condition that one can provide evidence that stimuli presented below that threshold produce qualitatively different effects from stimuli that are presented above it (see Section 13.3 below).
Nonconscious Processing in Brain-Damaged Patients
Neuropsychological studies have provided evidence showing that information that cannot be overtly recognized is processed outside consciousness. There are a number of examples in the literature of the way in which processing and consciousness can break down following brain damage (Farah, 1994; Köhler & Moscovitch, 1997; Styles, 1997). These are termed dissociations between preserved implicit (nonconscious) and impaired explicit (conscious) knowledge. Explicit knowledge is defined as knowledge of a particular aspect of stimulation that can be expressed as conscious experience. Implicit knowledge refers to knowledge of a particular aspect of stimulation that is revealed in task performance, but is not accompanied by a corresponding conscious experience.
Although the essence of many neuropsychological studies dealing with nonconscious processing is the indirectness of the testing procedure, an important distinction should be made between types of task and types of process (Young & De Haan, 1992). The use of the terms ‘explicit’ and ‘implicit’ is better reserved to refer to the nature, conscious or nonconscious, of the processing. In contrast, to characterizetasks, the terms ‘direct’ or ‘indirect’ should be used. A direct task directly inquires about a given ability, and performance is dependent on that ability. On an indirect task, assessment of the ability in question is incidental and the task ostensibly measures something else. In most patient studies, implicit processing was inferred on the basis of performance in indirect tasks.
Damage to the primary visual cortex causes blindness in parts of the visual field (scotomas). Complete lack of visual conscious experience is revealed by standard tests. However, it is now well established that some cortically blind patients possess preserved implicit knowledge about visual stimuli that fall in the scotoma. This preserved implicit knowledge is usually referred to as ‘blindsight’ (Weiskrantz, 1986).
Blindsight was initially tested by direct tasks that required the patients to perform forced-choice responses to stimuli shown in the scotoma. For example, a brief visual stimulus was presented in one of several locations within the scotoma. The patients were asked to guess where the target was presented by moving their gaze, or by manually pointing to the stimulated location, or by making a verbal response such as ‘top’ or ‘bottom’. Despite subjective impression of engaging in random guessing, and the absence of any reportable awareness of the stimulus, their localization judgements were surprisingly accurate. The forced-choice guessing procedure was also used to demonstrate that some patients are able to detect stationary or moving stimuli, to discriminate the direction of movement, and even to discriminate between different stimulus orientations.
Several criticisms were leveled at the forced-choice guessing procedure, the more damaging of which was that patients’ performance might reflect residual awareness of visual stimuli, which may be qualitatively different from normal conscious vision (Köhler & Moscovitch, 1997). The critical point is again that lack of consciousness is established solely on the basis of subjective reports, that is the patients’ claim that they do not see the stimuli. Because of that, a number of indirect tasks were introduced to examine the effects produced by information delivered to the scotoma on processing that occurs in the intact field.
When normal subjects are asked to respond to the second of two flashes that are presented in rapid succession, reaction time is slower than when a single flash is presented. In a cortically blind patient, Corbetta, Marzi, Tassinari, and Aglioti (1990) found that reaction time to the target flash presented in the intact part of the visual field was slower when a preceding flash was presented in the scotoma than when only the target flash was presented. The inhibitory effect occurred even though the patient was never aware of being presented with two flashes, and always perceived a single flash in the intact field. Similar indirect tasks showed a number of effects attributable to stimulation of the scotoma: interfield summation, inhibition of saccade execution, threshold resetting, tilt of the subjective vertical, color after-images, and perceptual completion.
Bilateral lesions of the occipitotemporal region may cause an impairment in visual processing of faces that is termed prosopagnosia. Proso-pagnosic patients are able to discriminate a face from a non-face but cannot identify specific individuals. However, a substantial body of evidence has been accumulating that indicates that prosopagnosic patients possess implicit knowledge about face familiarity and face identity.
De Haan, Young, and Newcombe (1987) asked a prosopagnosic patient to categorize a printed name as belonging to an actor or a politician. Simultaneously with each name, a task-irrelevant, to-be-ignored photograph of an actor’ or politician’ face was presented. Normal subjects are slower at categorizing the names when the face comes from a different category than when it comes from the same category. Although the patient was severely impaired at categorizing the faces overtly, he showed interference from different-category faces.
Neglect is a disorder that generally follows posterior parietal damage and results in patients’ failure to report or even orient to stimuli occurring on the contralesional side of space. Often the deficit manifests itself in the visual, auditory, and tactile modalities. There are two main hypotheses attempting to explain neglect. According to the representational hypothesis (Bisiach & Vallar, 1988) neglect is caused by a deficit in the ability to form a whole representation of space. Patients do not react to stimuli originating from the affected region because that region is impoverished or absent in their internal representation of space. The attentional hypothesis (Kins-bourne, 1987) maintains that each hemisphere is responsible for shifting attention in the contralateral direction in either region of space. Damage to one hemisphere (generally, the right hemisphere) causes an imbalance in favor of shifts contralateral to the intact side, thus rendering orienting towards the affected side (generally, the left side) difficult or impossible.
Standard tests show that patients are completely unaware of stimuli in the affected hemi-space. However, indirect tasks reveal that they may process to a remarkably high degree information presented in that hemispace. Làdavas, Paladini, and Cubelli (1993) reported a patient who could not read aloud words presented in the left field, nor judge their lexical status or semantic content. He could not even detect the presence of a string of letters. However, response to a word in the intact right field was faster when the word was preceded by a brief presentation of an associated word in the impaired field. The patient indirectly showed semantic priming caused by words presented in the affected field, even though he was unable consciously to process any of their attributes.
Amnesic patients, with either focal brain lesions in the medial temporal lobe or degenerative brain disorders, demonstrate retention of information about events for which they disclaim any memory (Moscovitch, Goshen-Gottstein, & Vriezen, 1994). They are catastrophically poor in direct tasks, which require explicit recollection, such as recognition and recall. In contrast, they can show near-normal performance in indirect tasks, which do not make direct reference to the past, but rather assess memory by showing changes in behavior with experience or practice, in the absence of a subjective experience of remembering.
An often-used indirect task is stem or fragment completion, which shows the increased likelihood that a word stem will be completed with a previously studied word (Schacter, Chiu, & Ochsner, 1993). In this task, the first letters of a word are presented (e.g., mot, for motel) and the patient is asked to complete the stem with the first word that comes to mind. Priming effects are obtained when probability of completion of studied words is above the baseline guessing rate and exceeds that of nonstudied words. Fragment completion is similar, except that a word fragment is presented (e.g., – t – i – g, for string) instead of the stem. Amnesic patients consistently show normal or near-normal priming, even though their memory for the studied words is severely impaired when it is tested directly. A similar pattern of results is obtained when pictures, rather than words, are used. Identification of degraded line drawings of familiar objects improves in amnesic patients, if they previously have seen the intact drawing, even though their explicit memory for the drawing is nearly null.
A controversial issue concerning dissociations between explicit and implicit memory is the assumption that direct and indirect tasks are ‘process pure’ (Rugg, 1995). A task may ostensibly be indirect, yet performance on it may be contaminated by explicit knowledge. Similarly, implicit knowledge may influence performance on what is ostensibly a direct task. An alternative approach has been developed by Jacoby, Toth, and Yonelinas (1993). It is termed ‘process dissociation procedure’ and requires a condition in which conscious and nonconscious processes have the same effects on performance, and a second condition in which the two processes operate in opposition to each other. Results from these two conditions can be combined so as to quantify the relative contributions of explicit and implicit memory to the ability to retrieve studied items. However, the procedure rests on the assumptions that conscious and non-conscious processes are independent and that their influence in the two conditions is invariant.
Models of the Dissociation
Three explanations have been proposed to account for the dissociation between explicit and implicit knowledge (Köhler & Moscovitch, 1997). One is that specialized neural mechanisms are disconnected from, and because of that do not convey the output of their processing to, the higher-level neural system(s) that subserve(s) consciousness. The explicit/implicit dissociation would reflect a disconnection between intact specialized mechanisms and intact consciousness system(s). Another explanation is that consciousness system(s) receive(s) degraded output from damaged mechanisms. Degraded information nonconsciously manifests its effects on indirect tasks, but is insufficient for being consciously processed in direct tasks. Identical mechanisms would subserve both explicit and implicit knowledge, the only difference being that, when lesioned, these mechanisms produce low quality output, which cannot give rise to consciousness. Because explicit and implicit knowledge would be subserved by the same mechanisms, they should be qualitatively similar. The third explanation is that the lesion destroys mechanisms that are necessary for demonstrating explicit knowledge through direct tasks, whereas performance in indirect tasks depends on different mechanisms, which do not have access to consciousness. Input would always be processed in parallel by mechanisms that send their output to consciousness and by mechanisms that do not. The lesion destroys the former, leaving the latter (nearly) intact. Because they are thought to be subserved by different mechanisms, explicit and implicit knowledge might be qualitatively different.
These accounts make alternative predictions and likely apply to different forms of the explicit/ implicit dissociation. The explanation based on disconnected but otherwise intact mechanisms predicts that patients’ performance in indirect tasks should very closely resemble that of normal subjects in the same tasks. The explanation based on degraded output from damaged mechanisms predicts that patients should perform worse (a quantitative difference) than normal subjects in indirect tasks. The explanation based on independent mechanisms for performance in direct and indirect tasks predicts the presence of qualitative differences between patients and normal subjects.
There is evidence that processes that are accompanied by conscious experience are qualitatively different from those that occur without any concomitant conscious experience. These qualitative differences were among the criteria that Cheesmann and Merikle (1985) adopted to substantiate the distinction between conscious and nonconscious processes.
In a study by Marcel (1980), subjects were shown three letter strings, one after another, and were asked to make lexical decisions to the first and third string. On some trials, the first string was a context word, which was followed by an ambiguous word (the prime), then by a target word. Examples are save – bank – money and save – bank – river. The lexical decision time to the third word indexed which meaning of the ambiguous word had been accessed. In one condition, the prime was presented unmasked, at an exposure duration of 500 ms, so that all words were consciously available. Results showed selective priming of the target: only the meaning of the prime related to the meaning of the context word was activated. The prime bank facilitated response to money but not to river. When the prime was presented for 10 ms and was pattern masked, so that subjects claimed to be unable to detect it, unselective priming occurred, and all meanings of the ambiguous word were accessed. The prime bank facilitated both money and river.
Negative priming refers to the slowing of responses to recently ignored stimuli. For example, with visually superimposed outline forms, subjects can be instructed to attend to the red figure and to ignore the superimposed green one, or vice versa. Negative priming manifests itself as a slowing of the response, relative to a control condition, when the previously ignored item is presented as a target on the next trial. Two explanations have been offered for negative priming, and both invoke some form of inhibition. One is that activation of the distracter representation is reduced and this inhibition is still effective when the next trial is presented. The other is that the distracter representation is denied access to the response system and its isolation from the control of action must be overcome when, on the next trial, the distracter requires a response.
In the flankers task (Eriksen & Eriksen, 1974), subjects are presented with a string of letters (e.g., SSFSS) and asked to give a discriminative response to the central letter (the target) and ignore the other letters (the distracters). The relation between the target and the distracters on the first, prime trial and the target on the subsequent, probe trial can be manipulated. When the second target is again F, the response to it is speeded up, indicating the usual positive priming effect. When the second target is instead S, the response is slowed down, indicating negative priming. Fuentes and Humphreys (1996) used a modified version of the flankers task to test the impaired and the intact visual fields of a patient with extinction, that is with a form of neglect that manifests itself when both visual fields are simultaneously stimulated. Results indicated a qualitative difference between processing of distracters in the two visual fields. In the intact field negative priming was found, whereas in the impaired field there was positive priming when the previously ignored distracter was re-presented as the target on the probe trial.
Function(s) of Consciousness
A crucial issue concerns what the function(s) of consciousness might be. If part of human cognitive processing is conscious, it is reasonable to assume, from a Darwinian stand point, that consciousness should have some function (Reber, 1992). However, this assumption is not universally accepted. Velmans (1991) has argued that consciousness plays no role in cognitive processing, not even in those higher order processes that render it possible to choose among alternative courses of action, to plan and time their execution, and to make decisions (executive processes). Of course, Velmans does not deny that we are conscious of those processes. However, he distinguishes between processes and the results of processes. Although one can have conscious access to some of the results of processes, one never has conscious access to the operations they perform. Therefore, conscious access cannot have a causal role in cognitive processing. What instead plays a role is focal attention. But consciousness and focal attention are confounded; processes that are at the focus of attention are accompanied by focal attention, and when focal-attentive processing is absent, consciousness also is absent.
No doubt attention is closely associated with consciousness: both are capacity limited and slow acting, both involve working memory, both operate in series and use processing resources, both intervene in planning and decision making. However, there are reasons for keeping them separated (Umiltà, 1992). Attention is not always necessary for the product of a process to access consciousness. It is a privileged route to consciousness, but not the only one available. There are cases in which consciousness is accessed bypassing attention. In a dichotic listening task, some ‘special’ unattended words (e.g., the listener’ name) gain immediate access to consciousness. In the case of well-learned routines, it is possible to be conscious of performing an action without paying attention to it.
The available evidence suggests that consciousness plays a causal role in mediating inhibition; inhibitory processes may not take place unless the stimuli are consciously processed. In the Marcel (1980) study, one of the meanings of a consciously perceived word was inhibited when the other meaning was selectively activated by the context word. In contrast, when the prime word did not access consciousness, both meaning were activated. The Fuentes and Humphreys (1996) study showed that the representations of distracters that did not access consciousness were not inhibited. It must be stressed that the inhibitory processes themselves are not necessarily conscious. It is the fact that the representations that need to be inhibited are conscious that allows inhibition to take place.
Also, a study by Libet (1985) provided evidence in support of the notion that consciousness is instrumental in allowing inhibition of ongoing cognitive processes. He developed a procedure that enabled subjects to establish the instant they experienced a wish to perform a spontaneous act involving no preplanning (a simple flexion of the wrist or fingers). The time of the conscious intention to act was obtained from the subject’ report of the spatial position of a revolving dot on a clockface when he/she became conscious of wanting to move. Simultaneously, Libet recorded from the scalp the electrophysiological readiness potentials that occur about 550 ms before execution of voluntary acts. The wish to act was consciously experienced about 200 ms before execution of the voluntary act, that is it followed the readiness potentials by about 350 ms. The conclusion was that activation of the neural mechanisms that eventually lead to an action begins non-consciously. Also, it was found that the decision to act could still be consciously inhibited during 150–200 ms after experiencing the conscious intention. Libet proposed that the role of consciousness would be that of either permitting or preventing the motor implementation of the intention to act. In the absence of conscious experience, execution of the movement could not be prevented.
Theories of Consciousness
Several theories posit that it is the activity of a unitary mechanism, which may have a unitary brain representation, that gives rise to consciousness (Umiltà, 1992). Other theories posit instead that consciousness arises as a result of interactions among various brain mechanisms (Dennett & Kinsbourne, 1992). Also, there are theories that posit that consciousness arises from the binding of disparate neural elements into an integrated pattern by an oscillating 35–75 Hz signal (Crick & Koch, 1990; Singer, 1993). Of the three most influential theories of consciousness, to be described next, Shallice’ is of the interactionist type, Schacter’ is of the unitary type, whereas Baar’ is ambiguous.
Baar’ (1988) theory relies on three components: expert processors, a global workspace, and contexts. Expert processors are specialized nonconscious processors, reminiscent of Fodor’ (1983) modules. They are extremely efficient in narrow domains, work autonomously, display their output in the global workspace, receive messages from the global workspace, and form coalitions. The global workspace is a device devoted to system-wide dissemination of information, which broadcasts global messages to all expert processors. In the global workspace new links between expert processors are formed, producing new coalitions. Contexts are pre-established coalitions of expert processors that can influence global messages without themselves entering the global workspace. They are devices that constrain conscious contents in the workspace without being conscious themselves. Only the contents of global workspace are conscious. Outputs from the expert processors that win the competition to reach the global workspace are broadcast throughout the system. If one thinks that representations become conscious by entering the global workspace, the theory is of the unitary type. If one thinks that representations become conscious by being disseminated throughout the whole system, the theory is of the interactionist type.
Shallice’ (1992) theory has several components, the most important among which are specialized processors, schemata, a contention scheduling system, and a supervisory attentional system (SAS). The specialized processors are akin to Baar expert processors. The schemata are well-learned, highly specific programs for routine actions and thoughts. Many schemata can be concurrently activated through the specialized processors and memory traces, and the competition among them is resolved by another specialist system, contention scheduling. Contention scheduling automatically selects among activated schemata by means of excitation and lateral inhibition. Routine actions and cognitive processes depend on the selection operations that are carried out by contention scheduling. When nonroutine actions and cognitive processes are needed, top-down biasing of schemata is provided by SAS. It operates by the application of additional excitation and inhibition to schemata, which changes the probability of selection by contention scheduling.
One might think of one of these components as the conscious part of the model, thus transforming Shallice’ theory into one of the unitary type. That is what Baddeley (1992) did by equating his central executive with SAS. However, Shallice maintained that the contents of consciousness do not correspond to the contents of SAS, but rather they correspond to the flow of information between the control components when there is concurrent and coherent operation of them on information from schemata.
Schacter’ (1990) theory shares most of its components with Baar’ and Shallice’ theories. These are special-purpose modules, an executive system, a declarative/episodic memory system, and a procedural/habit system. The key feature of the theory is the existence of a phenomenal consciousness component. It acts as the link between the special-purpose modules and the executive system, which controls and guides the activities of the special-purpose modules. The executive system would not operate if its contents had not been mediated by consciousness. The theory posits that representations become conscious because of what happens inside the phenomenal consciousness component. Because what matters for consciousness is the fact of being inside a specific component, rather than informational exchange among components, Schacter’ theory is clearly of the unitary type.
A Privileged Brain Locus for Consciousness?
The theories outlined above propose different solutions to the problem of how consciousness is represented in the brain. The main difference concerns whether consciousness is linked to activity in one, privileged part of the brain, or instead it originates from interactions among several brain structures. Dennett and Kinsbourne (1992) in particular have criticized the idea that consciousness depends on the activity of a single brain mechanism, which they called the Cartesian theater model, after the pineal gland in Descartes’ brain. A modern version of this model can be found in Gray (1995), who has proposed that the unitary brain mechanism of consciousness is the subicular area, which belongs to the septohippocampal system.
The explicit/implicit dissociations described earlier are domain specific because they occur in one single domain. That seems to militate against the possibility that there exists a unitary neural structure subserving consciousness. If there were a general-purpose consciousness system, then a focal lesion to it should produce patients showing the explicit/implicit dissociations across a number of different domains, possibly all domains. No such patient has been described so far. That supports the notion that consciousness arises from many independent consciousness systems that may be separately lesioned (Umiltà & Zorzi, 1995). Note that the three explanations of the explicit/implicit dissociations (see above) apply to either a unitary consciousness system or multiple consciousness systems.
Those states in which there is a general severe alteration in consciousness seem instead to contradict the notion of multiple consciousness systems. In this context, one should consider a basic distinction between the capacity for consciousness and the contents of consciousness (Zeman, 1997). The former depends on the normal activity of the upper brainstem and related areas of the thalamus (the reticular activating system), whereas the latter are supplied by cortical areas. The domain-specific dissociations between explicit and implicit knowledge occur when some of the contents of consciousness are lacking because of focal cortical lesions. Coma, a state of unconsciousness in which the eyes are closed and the sleep—wake cycle is absent, occurs when the reticular activating system is lesioned and the capacity for consciousness is lost. Of particular interest is the vegetative state (Giacino, 1997; Zeman, 1997), which, in a sense, is the converse of coma because the reticular activating system is still functional whereas all cortical functions are lost or gravely impaired. The vegetative state shows that there can be wakefulness without consciousness. Patients have the eyes open and are awake, but demonstrate no evidence of being conscious. Importantly, the contents of consciousness are lacking across all domains, even though reports hint at the possibility of preserved implicit knowledge (Barbur, Watson, Frankowiack, & Zeki, 1993; Menon et al., 1998). However, that is not necessarily evidence against the presence of multiple consciousness systems. The vegetative state results from acute insults or chronic diseases that affect the entire cortex. Therefore, it might well be that consciousness is lacking in every domain, not because a unitary consciousness system is lesioned but rather because all consciousness systems are damaged.
Other altered states of consciousness (minimally conscious state, akinetic mutism, and locked-in syndrome; Giacino, 1997) are less interesting from this point of view because in them the contents of consciousness are, to a certain extent, preserved, whereas in the vegetative state, the contents of consciousness are completely absent. It is possible that, as happens in the vegetative state, in the case of less severe alterations of consciousness, all consciousness systems are damaged, although to a lesser degree.