How can we know exactly when a mental process is unconscious? In examining this question it is useful to make a distinction between two stages of information processing: encoding and retrieval. The importance of this distinction is that a suitable methodology for demonstrating the implicitness at one point is not applicable at another. Consider research with amnesiacs in which the evidence for tacitly held knowledge is found at the retrieval stage (i.e., a patient performs better over time but does not remember the training session) but not at encoding (i.e., amnesiac patients are consciously aware of what they are learning while they are learning it). Demonstrating that memory is not used consciously at the time of retrieval entails a different methodology than that used to demonstrate that the encoding process was unconscious. For encoding, it is necessary to demonstrate that at the moment of the presentation of the stimulus, the subject's awareness of that stimulus was deficient.
It turns out that demonstrating a lack of awareness at encoding is deeply problematic. How do we know that the stimulus was really not consciously perceived in some, perhaps minimal, way? Because the decision as to whether the subject consciously perceived the stimulus relies on one or another form of subjective self-report, we are stuck with having to rely on nonverifiable measurement.
Imagine a typical subliminal perception experiment in which subjects are given brief or masked exposure to objects or words. Later, knowledge of the target is assessed either by direct tests (what word did you see?) or by indirect tests (changes in response times over time). Note that there are actually two constructs here: the accessibility of the information at encoding and the availability of the stored knowledge sometime after presentation. Some (Brody, 1989; Eriksen, 1959) maintain that unconscious perception can only be reliably established if the accessibility of the stimulus at the time of encoding is zero. In other words, there should be no difference in accessibility of a stimulus between the subject and a blind person. This criterion is very difficult (if not impossible) to achieve because it is always possible that something was perceived consciously and chance performance is attributable to subjects' not being sufficiently confident to make a response based on what little they did see.
Alternatively, as suggested by Erdelyi (1986, 2004), we can require that the availability of the stimulus to consciousness be greater than the extent to which that stimulus was consciously accessible at the time of encoding. This approach has an important advantage in that it reveals a critical but often unrecognized fact: the impact of unconscious knowledge on behavior is a continuum and not an either/or issue.
The Implicit/Explicit Continuum
It is erroneous to say that a behavior has to be explained in its entirety by either conscious or unconscious input. Most cognitive tasks, including perception, memory, problem solving, and creativity are products of the influences of both conscious and unconscious processes. The existence of conscious factors does not in any way preclude the further influence of unconscious ones. The findings of two important experiments help make this point.
In the first, Mathews et al. (1989) had experimental subjects engage in an implicit learning task over a four-day period. The study used what is known as an artificial grammar (AG). An example of a typical AG is given in Figure 18.1 along with several letter strings that it can generate and a number of nongrammatical or not well-formed strings that contain a single letter violation. It is apparent that the system is complex and, as Mathews et al., found, not easy to describe. In the canonical AG learning study, subjects memorize a number (perhaps 15 or 20) of exemplary letter strings and then, using what knowledge they acquired from the learning phase, attempt to distinguish whether new letter strings are "well-formed" or not - that is, whether or not they conform to the rules that generated the original set.
The clever twist that Mathews and his colleagues used was to stop their participants from time to time during the "well-formedness" phase and ask them to explicate, in as much detail as possible, what they knew and how they were making their decisions. Transcripts of their responses were then given to four different yoked-control groups who, without having any learning experience, were asked to classify the same strings. If subjects were consciously aware of the knowledge they had acquired and could communicate it, the yoked subjects should perform at the same level as the experimental participants.
Mathews et al. found that their experimental subjects could make reliable decisions on the very first day of the study but were remarkably inept in communicating what they had learned - yoked subjects working with the Day 1 transcripts performed at chance. However, as the experiment progressed, the experimental subjects' ability to verbalize their knowledge improved dramatically. In fact, the yoked subjects who received the Day 4 transcripts made decisions nearly as well as the experimental participants. Interestingly, the experimental subjects' performance on the primary task didn't improve significantly after the second day although their ability to explicate what they knew did. This is an example of how knowledge is encoded implicitly but over time becomes explicit and can then be retrieved consciously. Another implication of Mathews teams' study is that the implicit and the explicit are bound up in a delicate synergy and we would be wise to refrain from all-or-none distinctions.
In the second study, Knowlton and Squire (1 994) showed that even when conscious knowledge is fully available (i.e. in Erdelyi's  terms, accessibility equals availability) it doesn't mean that it is necessarily being utilized at all times. They found that amnesic patients' performance was indistinguishable from normal controls on a standard AG learning task, as described
Figure 18.1. A typical artificial grammar used in many studies of implicit learning. The grammar generates letter strings by following the arrows from the input state (S1) to the terminal state (S6). Several examples of "well-formed" strings are presented along with others that contain a violation of the grammar.
Figure 18.1. A typical artificial grammar used in many studies of implicit learning. The grammar generates letter strings by following the arrows from the input state (S1) to the terminal state (S6). Several examples of "well-formed" strings are presented along with others that contain a violation of the grammar.
previously, suggesting that representations of knowledge need not be held in a conscious form to be used to make decisions. However, when both groups were encouraged to make decisions by utilizing any similarities between the test stimuli and those used during learning, the two groups differed. The normal controls showed a small but significant improvement, whereas the patient group's performance, perhaps not surprisingly, actually diminished.
There are two implications of these studies. First, in these tasks, normal subjects possess a delicate balance of implicit and ex plicit knowledge and how each is manifested depends as much on the task demands as on the accessibility of conscious knowledge. Second, amnesic patients, whose neurological injuries have compromised their ability to form consciously accessible long-term memories, can still carry out complex implicit learning tasks. The balanced synergy is largely missing in this population, but the implicit system appears to be relatively intact.
In short, it is both theoretically sounder and methodologically more plausible to look at the impact of implicit knowledge as operating along a continuum. Rather than ask whether or not a particular task was implicitly or explicitly performed, one should examine the extent to which both implicit and explicit factors are playing a role in the behavior in question. With this framework in mind, let's explore the several domains in which unconscious mechanisms have been examined.
One interesting aspect of attention, at least for the purposes of implicit processes, is that attention and consciousness are highly correlated. When something is being attended to, for example, the words of this sentence, the object in the focus of attention becomes conscious. Of course, because of the limits on attention, some, if not most, of the sensorial events in the outside world are not within the focus of our attention. When reading a book, we "tune out" much of the outside world such as conversations and, traffic. Indeed, we ignore most of the events that are outside of the attentional focus. The question that almost asks itself is, What effect do the unattended, nonconscious events in our environment have on us? Do they get registered unconsciously in some fashion without our awareness, or are the effects of unattended events trivial and only become important when and if they are consciously attended to?
The effects of diverting attention from the stimulus at encoding are usually studied in the context of a dual-task paradigm in which attention is diverted by a secondary stimulus (Morrison, Chap. 19). For example, in the classic dichotic-listening task two different messages are played, one to each ear. One message is attended to, the other not -although the secondary message often contains important information. Afterwards, a simple memory or priming task is used to discover the effects of diverted attention.
The initial findings here suggested that, when attention is diverted from a stimulus, the effect of that stimulus is greatly reduced (Broadbent, 1958; Cherry, 1953;
Moray, 1959). For example, Johnson and Wilson (1980) presented ambiguous words like "sock" to one ear while a disambiguat-ing word ("foot" or "punch") was presented in the other. They found that the amount of attention allocated to the encoded stimulus was critical. When the instructions were to attend to both channels, the word "foot" facilitated the interpretation of the ambiguous homophone "sock." But when attention was directed to the channel in which the target words were presented, items in the unattended channel did not influence the perceived meaning of the targets.
Later studies, however, questioned this conclusion. Eich (1984) presented subjects with homophones such as "fare/fair" in one ear and a modifier of the less frequent meaning (e.g., "taxi") in the other. Subjects were then given a recognition test for the modifiers and were asked to spell the target words ("fare" or "fair"). Eich found a clear impact of implicit memory. Despite being virtually at chance on the recognition task, subjects showed a strong tendency to spell the test words according to their primed, but less common, meaning. Similar findings were reported in a series of studies by Mulligan (1997, 1998) in which subjects memorized word lists while their attention was diverted by a secondary task involving repeating strings of digits of varying lengths. Attentional load was manipulated by varying the length of the digit string from three to seven. Mulligan found that increasing the attentional load impaired explicit memory performance for the original words (using cued recall) but had essentially no effect on implicit memory (as measured by stem completion). It seems therefore that at-encoding stimuli have an impact on subsequent performance even when they are not consciously perceived.
Although diverting attention certainly reduces the likelihood of the stimuli's being consciously encoded, Kihlstrom et al. (1 990) made quite certain that their input stimuli were not being attended to. Their participants were surgical patients who were presented with a repeating list of stimulus items while completely anesthetized. After surgery, although patients had no explicit memory for the material, an implicit, free-association test showed that information presented during anesthesia was encoded. Shanks and St. John (1 994) criticized these and other studies on the grounds that they produced mixed findings and often used questionable methodology. However, recently, Merikle and Daneman (1 996) conducted a meta-analysis of 44 studies with several thousand participants and concluded that, taken as a whole, these studies support the argument that items presented during anesthesia can have an impact on postsur-gical tests. Questions still remain however with regard to the possibility that at least some of the subjects were partially conscious during stimulus presentation.
Historically, subliminal perception studies have been controversial. They have ranged from the embarrassing "eat popcorn - drink coke" hoax that was foisted on the public a half-century ago by an overzealous advertising agent (Pratkanis, 1992) to the vigorously debated use of subliminal messages in psychotherapeutic settings (Silverman, 1983; Weinberger, 1992). Admittedly, much of the early work was suspect and has been vigorously criticized (Eriksen, 1959; Holender, 1986; Shanks & St. John, 1994), and, as we noted earlier, there are numerous methodological traps that await the unwary. However, the impact of subliminally presented material on subsequent behavior has now been replicated in literally hundreds of experiments and the evidence appears to be convincing. Subliminal presentation can have an effect on emotional preferences (Kunst-Wilson & Zajonc, 1980; Murphy & Zajonc, 1993) and produce semantic priming (Draine & Greenwald, 1998). Moreover, similarly undetectable stimuli have been shown to activate appropriate brain regions -emotionally charged stimuli activate the amygdala (Elliot & Dolan, 1998; Whalen et al., 1998) and numerical presentations produce parietal activity (Naccache & Dehaene, 2001).
The studies with which we are most comfortable are those that follow Erdelyi's (1986, 2004) advice cited previously. Ensure that there are separate and reliable measures of accessible and available knowledge -with the critical inequality being those situations in which knowledge that is "accessible" by consciousness is less than knowledge that is "available" and can be shown to have some (indirect or implicit) impact on behavior. (For more on the issues of measuring unconscious knowledge, see Merikle and Reingold, 1992.) Two classic studies that appear to satisfy this condition (Marcel, 1983 and Kunst-Wilson & Zajonc, 1980) are worth a closer look.
In an extended series of experiments, Marcel (1983) showed that graphic (what the word looks like) and semantic (what the word means) information ofsubliminally presented words can affect choice behavior. One of Marcel's standard protocols involved presenting subjects with two words, one sub-liminally and the other supraliminally. After each presentation, subjects were asked whether or not they saw a word. After the pair was presented, they were asked whether the two words were physically and seman-tically related. By systematically varying the sub/supra-liminality of the stimuli, Marcel was able to explore the manner in which implicitly and explicitly encoded stimuli affected subjects' choices. The key finding was a threshold effect whereby subjects were at chance in determining the presence or absence of the subliminally presented word but could reliably report whether the two words were semantically and graphically similar. Marcel's conclusion, based on the full series of experiments, was that although there is a gradual effect of awareness on performance, importantly, a complete lack of awareness does not entirely remove that effect.
In Kunst-Wilson and Zajonc's (1980) classic study, subjects were subliminally presented with a set of irregular octagons. They were then shown pairs of octagons supral-iminally, one of which was from the set previously presented and one of which was new. Subjects were asked both to select the item they thought was presented before and to pick the one they preferred. KunstWilson and Zajonc found that, despite being at chance on the recognition task, subjects showed a preference for the subliminally presented octagons over the novel ones, demonstrating that affective preferences can be influenced by events that were not consciously noticed (Kunst-Wilson & Zajonc, 1980; Murphy and Zajonc, 1993).
Elliot and Dolan (1998) extended this "subliminal mere exposure" effect and showed that, in addition to preferring the previously presented items, different brain regions were activated when old and novel stimuli were later presented supraliminally. This finding is consistent with a large number of fMRI studies that suggest that implicit and explicit memory retrieval involves the activation of distinct brain regions (for a review see Cabeza & Nyberg (2000). Whalen et al. (1998) have also demonstrated that subliminally presented faces displaying fearful emotions activated the amygdala despite a lack of subjective awareness of ever having seen those faces. Happy faces presented for identical time periods had no effect on these structures.
Finally, Naccache and Dehaene (2001) presented evidence of abstract representation in subliminal priming. Subjects were asked to decide whether a target number was bigger or smaller than 5. Each target was preceded by a subliminal prime that was either spelled out (six) or presented in numeric form (6). Subjects displayed faster reaction times when the prime and target were the same number - regardless of the number's form. In addition, fMRI data revealed that the subliminal primes elicited the same parietal lobe activity as the supraliminal targets, suggesting similar cortical processing.
All of the studies above are open to the critique that some awareness of the stimuli might have contaminated the procedure (Holender, 1986; Shanks & St. John, 1994). To counter this criticism, Debner and Jacoby (1 994) used the process-dissociation procedure (Jacoby, 1991). In their study, words (e.g. MOTEL) were first presented sublimi-
nally. During testing, subjects were asked to complete word stems such as MOT with the restriction that they not use any word they thought might have been used in the subliminal presentation phase. The logic here is clever. If the word was consciously perceived, the subjects should have been able to refrain from using that word to complete the word-stem. However if they did not see the word, then the subliminally presented word should have been used as often as the others. The results showed that subjects were typically not able to follow this instruction and tended to use the subliminal primes.
In summary, the suggestion that attention and consciousness is needed for the encoding of complex, semantically sensitive events (Perruchet & Vinter, 2003; Shanks & St. John, 1994) is probably unwarranted. These studies, although not uniform in their conclusions, suggest that fairly sophisticated information about complex stimulus displays can be picked up under severe atten-tional load, when the material is presented subliminally and, possibly, under anesthesia. They also support the notion that this information is not simply logged in some inert form but has an impact on memorial representations, choice behavior, and decision making.
Virtually every complex living organism has the ability to store the products of experience to be accessed at some later time. People's ability to store a seemingly endless array of episodes, facts, motor skills, and linguistic and social knowledge and to retrieve the appropriate information rapidly and appropriately are remarkable phenomena -and one that still remains something of a mystery. Cognitive investigations ofmemory revealed early on that human memory is not a single, unified phenomenon. There are different kinds of memories and each is instantiated in a variety of ways. Our concern here is the extent to which memory processes are modulated by conscious intentions or are implicit and operate outside the spotlight of consciousness.
Conscious or explicit memory has traditionally been studied using direct tests in which participants are asked to consciously recall or recognize previously memorized items. The original assumption was that a failure to recall or an inability to recognize an item is diagnostic of that item having been forgotten. However, as we noted earlier, just because people cannot recall something does not necessarily mean the memory no longer exists. In some ways, it is surprising that it took cognitive psychologists so long to appreciate this aspect of human memory. Early reports by neurologists such as Claparede and Korsakoff implicated implicit representational systems and, lest we forget, Freudian psychoanalysis was founded on the existence of nonretrievable memories that play a role in human behavior (Erdelyi, 1985).
The renewed interest in implicit memory was largely attributable to the discovery that amnesiac patients, despite being compromised in their ability to form new explicit knowledge, can nevertheless acquire new information implicitly. The laying down of consciously retrievable, long-term memories has been compellingly shown to be dependent on structures in the medial temporal lobes (MTL), specifically the hippocampus (Squire, 1995). When the hippocampus and its associated areas are damaged or destroyed, it becomes difficult and, in extreme cases impossible, for new explicit memories to be formed. The discovery of the critical role the MTL structures play here was made in the case of HM, the first neurological patient to have his hippocampus surgically removed (see Corkin, 1968; Milner, 1962; Milner, Corkin, & Teuber, 1968; Squire, 1992; Warrington & Weiskrantz, 1968). HM suffered from severe, intractable epilepsy, the neural focal point of which was in the MTL. To alleviate his multiple, daily seizures surgeons extirpated bilaterally the affected brain regions. Although the surgery was successful in stopping the seizures, HM emerged from the procedure with profound, chronic antero-grade amnesia.
The standard interpretation of HM, based on the now rather large number of patients with similar neurological dam age (see Squire, 1 992 for a review), is that such people do not suffer from a learning deficit, per se, but rather from an inability to consolidate new explicit, or declarative, knowledge. Patients with MTL damage show no diminished ability to recall episodes that occurred prior to the trauma, they present a nearly normal short-term memory profile, and, importantly from our perspective, they show relatively intact implicit learning and memory. Indeed, a large literature has accumulated in recent years showing that the performance of anterograde amnesiacs is virtually indistinguishable from that of normals on a wide variety of memory tasks including word-stem completion (Warrington & Weiskrantz, 1968; 1974), fragment completion (Tulving, Hayman, &MacDonald, 1991), context sensitive memory (Schacter & Graf, 1986), memory for letter strings generated by an AG (Knowlton, Ramus, & Squire, 1 992; Knowlton & Squire, 1 994), and recall of dot patterns (Knowlton & Squire, 1994). As Seger (1 994) argued, amnesiac patients provide the best empirical support for the proposition that knowledge that is not consciously accessible can still have a profound influence on ongoing behavior.
These discoveries gave rise to a number of significant advances in our understanding of memory, both implicit and explicit. Reber (1992a,b), Schacter (1987) and Squire (1992) all argued that the human memorial system can be fruitfully viewed as though there were two distinct information-processing systems - one declarative or explicit, the other procedural or implicit. The explicit system was theorized to include declarative, conscious knowledge of episodes, facts, and events, whereas the implicit system was assumed to be operating largely outside of consciousness and to include implicit learning and memory, conditioning, and learning various skills and habits (sensorimotor learning). Although this distinction is probably a useful one in that it draws attention to the ways in which implicit and explicit functions can be dissociated, it is probably not the best stance to take from a functionalist point of view. As Reber (1993) argued, we need to be wary of falling into a "polarity" fallacy in which we treat two distinguishable systems that lie at the poles of a continuum as though they were onto-logically separate and distinct. It is almost certainly the case that virtually everything interesting that human beings entails a delicate synergy between the implicit and the explicit, the conscious and declarative, and the unconscious and procedural. If the implicit and the explicit systems ultimately are shown to be based on neuroanatomically distinct structures (as we suspect will be done), it will still be virtually impossible to find functionally pure instantiations of them.
In addition, Reber (i992a,b) argued that because human consciousness and its accompanying functions are late arrivals on the evolutionary scene, there should be particular patterns of dissociation between these two systems. The key predictions of the model for this discussion are:
(a) Storage and retrieval systems that serve the implicit system should be more robust and relatively undisturbed by insult and injury that compromise explicit functions.
(b) There should be relatively little in the way of developmental and life-span changes in implicit compared with explicit functions. This two-system model has garnered significant support over the past decade (see Reber, Allen, & Reber, 1999 and Squire & Knowlton, 2000 for reviews).
Taken together, this literature paints a clear picture. Human memory has distinct systems with distinct evolutionary histories and separate, although only partly understood, neurological underpinnings that map, on one hand, into conscious, subjective experience and, on the other, into a nexus of encoding, storage, and retrieval systems that function largely independently of awareness and consciousness. However, this picture is still incomplete, and appreciating the manner in which it operates in complex human thinking requires a deeper look at the topic of learning - specifically implicit learning in which knowledge about the complexities of the environment is acquired without benefit of consciously controlled processes.
Implicit learning is the process whereby organisms acquire knowledge about the regularities of complex environments without intending to do so and largely independently of conscious awareness of the nature of what was learned (Stadler & Frensch, 1998; Reber, 1967; Reber, 1993). The complex environments include virtually every facet of human life, including language learning, trait knowledge, categorization, acculturation, and the development of aesthetic preferences. The claim we are making is that people extract information about the world more often than they are aware and that this knowledge exists in a tacit form, influencing thought and behavior while itself remaining mostly concealed from conscious awareness.
implicit learning in infants
By the second month of life, infants can already distinguish between utterances spoken in their native language and those spoken in foreign languages. Infants can do this although they don't understand what the sentences mean in either language. Interestingly, this effect disappears when the sentences are presented backwards (Dehaene-Lambertz, & Houston, 1998; Mehler et al., 1988; Ramus et al., 1999). The implications here are that, despite not understanding the sentences backwards or forwards, the infants have become attuned to the natural flow of language. This natural flow is violated when the sentence is reversed. This sensitivity to the structure of linguistic sounds which seems to be the first stage of language acquisition, takes place implicitly and recruits brain regions similar to those of adults as shown by fMRI studies (Dehaene-Lambertz, Dehaene, & Hertz-Pannier, 2002).
Within a surprisingly short time, infants extract the phonetic regularities of their linguistic surroundings and can differentiate between sound sequences that are well formed and those that are not. The backwards sentences sound ill-formed to the infants because their sequential structure is discoordinate with the infant's experience and therefore seem as ill-formed as sentences in a foreign language. Note the similarity of this to the standard implicit learning procedure in Artificial Grammar studies discussed below.
These kinds of effects are not restricted to natural languages. Rovee-Collier and her colleagues (see Rovee-Collier, 1 990 for a review) report that infants rapidly pick up the relationships between their own motor actions and the impact that they have on the external world. Haith, Wentworth, and Can-field (1 993) showed that babies make anticipatory eye movements to regularities in the spatial patterns of visual displays. Saffran and her colleagues reported that infants as young as 8 months show a similar sensitivity to the arbitrary statistical nature of auditory patterns and can learn the rules governing artificial word segmentation (Saffran, & Aslin, & Newport, 1996). Interestingly, in Saffran's studies, the infants performed as well as a group of adults, a result that supports Re-ber's (1992b) prediction that implicit learning systems are present at a very early age and undergo little developmental change. Similarly, Gomez and Gerken (1999, 2000), using the AG learning procedure, showed that not only do one-year-olds learn the structural characteristics of these rather complex systems, they also transfer this knowledge to novel stimulus domains.
To date, this research has been restricted largely to sensorimotor, perceptual, and cognitive tasks. Surprisingly, little empirical work has been carried out on behaviors that are more reflective of social learning. However, given the existing database, we suspect that when processes of socialization are examined from this perspective, they will reveal a parallel set of operations in which infants gradually become inculcated with the social mores and ethical codes of their culture without conscious awareness of what has been learned and with little in the way of top-down control of the process.
implicit learning in adults
In recent years, a rather impressive array of specific tasks have been discovered to have dissociative elements in that either direct and indirect tests distinguish between im plicit and explicit memorial systems, or various patient populations manifest distinct patterns of loss of explicit acquisitional functions while maintaining those based on the implicit processes. Included here are studies on motor learning (P.J. Reber & Squire, 1998), AG learning (Knowlton & Squire, 1994, 1996; Reber, 1967, 1989), category learning (Knowlton & Squire, 1993; Squire & Knowlton, 1 996), Pavlovian conditioning (Daum & Ackerman, 1994; Gabrieli et al., 1995), decision making in social settings (Lewicki, 1986a; any of several contributions to Uleman & Bargh, 1989), the sequential reaction time task (see Hsiao & Reber, 1 998 for a review), the hidden covariation task (Lewicki, 1986b), preference formation (Gordon & Holyoak, 1983; Manza, Zizak, & Reber, 1998), the production control task (Berry & Broadbent, 1988), and dot pattern classification (P.J. Reber, Stark, & Squire, 1 998). The various chapters in Stadler and Frensch's (1998) edited volume Handbook of Implicit Learning are a good resource for a more detailed discussion.
These many reports are supplemented by additional findings that show that patients with damage to primary visual cortex learn to respond to objects in their blind fields (Weiskrantz, 1986), prosopagnosiacs who cannot consciously recognize the faces of family members show virtually normal implicit facial memory (De Haan, Young, & Newcombe, 1991), patients with neglect respond to the meaning of stimuli that they are unaware of processing (Berti & Rizzolatti, 1992), amnesic patients show improvement in solving problems (Winter et al., 2001) and learn to operate complex equipment (Glisky, Schacter, & Tulving, 1986) despite no conscious memory of the earlier training phases of the studies. Issues of the mechanisms underlying disordered thought are pursued in detail elsewhere in this volume by Bachman and Cannon (see Chap. 21).
The model that has emerged from this literature characterizes implicit learning as a mechanism the primary function of which is to monitor the environment for reliable relationships between events and to encode those patterns of covariation. In all likelihood, the underlying neurological mechanisms are diffuse neural nets that are linked to the modality of input of the stimulus display (Ungerleider, 1995). The underlying representations that are established are probably not as flexible or abstract as those that are under conscious control simply because the top-down modulation that comes with consciousness allows for deliberative shifts in representation and use of knowledge. However, this issue is a highly contentious one and we have more to say on it subsequently.
In addition, implicit acquisitional mechanisms appear early in life, well before conscious awareness has developed. They show relatively little change over the life span compared with explicit cognitive functions (Howard & Howard, 1 992, 1 997) and relatively little in the way of individual-to-individual variation (Reber & Allen, 2000). As noted previously in several places, the implicit system demonstrates a rather remarkable robustness and continues to function effectively in the face of a wide variety of neurological and psychiatric disorders that severely compromise functions based on explicit, consciously modulated mechanisms. It seems clear that the implicit system is the critical mental component that enables the infant and child to learn to navigate the world. Virtually all the essential knowledge of the perceptual, sensorimotor, linguistic and social patterns, that make up the environment and eventually become the epistemic foundations of adulthood is acquired through this nondeclarative, procedural mechanism. This is, indeed, how we learn about the world around us. For further explorations of this and related developmental mechanisms, see Halford (this volume).
Although these aspects of implicit thought are fairly well established, there are two issues that remain deeply problematical and need to be addressed: First, are (or better, perhaps, can) these implicitly formed representations be regarded as abstract? Second, what role might they play in complex cognitive processes such as problem solving that have been generally regarded as largely, if not completely, explicit and under conscious control?
Was this article helpful?