Max Velmans, Dept of Psychology, Goldsmiths, University of London, New Cross, London SE14 6NW; email m.velmans@gold.ac.uk;

British Journal of Psychology 90(4), 543-566.

The study of preconscious versus conscious processing has an extensive history in cognitive psychology, dating back to the writings of William James. Much of the experimental work on this issue has focused on perception, conceived of as input analysis, and on the relation of consciousness to attentional processing. The present paper examines when input analysis becomes conscious from the perspectives of cognitive modelling, methodology, and a more detailed understanding of what is meant by "conscious processing." Current evidence suggests that perception becomes conscious at a late-arising stage of focal-attentive processing concerned with information integration and dissemination. Reliable criteria for determining when perception becomes conscious combine the evidence of "first-person," phenomenological reports with "third-person" functional dissociations between preconscious and conscious processing. There are three, distinct senses in which a process may be said to be "conscious." It might be "conscious" (a) in the sense that one is conscious of the process, (b) in the sense that the operation of the process is accompanied by consciousness (of its results) and (c) in the sense that consciousness enters into or causally influences the process. Consciousness of familiar stimuli, rather than entering into input analysis, appears to follow it, in human information processing. Processes closely associated with the appearance of consciousness such as information integration and dissemination appear to operate unconsciously. Consequently, perception appears to be "conscious" only in sense (b).

 Keywords. Perception, conscious, preconscious, automatic, flexible, novel, unconscious, information processing, criteria, methodology, causal role.

It is common to think of perception as being conscious. Reading this sentence is conscious in the sense that the sentence enters conscious awareness. However, deciding when a mental process is conscious is not as easy as it seems. Reading, for example, requires the identification of visual patterns, involving input analysis. To decide when input analysis becomes conscious we need

1. an understanding of where in the information processing sequence (from input to output) consciousness of the input arises.

2. criteria for deciding when a process is conscious and when it is not

3. an understanding of the different senses in which a process may be said to be "conscious".

Subjectively, we seem to be immediately aware of what we attend to. However, experiments on the timing of conscious awareness suggest that consciousness of input does not arise until at least 200 milliseconds (msec) after stimuli arrive at the cortical surface (see Libet, 1996 for a review). Libet, Wright, Feinstein & Pearl (1979), for example, found that direct microelectrode stimulation of the somatosensory cortex required a pulse train of at least 200 msec duration before any conscious awareness of the stimulus was reported (pulse trains 10% shorter than this were not subjectively experienced). Libet et al. also found that tactile stimuli applied to a finger were masked (prevented from entering consciousness) by microelectrode stimulation of the somatosensory cortex applied up to 200 msec after the arrival of the tactile stimuli. On the grounds that one cannot prevent a stimulus from entering consciousness after it has done so, they concluded that at least 200 msec of processing are required to produce neural conditions adequate to support consciousness. The reason we do not experience any mismatch between experienced and actual stimulus arrival time appears to be that the brain compensates for its own processing time in the representations of input that it constructs. Libet (1996) reviews evidence that the brain records the time of tactile stimulus arrival with a "time marker" in the form of an early evoked potential at the somatosensory cortical surface. However microelectrode stimuli applied directly to cortical areas such as the medial lemniscus (LM) do not produce such early time markers. By contrasting the subjective timing of stimuli with and without such time markers he found that the former but not the latter are subjectively referred "backwards in time" (to the time of occurrence of t he marker); for example, tactile stimuli applied 100 msec after the LM cortical stimuli appeared, subjectively, to precede them (by around 100 msec). Consequently such tactile stimuli do not appear to be subjectively delayed (by 200 msec).

These surprising findings and conclusions about "subjective referral" have not gone unchallenged (see for example the open peer review accompanying Libet, 1985). However, the suggestion that consciousness of input is preceded by a period of preconscious processing is broadly supported by cognitive research - and a common estimate of preconscious processing time is in the order of 250 msec (e.g. Neeley, 1977; Posner & Snyder, 1975). In information processing terms there is much to do before one can identify a stimulus, for example stimuli must be transformed into neural code, analyzed, and matched to memory traces before they can be identified. Complex stimuli such as sentences also require syntactic and semantic analysis and an interpretation of meaning in the light of prior verbal context, current physical context, and accumulated "global knowledge" of the world. Such processing requires time. It makes evolutionary sense for mental models of stimulus arrival time to compensate for the processing time required to make those stimuli conscious.

Given that a period of preconscious processing precedes the entry of input into conscious awareness, what is the nature of preconscious processing and how does it differ from "conscious" processing? Efforts to clarify the functional differences between preconscious and conscious processing date back to William James. According to James (1890), stimuli which enter consciousness are at the focus of attention, having been selected from competing stimuli to enable effective interaction with the world. The current contents of consciousness also define the "psychological present" and are contained in "primary memory" (a form of short-term working store). The contents of "secondary memory" (a long-term memory store) define the "psychological past," and while they remain in secondary memory they are unconscious. These ideas, developed around 100 years ago are still the focus of much psychological research.

For example, following the seminal investigations of the "cocktail party problem" by Cherry (1953) and the development of a selective attention model by Broadbent (1958), the transition from the preconscious to the conscious state, and James's linkage of conscious content to what is at the focus of attention became major topics of enquiry. This is illustrated in a cocktail party situation, where the conversation one attends to enters consciousness, while the competing conversations seem to form a relatively undifferentiated background noise. However, all speech waveforms arrive concurrently at the eardrum. So how does the brain select the required message from such complex auditory stimuli? At the same time, if one's name is mentioned by someone across the room, one's attention might switch, suggesting that, to some extent, even non-attended messages are analyzed - but to what extent?

The extent of preconscious analysis.

Early experiments by Cherry (1953) and Broadbent (1958) used a shadowing task in which subjects were required to attend to and repeat a message presented to one ear while another message was simultaneously presented to the other, non-attended ear. After the task, subjects were required to report what they could remember of the non-attended message. Early findings indicated that subjects could not report the identity or meaning of stimuli on non-attended channels, although they could report some of their physical features. On the basis of such findings Broadbent (1958) proposed an "early selection" model in which all input stimuli receive a physical analysis in an automatic, parallel, preattentive fashion. Only stimuli which are selected for more detailed focal attention receive an analysis for meaning, update long-term memory and enter consciousness.

However later experiments demonstrated that the meaning of non-attended stimuli affected the processing of attended stimuli, indicating that analysis of meaning of non-attended stimuli can take place without conscious awareness of them or subsequent recall. Lewis (1970) for example, demonstrated that the shadowing latency of attended-to words was influenced by the simultaneous appearance of semantically related words in the non-attended ear; however, shadowing latency was not affected by unrelated words. Corteen & Wood (1972) found that changes in GSR, which accompanied target words conditioned to electric shocks, continued when those target words appeared in the non-attended ear although subjects were unable to identify the words themselves. This occurred also with words which were semantically related to the conditioned word. Similar findings have been obtained by Corteen and Dunn, 1974; Forster & Govier, 1978; Von Wright, Anderson & Stenman, 1975; but not by Wardlaw & Kroll, 1976. Mackay (1973) also found that disambiguating words in the non-attended ear biased the meanings of ambiguous sentences in the attended ear which subjects were required to paraphrase. For example, the cue word "money" would encourage subjects to paraphrase "They threw stones towards the bank" as "They threw stones towards the savings and loan association" whereas the cue word "water" would encourage the paraphrase "They threw stones towards the side of the river."

Such studies have often been cited as evidence that under some circumstances the preconscious processing of stimuli outside the focus of attention includes the analysis of meaning. Indirect effects of the kind mentioned above, however, appear to be sensitive to small perturbations in experimental design. Consequently, in recent years the interpretation of these studies has become controversial.

The Lewis (1970) study, for example, was repeated by Treisman, Squire & Green (1974), using a list of 10 dichotically presented word pairs. They replicated Lewis's findings for related pairs when these occurred early in the list (position 3) but not when they occurred later in the list (position 7). According to Holender (1986) this suggests that subjects who are accustomed to the shadowing task no longer analyze the nonselected message - in which case Lewis's findings may have resulted from subjects, unaccustomed to shadowing, switching their focal attention to the nonselected ear. Alternatively, it may be that once subjects become accustomed to the shadowing task, their speed of responding to shadowed words is no longer affected by the results of analyses that continue to take place on the nonselected ear.

The various replications of Corteen & Wood's (1972) study indicate that their results are reliable. Again, however, the findings are subject to more than one interpretation. Dawson & Schell (1982), in a similar study, found that if subjects were told beforehand that they would be required to name the conditioned word in the nonselected ear, they could sometimes (but not always) do so. According to Holender (1986), this suggests that subjects had been momentarily aware of the nonselected, conditioned words in the earlier studies - a possibility admitted by Corteen (1986). Dawson & Schell's procedure, however, required subjects to divide their attention between the selected and nonselected ear. It is therefore not comparable to earlier studies on where subjects were simply asked to shadow the message in the attended ear. Nevertheless, their finding highlights the difficulty of assessing the awareness of nonselected words in dichotic listening studies.

Mackay's (1973) finding that single, concurrent words in the nonselected ear disambiguated sentences in the selected ear was replicated by Newstead & Dennis (1979). However, if the critical words were embedded in sentences, the disambiguating effect did not occur. Again, according to Holender (1986), this implies that isolated words in the nonselected ear momentarily attract attention, leading to focal-attentive switching and conscious identification - an effect which does not occur if the words form part of a relatively continuous stimulus presented to the nonselected ear. Holender accordingly argues that findings from dichotic listening experiments do not demonstrate semantic analysis without conscious identification.

However, focal-attentive switching cannot account for the evidence of preconscious semantic analysis (in nonselected channels) found by Groeger (1984a, 1984b, 1988). Groeger demonstrated that the effects of disambiguating words in the non-attended ear were different if they were above threshold (consciously detectable) versus below threshold. For example, in one experiment subjects were asked to complete the sentence "She looked ___ in her new coat" with one of two completion words, "smug" or "cosy". Simultaneous with the attended sentence the word "snug" was presented to the nonselected ear (a) at threshold, or (b) below it. With "snug" presented at threshold, subjects tended to choose "smug" which could be explained by subjects becoming momentarily aware of the physical form of the cue. With "snug" presented below threshold, subjects tended to choose "cosy" indicating semantic analysis of the cue word without accompanying awareness.

One cannot assume from the findings above that semantic analysis of nonselected messages always takes place in dichotic listening studies and it is often difficult to be certain that subjects have no awareness of stimuli presented to the nonselected ear. Nevertheless, such studies have produced diverse evidence of semantic analysis of nonselected words, under conditions where subjects claim to have no awareness of those words and are unable to report them afterwards. This suggests that under some circumstances a preliminary analysis for meaning can take place outside the focus of attention, without reportable consciousness.

Such findings have been used to support a "late selection" model (Deutsch & Deutsch, 1963; Norman, 1969) in which the meaning of all familiar input stimuli are preattentively analyzed. As Norman pointed out, unless one does analyze the meaning and significance of input stimuli on non-attended channels it would be difficult to judge when it is pertinent to switch one's focal attention to them. If so, the analysis of meaning (of simple familiar stimuli) may not require focal attention, or entry of the stimuli into consciousness.

Posner & Snyder (1975) extended this late-selection model into a two-process model in which preattentive, preconscious processing is thought of as a fast, automatic, spreading activation in the central nervous system. This activates not only memory traces of a given input stimulus but also related traces that share some of its features. For example, reading the word "DOCTOR" also activates or "primes" semantically related features in the word "NURSE," making the latter easier to recognise (Meyer, Schvaneveldt & Ruddy, 1975). However, this process has no effect on unrelated traces (for example, "DOCTOR" does not prime "BREAD"). By contrast, attentional processing occurs only after such spreading activation, it is relatively slow and serial in nature, and it cannot operate without intention and awareness. This process not only activates the traces of related stimuli but also inhibits the activation of unrelated stimuli (making them harder to recognise). Evidence for this complex theory was gathered by Neely (1977). Evidence for the preconscious, parallel activation of traces which share features with an input stimulus, followed by selection of the most pertinent traces (and inhibition of nonpertinent traces) has also been found in studies of speech perception (Pynte, Do & Scampa, 1984; Swinney, 1979,1982) and visual masking - a procedure where visual stimuli are prevented from reaching consciousness by the presentation of a subsequent visual stimulus or "mask" (see discussion of Marcel, 1980, and Greenwald, Klinger & Liu, 1989, below).

However, it seemed unlikely that preattentive input analysis is limited to spreading activation, and theories of preattentive and focal-attentive processing continued to develop. For example, La Berge (1981) and Kahneman & Treisman (1984) pointed out that different forms of attention may have to be devoted to different stages of input analysis. Processing resources may be devoted to the identification of physical features if one is searching for a target input stimulus, but other resources may be required to integrate the set of features at the location found by the search. In addition, if any consequent action is to follow input analysis, its results need to be disseminated to other processing modules (see also Baars, 1988; Baars & McGovern, 1996). Posner, DiGirolama & Fernandez-Duque (1997) further subdivide attention into three functions: orienting to sensory stimuli, executing control (including target detection and response selection), and maintaining an alert state.

While the details of focal-attentive processing are still being researched (see, for example, Styles, 1997; and the whole of Consciousness and Cognition, 1997, vol 6, no 2/3) there appears to be some consensus within the experimental literature that input stimuli in different channels are preattentively analyzed in a fast, parallel, automatic, preconscious fashion, with little mutual interference, up to the point where each stimulus is matched to its previous traces in long-term memory, enabling a simple analysis of its meaning or significance (Norman 1969; Posner 1978; Posner & Boies 1971; Shiffrin & Schneider 1977). Whether non-attended processing can extend to more complex analyses is uncertain. Underwood (1977) for example found that the effect of non-attended words on attended words in a shadowing task was not influenced by placing the non-attended words in a sentence context. This suggested that without attention there may only be limited integration of words into sentences. In a study which investigated the effects of visual, masked primes on the speed at which subjects could evaluate visually presented target words as "positive" or "negative", Greenwald and Liu (1985) found that single, subliminal words primed evaluatively congruent meanings, but two-word phrases did not. That is, a negative prime speeded the subject's response to a negative target, but not to a positive target (and vice versa). As one would expect from single word priming, a two-word prime such as "enemy loses" speeded the response to negative targets, in spite of the fact that the phrase as a whole is evaluatively positive. Greenwald (1992) called this apparent upper limit on the complexity of preattentive, preconscious processing the "two-word challenge."

It would be misleading to suggest that all the evidence relating to preattentive and focal-attentive processing fits into this relatively neat picture. Nevertheless, the transition from processing single, familiar words to processing more complex or novel input stimuli is often thought to mark the transition from preattentive to focal-attentive processing. The latter is thought to be more flexible, relatively slow, serial, voluntary, limited in capacity, and conscious.

Whatever the precise differences between pre- and focal-attentive processing turn out to be, few cognitive theorists would disagree with William James that there is a close association between "attention" and "consciousness." Indeed, many psychologists have explicitly or tacitly assumed that "preconscious" processing is identical to "pre-attentive" processing, whereas "conscious" processing is identical to "focal-attentive" processing (e.g. Baars, 1991; Mandler, 1975,1985,1991; Merikle & Joordans, 1997; Miller, 1987). However, these assumptions need to be treated with caution. As Kahneman & Treisman (1984) note, the question of how attentional resources are allocated is in principle distinguishable from the question of what is or is not conscious. A close association of consciousness with focal-attention does not establish their ontological identity (Velmans, 1998). Such caveats aside, we can still ask, "What is it about attentional processing that relates most closely to consciousness?"

The functional correlates of consciousness.

Clues about the functional correlates of consciousness are offered by situations where attentional processing is partially dissociated from consciousness, for example where subjects focus their attention on an input stimulus but consciousness of the stimulus does not arise. It seems reasonable to assume in such situations that some aspects of attentional processing are operating but other aspects (associated with consciousness) are not (cf Velmans, 1991a for a review). A classic example occurs in "blindsight" produced by striate cortex lesions (Weiskrantz, 1986). Blindsighted subjects can direct their attention to an input stimulus, identify some of its properties and make appropriate identification responses, but are unable to experience the stimulus to which they attend. Such subjects, however, need to be forced to make decisions about stimuli they believe they cannot see, indicating that information about the stimulus is not readily available to all parts of the information processing system. Similarly, Marcel (1986) found that blindsighted patients make no attempt to grasp a glass of water in their blind field even when thirsty, suggesting that information about the input remains dissociated from systems subserving voluntary control.  

Partial dissociations also occur in implicit learning and memory studies, where information which is not present to consciousness at the time of learning (according to subjective reports) may update long-term memory and influence performance, but may not be available for later explicit recognition and recall (Gardiner, 1996; Berry and Dienes, 1993; Reber, 1993; Schacter, 1992).

Another dissociation of attention from consciousness occurs in hypnotic analgesia, where patients are induced to direct their attention away from the painful stimulus. However, during hypnosis the patient may be told that a hidden observer will continue to monitor everything that is happening although the patient will experience no pain (Hilgard, 1986). In subsequent surgery, the awake patient may report no experience of pain and this may be accompanied by an absence of physiological indices of pain along with reduced bleeding and salivation (Oakley & Eames, 1985) indicating that information about the painful input is not generally available to other parts of the system. But the hidden observer continues to attend to the pain and to enter information about it into memory. After surgery, with the subject still under hypnosis, one can ask to speak to the "hidden observer" in which case it gives a vivid report of the pain it has experienced.

What such findings demonstrate is that partial dissociations of attentional processing from consciousness result in different forms of information "encapsulation." As Kahneman & Treisman (1984) suggest, the dissemination of currently processed information to other information processing modules may be one of the functions of focal-attentive processing, enabling greater resources to be devoted to the input and allowing the system as a whole to respond to input at the focus of attention in a coherent, global way. This would account for the greater flexibility and sophistication of "conscious", focal-attentive processing (compared with "preconscious", preattentive processing). When information dissemination is disrupted, disruption of consciousness (of that information) also occurs. This would suggest that input analysis becomes conscious when its products are being disseminated - a late-arising stage of focal attentive processing.

Other conditions for consciousness, specifiable in information processing terms, also need to be met. For example, disseminated information needs to be sufficiently well-integrated to support an integrated conscious experience (the "binding problem"). But in the sequence of attentional processes, the information dissemination stage appears central. Through an extensive review of the contrasts between conscious and nonconscious processes, Baars (1988) and Baars & McGovern (1996) come to similar conclusions (via a different route), although the term they use for "information dissemination" is "broadcasting."

It should be apparent that the broad acceptance of preconscious semantic processing of simple, familiar stimuli followed a change in methodology from explicit to implicit tests. Early studies assessed the extent of processing in non-attended ears by asking subjects to report what they could remember of the non-attended stimuli after the shadowing task was complete. Later experiments assessed the extent of non-attended processing indirectly, by effects on GSR or on the response to messages in the attended ear. It is clear that the latter technique provides a more sensitive measure of non-attended processing, apparently dissociating semantic processing from consciousness.

Sceptics, however, remained unconvinced on methodological grounds. According to Holender (1986) the inability to report a word in a nonshadowed ear does not guarantee that it has not been attended to or entered consciousness. For example, in the shadowing tasks described above subjects might switch attention to the nonshadowed ear momentarily, become conscious of the stimulus and subsequently forget it (shadowing would interfere with recall). Similar doubts about the reliability of subjective reports had been raised about earlier experiments on the influence of subliminal stimuli by Eriksen (1956, 1958, 1960). Both Eriksen and Holender argued for the replacement of "subjective" reports of conscious awareness with more "objective" discrimination measures. That is, a stimulus can be said to be "conscious" only if one can distinguish it from competing stimuli in a discrimination task. They claimed that, using this criterion, there is no convincing evidence for meaning analysis without consciousness.

Unfortunately, such a criterion would rule out preconscious identification of any stimulus property by fiat, as the means for establishing such identifications is precisely the ability to make a discriminative response to the relevant property in the absence of any reportable awareness. Kihlstrom (1996) concludes that for this reason alone Holender's and Eriksen's suggestion should have been rejected out of hand. Their criterion is also counterintuitive. As early as 1885, Pierce & Jastrow demonstrated that subjects with zero confidence about which of two stimuli was brighter or heavier, had better than chance performance when forced to guess. That is, information about brightness and weight differences was available, but not to the subjects' conscious awareness.

In fact, the evidence that stimulus meaning can affect a subject's response in the absence of any reportable awareness (of that stimulus) is extensive (see Dixon, 1971, 1981; Kihlstrom, 1996; Merikle & Daneman, 1998; Velmans, 1991a for reviews), and as shown by Groeger (1984, 1988) not all the effects in dichotic listening experiments can be explained by focal-attentive switching with accompanying consciousness (see above). Greenwald, Klinger & Liu (1989) also demonstrated preconscious analysis of meaning in a paradigm that even appears to meet Holender's overly-stringent criteria. In their experiment, subjects were asked to judge whether a visually presented target word had a positive or negative evaluative meaning (as in Greenwald & Liu, 1985 discussed above). Target words were preceded by either a positive, negative, or neutral prime. To prevent conscious awareness of the prime, a dichoptic, backward-masking technique was used, in which the prime is presented to one eye and a pattern mask to the other eye (this produces central rather than peripheral masking). Prevention of conscious awareness is usually assessed by subjects' inability to report the presence of a prime. But in this experiment, following Holender's criteria, subjects were asked to make a discriminative response, in the form of a judgement about whether the prime appeared to the left or right of a central fixation point. In confirmation of earlier findings, they found that reaction times to the target words were speeded when preceded by evaluatively congruent primes, and slowed by incongruent primes, even when subjects were unable to detect the location in which the primes had appeared.

An understanding of how a "subjective" criterion of conscious awareness might relate to "objective" measures of discriminability has been developed by Merikle (1982, 1992) and his colleagues (e.g. Cheesman & Merikle, 1984, 1986; Reingold & Merikle, 1993). Merickle defined the subjective threshold as the point at which a subject's confidence in their ability to discriminate drops to zero. By contrast, the objective threshold is the point where actual discrimination performance is at chance levels. As noted above there are various situations in which the objective threshold is lower than the subjective threshold, although intuitively, it is the subjective threshold which corresponds to the presence or absence of conscious experience.

Merikle and his colleagues demonstrated the importance of the subjective threshold to cognitive theory, by presenting evidence of qualitative differences in subjects' performance at the subjective versus objective threshold, indicating that these two thresholds correspond to functional differences which can be specified in information processing terms. For example, in masked priming experiments, the prime-mask interval at which subjects claim they cannot see the prime (the "subjective" threshold) is generally longer than the interval at which subjects cannot make a better than chance judgement about the prime (the "objective" threshold). Qualitative differences in performance at these two prime-mask intervals have been demonstrated by Cheesman & Merikle (1984,1986), Dagenbach, Carr & Wilhelmsen (1989), Forster & Davis (1984), and Marcel (1980, 1983 experiment 5). Marcel (1980), for example, used polysemous primes (with multiple meanings) embedded within three word sequences, each beginning with a context word followed by the polysemous prime, and ending with a target word (for example, "save - bank - money," or "save - bank - river"). In one condition, the primes were presented unmasked, at an exposure duration of 500 msec. In this suprathreshold condition, selective priming of the target occurred; for example, in the above three word sequences, the cued meaning of the prime "bank" facilitated the response to the target "money," but not "river." In the other condition, the primes were presented for 10 msec. and followed by a pattern mask at an interval set to produce chance detection of the prime. With masked primes, unselective priming occurred; for example, in the above sequences, the word "bank" primed both "money" and "river."

It should be apparent that it is particularly important to isolate such functional differences between preconscious and conscious processing at the preconscious/conscious borderline. Stimuli that are below the "objective" threshold, or confidently above the "subjective" threshold are relatively easy to categorize. Holender (1986) for example, directs his critique at the supposed preconscious nature of stimuli in non-attended channels. If subjects can discriminate these stimuli in implicit tests, how can one be certain that the stimuli were not briefly conscious - but so fleeting and vague that subjects cannot remember them, even immediately afterwards?

In fact we cannot be certain. But this argument overlooks the fact that the same thing can be said of liminal attended-to stimuli. If, for example, we lower the intensity of an auditory stimulus to the point that subjects claim not to be able to hear it, how do we know for certain that it was not so fleetingly audible that they cannot remember it even immediately afterwards?

In practice, hearing thresholds are determined by reliance on subjective judgements. Likewise, it is not clear how one can avoid subjective judgements when deciding the point of transition from preconscious to conscious states in shadowing, visual masking and other similar tasks. As Baars & McGovern (1996) note, if one cannot describe, recall, report, think about or make decisions about some input stimulus then, functionally, it is not conscious. With little theoretical cost, "consciousness of a stimulus", in these situations, can be taken to mean "reportable consciousness."

Baars & McGovern (1996) take a similarly pragmatic view:

"Conscious representations can be operationally defined as those which:

(a) are claimed by people to be conscious; and which

(b) can be reported and acted upon,

(c) with verifiable accuracy,

(d) under optimal reporting conditions." (p66)

Criteria (a) and (b) are clearly necessary and sufficient conditions for reportable consciousness. Given the private nature of conscious experience, criterion (c) is a recommended, added condition in situations where independent verification is possible, for example in tests of veridical perception and memory (where there is public access to the test stimuli). Nearly all the studies considered in the present paper are of this kind.

Criterion (c), however, may be overly stringent in situations where veridical perception or memory is not of central interest. In studies of visual illusions (such as the Muller-Leyer) for example, it is standard practice to measure the size of the illusion by asking subjects to adjust a comparison stimulus until it appears the same as the test stimulus. The difference between the actual dimensions of the test stimulus and the adjusted dimensions of the comparison stimulus measures the size of the illusion. While the reliability of the procedure can be assessed and improved by repeated measures, and the generalisability of the illusion assessed by testing different subject groups, there is a sense in which the size of the illusion on a given occasion, for a given subject does not have a verifiable accuracy which can be independently assessed (it is up to the subject to decide when the comparison stimulus matches the test stimulus). The absence of some "objective" measure of the size of illusions has not prevented their extensive investigation in experimental psychology.

Criterion (d) (optimal reporting conditions) has to do with good experimental practice. Baars & McGovern suggest a minimum delay between the conscious event and the report and freedom from distraction - which are standard practice in the study of perception, memory, problem-solving, imagery and so on. A more extensive discussion of good practice in the use of subjective reports is given in Ericcson & Simon (1984).

By contrast,

"..mental events can be defined as unconscious for practical purposes if:

(a) Their presence can be verified (through facilitation of other observable tasks, for example); although

(b) they are not claimed to be conscious;

(c) and they cannot be voluntarily reported, operated upon, or avoided;

(d) even under optimal reporting conditions." (Baars & McGovern, p67)

For events to be judged functionally unconscious all these conditions have to be met. As noted above, independent verification is recommended, but not a necessary condition for conscious events. For unconscious events, however, if the subject cannot voluntarily report, operate on, or avoid the event, the verification of its existence through some other means becomes a necessary condition.

In sum, there seems little doubt that criteria to decide the transition from preconscious to conscious processing must, in the initial instance, rely on subjects' "first-person" experience. This requires some form of subjective judgement to be made which can be externalized, for example translated into a verbal report, a confidence measure, a yes/no response and so on. Because an experimenter cannot directly access a subject's experience, no "third-person" measure, unsupported by subjective evidence, can be relied on (cf Libet, 1996; Marks, this issue). That said, external correlates of conscious experience, for example functional dissociations between "subjective" and "objective" thresholds, can provide greater confidence in subjective reports (see also Valentine, this issue); if they turn out to be reliable, they may eventually become accepted indicators of subjective effects; in principle, they also allow subjective experience to be related to its functional correlates, specified in information processing terms.

IN WHAT SENSE IS INPUT ANALYSIS "CONSCIOUS"?

As Kihlstrom (1996) notes, psychology now embraces a distinction between automatic (unconscious) processing, and controlled (conscious) processing:

"Whether they are innate or routinized by extensive practice, automatic processes are inevitably engaged by specific inputs, independent of any intentionality on the part of the subject, and they cannot be controlled or terminated before they have run their course. We have no conscious awareness of their operation, and we have little or no awareness of the information which they process. Thus automaticity represents unconscious processing in the strict sense of the term: we have no introspective access to automatic procedures, or their operations; these can only be known to us indirectly, by inference." (Kihlstrom, 1996, p23).

It will be apparent from the review above that input analysis of information on non-attended channels is thought to be unconscious or preconscious in this sense. By contrast, input analysis on attended channels is thought to be voluntary, controlled, and conscious. But in what sense is the analysis of attended input "voluntary, controlled and conscious"?

In ordinary usage, consciousness refers to phenomenal consciousness, and the contents of consciousness are simply the events we experience. This usage is also common in psychology (cf readings in Velmans, 1996a); in psychophysics for example subjects are typically required to report on whether or not they experience a stimulus (in setting thresholds), or experience the difference between two stimuli (in establishing difference limens) and so on. In such experiments conscious experiences are first-person psychological data, which form the basis for subjects' responses or verbal reports. In early theories of selective attention such as Broadbent (1958) and Norman (1969) adjectival or adverbial use of the term "conscious" suggest that it is also a property of a process, for example a property of the LCDC, focal-attentive processing, or primary memory. This associates conscious experiences with certain forms of processing, but entails no commitment about whether consciousness as such actually does anything in the activities of the brain. Consciousness might, for example, be an epiphenomenal property which accompanies, emerges from or is produced by certain forms of processing.

Many theorists, however, have taken a bolder position. In 1962, for example, George Miller claimed that "the selective function of consciousness and the limited span of attention are complementary ways of talking about one and the same thing" (see Miller, 1987, p 65), and from around 1970, phenomenal consciousness began to be redefined in information processing terms. Posner and Warren (1972) asserted that the use of a limited capacity central processing system "... becomes the central definition of a conscious process and its non-use is what is meant by a process being automatic".

Comparisons were also made between consciousness, the operations of the limited capacity central processor and an "executive monitor programme" sometimes used in large computing installations to allocate processing resources efficiently to the many simultaneous tasks in which the system is engaged (Shallice 1972; Bower, 1972; Bjork, 1975). Bjork (1975) for example, outlined a model of human information processing in which

" ... an explicit central processor is proposed as a kind of executive consciousness that controls and governs the system; without the involvement of the central processor, nothing happens in the system beyond the formation of input traces".

Theories of consciousness that give it selective functions (in attentional processing), or identify it with a "central processor" or "central executive system" treat it as a distinct, functional "module" which clearly does something useful in the activities of brain. For example, if nothing happens without it other than the formation of input traces, consciousness must be necessary for the analysis of complex, novel stimulus combinations which occur, for example, in reading or the perception of connected speech. This would be consistent with the evidence that preconscious analysis (in nonattended channels) may be limited to the meanings of individual words (see Kihlstrom, 1996; Greenfield, 1992; Underwood, 1991).

In this connection, Mandler (1975) has argued that

".. relational processes operate primarily if not exclusively on conscious content. In addition to choice, these include evaluation, comparison, grouping, categorization and serial ordering. In short, practically all novel relational orderings require that the events to be ordered must be simultaneously present in the conscious field ... Once relations have been established and stored, subsequent evaluations are frequently unconscious".

For similar reasons, Baars (1988) has tried to position consciousness within a "global workspace" architecture of the brain. In their review of cognitive models of consciousness, Baars & McGovern (1996) point out that the brain has hundreds of different types of unconscious specialised processors such as feature detectors for colours, line orientation and faces, which can act independently or in coalition with one another, thereby bypassing the limited capacity of consciousness. These processors are extremely efficient, but restricted to their dedicated tasks. The processors can also receive global messages and transmit them by 'posting' messages to a limited-capacity, global workspace whose architecture enables system-wide integration and dissemination of such information. Such communications allow new links to be formed between the processors, and the formation of novel expert 'coalitions' able to work on new or difficult problems.

As do prior theories which identify consciousness with information "at the focus of attention", "in a working store", "in a limited capacity decision channel" an so on, Baars & McGovern (1996) assert that "information in the global workspace corresponds to conscious contents" (p 89). Accordingly, they give consciousness a central role in the economy of mind which corresponds to the functions of the global workspace. Within their model, the global workspace is essential for organising novel, complex activities. Consequently, consciousness has many things to do. For example, in perception (our present concern), "By relating input to its context, consciousness defines input, removing its ambiguities in perception and understanding" (p 91). But they list many additional functions. Overall, "..consciousness appears to be the major way in which the central nervous system adapts to novel, challenging and informative events in the world" (Ibid, p92).

If this widely held view is correct, preconscious analysis of complex, novel information should be impossible. According to Greenwald (1992), even the preconscious analysis of two connected words poses "a challenge" (see above). Perhaps this is so for nonattended input. However, the evidence for preconscious analysis of complex, novel messages in attended input is clear.

In psychological tasks, the "attended" channel is operationally defined by combining instructions to subjects to attend in a given way with appropriate forms of stimulus presentation. For example, the subject might be asked to focus on material in one ear rather than the other, or to fixate a particular point on a screen, and then the stimulus is presented to the point of focus. In the sense that subjects can choose whether or not to follow instructions, their attention may be said to be voluntary, controlled and conscious. It has to be borne in mind, however, that most models of attentional processing assume that input stimuli receive some initial, preconscious analysis (preliminary attention) whether or not they are in the attended channel. This applies to both early-selection models (e.g. Broadbent, 1958) and late-selection models (e.g. the "two-process" model of Posner & Snyder, 1975, discussed above). Stimuli in the attended channel differ in that they are normally selected for further "focal-attentive processing" and it is only when this happens that they enter consciousness. In principle, therefore, it might be possible for input in an attended channel to be given a preliminary, preconscious analysis without being subject to "conscious" focal-attentive analysis - for example, if the input is "masked" before it can be fully analyzed, as in the studies of "masked priming" discussed above.

Suppose, however, that focal-attentive analysis is not disrupted in any way. In what sense, under these circumstances, is the analysis of complex stimuli "voluntary, controlled and conscious"?

Marslen-Wilson (1984) reviews evidence that the analysis of words in attended-to connected speech is both "data-driven" and "cognitively-driven," combining knowledge of the stimulus with knowledge of its context. For example, in Grosjean's (1980) word recognition task, successively longer fragments of a word were presented. If the words were presented in isolation, subjects required fragments of 333 msec. (on average) to identify them (total word length was in excess of 400 msec.). But if the words were presented in normal verbal contexts, a fragment of 199 msec. (on average) was sufficient to identify them. In a related experiment, Marslen-Wilson and Tyler (1980) found that the average reaction time to detect target words (in context) was 273 msec., although their mean length was 370 msec. Allowing around 75 msec. execution time (the time to press a button) this again suggests a word identification time of around 200 msec.

Now, a word fragment of 200 msec. is large enough to contain just the first two phonemes and, according to Marslen-Wilson (1984), these convey useful information. Assuming a dictionary of 20,000 American-English words, knowledge of the first phoneme reduces the set of possible words to a median of 1,033, knowledge of the first two phonemes reduces the set size to a median of 87, and so on (Kucera & Francis, 1967). In this way, sensory analysis (a largely "data-driven" process) contributes to word identification. After two phonemes, however, a large number of possible words remain (a median of 87). Hence subjects who can identify the word on the basis of the first two phonemes must use their knowledge of the context to decide which of the remaining words is the correct one (a "cognitively-driven" process).

On the basis of this and other evidence Marslen-Wilson (1984) concludes that to cope with a complex acoustic waveform developing over time the speech processing system moves the analysis of the sensory signal as rapidly as possible to a domain where all possible sources of information (semantic as well as phonemic) can be brought to bear on its further analysis and interpretation - a process of "on-line interactive analysis" of considerable sophistication and flexibility.

The stimuli to be identified in these experiments are in the attended channel. Yet if words (in context) are identified within 200 msec., this confluence of data-driven and cognitively driven processing cannot be conscious, for according to the evidence reviewed earlier (Libet, et al., 1979; Posner & Snyder, 1975; Neeley, 1977), consciousness of a given stimulus does not arise until at least 200 msec. after the stimulus arrives at the cortical projection areas, i.e. after the identification of a word (in context) has been achieved!

In these experiments spoken words in the attended channel are therefore analyzed in preconscious fashion. Rather than consciousness entering into input analysis of well-known stimuli, consciousness of those stimuli appears to follow sophisticated, preconscious analysis and identification. If this is the case, consciousness cannot be necessary for the analysis and identification of such stimuli even when they occur in novel, complex combinations. This conclusion may seem counterintuitive. It is, however, easy to illustrate. Consider how one silently reads the following sentence:

"If we don't increase the dustmen’s wages, they will refuse to take the refuse."

Note that on its first occurrence in your phonemic imagery or "covert speech", the word "refuse" was (silently) pronounced with the stress on the second syllable (refuse) while on its second occurrence the stress was on the first syllable (refuse). But how and when did this allocation of stress patterns take place? Clearly, the syntactic and semantic analysis required to determine the appropriate meanings of the word "refuse" must have taken place prior to the allocation of the stress patterns; and this, in turn, must have taken place prior to the phonemic images entering awareness.

Note too, that while reading, one is not conscious of any pattern recognition processing to identify individual words or of any syntactic or semantic analysis being applied to the sentence. Nor is one aware of the processing responsible for the resulting covert speech (with the appropriate stress patterns on the word "refuse"). In this case, not just an individual word, but an entire attended-to sentence appears to be analyzed in preconscious fashion. The same may be said of the paragraph you are now reading, or of the entire text of this paper. You are conscious of what is written, but not conscious of the complex input analysis involved.  

Note finally that the analysis of well-known stimuli proceeds in a largely involuntary fashion, whether or not the stimuli are in the attended channel. Even if one "consciously attends" to a given stimulus, it may be difficult to prevent certain analyses from being carried out. In this sense, the analysis is automatic. This point was demonstrated by Stroop (1935), who observed that subjects instructed to name the colour in which a word is printed found the task far more difficult if the word was itself a colour name, but of a different colour. For example, subjects presented with the word "red" printed in orange cannot restrict their analysis to the colour of the print (orange) because they cannot prevent themselves from reading the word ("red").

On the basis of this and other evidence, Kahneman (1973) concludes that "subjects cannot prevent the perceptual analysis of irrelevant attributes of an attended object." Even if a stimulus is consciously attended to, what is analyzed may not be under conscious voluntary control. However, an "involuntary" process is not necessarily "inflexible" (see discussion of speech perception above). Nor need it be "effortless". For example, studies of the Stroop effect indicate that while input analysis may be automatic in the sense of "involuntary," it nevertheless draws on limited processing resources (Kahneman & Treisman, 1984).

Conventionally, "preconscious" analysis is thought to be automatic (in the sense of being involuntary), and restricted to simple, familiar stimuli whose long-term memory traces are accessed in data-driven fashion. The terms "preconscious analysis", "preattentive analysis", or "preconscious preattentive analysis" are often used interchangeably. "Conscious" analysis or "focal-attentive" analysis is thought to be voluntary and flexible (involving cognitively-driven as well as data-driven processing) and, again, the terms "conscious analysis", "focal-attentive analysis", or "conscious focal-attentive analysis" are often treated as if they are synonymous.

The evidence reviewed above suggests that this rigid linkage of "preconscious" versus "conscious" processing to "preattentive" versus "focal-attentive" processing requires re-examination. Stimuli in attended channels are subject to a far more sophisticated analysis than those in nonattended channels. But, if the meanings of attended-to phrases and sentences can be analyzed before they enter consciousness this attentional analysis cannot be conscious. Conversely, preconscious analysis in attended channels cannot be restricted to simple, familiar words. Reading, and the on-line analysis of speech are amongst the most sophisticated of human pattern recognition tasks, involving both cognitively-driven and data-driven processing. If the input analysis of text and speech operates preconsciously, then preconscious, attentional analysis might be automatic (in the sense of being involuntary) but it cannot be inflexible.

To put the point another way, by the time perceived text or speech enters consciousness the analysis of words in context (including both semantic and syntactic analysis) has already been achieved. If so, consciousness (of the input) arises too late to affect the processing with which it is most closely associated. Reading and speech perception of attended to messages are universally thought of as "conscious processes". Yet, the processes which enable reading and speech perception are, strictly speaking, preconscious.

According to Velmans (1991a), ".. in its ordinary usage "consciousness" refers to something other than "focal-attentive processing." It refers primarily to "awareness," whereas "focal-attentive processing" refers to a functional subdivision within an information processing model of the brain. Focal-attentive processing is thought to be a necessary condition for conscious awareness. Operationally, however, they are distinct (Nissen & Bullemer, 1987; Kahneman & Treisman, 1984). Conscious contents are typically investigated by the use of subjective reports (of subjective experience) - usually verbal reports, although various other means of communicating experience exist (Ericsson & Simon, 1984; Pope & Singer, 1978). By contrast, human information processing and functional divisions within such processing are typically inferred from performance measures such as reaction time, error score, and so forth"(p 665).

It is interesting to note that Shiffrin (1997) has recently come to similar conclusions about the need to distinguish consciousness from focal-attentive processing. Studies of capacity limitations, automaticity, depth of processing, effort, and speed of processing suggest that "there are incontrovertible links between attention and consciousness" (p 53). Nevertheless, a careful examination of the literature reveals that, "Despite the positive correlation between attention and consciousness, the mapping between the two conceptual frameworks is quite poor." (p 63). By contrast, Baars (1991) resisted any consciousness - focal-attention dissociation. Rather, personal experience and focal attention "covary so perfectly, we routinely infer in our everyday life that they reflect a single underlying reality." (p 669) The attempt to dissociate these (in Velmans, 1991a) is just one of a series of misguided attempts (by philosophers, psychologists, and neuroscientists) to deny the "common-sense and scientifically useful idea that reports of conscious experience, focal-attention, and wakefulness reflect an internal but nevertheless knowable aspect of our nervous system." (Ibid) However, Baars (1997) now agrees that consciousness needs to be dissociated from focal-attention both conceptually and operationally. As he points out,

"English makes a clear distinction between "looking" and "seeing", "listening" and "hearing", and "touching" and "feeling". The first word of each pair describes a way of gaining access to a conscious perceptual experience (looking, listening, touching), while the second refers to the resulting experience itself (seeing, hearing, feeling). We use the first verb of each pair in order to gain access to the second."(p364) Furthermore, "Attentional operations include instructions to attend and disattend, effortful control of attention against competing input, and experimental manipulations of attentional selection priorities ... In contrast, our most obvious index of consciousness involves people describing their experiences in some verifiable way, under conditions that maximise accuracy." (Ibid, p364 - my italics).

However, rather than rejecting the ontological identification of consciousness with information processing, Baars then goes on to identify consciousness with a slightly later stage of information processing (as does Mandler, 1991, 1997) in terms that have very little to do with people's descriptions of what they experience. Attention now becomes the "gatekeeper" for the global workspace and, as noted above, the contents of the global workspace are equated with consciousness. Thus, "attention creates access to consciousness", but "consciousness is needed to create access to unconscious processing resources" (Baars, 1997, p370). Consequently he claims that consciousness carries out the many functions which require global access to unconscious processing resources such as system-wide integration and dissemination of information, the formation of new links between unconscious processors, and so on.

Unfortunately, this identification of consciousness with the operations of a "global workspace" runs into difficulties of its own. Consider again how one silently reads, "If we don't increase the dustmen’s wages, they will refuse to take the refuse." As noted above, while reading, one is not conscious of any pattern recognition processing to identify individual words or of any syntactic or semantic analysis being applied to the sentence. Nor is one aware of the processing responsible for the resulting covert speech (with the appropriate stress patterns on the word "refuse"). But the same applies to the supposed "conscious operations" of the "global workspace". While reading, on is not aware of consciously carrying out any system-wide integration and dissemination of information, or of forming new links between unconscious processors! Rather, information which enters consciousness has already been integrated and appears to be generally available to the system as a whole. So, if consciousness does carry out such functions, in the manner Baars (1997) suggests, it must do so unconsciously.

Cognitive psychology has made considerable progress in locating those aspects of information processing most closely associated with phenomenal consciousness. But deep problems follow from the reductionist identification of phenomenal consciousness with information processing, which has become common in functionalist, cognitive accounts. For consciousness to be a "global workspace" it must carry out the functions of the global workspace. But if we are not aware of carrying out such functions how can they be conscious? For those who are not yet convinced that there is a problem, I leave the following conundrum:

Question: Is it possible for consciousness to do something to or about something that it is not conscious of?

If the answer is NO.

We are not aware of the activity of our own brains.

So we conclude that consciousness as such does not influence brain activity.

If the answer is YES.

We are not aware of the activity of our own brains.

So consciousness must influence brain activity unconsciously.

So we conclude that consciousness as such does not influence brain activity.

Yet consciousness is central to human being.

Without it our existence would be like nothing.

So the notion that consciousness does nothing makes no sense.

According to Velmans (1991a) a first step in addressing such problems must be a re-evaluation of what is meant by "conscious processing." On closer examination a process might be said to be "conscious" in three distinct senses. It might be "conscious"

Some processes (problem solving, thinking, planning, and so on) are conscious in sense (a) but only in so far as their detailed operation is accessible to introspection. Perception, by contrast, is conscious only in sense (b). Whether any human information processing is conscious in sense (c) is open to question (see Velmans, 1991a,b, 1993b, 1996b and accompanying commentaries). In short, the subtle, differing ways in which phenomenal consciousness relates to information processing needs to be specified with care. Theories which simply redefine consciousness to be a form of processing (such as focal-attention, information in an LCDC, a global workspace etc.) confound these relationships, thereby begging the question about the functional role of phenomenal consciousness in the economy of mind.

When does perception become conscious? On present indications, only once analysis is complete, and attended-to information is sufficiently well-integrated to be disseminated throughout the brain. How can we decide when perception becomes conscious? By combining third-person measures of functional dissociations between preconscious and conscious states with first-person reports of subjective experience. In what sense does perception become conscious? Only in the sense that analysis of input can result in a conscious experience. Consciousness of familiar stimuli, rather than entering into input analysis, appears to follow it, in human information processing. Information processing most closely associated with conscious awareness of input appears to operate unconsciously in the economy of mind.

Baars, B.J.(1997) Some essential differences between consciousness and attention, perception, and working memory. Consciousness and Cognition, 6(2/3), 363-371.

Baars, B.J.(1991) A curious coincidence? Consciousness as an object of scientific scrutiny fits our personal experience remarkably well. Behavioral and Brain Sciences, 14(4), 669-670.

Baars, B.J.(1988) A Cognitive Theory of Consciousness. Cambridge: Cambridge University Press.

Baars, B.J. & McGovern, K.(1996) Cognitive views of consciousness: What are the facts? How can we explain them? In M. Velmans (Ed.) The Science of Consciousness: Psychological, Neuropsychological, and Clinical Reviews. London: Routledge.

Berry, D.C. and Dienes, Z. (Eds.)(1993) Implicit Learning: theoretical and empirical issues. London: Lawrence Erlbaum.

Bjork, R.A.(1975) Short-term storage: the ordered output of a central processor. In F.Restle, R.M.Shiffrin, N.J.Castellan, H.R.Lindman & D.B.Pisoni (eds.) Cognitive Theory Vol.1. Hillsdale, New Jersey: Lawrence Erlbaum Associates.

Bower, G.H.(1972) A selective review of organizational factors in memory. In E. Tulving & W. Donaldson (eds) Organization in Memory. New York: Academic Press.

Broadbent, D.E.(1958) Perception and communication. London: Pergamon Press.

Campion, J., Latto, R. & Smith, Y.M.(1983) Is blindsight an effect of scattered light, spared cortex, and near-threshold vision? Behavioral and Brain Sciences, 6, 423-486.

Cheesman, J. & Merikle, P.M.(1984) Priming with and without awareness. Perception and Psychophysics, 36, 387-395.

Cheesman, J. & Merikle, P.M.(1986) Distinguishing conscious from unconscious perceptual processes. Canadian Journal of Psychology, 40, 343-367.

Cherry, C.(1953) Some experiments on the reception of speech with one and with two ears. Journal of the Acoustical Society of America, 25, 975-979.

Corteen, R.S.(1986) Electrodermal responses to words in an irrelevant message: A partial reappraisal. Behavioral and Brain Sciences, 9, 27-28.

Corteen, R.S. & Wood, B.(1972) Autonomic responses to shock-associated words in an unattended channel. Journal of Experimental Psychology, 94, 308-313.

Corteen, R.S. & Dunn, D.(1974) Shock associated words in a non-attended message: a test for momentary awareness. Journal of Experimental Psychology, 102, 1143-1144.

Crick, F.(1994) The Astonishing Hypothesis: The scientific search for the soul. London: Simon & Schuster.

Dagenbach, D., Carr, T.H. & Wilhelmsen, A.(1989) Task-induced strategies and near-threshold priming: Conscious influences on unconscious perception. Journal of Memory and Language, 28, 412-443.

Dawson, M.E. & Schell, A.M.(1982) Electrodermal responses to attended and unattended significant stimuli during dichotic listening. Journal of Experimental Psychology: Human Perception and Performance, 8, 315-324.

Deutsch, J.A. & Deutsch, D.(1963) Attention: some theoretical considerations. Psychological Review, 70, 80-90.

Dixon, N.F.(1971) Subliminal perception: the nature of a controversy. New York: McGraw-Hill.

Dixon, N.F.(1981) Preconscious Processing. Chichester: Wiley.

Ericsson, K.A. & Simon, H.(1984) Protocol Analysis: verbal reports as data. Cambridge, Mass.:MIT Press.

Eriksen, C.W.(1960) Discrimination and learning without awareness: a methodological survey and evaluation. Psychological Review, 67, 279-300.

Eriksen, C.W.(1958) Unconscious processes. In M.R.Jones (Ed.) Nebraska Symposium on Motivation. Lincoln, NE: University of Nebraska Press.

Eriksen, C.W.(1956) An experimental analysis of subception. American Journal of Psychology, 69, 625-634.

Forster, K.I. & Davis, C.(1984) Repetition Priming and frequency attenuation in lexical access. Journal of Experimental Psychology: Learning Memory and Cognition, 10, 680-698.

Forster, K.I. & Govier, E.(1978) Discrimination without awareness? Quarterly Journal of Experimental Psychology, 30, 289-295.

Gardiner, J.(1996) On consciousness in relation to memory and learning. In M. Velmans (ed.) The Science of Consciousness: Psychological, Neuropsychological, and Clinical Reviews. London: Routledge.

Greenwald, A.G.(1992) New look 3: unconscious cognition reclaimed. American Psychologist, 47, 766-790.

Greenwald, A.G., Klinger, M.R. and Liu, T.J.(1989) Unconscious processing of dichoptically masked words. Memory and Cognition, 17, 35-47.

Greenwald, A.G. and Liu, T.J.(1985) Limited unconscious processing of meaning. Paper presented at the annual meeting of the Psychonomic Society, Boston, MA, November.

Groeger, J.A.(1984) Evidence of unconscious semantic processing from a forced error situation. British Journal of Psychology, 75, 305-314.

Groeger, J.A. (1988) Qualitatively different effects of undetected and unidentified auditory primes. Quarterly Journal of Experimental Psychology, 40A, 323-329.

Grosjean, F.(1980) Spoken word recognition processes and the gating paradigm. Perception and Psychophysics, 28, 267-283.

Hilgard, E.R.(1986) Divided consciousness: Multiple controls in human thought and action. Chichester: Wiley.

Holender, D.(1986) Semantic activation without conscious identification in dichotic listening, parafoveal vision, and visual masking. Behavioral and Brain Sciences, 9, 1-66.

James, W.(1890) The principles of psychology. New York: Dover. (Reprinted 1950).

Kahneman, D.(1973) Attention and Effort. Englewood Cliffs, NJ: Prentice-Hall.

Kahneman, D. & Treisman, A.(1984) Changing views of attention and automaticity. In R. Parasuraman & D.R. Davies (Eds.) Varieties of Attention. London: Academic Press.

Kihlstrom, J.F.(1996) Perception without awareness of what is perceived, learning without awareness of what is learned. In M. Velmans (Ed.) The Science of Consciousness: Psychological, Neuropsychological, and Clinical Reviews. London: Routledge.

Kucera, H. & Francis, W.M.(1967) Computational analysis of present-day American English. Brown University Press.

La Berge, D.(1981) Automatic information processing: A review. In J. Long & A. Baddeley (Eds.) Attention and Performance IX. Hillsdale, NJ: Erlbaum.

Lackner, J. & Garrett, M.F.(1973) Resolving ambiguity: Effects of biasing context in the unattended ear. Cognition, L, 359-372.

Lewis, J.L.(1970) Semantic processing of unattended messages using dichotic listening. Journal of Experimental Psychology, 85, 220-227.

Libet, B.(1996) Neural processes in the production of conscious experience. In M. Velmans (Ed.) The Science of Consciousness: Psychological, Neuropsychological, and Clinical Reviews. London: Routledge.

Libet, B.(1985) Unconscious cerebral initiative and the role of conscious will in voluntary action. Behavioral and Brain Sciences, 8, 529-566.

Libet, B., Wright Jr., E.W., Feinstein, B. & Pearl, D.K.(1979) Subjective referral of the timing for a conscious experience: A functional role for the somatosensory specific projection system in man. Brain, 102, 193-224.

MacKay, D.G.(1973) Aspects of the theory of comprehension, memory and attention. Quarterly Journal of Experimental Psychology, 25, 22-40.

Mandler, G.(1997) Consciousness redux. In J.D. Cohen & J.W. Schooler (eds) Scientific Approaches to Consciousness. Mahwa, NJ: Lawrence Erlbaum Associates.

Mandler, G. (1991) The processing of information is not conscious, but its products often are. Behavioral and Brain Sciences, 14(4), 688-689.

Mandler, G.(1985) Cognitive Psychology: an essay in cognitive science. Hillsdale, NJ: Erlbaum.

Mandler, G.(1975) Mind and emotion. Chichester: Wiley.

Marcel, A.J.(1980) Conscious and preconscious recognition of polysemous words: Locating the selective effects of prior verbal context. In R.S. Nickerson (Ed.) Attention and Performance VIII. Hillsdale, NJ: Erlbaum.

Marcel, A.J.(1983) Conscious and unconscious perception: Experiments on visual masking and word recognition. Cognitive Psychology, 15, 197-237.

Marcel, A.J.(1986) Consciousness and processing: Choosing and testing a null hypothesis. Behavioral and Brain Sciences, 9, 40-41.

Marslen-Wilson, W.D.(1984) Function and process in spoken word recognition-A tutorial review. In H. Bouma & D.G. Bouwhuis (Eds.) Attention and Performance X. Hillsdale NJ: Erlbaum.

Marslen-Wilson, W.D. & Tyler, L.K.(1980) The temporal structure of spoken language understanding. Cognition, 8, 1-71.

Merickle, P.M.(1992) Perception without awareness: critical issues. American Psychologist, 47, 792-795.

Merikle, P.M.(1982) Unconscious perception revisited. Perception and Psychophysics, 31, 298-301.

Merikle, P.M & Daneman, M.(1998) Psychological investigations of unconscious perception. Journal of Consciousness Studies, 5 (1), 5-18.

Merikle, P.M. & Joordans, S.(1997) Parallels between perception without attention and perception without awareness. Consciousness and Cognition, 6 (2/3), 219-236.

Meyer, D.E. & Schvaneveldt, R.W. & Ruddy, M.G.(1975) Loci of contextual effects on visual word recognition. In P.M.A. Rabbitt & S. Dornic (Eds.) Attention and Performance V. Hillsdale NJ: Erlbaum.

Miller, G.A.(1987) Psychology: the science of mental life. Harmondsworth: Penguin.

Neely, J.H.(1977) Semantic priming and retrieval from lexical memory: Roles of inhibitionless spreading activation and limited capacity attention. Journal of Experimental Psychology: General, 106, 226-254.

Newstead, S.E. & Dennis, I.(1979) Lexical and grammatical processing of unshadowed messages: A re-examination of the Mackay effect. Quarterly Journal of Experimental Psychology, 31, 477-488.

Nissen, M.J. & Bullemer, P.(1987) Attentional requirements of learning: Evidence from performance measures. Cognitive Psychology, 19, 1-32.

Norman, D.(1969) Memory and attention: an introduction to human information processing. Chichester: Wiley.

Oakley, A.D. & Eames, L.C.(1985) The plurality of consciousness. In D.A. Oakley (Ed.) Brain and mind. London: Methuen.

Perenin, M.T. & Jeannerod, M.(1978) Visual function within the hemianopic field following early cerebral hemidecortication in man: 1-Spatial localization. Neuropsychologia, 16, 1-13.

Pierce, C.S. & Jastrow, J.(1885) On small differences in sensation. Memoirs of the National Academy of Sciences, 3, 317-329.

Pope, K.S. & Singer, J.L.(1978) The Stream of Consciousness. New York: Plenum Press.

Posner, K.R.(1978) Chronometric Explorations of Mind. Hillsdale, NJ: Erlbaum.

Posner, M.I. & Boies, S.W.(1971) Components of attention. Psychological Review, 78:391-408.

Posner, M.I., DiGirolamo, G.J. & Fernandez-Duque, D.(1997) Brain mechanisms of cognitive skills. Consciousness and Cognition, 267-290.

Posner, M.I. & Snyder, C.R.R.(1975) Facilitation and inhibition in the processing of signals. In P.M.A. Rabbitt & S. Dornick (Eds.) Attention and Performance V. London: Academic Press.

Posner, M.I. & Warren, R.E.(1972) Traces, Concepts, and conscious constructions. In A.W.Melton & E.Martin 9eds) Coding Processes in Human Memory, Washington, DC: V.H. Winston & Sons.

Pynte, J., Do, P. & Scampa, P.(1984) Lexical decisions during the reading of sentences containing polysemous words. In S. Kornblum & J. Requin (Eds.) Preparatory States and Processes. Hillsdale NJ: Erlbaum.

Reber, A.S.(1993) Implicit Learning and Tacit Knowledge: an essay on the cognitive unconscious. Oxford: Oxford University Press.

Reingold, E.M. and Merikle, P.M.(1993) Theory and measurement in the study of unconscious processes. In M.Davies and G.W.Humphries (Eds.) Consciousness. Oxford: Blackwell.

Schacter, D.L.(1992) Consciousness and awareness in memory and amnesia: Critical issues. In A.D.Milner and M.D.Rugg (Eds.) The Neuropsychology of Consciousness. London: Academic Press.

Shallice, T. (1972) Dual functions of consciousness. Psychological Review, 79(5), 383-393.

Shanks, D.R. & St.John, M.F.(1994) Characteristics of dissociable human learning systems. Behavioral and Brain Sciences, 17(3), 367-447.

Shiffrin, R.M.(1997) Attention, automatism, and consciousness. In J.D. Cohen & J.W. Schooler (eds) Scientific Approaches to Consciousness. Mahwa, NJ: Lawrence Erlbaum Associates.

Shiffrin, R.M. & Schneider, W.(1977) Controlled and automatic human information processing: II. Perceptual learning, automatic attending, and a general theory. Psychological Review, 84, 127-190.

Stroop, J.R.(1935) Studies of interference in serial verbal reactions. Journal of Experimental Psychology, 18, 643-662.

Styles, E.A. (1997) The Psychology of Attention. Hove, East Sussex: Psychology Press.

Swinney, D.A.(1979) Lexical access during sentence comprehension: (Re)consideration of context effects. Journal of Verbal Learning and Verbal Behaviour, 18, 645-659.

Swinney, D.A.(1982) The structure and time-course of information interaction during speech comprehension: Lexical segmentation, access, and interpretation. In J. Mehler, E.C.T. Walker & M. Garrett (Eds.) Perspectives on Mental Representation. Hillsdale, NJ: Erlbaum.

Treisman, A.M.(1964) Verbal cues, language and meaning in attention. American Journal of Psychology, 77, 206-214.

Treisman, A.M., Squire, R., & Green, J.(1974) Semantic processing in dichotic listening? A replication. Memory and Cognition, 2, 641-646.

Underwood, G. (1991) Attention is necessary for word integration. Behavioral and Brain Sciences, 14(4), 698.

Underwood, G.(1977) Contextual facilitation from attended and unattended messages. Journal of Verbal Learning and Verbal Behaviour, 16, 99-106.

Velmans, M. (1998) Goodbye to reductionism. In S. Hameroff, A. Kaszniak, and A. Scott (eds) Toward a Science of Consciousness: The Second Tucson Discussions and Debates. Cambridge, Mass: MIT Press (in press).

Velmans, M. (Ed.) (1996a) The Science of Consciousness: Psychological, Neuropsychological, and Clinical Reviews. London: Routledge.

Velmans, M. (1996b) Consciousness and the "causal paradox." Behavioral and Brain Sciences, 19(3), 537-542.

Velmans, M.(1993) Consciousness, causality and complementarity. Behavioral and Brain Sciences, 16(2), 409-416.

Velmans, M.(1991a) Is human information processing conscious? Behavioral and Brain Sciences, 14(4), 651-669.

Velmans, M.(1991b) Consciousness from a first-person perspective. Behavioral and Brain Sciences, 14(4), 702-726.

Von Wright, J.M., Anderson, K. & Stenman, U.(1975) Generalisation of conditioned GSR's in dichotic listening. In P.M.A. Rabbitt and S. Dornic (eds) Attention and Performance V, London: Academic Press.

Wardlaw, K.A. & Kroll, N.E.A.(1976) Automatic responses to shock associated words in a non-attended message: A failure to replicate. Journal of Experimental Psychology: Human Perception and Performance, 2, 357-360.

Weiskrantz, L.(1986) Blindsight: a case study and implications. London, Open University Press.

Weiskrantz, L.(1988) Neuropsychology of vision and memory. In A.J. Marcel and E. Bisiach (eds) Consciousness in Contemporary Science, Oxford: Oxford University Press.

Weiskrantz, L. (1997) Consciousness Lost and Found. Oxford: Oxford University Press.