CONSCIOUSNESS FROM A FIRST-PERSON PERSPECTIVE

Email: m.velmans@gold.ac.uk
URL: http://www.gold.ac.uk/academic/ps/velmans.htm



KEYWORDS: attention, complementarity, consciousness, functionalism, epiphenomenalism, information processing, mind, unconscious, first person, third person, ontological monism, epistemological dualism



The sequence of topics in this reply roughly follows that of the target article. The latter focused largely on experimental studies of how consciousness relates to human information processing, tracing their relation from input through to output. The discussion of the implications of the findings both for cognitive psychology and philosophy of mind was relatively brief. The commentaries reverse this emphasis, and so, correspondingly, does the reply. Sections 1.1 to 1.9 begin with details of the empirical findings (Underwood, Inhoff, Mangan, Van Gulick, Shevrin, Dagenbach, Spiegel, Rey, Lloyd, Wagstaff, Block, Carlson, Mandler, Libet, and Keane). Sections 2 & 3 deal with general comments about the laboratory-based approach of the review (Kinsbourne, Reznick & Zelazo, Bowers, and Foulkes) and the status of subjective reports (Gregson, Economos, Gray, and Lloyd). Sections 4 to 6 discuss the nature of "conscious processing," beginning with differences between conscious and nonconscious processing in section 4 (Wilson, Bowers, Carlson, Block, and Inhoff), followed by a detailed treatment of the relation of consciousness to focal-attentive processing in sections 5.1 to 5.4 (Mandler, Baars, Carlson, Rey, Mangan, Klein, Van Gulick, and Block) and a further discussion of what is meant by a "conscious process" in section 6 (Van Gulick, Dretske).

Sections 7 to 9, deal with broad philosophical and theoretical implications, starting (in 7.1 to 7.3) with the mistaken assumption (of Baars, Block, Lloyd, Mangan, Van Gulick, Kinsbourne, Corteen, Rey, and Hardcastle) that I support epiphenomenalism, and the problems of dualist-interactionism (MacKay, and Mangan) and reductionism (Sloman, Hardcastle, Rey, and Stanovich). Section 8 gives a fuller account of "first-person" vs "third-person" perspectives (Stanovich, Schaeken & d'Ywalle, Shevrin and Gardiner), and section 9, of "complementarity." 9.1 replaces the privileged status of "third-person" accounts with a more balanced view of the two perspectives (Lloyd) and 9.2 moves on to a proposed resolution of the paradoxes surrounding the causal interactions of consciousness and the brain (Libet). A "complementary" account is given of the relation of consciousness to its neural correlates in 9.3 (Economos, Koch & Crick, Van Gulick, and Navon) and, finally, of the role of consciousness in evolution in 9.4 (Mangan, Corteen, Klein, editorial, and Schaeken & d'Ywalle).

1. The review of experimental findings: points of detail.

1.1 The limits of input analysis in nonselected channels.

Underwood, and Inhoff focus on the limits of input analysis in nonselected channels. "Early-selection" models of attention assume that preliminary analysis of verbal stimuli is restricted to physical properties. "Late-selection" models such as that of Underwood (1979), assume it to be restricted to accessing the memory traces (and meanings) of individual words (see TA section 1.4). In TA section 2.1, I argue for a further loosening of restrictions. Drawing on evidence from Treisman (1964a, 1964b) and Lackner & Garrett (1973), I argue that it is possible for analysis in the nonselected ear to extend to the meanings of spoken phrases and sentences - a complex task requiring a combination of phonemic, syntactic and semantic analysis.

It is important to note, however, that I do not claim such an analysis always takes place (see TA 1.2, end); in Treisman's (1964a) study, just over half the bilingual subjects did not recognise the nonselected French translation of the shadowed message. Inhoff provides further examples (Inhoff & Briihl, 1991; Yantis & Johnston, 1990; Kahneman et al, 1983).[1] The extent to which preliminary analysis is carried out appears to be both task and resource limited. Inhoff & Briihl found that semantic analysis of nonattended, visually presented text did not take place without (momentary) eye fixations on the nonattended material. On the other hand, in Treisman's (1964b) experiment, the more skilled subjects were at French, the more difficult it was for them not to analyze the message in the nonselected ear, often giving a shadowing response in mixed English and French. Limits on preliminary analysis are also set by the complexity of the material. As Underwood rightly points out, without focal attention while reading, the comprehension of difficult text is likely to meet with mixed success. Focal-attentive processing provides options additional to preliminary analysis. For example, information given focal-attentive processing may become generally available throughout the processing system (TA section 8). This makes it possible both to assimilate input and to contextualise it, enabling comprehension of difficult or novel material.

In short, while I suggest that in some circumstances preliminary analysis may be more extensive than hitherto thought, I do not make the general claim that (focal) "attentional processing is unnecessary for the recognition of complex environmental events," or dismiss "the premise that the processing of novelty requires attention," as Underwood claims. Nor do I regard the difference between preliminary and focal attention to be "negligible," as Inhoff implies. Although the sentences in Treisman and Lackner & Garrett's studies formed complex, potentially novel patterns, when preliminary analysis was possible, both the component words and the rules for sentence formation were well known. Focal attention is required when novel stimuli or stimulus combinations that do not follow known rules are being learnt. Whether consciousness is required for such learning is a different matter (see TA section 4).

According to Underwood, however, preliminary analysis of nonselected verbal material never proceeds beyond accessing the meaning of individual words. Accordingly, he attempts to argue that in the experiments I cite, a more complex analysis did not take place. He suggests that subjects' ability to recognise the French translation in Treisman's experiments might have come about from corresponding English and French words occurring in temporal proximity, in which case the interference effect (Treisman, 1964b) might have come about through simultaneous word recognition. However, the effects found by Treisman occurred even when the selected passage lagged behind the nonselected passage by one or two seconds; given the further variations produced by differences in French and English syntax, it is difficult to know how Underwood defines "proximity." Even if (decontextualised) French synonyms had been simultaneous with their English counterparts, it is difficult to see why they should intrude; no similar intrusions are reported in Lewis' (1970) shadowing task using English words, when English synonyms were simultaneously presented to the attended ear. Given this, it seems more plausible to suggest that the interference was caused by a deeper analysis of the nonselected message producing the realisation that the messages (encoded in French and English) were the same, leaving subjects confused about whether they should respond in English or French.[2]

Lackner & Garrett (1973) found that context sentences presented to the nonselected ear, biased the paraphrases of ambiguous sentences in the attended ear. They also point out that for most of the context sentences they used, at least a phrasal analysis would have been necessary to produce the required disambiguation (see TA section 2.1). Underwood challenges Lackner & Garrett's (1973) interpretation of their own findings, on the grounds that "Mackay (1973) and Newstead & Dennis (1979) report(ed) very limited effects of unattended messages upon the interpretation of shadowed messages." He concludes that, "At best it seems that individual words might have an effect upon the lexical ambiguity in the attended message, but these experiments together do not provide convincing evidence of deep structural analysis without attention." What Underwood obscures in this argument is that the disambiguating cues used both by Mackay (1973) and Newstead & Dennis (1979) were restricted to individual words (although in one of the conditions in the latter experiment, these were embedded in sentences). Consequently, it is hardly surprising that these two experiments provide evidence only of the disambiguating effect of individual words. They have no bearing whatsoever on the evidence for phrasal analysis (and consequent disambiguation) in the Lackner & Garrett study.

Underwood goes on to describe two experiments (Underwood, 1977; Kleiman, 1975) which "suggest that without attention there is limited integration of words in a sentence." In Underwood's dichotic listening task, shadowing latencies of attended words progressively decreased as prior, attended context increased. While prior context words on the nonselected ear also produced some reduction in shadowing latency to attended words, there was no further progressive reduction as nonselected context increased. He concludes that the additional nonselected context words must either have been forgotten or (without attention) they could not be contextually integrated. It is not obvious, however, that successful integration of nonselected word strings would produce decreasing shadowing latencies to words in the attended ear.

Focal-attentive processing enables full assimilation and contextualisation of input in the light of current knowledge, expectations, needs, goals and so forth. According to Posner & Snyder (1975) and Neeley (1977) it both activates related traces and inhibits unrelated traces as expectations regarding subsequent stimuli increase. In attended-to connected speech, for example, it appears that all meanings associated with a given word are initially activated in parallel, but once context is taken into consideration, irrelevant meanings are inhibited (Pynte, Do & Scampa, 1984; Swinney, 1979). This combination of activation and inhibition would account for the progressive decrease in shadowing latencies, with increasing, attended context, in Underwood's experiment.[2] Nonselected stimuli, on the other hand, receive only preliminary analysis. This activates traces of related stimuli, but produces no combined activation of related traces and inhibition of unrelated traces as context increases. This would apply irrespective of the complexity of preliminary analysis. Consequently, Underwood's finding that shadowing latency does not progressively decrease as context in the nonselected ear increases, may have no bearing on whether a relatively complex analysis of nonselected word strings took place.

Kleiman's (1975) divided attention study demonstrates that digit shadowing reduces subjects' ability to read text, and that reading sentences requires more processing capacity than the identification of individual words. Given the relatively complex nature of sentence integration when reading, this is hardly surprising; both preliminary and focal-attentive processing are thought to be subject to capacity constraints (see TA section 2.4). But this does not establish that preliminary analysis is invariably restricted to individual words, either in reading or in dichotic listening tasks, as Underwood claims.[2] Currently, no definitive answers are available on these issues. The relation of preliminary processing to focal-attentive processing is extremely complex and requires more detailed exploration both in different sense modalities, and different task and resource limited conditions.

The central theme of my target article however concerns the relation of consciousness to human information processing. My argument that preconscious analysis in nonselected channels may sometimes be more extensive than hitherto thought, is preliminary to the stronger point that even when stimuli are subject to focal-attentive processing, awareness follows the processing to which it relates and does not enter into it. These arguments, developed in relation to input analysis in TA sections 2.2 to 2.5, and more generally in sections 3 to 8, are ones that Underwood chooses not to address.

Instead, he offers the opinion that "the suggestion that we become aware of a stimulus or a response only after processing is...misguided." He supports this by asserting that focal-attentive processing "associated with the vividness of awareness" depends on novelty, and by elaborating on the theme that in different situations we become aware of our processing, although sometimes we are not. Neither of these points has any bearing on my arguments.

1.2 The processing of complex, attended information.

According to Mangan, consciousness is particularly necessary "in processing complex, novel information to be used to plan, reflect and create," and he complains that my treatment of these complex tasks is relatively thin. In a sense, this is fair. In comparison, say, to input analysis of individual words, the relation of consciousness to more complex tasks has been less well explored; consequently, less evidence is available.

The available evidence nevertheless supports a dissociation of consciousness from functioning; and the evidence does extend to complex tasks. Treisman's (1964) shadowing experiments, for example, appear to show analysis of prose passages on the nonselected ear (TA 2.1). TA 2.3 deals with preconscious word identification in attended, connected speech. But, Mangan chooses to focus on my example of preconscious analysis in silent reading of a single sentence, at the focus of attention (end of TA 2.3); as he points out, "This is not exactly Proust." I agree. But what he needs to show is that the principles which apply to the (silent) reading of individual sentences do not apply to more complex passages. Consider, for example, your own silent reading of this text. My claim is that "while reading, one is not conscious of any pattern recognition processing to identify individual words or of any syntactic analysis being applied to the sentence. Nor is one aware of the processing responsible for the resulting covert speech" (TA 2.3). If my own silent reading is representative, not just individual words and sentences, but entire complex arguments (like the present one) are processed in preconscious fashion.

According to Van Gulick, evidence from the comprehension and production of speech is irrelevant to the analysis of novel stimuli, or the planning and control of novel responses, on the grounds that these skills "are generally agreed to be the result of task-specific modular processors which operate in an automatic and informationally encapsulated way." As noted in TA 2.1, 2.3 and 5.4, language processing is amongst the most complex of human cognitive skills, and there is no definable limit to the novel sentences we can comprehend and produce. Processing may be automatic (in the sense of involuntary) without being effortless or inflexible (TA 2.4, 2.5). Much of the processing may be modular, but the extent to which it is "informationally encapsulated" depends on the nature of processing. Full sentence comprehension, for example, requires the integration of input strings with stored knowledge of the world. And the design of utterances requires the speaker not only to arrange words in a syntactically appropriate way, but to take into account the physical and social context (as well as the prior verbal context) of the utterance. Mangan suggests that to establish my case I need to consider the role consciousness might play at the upper end of the cognitive spectrum. Language is at the upper end of this spectrum. If consciousness does not enter into functioning, added complexity will make no difference.

1.3 Memory without prior experience.

In TA 4.2, I suggest that preconscious contents can influence long-term memory (e.g. Eich, 1984). Further evidence of memory without prior experience is mentioned by Shevrin (1986a, 1990); subjects given subliminal stimuli, without any memory of having seen them were nonetheless shown to have been influenced by them in subsequent responses, associations, images, and dreams. But, as Dagenbach notes, such studies do not demonstrate episodic memory of the nonexperienced stimulus (at the time of its occurrence). For this, one has to turn to the "hidden observer."

According to Spiegel, recall of painful stimuli by the "hidden observer" involves consciousness at the time of retrieval. However, there seems to be little that is conscious, about "conscious retrieval." Only the end product of such processing enters consciousness (Miller, 1987). In any case, in TA 4.3, this phenomenon was intended to illustrate entry of information into long-term memory, without consciousness, at the time of encoding.

It is, of course, difficult to be certain about what, if anything, the hidden observer might be aware of at the time of encoding; Spiegel, Rey, Lloyd and Dagenbach suggests that it is aware of the pain. According to Spiegel, the subject is simply not "aware of being aware." The latter involves an added act of self-reflection. As Wagstaff makes clear, this issue is a continuing source of confusion in the field. Hilgard (1978), for example, writes that "The hidden part represents data that are processed and not consciously perceived," yet elsewhere implies that "the effect occurs, not so much because information is processed entirely 'without consciousness', but rather because it is perceived in 'another part of consciousness'." Cases of multiple personality present similar problems.

In TA 4.3, I suggest the hidden observer is "cognizant" (rather than "conscious") of what is going on - that under special circumstances, painful stimuli can be preconsciously processed and entered into episodic memory without first entering consciousness. Subjects' claims not to experience pain supports this interpretation, particularly when combined with evidence of reduced bleeding and salivation. This interpretation also has the advantage of parsimony. In place of multiple consciousnesses, which somehow escape the subject's awareness, nondominant (multiple) personalities and hidden observers remain unconscious until the time they become dominant. In this, they might resemble repressed memories that remain active, although unconscious.

Wagstaff attempts to be even more parsimonious. For him, there is no hidden observer to explain; hidden observer effects result entirely from social compliance. Suggestion-induced pain reduction, he argues, may be explained in socio-cognitive terms, without invoking a "hidden observer." In subsequent recall, some subjects may be induced (by task demands) to exaggerate the pain they have experienced (Spanos, 1989). My argument regarding pain encoding without awareness, however, is neutral as to why pain reduction (during encoding) takes place. In TA 4.3, the hidden observer exemplifies the ability to remember pain, without prior experience. That subjects might sometimes be induced to exaggerate, does not imply they recall no pain at all.[3]

1.4 Implicit learning and memory: a case of focal-attentive processing without awareness?

The evidence for implicit learning and memory is extensive (Schacter, 1987; Reber, 1989a). In TA 4.4, I focused on Nissen & Bullemer (1987), for the reason that their findings demonstrate implicit sequence learning in amnesics, with focal-attention, but without conscious awareness.

Block points out that amnesics might simply have forgotten[4] - but misses the fact that I have considered the same point and rejected it, on the grounds that sequence learning with the same task has been found in healthy young adults (Willingham, Nissen & Bullemer, 1989)[5] (see TA 4.4, for extensive, additional evidence). Carlson points out, however, that in implicit learning tasks, such as that used by Reber, subjects sometimes have fragmentary knowledge which might have sufficed to support their performance, and the same might have applied in Willingham, et al's (1989) study. Reber (1989a, 1989b) reviews the evidence (with normal subjects) relating to this point, and concludes that while knowledge in earlier studies might not have been completely unavailable to consciousness, knowledge acquired from implicit learning procedures is always ahead of subjects' explicit knowledge. Thus, as suggested in the target article, consciousness follows the information processing to which it relates. In any case, there appears to be no evidence of fragmentary knowledge in Nissen & Bullemer's amnesics.

1.5 Is consciousness necessary for declarative knowledge?

According to Mandler, my analysis of implicit learning ignores the implicit/procedural - explicit/declarative distinction. I agree that this distinction is a useful one, although one must be cautious about bracketing (implicit/procedural) and (explicit/declarative) together; procedural information is sometimes explicit and declarative information is sometimes implicit; sometimes we are conscious of how to do something, and sometimes knowledge that something is the case remains concealed in long-term memory. I also agree with Mandler, that focal-attentive processing is required for learning declarative information (as well as procedural knowledge). But I do not ignore this distinction in my treatment of implicit learning in section 4.4, as he claims. Nissen & Bullemer's (1987) demonstration of implicit (procedural) sequence learning in amnesics without explicit awareness (of what they have learnt) relies on this distinction, as do the many other studies cited in this section. How else could one demonstrate implicit learning without explicit knowledge? Mandler's point that consciousness is always present when declarative knowledge becomes explicit does not establish a special role for consciousness in the acquisition of such knowledge; when knowledge (declarative or procedural) becomes explicit, consciousness is always present by definition. But, as Baars (1989) notes, the contribution of consciousness to learning remains a mystery (see section 4.1).

1.6 Does consciousness exercise a veto?

Libet's (1985) experiment demonstrated that the initiation of a voluntary act develops unconsciously in the brain, about 350 msec. before the subject is consciously aware of the urge or intention to act. But the awareness preceded the motor act by about 150 msec. Libet therefore opposes my contention that consciousness does not enter into brain processing; he argues that "the subject could consciously control the outcome by vetoing the intention to act and not moving at all, or by passively or actively promoting its completion." Consequently, "There is no experimental evidence that would deny a causal role for consciousness here, although admittedly there is none to prove such a role either" (Dixon makes a similar suggestion).

I agree. Nevertheless, my generalisation is consistent with the existing experimental evidence. Viewing the brain from the outside, Libet has shown that the experienced intention to perform an act is preceded by cerebral initiation. Why should the experienced decision to veto that intention, or to actively or passively promote its completion, be any different?[6]

1.7 Is blindsight conscious?

I have argued (TA 5.2) that blindsight exemplifies overt identification and discrimination without awareness. According to Spiegel, however, when subjects are persuaded to guess about stimuli they cannot see, "The initial reports of ignorance are overridden through the addition of conscious effort...Thus conscious effort makes up for some of what the missing cortex did, and in combination with non-conscious processing, provides new and accurate information." For Spiegel, this implies that the information subjects give is partly conscious.

I do not deny that guessing requires effort. But in what sense is the effort "conscious"? Once the guess is made, the verbal response enters consciousness. But subjects remain convinced they cannot see the stimulus - a clear case of identification (of a stimulus) without awareness (of that stimulus).

1.8 Is consciousness required for creativity?

According to Keane my case for creativity outside of consciousness (in the "incubation period") is based on relatively loose or anecdotal evidence; recent experimental evidence, he suggests, gives a different picture. He argues that creative problem solving involves retrieval of an appropriate analogy followed by "a suitable analogue mapping" between the domain of the analogy and that of the problem. Recent studies with Duncker's radiation problem show that subjects previously given analogous situations (in the form of stories) did not draw on the stories unless explicitly instructed to do so (Gick & Holyoak, 1980) and that the difficulty of retrieving a given analogy increases as its semantic distance from the problem increases (Keane, 1987; Holyoack & Koh, 1987). He speculates that if these results apply generally to creativity, "there is no evidence in the literature of subjects spontaneously retrieving analogues without explicit instructions to do so....This indicates that the spontaneous retrieval of analogues, when it does happen, only happens when a conscious search is instigated." The unconscious incubation period may be devoted to the difficult business of retrieving remote analogies.

The extent to which such results do 'capture' the creative process is, of course, open to question. The fact that there is no evidence of subjects spontaneously retrieving analogies (without explicit instructions to do so) in the few artificial situations Keane describes, hardly rules out the spontaneous emergence of self-generated inspirations, intuitions and the like reported by creative artists and scientists. In any case, the creative ability to construe hypothetical universes as well as actual ones (past or present) may not be entirely reducible to analogy formation. But these caveats are largely tangential. Keane's central point is that unconscious incubation does not occur without conscious instigation (of a memory search), and consciousness is therefore necessary for creativity. In part, I agree. Generally, unconscious incubation appears to be preceded by a conscious "preparation period" involving extensive focal-attentive processing, accompanied by consciousness (of the problem to be solved, of trial solutions, and so on - see TA section 6). Consequently, from the first-person perspective of the creative artist or scientist, consciousness often seems essential.

But this does not clarify the relation of consciousness to the processes that initiate and enable creativity. As noted in TA section 9.1, a process might "be conscious" (a) in so far as one is conscious of it, (b) in that it is accompanied by consciousness (of its results), and (c) in that consciousness enters into the process. My claim is that information processing may be conscious in senses (a) and (b), but not in sense (c). In his analysis of "conscious instigation", Keane ignores these distinctions.

Keane accepts (as I do) that creative problem solving requires focal-attentive processing, at least in its initial stages. Focusing attention on a problem is usually accompanied by awareness of the problem, possible routes to solution, trial solutions, and so on - making the processing partly "conscious" in senses (a) and (b). But one cannot assume from this that the "awareness" which accompanies (or follows from) such processing is just another form of processing which "initiates" subsequent pattern recognition, analogue memory search and retrieval, information transformation and so on. Focal-attentive processing itself initiates such functioning (see TA section 7). What enters awareness results from such processing and cannot therefore enter into it ("conscious initiation" is not "conscious" in sense (c)).

Keane accepts that in some situations awareness can be operationally dissociated from focal-attentive processing (see TA section 8); but he argues that this does not apply to creativity - because "in creativity involving analogy there is.. no evidence ..(at present).. of such a dissociation." However, once one accepts that "awareness" can be dissociated from "focal-attentive processing" both conceptually and (in some situations) operationally, the burden of proof falls not on those who presume them to be dissociable (in principle) in other situations, but on those who presume them to be one and the same.

1.9 Miscellaneous mistakes.

According to Rey I have not even established the case for semantic analysis, without consciousness. The only evidence he challenges, however, relates to priming, and silent reading. Priming, he writes, "might be modularised, purely syntactic processing, involving, e.g. lexical associates." This is a non-sequitur. Priming does involve lexical associates, but the associates are semantic ones, not syntactic ones (doctor will prime nurse, but not bread, in spite of the fact that all three words are nouns). Rey writes that I need to show semantic processing in silent reading of "The forest ranger would not permit us to enter the park without a permit." But he does not explain how the appropriate meanings of "permit," with the appropriate stress patterns, could be obtained without semantic processing. He also omits the further extensive evidence for semantic analysis without consciousness reviewed in TA 1.2, 1.3, 2.1, 2.2, and 2.3 (see also Dixon, 1981).

Rey misinterprets my analysis of "sleeping on a problem" to imply that dreams are unconscious. They are not. Solutions emerge spontaneously into consciousness after unconscious incubation, whether (symbolically) in dreams or in the waking state. He writes that "it is extremely odd" I do not discuss the computational processes involved in introspection (I frequently cite Ericsson & Simon, 1984) and that I entirely ignore "second-order" propositional attitudes (e.g. thinking about thoughts). I have ignored self-awareness (also mentioned by Klein) along with many other candidates for a conscious function, not because they present difficulties, but because they can be dealt with in the same way as the many cases which are discussed. In functional terms, any system capable of self-awareness must be capable of self-representation. But if it already has self-representation, what need has it of self-awareness? As before, the functioning implicit in a first-person account, can be translated into a third-person description.

According to Rey, the dissociation of consciousness from focal-attentive processing, "is the only argument Velmans (is) prepared to raise against such proposals." It should be apparent from the above, that this distinction is irrelevant to Rey's varied proposals (and I don't raise it).

2. Are the target article conclusions an artifact of laboratory -based research?

Kinsbourne, Reznick & Zelazo, and Bowers suggest that the conclusions drawn in the target article regarding consciousness, may be an artifact of laboratory-based research. Foulkes suggests that consciousness may relate differently to dreams.

According to Kinsbourne, in laboratory experiments, subjects are instructed how to respond, told what displays to expect, practised, and given "unnatural" forced-choice paradigms. Consequently, "the results reflect performance that is to a variable degree automatic."[7] On reading further, however, it is difficult to know what Kinsbourne intends with this point. His implication seems to be that under natural conditions, consciousness does play a role in information processing (when behaviour is not automatic). However, in the following paragraph he writes "consciousness is not thought to have a role in information processing anyway. It is not process but representation."[8] This is followed by the claim that representation in consciousness is a necessary antecedent to action "when performance is not automatized." But if conscious representation is necessary for action, Kinsbourne's distinction between process and representation loses its force; in so far as representations enter into the determination of action, they do "have a role in information processing."

In any case, Kinsbourne's description of my own position on these issues, is a complete misrepresentation. I do not deny that integrated representations (of inner and outer events) produced by focal-attentive processing, are necessary for subsequent action. What I do resist is the ontological identification of such representations with "conscious awareness" (see section 9.3 below). According to Kinsbourne, "Velmans' commitment to the psychological laboratory reaches its acme when he identifies consciousness with focal attention." In TA section 8, I explicitly dissociate consciousness from focal attention. While I respect laboratory-based approaches, they are but one of a number of methods for the exploration of mind.

Bowers also criticizes the research I review, on the grounds that "it relies heavily on experimental methodology." This is certainly an unusual criticism of research. Bowers goes on to explain that experimental "effects" can be attributed to the independent variable (a stimulus) with less ambiguity than to whether the stimulus is consciously perceived (the latter is not manipulable). Consequently, whether subjects are conscious or not is "very much a matter of interpretation" and "underdetermined by data." Dagenbach also notes the "inherent difficulties in assessing processing without awareness."

As Gregson points out, around threshold, consciousness does not have a sharp edge. In psychophysics it is well established that threshold reports are influenced both by stimulus level and response criterion. The reliability of subjective reports (around threshold) has also been challenged by Lupker (1986). We can be wrong about what we experience, and at threshold levels we may be uncertain about whether we experience. Well above and below threshold, however, there seems little reason to doubt subject's reports about whether or not they are conscious of a stimulus. Attended-to connected speech, for example, is consciously perceived, although word identification takes place too quickly for consciousness to play a role (TA 2.3); in blindsight, there seems little reason to disbelieve subjects' claims that they cannot see stimuli, in spite of being able to identify them (TA 5.2), and so on.

In any case, interpretations of experimental findings are always underdetermined by data. This is particularly true in psychological research, where the bridging assumptions linking data to theory are frequently extensive (witness the ubiquitous use of reaction times in cognitive research). That alternative explanations might be available for some experimental findings is hardly a criticism of a research area. On the contrary, it is the staple diet of research.[9]

Bowers also criticises the typical research paradigm for the study of (un)conscious processing for its focus on "basic" processes; research should also focus on "whether noticed information is influential." This may be so, but this point has little relevance to the issues addressed in the target article.

Reznick & Zelazo suggest that once consciousness is recognised to be a continuous stream, one opens up the possibility "that it enters into and causally influences information processing by altering the subsequent perception of a stimulus and the probabilities of various responses." I agree that it is important to study how information is integrated over time to produce a "conscious stream" (see Blumenthal, 1977, for a review). But this is tangential to whether consciousness enters into processing. As argued in TA section 7, focal-attentive processing enables integration and dissemination of information throughout the processing system. This, and not the accompanying (dissected or integrated) consciousness, alters the subsequent perception of stimuli and production of responses. This argument does not "murder consciousness" as Reznick & Zelazo maintain. It is the elimination of a first-person perspective from psychology that achieves that dire end.

According to Foulkes, dreams integrate currently active information with prior information in a way that is both internally consistent and faithful to the diversity of the material. This reveals "the integrative, model-building, constructive, and narrative nature of the nonconscious processes that generate conscious experience" more clearly than cognitive studies of the waking conscious state (which involve "mere stimulus identification"). Given this "ambitious", constructive processing, Foulkes concludes that, "it seems less probable that consciousness, either as contents or as system (i.e. the processes generating those contents) can be epiphenomenal, even from a third-person viewpoint." I agree that dreams form an important area of study, although constructive processes operate similarly in the waking state (perceptual processing involves not only stimulus identification but also the construction of an entire phenomenal world). But, it is not clear that Foulke's conclusions follow from his premises.

As Foulkes goes on to note, "Dreaming establishes that highly elaborate conscious contents can be experienced through processing that is wholly nonconscious." That is precisely my argument. From a third-person perspective, conscious contents appear as output. Foulkes calls the "wholly nonconscious (involuntary, unintended)" system responsible for such output, the "consciousness system." I agree that there is such a system; but it is a (nonconscious) system for producing conscious experience; consciousness cannot enter into its own production.

3. The status of subjective reports.

Economos wonders about the legitimacy of subjective reports. But her judgements of legitimacy presuppose a psychology dominated by radical behaviourism. They might be acceptable as barks and grunts, she writes, but they are "not to be interpreted anthropomorphically" as being about something. If they are, psychology "is not at all like other sciences which do not ever deal in intentional matters." Thankfully, following the cognitive revolution, and the emergence of the psychology of language, psychologists no longer have to regard the efforts of other people to communicate as nothing more than fancy barks and grunts ("verbal behaviour"). Economos' anxiety about anthropomorphizing the utterances of other people just needs a little semantic therapy. If what others tell us is not about anything, conversations would be dull indeed.

Economos is right, of course, to imply that the status of subjective reports for cognitive modelling requires further clarification (Ericsson & Simon, 1984), as does the question of how utterances become intentional in humans and other animals (see Gauker, 1990; Harnad, 1991; Savage-Rumbaugh, 1990). It is also fair for her to ask why human subjective reports should be any different to verbal "protocols" produced by computers.

My own approach to this is implicit in the target article. In humans, sentience is dissociable from functioning. Human information processing operates nonconsciously. Computer models that are designed to simulate or emulate functioning, may therefore be presumed to operate in a similar, nonconscious fashion. At present the sufficient conditions for consciousness in humans are unknown. While it might have arisen in a computer by accident (or, perhaps, through being a natural accompaniment of matter of any kind) no living engineer knows how to create it by design. Consequently, verbal protocols produced by computers are nothing more than the output of nonconscious processing operations.[10]

On the other hand, we know that the sufficient conditions for consciousness are met in our own case, and that our reports may (to the best of our knowledge) be reports of what we experience. Given our shared heredity with others, we have good reason to believe that their reports (in combination perhaps with corroborating overt behaviour) may also be reports of what they experience. That we cannot be certain is irrelevant (there is no certainty in science). It is simply the most plausible inference, given the available evidence.

Gray notes that I stress the primacy of subjective reports in determining what enters consciousness - but asks whether there can be, or whether we can judge there to be, consciousness without such reports either in humans or other animals. The dissociation of consciousness from functioning, he suggests, seems to rule this out. But if we cannot make inferences about consciousness based on observations of functioning, are we not committed either to the view that consciousness in non-human animals (without language) is absent, or that we must remain forever agnostic about its presence? According to Gray, both conclusions are unpalatable, and I agree.

The thrust of my argument on this point was to establish that when subjective reports are available they provide a valid basis for assessing whether or not an item enters consciousness that is operationally separable, in principle, from measures of functioning such as discrimination, reaction time, error score and so forth. TA 2.2, for example, reviews findings that stimuli at the "subjective" threshold (at which subjects report they cannot see the stimulus) are operationally distinct from stimuli at the lower "objective" threshold (at which subjects cannot make a better than chance discriminative judgement about the stimulus); stimuli at these respective thresholds have different effects on subsequent processing (Cheesman & Merikle, 1984,1986; Dagenbach, et al, 1989; Forster & Davis, 1984; Marcel, 1980,1983 experiment 5). Cheesman & Merikle stress that in masked priming studies it is the "subjective" threshold, obtained from subjective reports, which defines the transition between whether a stimulus is or is not experienced. Below the subjective threshold (but above the objective threshold) discriminative responding to the stimulus may remain possible in the absence of reportable experience (TA sections 1 & 2 review analogous findings with other input analysis tasks). At or below the subjective threshold, therefore, discriminability cannot be the sole criterion of whether an item enters consciousness. For normally functioning human beings, able to communicate what they experience through language, subjective reports provide a more reliable criterion.

It does not follow, of course, that verbal reports provide the only indicator of what is or is not experienced, particularly when stimuli are above the subjective threshold. Pope & Singer (1978) review a range of experimental, non-verbal reporting techniques, and there are many natural forms of non-linguistic communication. The cries, chuckles, smiles, etc. of pre-linguistic babies no doubt signal something of how they experience the world and this is no less true of adults; behavioural techniques have also been developed to explore the experiences of other animals (see in particular, Dawkins, 1980, 1990). That consciousness does not enter into cerebral functioning (viewed from a third-person perspective) is tangential to this point. As I note in section 9.1, many human processes (input analysis, motor control, etc.) are accompanied by consciousness of their results - and the results of processing determine overt behaviour. Consequently, overt behaviour can provide useful evidence about what is experienced, both in other animals and other human beings.

Lloyd has similar worries about the stress the target article places on subjective reports, which he thinks, "implicitly identifies consciousness with introspection." Introspections necessarily follow what they are about, and that he argues, is the basis for my claim that consciousness follows the information processing to which it relates.

I disagree. For subjective reports to be possible, there must be something to report, i.e. a conscious experience. Subjective reports provide one method of communicating about such experience. The thrust of my argument is not that subjective reports follow the processes to which they relate, as Lloyd claims, but that experiences do so (see, for example, TA section 2.3). To argue that reports follow what they are reports of would be trivial.

Nor do I identify subjective with introspective. The target article focuses on how consciousness relates to human information processing. Consequently, much of the discussion focuses on the awareness of inner events (stimulus analysis, learning, memory, and the like). But I agree with Lloyd that under normal waking conditions, awareness is dominantly "awareness of the world." Reports of such awareness are simply reports of the phenomenal world. These remain subjective but are not introspective. I have not focused on this theme for the reason that it raises issues well beyond the scope of the target article (but see Velmans, 1990a, for an extensive discussion).

I also agree with Lloyd that experiences often cannot be fully translated into words. Nevertheless, subjects remain the best judge of whether they have or have not had an experience. As should be clear from the discussion of masked priming (TA section 2.2) and blindsight (TA section 5.2), Lloyd is wrong to "consider experience to be signalled by any discriminative response."

4. Conscious processing is different.

A number of commentators have argued that consciousness must play some role in information processing, because conscious processing is different. Wilson, for example, agrees that evidence of nonconscious processing in realms previously thought to be the exclusive domain of conscious processing requires the boundaries of conscious vs. nonconscious processing to be redrawn; but it does not demonstrate that consciousness plays no role in information processing, or that we do not need the concept of consciousness to understand how such processing occurs. He goes on to list a range of interesting social psychological studies which demonstrate that subliminal stimuli have different effects to supra-threshold stimuli and that when people consciously reflect on their feelings they respond differently from when they do not. Bowers reviews studies of problem solving which demonstrate differences in processing strategy before and after subjects have a (conscious) "hunch" about the solution. As he notes, it is difficult to understand how scientific hypotheses could be tested, if they remained unconscious. Carlson points out that explicit knowledge of patterns is typically associated with better performance than implicit knowledge in implicit learning tasks. Block, and Inhoff make the point that without conscious information, blindsight patients do not initiate voluntary action. To this one might add the long history of research into subliminal effects (see Dixon, 1971,1981 for extensive reviews), along with many other demonstrations of how processing that is accompanied by consciousness differs from nonconscious processing (see the target article). Such studies demonstrate that people often respond in a different way when they are conscious (as opposed to not conscious) of input stimuli, of their own thoughts, of their overt response, and so forth.

But this does not address the question of what role awareness plays in human information processing. The target article ranges over all the main phases of information processing (from input to output) and demonstrates that consciousness (of a stimulus, response, or intervening process) follows the processing to which it relates and cannot therefore enter into it. What enters consciousness is already a form of integrated output; perceived input has been subject to pattern recognition, thoughts that pop into consciousness have been formed by prior cognitive processing, one is aware of what one wants to say after phonemic imagery has been formed, after one has said it, and so on (TA 9.1).[11]

Few would doubt that when consciousness is present something will be different in the activities of the brain and that this might have consequences for overt behaviour. If focal-attentive processing integrates the results of prior processing with stored knowledge, expectations, goals, etc. and disseminates those integrated results throughout the processing system, this will have powerful effects not only on resulting conscious states, but also on consequent cerebral functioning and overt response.[12] Demonstrating processing with consciousness to be different is therefore tangential to whether consciousness enters into processing.[13]

Of course, one cannot rule out the possibility that a genuine counterexample may be found. To reject this possibility would require an infinite number of falsifying observations. But this has to do with the open-ended nature of science, rather than with any weakness in my case. A genuine counterexample would have to show that consciousness enters into processing (viewed from a third-person perspective), as opposed to merely accompanying it, or resulting from it (see TA 9.1) - and that the functions it performs cannot be explained more parsimoniously, in purely neural (or other physical) terms.[14]

Wilson counterargues that attributing the apparent functions of consciousness to focal-attentive processing will not do, as "without a clearer way of distinguishing between focal-attentive processing and consciousness...it is not clear what this distinction buys us." I would argue that the differences between "awareness" and properties such as "limited capacity", "seriality" and the many other operations required to produce the integrated representations which are manifested in conscious states are clear enough. It is the misidentification of "awareness" with such properties that is problematic. The ancient problem of how consciousness (understood as "awareness") could causally influence brain functioning is one Wilson does not address (see section 9.2, below). A similar problem led Leibnitz and Spinoza to reject the dualist-interactionism of Descartes.

5. The relation of consciousness to focal-attentive processing.

5.1 Confusions.

According to the target article, consciousness results (in part) from focal-attentive processing, but is not ontologically identical to it. Nevertheless it remains open to scientific investigation (TA 9.2). I found it difficult to recognize my views in some of the commentaries on this issue. Mandler, for example, accuses me of trying to "proscribe the use of the concept of consciousness from psychological science" by replacing it with "attention." Baars accuses me of trying to remove consciousness from scientific discussion by dissociating it from "attention." Carlson accuses me of "replacing the notion of consciousness with the notion of focal-attentive processing," yet notes that I reject "the identification of consciousness with focal attention." Rey, on the other hand, opposes my attempt to dissociate consciousness from attention, for the reason that he wishes to remove the concept of consciousness from psychological science. Their detailed commentaries are similarly confusing.

According to Mandler, I merely note, and do not consider further "the position exemplified by Kahneman & Treisman (1984) that attention and consciousness are not coextensive and separable." Section 8 explicitly distinguishes the concept of consciousness (understood as "awareness") from focal-attentive processing; the work of Nissen & Bullemer (1987) which operationalizes this distinction is also extensively considered in section 4.4.

Mandler goes on to list statements about the relationship of consciousness to focal-attentive processing, drawn out of context from the target article, and suggests that they are "if not contradictory, at least inconsistent"; he suggests that questions such as "is focal attention a necessary or a sufficient precursor of consciousness," have not been addressed. What is inconsistent however is not made apparent, and some of the questions he asks are explicitly answered in the target article.

The relationship of consciousness to focal-attentive processing is summarized in section 8, where I state that focal-attentive processing is a necessary (but not sufficient) condition for consciousness in humans; in some circumstances (blindsight, implicit learning, etc.) focal-attentive processing is not accompanied by consciousness. Usually, however, both the necessary and the sufficient conditions are met, and focal-attentive processing is accompanied by consciousness. Consequently, when consciousness is absent, focal-attentive processing is usually (but not always) absent. Normally, of course, the brain functions in an integrated way. Experimental disruptions of consciousness are therefore likely to disrupt those aspects of focal-attentive processing with which it is most closely associated (e.g. the dissemination of the results of prior processing throughout the processing system). This multifaceted relationship may complicate research, but it presents no logical inconsistencies.[15]

Mandler appears to agree that the mechanisms of selection and choice which determine what we attend to are preconscious.[16] Curiously, given the thrust of his commentary, he also agrees that "information processing is not conscious, but its products are." But he then asserts that "properties such as seriality, limited capacity, and relative slowness (are) usually assigned to consciousness"; he argues that if we assign them to focal attention, attention and consciousness become once again indistinguishable (which he opposes). I am equally opposed to such a misassignment of properties, but argue the opposite. Consciousness primarily refers to "awareness". Seriality, limited capacity, slowness, and the like, are properties of human information processing. This, Mandler admits, is not conscious. If we assign these properties to consciousness attention and consciousness become indistinguishable.

Rey describes my attempt to (partially) dissociate consciousness from focal-attentive processing is "oddly philosophical" and "invalid." I point out that consciousness primarily refers to "awareness," whereas "focal-attentive" processing refers to a functional subdivision in an information processing model of the brain. Rey points out that "differences in..definition or reference fixer do not entail differences in reference." I agree. But neither can ontological identity be assumed. Rey notes that we can imagine consciousness and focal-attentive processing coming apart, "but it doesn't follow that it is genuinely possible that they do." Given that my case makes no reference to what can be "imagined," I fail to see the force of this. In TA 8, I argue (a) that consciousness is normally dissociated from focal-attentive processing conceptually, (b) that they are dissociable operationally, and (c) that various cases exist where they are actually dissociated, at least in part. Rey claims he doesn't "see the evidence for it, either in this paragraph (TA 8, para 7) or elsewhere." But, the evidence is given in TA 8, paras 4 and 6, TA sections 4.3, 4.4 and 5.2 - and added evidence is given in the commentaries by Libet and Stanovich.

According to Carlson, my analysis should start with an "explicit theoretical description of the term consciousness," using cognitive science concepts such as representation, intentional state, and working memory. Given that target article aims to distinguish consciousness (in the sense of "awareness") from such concepts, this does not seem to be very useful suggestion. He criticises my attempt to relate consciousness to focal-attentive processing on the grounds that the latter concept is "not developed, and is hardly less vague than the concepts of consciousness or awareness." The study of focal-attentive processing, however, has been central to psychological investigations for almost 40 years (following the seminal work of Cherry, 1953, and Broadbent, 1958 - see TA 1.1 to 2.5). Carlson goes on to complain that "it is very difficult to know just what hypothesis about consciousness is being rejected." In TA 9.1, I suggest that a process may be "conscious" in so far as one is conscious of it or of its results, but not in the sense that consciousness enters into the process. Carlson misreads this as: "information processing (is) unconscious because outputs or contents but not processes can be reported," which he describes as, "either trivial or absurd." He argues that inferred relations amongst conscious states, describing their succession and consequences for behaviour, do not themselves enter consciousness. So why should theoretical descriptions of cognitive processes do so?

This combines a misreading of my text with a non-sequitur. Inferred relations or theoretical descriptions are abstractions. They may be said to be "conscious" only in so far as they are exemplified in experience. Relations amongst conscious experiences are by definition conscious in this sense. Cognitive processes such as problem solving, planning, and the like may also be (partly) conscious in this sense, and can be reported (conscious in sense (a) - see TA 9.1). However, being conscious of a process (or of its results) does not establish a causal role for consciousness in human information processing.

Carlson goes on to claim that my references to (unconscious) information processing being sophisticated or complex is "theoretically loaded." If our theories of intervening processing are incorrect, he argues, it is hardly surprising that we are not aware of them. A connectionist theory of word recognition, for example, would have "very different implications for arguments about consciousness" than one postulating intervening representational states.

This is another non sequitur. We are no more aware of connectionist processes at work in our own brains, than of representational ones (whether simple or complex). In any case, according to TA 9.1 (note 15), without a first-person perspective, cognitive models that deal only with the way input is transformed to behavioural output remain incomplete whatever their internal mechanisms.

5.2 Misidentifications.

For Baars, awareness and focal attention "covary so perfectly, we routinely infer in our everyday life that they reflect a single underlying reality." My target article 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," and to "demonstrate that consciousness cannot be associated with all of its obvious correlates --- in this case with "focal attention."

This is a complete misreading of the text, which emphasises throughout that consciousness is associated with focal attention (it results from focal-attentive processing). I merely deny their ontological identity (causes are not ontologically identical to their effects). If Baars means to imply that "reports of conscious experience," "focal-attention," and "wakefulness" reflect a single aspect of our nervous system, he is evading the complexities involved.

Baars goes on to complain that "we can in principle call human experience anything we like: focal attention, wakefulness, awareness, consciousness..and so on..The trouble, if we never go beyond these words, is that they do not allow us to call a spade a spade."(my italics). I also argue that we should call a spade a spade. Consequently we cannot call human experience anything we like. Baars (like Mandler) accuses me of behaviourism (which I oppose in TA 9.2). Unfortunately, he seems blind to the similar dangers of unreconstructed cognitivism. For radical behaviourists, all talk of mind could be translated, without scientific loss, into talk about behaviour. For the new "radical cognitivists" all talk of mind (including consciousness) can be translated, without scientific loss, into talk about information processing. But the effect of calling "consciousness" (or "experience") "focal attention," is to remove the subject's experience from science (collapsing the subject's first-person perspective to the external observer's third person perspective - see TA 9.3).

Curiously, Baars appears to agree. He writes that, "denial of first-person conscious experience in other people may lead to a profound kind of dehumanization. It comes down to saying that other people are not capable of joy or suffering, that in fact, as far as the outside observer is concerned, we are not to see others as they see themselves. The consequence of this prohibition against the first-person perspective is a kind of mechanization of other people. Psychology under the thumb of behaviorism did indeed display this kind of dehumanizing, mechanistic thinking. It is only when we acknowledge the reality of conscious experience in the minds of others, that we can recognize their full humanity" (my italics). I can do no better than to commend Baars to his own eloquent writing.

Mangan accepts that neurological processes responsible for focal attention also produce conscious effects. But he asks why, if they are in principle dissociable, do they remain so closely correlated? Consciousness might, for example, contemplate the fine points of digestion, while focal-attentive processing played the stock market, etc. He suggests that the straightforward answer is a functional one. Consciousness performs the functions it appears to perform, which overlap at times with some of the nonconscious processing mechanisms which support it. I think there is a more straightforward answer. The conditions for consciousness in the human brain are (in part) produced by focal-attentive processing. That is why they correlate so closely.

Spiegel also stresses the close relation of consciousness to give brain events; the P300 component of the ERP, for example, has been found to increase with conscious awareness of stimuli (Posner, 1978); consciousness also seems to relate closely to the dissemination of information throughout the processing system. From this, he concludes, that consciousness cannot be "easily extracted from cognition." However, few would deny that consciousness is closely related to brain events. The question Spiegel does not address is what is the nature of this close relationship.

5.3 Prevarications.

A number of other commentators prevaricate over the relation of consciousness to focal-attentive processing. Klein, for example, agrees that consciousness is not identical to focal-attention and primary memory, although it is closely related to these functions. Yet later, he writes he is "not convinced" by the evidence for their dissociation (but he doesn't give any reasons). Van Gulick agrees with me that consciousness in the sense of "awareness" should "not just be equated with, or used as an alternative name for focal-attentive processing." But he goes on to equate consciousness with (aspects of) focal-attentive processing in his claim that, "many of the states of brain activity that constitute focal attentive processing are in fact identical with states of conscious awareness."

5.4 Reversals.

In the target article, I suggest that focal-attentive processing provides at least some of the conditions for conscious experience. What enters consciousness follows the processing to which it most closely relates (and cannot therefore enter into it). By contrast, Block argues that information cannot pass to the executive without first entering consciousness. Consequently, consciousness is necessary for executive processing (including focal-attentive processing). He claims the TA shows that some things can be done without consciousness, but that these cases all deal with processing in specialised modules. However, he selects only those aspects of my review that fit in with his claim. In fact, much of the target article focuses on how preliminary ("specialised module") processing relates to focal-attentive processing (TA sections 1 & 2) and more importantly, how consciousness relates to (and can be dissociated from) focal-attentive processing (TA sections 8 & 9.1).

Block goes on to list some of the cases I consider (Nissen & Bullemer's experiments, avoidance of motor car accidents, creative inspirations, and disconnection syndromes such as blindsight). But his intention is to dismiss them, or argue that they are evidence for his case. For example, Block accepts that creative thought is a function of executive processing "which is unconscious except for the conscious result of the processing, i.e. the sophisticated idea itself. So we have executive activity without consciousness." He notes that this seems to pose a problem for his theory.

I agree. In creative problem solving, unconscious "incubation" is preceded by conscious "preparation" (see 1.7 above); but the time interval between conscious preparation and the emergence of a solution may be considerable. Before "sleeping on a problem," for example, focal attention is accompanied by consciousness. But it seems perverse to suppose that during sleep no further information reaches the executive - else how could a solution "pop into consciousness" on waking? It is central to Block's model that information does not reach the executive without first entering consciousness. But his only defence is that his model, "does not dictate that everything that happens in the executive is passed to consciousness." Given that spontaneous inspirations are cases where things happening in the executive are passed to consciousness, this defence is obscure in the extreme.

Block's use of blindsight as evidence of "no executive action without prior experience" is equally obscure. This condition is called "blindsight" for the reason that subjects are able to identify shapes projected to their blind hemifield, reach for glasses, etc., in spite of the fact they cannot see them, no matter how closely they attend to what is going on, or how well they are in control of their decisions and motor movements. This appears to be a clear case of executive action without prior conscious experience. Block counters that blindsight patients have not as yet been trained to spontaneously use the nonconscious information. Why this should be evidence of the priority of conscious information is not clear. As suggested in TA section 8, blindsight subjects may not be able to use input information because it is not generally available throughout their processing system, preventing its full integration and contextualisation (even though it remains available in a relatively "raw" form to the executive).[17]

Block's model 1 provides a clear contrast with my own. For Block, consciousness occupies a box within an information processing model of the brain. That is exactly what I oppose. But Block's description of my case misrepresents it, and Block's defence of his own case misrepresents the evidence.

6. What is a conscious process?

In my analysis of what is meant by a "conscious process" (TA 9.1), I argue that one may be conscious of a process (or of the results of a process) without consciousness entering into processing. A fortiori if one is not even conscious of a process, it makes no sense to speak of consciousness entering into that process. One has no awareness, for example, of how information is integrated and disseminated throughout the processing system. Consequently, such functioning cannot be conscious.

However, according to Van Gulick, this conflates "control access with introspective access." Consciousness might provide high level control over low level processes to which it does not have introspective access in the same way that high level computer instructions control machine events, via a compiler. "Might not conscious awareness exercise a similar control over memory processes or over the processes coordinating muscle movements despite my introspective ignorance of the details of those processes?" In principle, this might be so, but the argument is only as good as the analogy.

Unlike computer programs, which provide a high level functional description of machine language operations, an experienced wish to learn or recall something gives no indication of how encoding or retrieval processes function. To describe such processes as under conscious control seems like a contradiction in terms. Motor processes are more complex in that one may be conscious of motor movements at the focus of attention. Again, however, one needs to ask, just how conscious is "conscious control"? This is extensively discussed in TA 5.1 to 5.4. To exercise control, program instructions must precede machine language operations. What enters consciousness, however, appears to be the result of focal-attentive processing. In speech, for example, there is a sense in which one is only aware of what one wants to say, after one has said it. Even the experienced urge or intention to act, appears to be preceded by (and to be the result of) preconscious neural events (see 1.6 above, and TA 3).

Dretske also elaborates on what is meant by a "conscious process." He distinguishes consciousness of a process, or consciousness of the results of processing from an active role for consciousness in processing, as I do in TA section 9.1. According to Dretske, however, a conscious process should be redefined as, "the act of awareness, the process by which we are made aware of whatever objects we are aware of. P is a conscious process (in S) if there is something, X, that P makes us aware of." This presents difficulties.

The process (or processes) which make us aware are usually thought of as being neurophysiological ones (involved in perception, emotion, thinking, and so on). They may cause awareness or (in the case of the neural correlates of consciousness) accompany awareness, but they are not acts of awareness at all! Awareness does not play an active role in the processes which make us aware.

Dretske rightly points out that objects or events are not themselves conscious, simply because we become conscious of them. This applies equally to internal events and processes (heartbeats, food in the throat, etc.). But he says some internal events and processes (pains, visual and auditory sensations, etc.) are conscious (in the relevant sense) because they are acts of awareness, and constitute awareness - not by virtue of us being conscious of them. I agree that pains, auditory sensations and so on are "conscious" by virtue of constituting consciousness (they are amongst the contents of consciousness). Indeed, this applies to the entire phenomenal world.[18] But, this does not make them "acts" or "processes," as Dretske claims. Nor, as he later suggests, can we locate consciousness in "the processes that underlie and constitute our consciousness of things, for the simple reason that brain processes which underlie (in the sense of cause or correlate with) consciousness, do not at the same time constitute its contents.

In his attempt to find a role for consciousness in human information processing, Dretske redefines awareness as an act or process by which we become aware. This confounds conscious contents with antecedent neurophysiological processing, rather than clarifying their relationship.

7. Philosophical implications.

7.1 Epiphenomenalism

According to Baars, Block, Lloyd, Mangan, Van Gulick, Kinsbourne, Corteen, Rey, and Hardcastle, I support Epiphenomenalism and their commentaries are largely devoted to the case against it. Let me repeat at the outset that I do not support epiphenomenalism. In TA 9.3, I argue that from a third-person perspective epiphenomenalism appears true, but from a first-person perspective it appears false - an apparent paradox, that any complete theory of the mind must offer to resolve. For these commentators however, the first-person perspective is not to be taken seriously. If consciousness plays any causal role that matters, it must do so from a third-person perspective.

My argument is that the instances of consciousness not playing a causal role reviewed in the target article, exemplify a general relationship. Viewed purely from a third-person perspective, consciousness is (inferred) output, and consequently epiphenomenal. The manifest implausibility of this (given the many cases where we cannot function without relevant experience) arises from our first-person perspective. Consequently, any complete account has to take the first-person perspective seriously (I return to this in 9.1 below).

7.2 Dualist-interactionism.

MacKay is a dualist-interactionist (in the tradition of Eccles, 1980), and claims consciousness to be the "king of the neuronal processors." It "sets the goals," "issues commands, "plays power broker," and has the ultimate "veto." It also "synthesizes the highly distributed information from neuronal processors into unified percepts." "Consciousness coexists with the nervous system in symbiotic relationship."

What I share with MacKay, is his anti-reductionism, his intuition that it is no recent arrival, and his conviction that the phenomenon is basic. I also agree that there are mysteries here. There is something happening in the brain which both enables functioning and awareness. But, in allocating the highest brain functions to awareness, MacKay attempts to straddle this divide, with all the usual, attendant problems. For example, he does not address how a symbiotic consciousness could set goals, issue commands, etc., or why such functions could not be performed by the brain itself. Nor does he address the evidence (reviewed in TA) that consciousness follows the processing to which it relates - that once information enters consciousness it is already integrated, and that the operation of human information processing is unconscious. Dixon's suggestion that hydrocephalus cases might show "how consciousness can take over when the usual machinery for complex cognitive activity is grossly impaired" faces similar problems. For these reasons, I adopt a monist view. Functioning and awareness are the manifestations of one process viewed from different perspectives (see 9.3 below).

Mangan is also a dualist-interactionist, but accepts that there is a mystery about how (a nonphysical) consciousness could interact with the physical brain. He rightly points out that epiphenomenalists face the same problem. This is one of the reasons I reject both dualist-interactionism and epiphenomenalism, in favour of a "complementary perspectives" approach.

7.3 Functionalist reductionism.

Many commentators have tried to show that consciousness deserves its own "box" within a functionalist model of the mind. By contrast, Sloman, Hardcastle, Rey, and Stanovich attempt to "deal" with consciousness by eliminating it.

If consciousness is distinguishable from functioning, yet open to scientific investigation, as my target article claims, this eliminative programme fails. Accordingly, Sloman seeks to undermine my case by arguing that, "people who discuss consciousness delude themselves in thinking that they know what they are talking about...it's not just one thing but many things muddled together" - rather like our "multifarious uses of "energy" (intellectual energy, music with energy, high energy explosion, etc.)." Stanovich likewise points out that "the term "consciousness" fractionates into half a dozen or more different usages." Given this, they argue, one can make no generalizations about it.

I disagree. There is nothing to prevent organized discussion of a specific usage of a term. The target article, for example, picks out consciousness in the sense of "awareness." The contents of consciousness may be indefinitely varied, but this does not mean their dependence on physical stimuli, sensory physiology, and the like, cannot be studied systematically within perception, psychophysics, and so forth. Nor does this prevent investigation of the conditions under which each of these contents become conscious. There is extensive evidence, for example, that events enter consciousness only if they are at the focus of attention, irrespective of their variety (see TA 1 and 2). The neural conditions for consciousness can also be experimentally explored (see, for example, Libet; Koch & Crick; and TA 9.1). In short, the conditions for the existence of consciousness can, in part, be distinguished from the added conditions required to produce its varied contents. The dissociation of awareness from information processing (proposed in TA 9.1) applies in every case (irrespective of the object of awareness). The analogy Sloman draws between metaphorical uses of the term "energy" and nonmetaphorical use of the term "consciousness" is therefore a false one.[19]

Sloman's attempt to fragment consciousness is followed by an attempt to eliminate it from the analysis of mind altogether, to be replaced by a study of capabilities. "If we give up the idea of a unique referent, we can instead survey relevant phenomena, analyze their relationships to other capabilities ... and try devising mechanisms capable of generating all these capabilities, including self-monitoring capabilities." He goes on to discuss architectures that might support monitoring, information integration and higher level control. I have no objection to the study of capabilities or to the investigation of architectures that instantiate them, but if my TA argument is right, a psychology that speaks only of capabilities and their embodying architectures has nothing to say about consciousness at all (whether fragmentary or unified)!

This is neatly illustrated by Sloman's final conjecture, "That this..design-based strategy for accounting for phenomena supporting talk of consciousness will eventually explain it all. We'll have evidence of success if intelligent machines of the future reject our explanation of how they work, saying it leaves out something terribly important, something that can only be described from the first-machine point of view.."

I agree. If consciousness is dissociable from functioning, a nonconscious intelligent machine would accept a complete explanation of how it functions as having left out nothing. Conversely, if the machine were conscious, and consequently rejected a purely functional account, that would be because this said nothing about its "first-machine" perspective.

For Hardcastle, such consequences are of little concern. If consciousness is not captured by (a third-person) psychology, so be it; "consciousness could simply be outside the domain that psychologists are trying to capture ... Whether an information processing model is complete depends on what it is explaining." In any case, she argues, reduction between perspectives is relatively straightforward, "since science regularly and nonproblematically redescribes the way the world seems to us from a first-person perspective in third-person objective terms. To wit, objects which appear red to us do so because they reflect a certain wavelength of electromagnetic radiation. Surfaces which seem warm to us do so because their mean molecular kinetic energy is above a certain level relative to the MMKE of our skin. There is no reason why consciousness should not be reducible in the same way."

However, it should be clear from the target article and the commentaries, that many psychologists are concerned to capture consciousness; and Hardcastle's old reductionist argument (see, for example, Place, 1956) is based on a simple error. It assumes that if C is shown to cause E, then E reduces to C. A sensation of redness might be caused by certain electromagnetic wavelengths interacting with the colour coding mechanisms of the visual system, but that is not to say that the resulting sensation is nothing more than "electromagnetic radiation." On the contrary, such cases exemplify the need to incorporate first- and third person perspectives into any complete cause-effect description. From a third-person perspective, electromagnetic energy can be observed to innervate the eye and visual system; but there is no way of knowing that this results in a "red" sensation, without incorporating the subject's point of view. And the same applies to all other contents of consciousness.

To be sure, the red sensation is (the subject's experience) of something going on in the world; and for the purposes of physics it may be useful to redescribe that event (in the world) as electromagnetic radiation. The ability of the visual system to translate electromagnetic frequencies into colour sensations, however, is what is of interest to psychology.

Hardcastle goes on to dismiss consciousness as cognitively irrelevant "noise." She maintains, "we have simply talked ourselves into believing that consciousness has certain causal powers ... perceptual experiences tell us nothing." Ironically, one of the most remarkable facts about consciousness is its well-integrated nature. In spite of the complexity of antecedent preconscious processing, what we experience is not "noise," but a coherent phenomenal world. From a first-person point of view, such meaningful experience forms the basis of all our activities. One does not need to talk oneself into believing that the contents of consciousness have causal powers. If one ignores what one experiences one is likely to be run over by a bus.

Hardcastle argues (following Wilkes, 1988), that some languages do not have a word for "consciousness" (including ancient Greek) and that our assumptions about its causal power is peculiar to "our epistemic milieu." The notion that man is something more than a body, however, is evident in the funereal furniture and artifacts meant to accompany the souls of the dead in palaeolithic graves, and in Egyptian and Assyrian mythology. Dualist-interactionism, furthermore, originates not with Descartes, but with Plato's deliberations on the nature and activities of nous. That some languages do not have an exact word for consciousness is irrelevant. As Block, Sloman and Stanovich point out, one would have to include English amongst their number. What is important, is that all languages describe causal sequences from a first-person perspective. It is the attempt to eliminate the first-person perspective that is peculiar to "our epistemic milieu."

Rey and Stanovich agree with me that consciousness as we normally understand it, cannot be identified with any psychological process. However, they draw entirely different conclusions. I argue that if information processing models do not encompass consciousness they are incomplete. Rey argues that this is "a reason to disbelieve in consciousness altogether." Stanovich calls consciousness a "botched concept"; a psychiatric institution is too good for it; it deserves the "death penalty." Not content with throwing the baby out with the bathwater, they throw out the bath as well.

In TA 9.2, I point out that in studies of perception, psychophysics and so on, the contents of a subject's experience may be thought of as a form of output (of perceptual and other processing); as such, it is (and always has been) a legitimate object of study. According to Rey, "this amounts to reneging on epiphenomenalism: for how can something without causal consequences be legitimately inferred from such data?" This is a non-sequitur. Epiphenomenalism, as defined by Thomas Huxley, claims consciousness to be a form of output, caused by brain processing, which in turn has no causal influence on that processing. In any case, I do not adopt epiphenomenalism, for the reason that it has nothing to say about causality viewed from a first-person perspective.

Rey's opposition to this is deep. He writes that consciousness might seem to make a difference, but, "So does God, the soul and contra-causal freedom." According to him, the very idea that causal claims are perspective-dependent is suspect, because "Causal claims are about the world independently of perspective." One does not need to be a relativity theorist to doubt this. Causal claims may about the world, but they are never independent of perspective. They are always about events viewed under given conditions, conceptualized from a given point of view (this point is so well-established that I won't labour it).

Rey concludes by denying the existence of consciousness, comparing my faith in the existence of consciousness to a theologian's faith in the existence in God: "Why in the world should one believe in such a God? Why should one believe in such a consciousness? In both cases, of course, people have been tempted to say, "Because I have direct access to it." But such first-person breast beating begs the question...the challenge..is to come up with some non-question-begging reason to believe consciousness exists. I doubt there is any to be had."

This departure into rhetoric is a poor substitute for common sense. In denying the existence of consciousness, Rey is denying the existence of all its contents. Not just love and hate, pleasure and pain, and other inner events such as thoughts, images and dreams - but also the experienced body and the entire phenomenal world (including visual experiences of meter readings, brain events in others, etc). Given a denial of this order, not even a direct line to the almighty will save his argument.

Stanovich focuses on the utility of "consciousness" for psychological theory. He notes that "there does not exist one single model in cognitive psychology where "consciousness" plays a necessary theoretical role." In so far as such models deal solely with human functioning viewed from an external observer's perspective, I agree. But that is not a reason for eliminating consciousness. In models of perception, for example, consciousness is one (inferred) output of human information processing. While resource use, obligatory vs. optional execution, controlled processing, and the like should not be confused with whether or not information enters consciousness, the relation of consciousness to focal-attentive processing is close, and deserves more detailed experimental examination (see TA section 8). Processing accompanied by awareness, differs from processing not so accompanied, in many experimental domains. While this does not establish the causal efficacy of consciousness within such processing, what enters awareness is of interest in itself. It can also provide a valuable indicator of functional differences in processing. According to Stanovich, talk of consciousness is "ineffable" and invariably confuses discussions of functioning. There is nothing confused, however, in Gardiner's review of the way differences in phenomenal awareness relate to functional differences in memory, in Libet's comments on how the relation of first-and third-person perspectives can be investigated experimentally, or in Shevrin's point that therapy primarily relies on the client's first-person perspective. Worries about the "ineffability" of first-person accounts presuppose consciousness to be a form of "ghost" lurking around the machinery of the brain. Reductionists, such as Stanovich, understandably call for the exorcist. Elsewhere (Velmans, 1990a), I have argued that this classical dualist vs. reductionist way of viewing consciousness, is grounded in fictitious presuppositions about its "ghostly" nature that the protagonists share. Space prevents me recapitulating here an alternative, reflexive model of how consciousness, brain and the physical world relate to each other. Given the thrust of the commentaries, however, it is essential to elaborate on my brief introduction (in TA 9.3) to first- vs third-person perspectives.

8. First-person vs third-person perspectives.

By "first-person perspective," I simply mean how things appear from the subject's point of view; "third-person perspective," refers to how events relating to the subject appear to an external observer (e.g. an experimenter). Included in the subject's first-person perspective are not just inner experiences (thoughts, images, emotions, and so forth) but also the body as-experienced, and the physical world as-experienced, extended in three-dimensional space beyond the body surface. All these contents are included in what the subject experiences, and are no more "ineffable" than what the experimenter experiences while observing the subject. Whether a perspective is "first-person" or "third-person" depends solely on the relation of the observer to the event of interest. A subject, adopting a first-person perspective, is focused on his own experience and consequent behaviour; an experimenter, adopting a third-person perspective, is focused on the brain functioning, behaviour, and (inferred) experience of a subject.

In TA 9.3, I focus on first- and third-person perspectives of a subject's information processing; consequently I suggest that the experimeter's and subject's access to that processing is asymmetrical - via exteroceptors for the experimenter and interoceptors for the subject. Symmetries and asymmetries of access are complex, however. Subject and experimenter, for example, have symmetrical access to events out-there in the world (via exteroceptors). Viewed in this way, differences between first- and third-person perspectives are far more subtle than generally thought.

Stanovich, for example, argues that first-person accounts are largely a cultural product and contaminated by our folk theory of the mental. This may be. But reports of external objects and events, are equally a cultural product, paradigm-driven and theory-laden. To be sure, first-person accounts are corrigible, and need to be assessed in the light of all available evidence (including third-person evidence). But third-person accounts are also corrigible (if they were not, there would be nothing for cognitive scientists to discuss).[20] Whether first-person or third-person accounts are more useful, depends entirely on the explanatory context (TA 9.3).

Stanovich cites Churchland (1985), who casts doubt on the content of first-person (vs third-person) reports, on the grounds that "the introspective discriminations we make are for the most part learned; they are acquired with practice and experience, often quite slowly. And the specific discriminations we learn to make are those that it is useful to make." But the partial dependence of discrimination on learning, applies to perception of inner and outer events alike. It does not make first-person reports more suspect. In any case, it seems perverse to claim that our ability to distinguish pleasure vs pain, images vs percepts, auditory vs. visual experiences, and so on, are "contentless", slowly acquired, culture-bound discriminations. Babies seem to manage quite well! It is trivially true that one cannot report on inner events prior to learning the rules of the language game (this applies equally to external objects and events). But this hardly means there is nothing to report! - or that, (following Dennett, 1988), "qualia" have "no properties or features at all."

Schaeken & d'Ydewalle accept that whichever perspective one adopts, there is always an observer and an observed. But they go on to assert that what distinguishes these perspectives, is the object of observation - that the first-person perspective enables only the investigation of consciousness, whereas the third-person perspective enables investigation of the entire mind. Consequently, they claim that, "The small focus of the first-person perspective can be the cause of small and distorted theories, whereas the broadness of the third-person perspective can take into account as much factors or objects as needed, including the one used in the first-person perspective." I disagree.

The "object of observation" depends entirely on the focus of interest. From a first-person perspective, consciousness itself might be the focus of interest, but more often the focus will be on what one is conscious of. Whichever is the case, subject and experimenter may have the same focus of interest (the experimenter, for example, might be interested either in the subject's awareness (its qualities, causes, etc.) or in the processing of which the subject has some awareness (in problem solving, planning, and so forth).

Whether a given perspective is "broader" or relatively "narrow and distorted" also depends on the focus of interest. First-person accounts reveal little about human information processing (the subject's access to his own processing is limited). Third-person accounts, however, reveal little about how subjects experience themselves in the world - and for psychology this is also of theoretical and practical interest.[21] A client's depression, for example, might be "broadly" understood in terms of loneliness, or lack of love - and therapy, as a further exploration of experience. By contrast, a third-person focus on brain chemistry, antidepressants, and the like, may give a relatively "narrow and distorted" picture.

As Shevrin notes, "Psychoanalysis is entirely based on introspective first person reports recounting the individual's own understanding of his/her actions." These reasons or causes may not be currently available to introspection. But they may be explored from a first-person point of view, to produce a deeper awareness of ones "own beliefs, desires, wishes, memories, judgments, etc.," which does not reduce to "the nonconscious information processing associated with these beliefs, desires, etc." Thus, there may be "An unconscious from a first-person perspective."[22]

This is inconsistent with Schaeken & d'Ydewalle's conclusion that "In psychology, there is no need for an additional first-person perspective, but there is a need for a more complete third-person perspective which incorporates the first-person perspective." Accounts of how subjects experience themselves in the world can be translated into accounts of brain processing, but there will always be something lost in the translation. While first-person accounts are complementary to third-person accounts, they cannot be reduced to them.

9. Complementarity.

9.1 Neither first-person nor third-person accounts are privileged.

It should be clear from the above, that I not wish to remove consciousness from models of mental functioning, as Wilson implies, and that in my analysis consciousness does not "shrink dramatically" as Lloyd maintains; nor do I view first-person and third-person accounts as "mutually contradictory" as Bowers (falsely) claims. I simply argue that information processing models which view humans only from a third-person perspective are incomplete. As noted in section 9.3, first-person and third-person accounts are complementary, and mutually irreducible. A complete psychology requires both (see also Velmans, 1990b).

But, according to Lloyd, the complementarity I propose, "is an odd one since one main conclusion is that subjective impressions of the efficacy of consciousness are wrong." This misreads TA sections 9.1 to 9.3, but Lloyd cannot be entirely blamed for this interpretation. The bulk of my review necessarily addresses the relation of consciousness to human information processing from within the current information processing paradigm. This views what is going on in the brain from an exclusively third-person perspective. My argument is that if one views matters only from this perspective, not only does consciousness appear to "shrink," if one pursues it closely enough, one chases it out of the system altogether! However, as consciousness "shrinks" from a third-person perspective (as the attempt to reduce it to an information processing function progressively fails), it regains its true, first-person perspective status. Lloyd rightly points out that the contents of consciousness encompass the entire experienced world. And, from a first-person perspective, instances of the causal efficacy of conscious states are innumerable (TA section 9.3). These 'subjective impressions' would only be "wrong" if the third-person perspective were somehow privileged. This, I reject.

The phenomenal world, experienced body, and experienced inner states such as thoughts, emotions, and so on, are just as "real" (viewed from a first-person perspective) as are observations of the brain states of others (viewed from a third-person perspective). The complementarity I propose is genuine. Lloyd proposes an alternative 'complementarity' which identifies "states of consciousness with causally active cognitive states." This reduces the first-person perspective to the third-person perspective and is not complementarity at all.[23]

9.2 A resolution of the causal paradox.

Libet agrees with my proposal that first- and third-person perspectives are complementary and mutually irreducible (the need to relate these perspectives has been central to his own experimental work). But he thinks I use complementarity to "eliminate the need even to raise the question of whether there is a causal relationship between a brain state and a conscious experience," thereby begging the question at issue. This is not my intention.

To begin with, I agree with Libet that there appear to be many cases of brain modifications causing modifications in experience. As Thomas Huxley observed, "one has only to stick a pin into oneself to perform a sufficient demonstration." I also agree with Libet (against Huxley) that there appear to be many cases of experiences causing brain and bodily changes. At the same time, viewed from a third-person perspective, there seems to be no need to invoke consciousness to explain the activity of neurons (see TA). Nor, from an everyday first-person perspective, does one need neurophysiology to explain what one experiences.[24] Indeed, it not obvious how events as apparently different as brain states and experiences could have causal influences on each other. As Mangan points out, this problem applies to epiphenomenalism as much as to dualist-interactionism. This forms a "causal paradox" which any adequate theory of how consciousness relates to the brain must address.

I argue that this paradox can be resolved by recognising that such accounts view matters from different (first- and third-person) perspectives; some explanations view matters from both perspectives, although we seldom acknowledge that they do so. Scientific accounts (of other people) are normally based on what can be observed of them, viewed from an external, third-person perspective. Everyday accounts of ourselves are normally based on how we experience ourselves. But we take it for granted that others have a first-person perspective, and that a third-person perspective can be adopted toward ourselves (that we can be viewed as others see us, as bodies, brains, and so on). Accordingly, when explaining others or ourselves we sometimes give mixed perspective explanations.

Explanations of causal interactions between brain states and conscious experiences are of this kind. Libet's (1985) study, for example, investigates how the readiness potential (viewed from a third-person perspective) relates to an urge or intention to act (as experienced by the subject). In the studies of Lassen, et al. (1978), and Roland & Friberg (1985), imagined movements (experienced by the subject) were accompanied by increases in local cerebral rates of blood flow (observed by the experimenter). Such accounts of causal sequences switch from how things are experienced by the subject to how things are experienced by the external observer, but each perspective is legitimate (a complete psychology requires both).[25] How these perspectives relate can only be understood through experimental investigation. Accordingly, I do not aim to "inhibit further experimental exploration of the possibilities of causal interaction between neural and conscious states," as Libet fears. My aim is to encourage it. The target article demonstrates that first-person accounts can be translated into third-person accounts, but they cannot be reduced to them.

9.3 Complementarity at the interface of consciousness and brain.

The possibility of translating what is experienced into (third-person) functional terms, is in any case implicit in the widely-held assumption (which I adopt) that every distinct conscious experience has a distinct neurophysiological correlate[26]; information (about inner or outer events) experienced by the subject will therefore appear to the external observer to be coded in a neurophysiological form. According to TA section 8, this neural representation (like it's conscious accompaniment) is produced by focal-attentive processing; it is an integrated state, combining information relating to current events with expectations, needs, goals and so forth, in a way that enables its dissemination throughout the processing system.[27]

As Economos notes, this is a form of dual-aspect theory (in the tradition of Spinoza). In so far as conscious states and their neural correlates encode identical information this is also a form of "identity theory" as Libet (and Mandler) suggest. But as Economos rightly insists, it is not traditional reductionist identity theory. Conscious states no more reduce to neurophysiological ones (or the reverse) than pictures on television screens reduce to the magnetic codings on video tapes which encode identical information.[28]

This position contrasts with that developed by Van Gulick, and Navon, who have similar views about consciousness being related to the integration and dissemination of information, but claim consciousness to be identical to the brain processes involved.[29] Van Gulick, for example, suggests that conscious awareness is "identical with (or at least part of) the process by which information is broadcast." This collapses the subject's first-person perspective to the external observer's third-person perspective. The difficulty with such a reduction is illustrated by Van Gulick's subsequent comment that "We should not forget that our conscious awareness is typically of a phenomenal world in the Kantian sense." It is important to add that the phenomenal world constitutes (the major part of) the contents of consciousness (there is no extra experience of a phenomenal world, in the "head","brain" or "mind"). While information about this world as-experienced may be coded in (suitably integrated) correlated neural states, it does not make sense to reduce the phenomenal world to a state of the brain (see Velmans, 1990a, for a detailed discussion).

Navon's position is more complex. Whereas I identify consciousness with "awareness," and distinguish consciousness from its (integrated) neural correlates (with their attendant functions), Navon distinguishes between two forms of awareness. There is "awareness in the functional sense" which is (identical to) "the availability of information about the output of a process to other processes within a community of processes"; and there is "consciousness in its common usage" - which might be "the experiential facet of awareness defined in this (functional) way." Navon accordingly concludes that, "consciousness (as function) can be disentangled from awareness." I think this is a good way to tangle them up. As Navon himself points out, "the state of information dissemination can also be presented in terms of information processing without any reference to consciousness." So why should one call this "consciousness" or "awareness"? In any case, my claim is that conscious experience does not enter into information processing.

9.4 Complementary views of consciousness and evolution.

The interrelation of consciousness and evolution is a deep issue which cannot be adequately discussed within the limited available space. A few introductory comments may nevertheless be useful. As Mangan notes, "it would be very odd that something as remarkable as consciousness evolved and yet did not execute a cognitive function" (a sentiment shared by Corteen). Klein writes, "If..consciousness does not enter into..human information processing - then it is awfully hard to imagine what the adaptive functions of consciousness might be that would have conferred a selective advantage leading to its evolution." I agree. Evolutionary theory is a functional theory. If sentience is dissociable from functioning (TA 9.1) then the emergence of sentience cannot be explained in entirely functional terms. But this too, in my view, has to be looked at from both a first- and third-person perspective.

Viewed from a first-person perspective, the adaptive advantages of conscious contents seem clear. There appear to be innumerable situations in which an integrated awareness of inner and outer events is necessary for adaptive interaction. The problem consciousness poses for evolutionary theory arises when matters are viewed from an exclusively third-person perspective (the traditional perspective of science). To an external observer, all the supposed functions of consciousness are fulfilled by its neural correlates, which embody integrated information (about inner and outer events) in a form that enables its dissemination throughout the processing system. The evolutionary advantages of such integrated information about inner and outer events for adaptive behaviour again seem clear. First-person and third-person accounts remain complementary, in that they both refer to the need for integrated representations.

Yet, given the dissociability of sentience from functioning, it is easy to imagine a machine (an android) that functioned (adaptively) just as humans do without awareness (editorial). If so, what is the added value of awareness? According to Schaeken and d'Ywalle, if consciousness does not enter into functioning (viewed from a third-person perspective) then, "considerations about the function or goal of consciousness can no longer be of help in clarifying its fundamental nature." This impasse, however, presupposes that the fundamental nature of consciousness can only be understood from a third-person perspective, and that the only function or goal worthy of consideration is survival fitness. To avoid the impasse one has to re-examine the presuppositions.

Viewed from a third-person perspective, consciousness does not enhance adaptive functioning. Rather, the brain functions, in part, to produce experience. From a first-person perspective, the difference this makes is obvious. Without consciousness there would be no experienced world. Denied the possibility of experience, survival appears without purpose. Placed in irreversible coma on a life support machine, few humans would choose to continue to survive. From this perspective, consciousness is the goal...but not one that can be understood in (third-person) information processing terms.

Additional references.

Dawkins, M.(1980) Animal suffering: the science of animal welfare. Chapman & Hall.

Dawkins, M.(1990) From an animal's point of view: Motivation, fitness and animal welfare. Behavioral and Brain Sciences 13:1-61.

Gauker, C.(1990) How to learn language like a chimpanzee. Philosophical Psychology 3:31-53.

Place, U.T.(1956) Is consciousness a brain process? British Journal of Psychology 47:44-50.

Reber, A.S.(1989b). More thoughts on the unconscious: Reply to Brody and to Lewicki and Hill. Journal of Experimental Psychology: General 118:242-244.

Savage-Rumbaugh, S.(1990) Language as a cause-effect communication system. Philosophical Psychology 3:55-76.

Note: Reber, A.S.(1989) cited in the target article (page 38), becomes Reber, A.S.(1989a)

Footnotes.

 Inhoff also points out that the Treisman's (1960) findings, often cited as evidence of semantic analysis on the nonselected channel, are ambiguous. I agree - but I did not cite this study.

[2.] Conveniently, this question can be settled experimentally. According to Underwood, deeper analysis of the nonselected message does not contribute directly to the interference effect Treisman found with bilinguals. If so, interference should be equally strong whether the French synonyms occur in syntactically well-formed, meaningful paraphrases (of the shadowed passage) or in random French strings, provided that in the two conditions, synonyms are matched for proximity to their shadowed counterparts.

[3.] A similar, progressive decrease in reaction time was found to accompany focal-attentive processing in Neeley's (1977) priming experiment.

[2.] A variation of the masked priming design used by Marcel (1980) experiment 5, could provide a relatively direct test of whether phrasal analysis prior to focal attention is possible in the visual domain. Marcel found that in three word sequences such as "save - bank - money," and "save - bank - river," if the word "bank" was masked, preliminary analysis unselectively primed following targets (both "money" and "river"); if the prime was not masked, focal-attentive processing selectively primed following targets ("money" but not "river") - see TA section 2.2. If phrasal analysis is possible, it might be that masked word combinations such as "venetian blind" would prime not only individual word associates such as "italian" and "deaf" but also associates of the combination, such as "window"; similarly, a phrase such as, "insect under foot," might prime "crushed."

[3.] According to Wagstaff, if the hidden observer is conscious, maybe all cases of preconscious processing are cases of hidden consciousness (in dichotic listening, implicit learning, and the like). Abandoning the hidden observer (in favour of social compliance theory) would avoid this unacceptable conclusion. However, Wagstaff's argument is equally unacceptable. If the hidden observer effect were paradigmatic of preconscious processing in general, then explanations of the hidden observer would apply to preconscious processing in general. That is to say, all the preconscious phenomena reviewed in TA 1 to 8, would just be the result of social compliance (a dubious suggestion). It seems more plausible to assume that the hidden observer is not conscious (while hidden); and that hypnotically induced dissociations differ in important ways from other forms of preconscious processing.

[4.] As Block (and Dennett, 1988) point out, in some situations, consciousness-plus-forgetting cannot be distinguished from no-consciousness. This point is fair - but only when there is some independent theoretical or evidential reason for invoking forgetting. Otherwise, all claims about subjects not being able to hear, see, and so on, are unfalsifiable (subjects might simply have forgotten). Block (and Rey) argue that a driver who exclaims "I don't know how I missed him" might simply have forgotten how he avoided an accident (TA 5.3). It seems more plausible, however, to treat this statement as an expression of the driver's amazement at his own preconscious (executive) motor skills.

[5.] My thanks to a vigilant reviewer for bringing this study to my attention.

[6.] This question is testable. But, to establish that consciousness, rather than antecedent neural events, exercised a veto, one would have to eliminate all anteceding neural events that might produce the required effects.

[7.] This is a rather biased description of the literature. In most experiments, subjects do not know exactly what stimulus to expect, and while the manner of responding might be determined (recall, reaction time, and so forth) the processing strategy need not be. In any case, the evidence reviewed also draws on clinical observation and observation in the field (e.g. the hidden observer, blindsight, studies of creativity, and the like).

[8.] One might ask, "thought by whom?" The list of theorists who think consciousness does play a role in information processing is extensive (see, for example TA note 2, and many of the commentaries).

[9.] Bowers and Dagenbach, mention Holender's alternative interpretation of some of the findings in TA sections 1 and 2. But these have already been critically discussed in the target article, and they suggest nothing further on this issue.

[10.] Even if the machine had a silicon consciousness, there would be no guarantee that its verbal protocols (designed by humans) were reports of that consciousness.

[11.] Mangan, Block, Libet, Van Gulick, Klein, and Spiegel point out that even if functions can be performed without consciousness, it might be that they could also be performed by consciousness; nature often duplicates functional systems. This might be true in principle. But if consciousness follows the processing to which it relates, it cannot enter into it in practice.

[12.] Wagstaff's demand that I demonstrate processing accompanied by consciousness to be no different to nonconscious processing is consequently unrealistic.

[13.] Inhoff also misses this point. As he notes, there are many relatively long tasks in which one becomes conscious of what is going on. But this does not establish that consciousness enters into processing. According to Inhoff, I suggest "that there is no significant difference between preattentive and attentive processes and that there is no significant difference between focused attention and awareness." In fact, I suggest the opposite (see TA 8).

[14.] This is attempted, for example, by Eccles (1980) and Sperry (1985), who both maintain that consciousness plays an integrating role in the activities of the brain (which cannot be explained in neural terms). They do not demonstrate however why this cannot be done (see, for example, note 21).

[15.] It should be clear from this that the precise relationship of consciousness to focal-attentive processing can only be determined empirically. I do not assert it "ex cathedra" as Navon maintains.

[16.] Earlier, Mandler tries to argue that what enters consciousness is not necessarily at the focus of attention; for example, when "thoughts come to mind unbidden or when we (often unexpectedly) feel (are aware of being) anxious, happy, etc..(where) there does not appear to be a prior focus of attention." Wilson makes the similar point "that recent experimental work by Wegner (1989) on thought suppression..shows that sometimes what pops into consciousness are precisely those thoughts we are trying not to think. These observations, however, simply exemplify the preconscious nature of the selective mechanisms involved.

[17.] This lack of integration (at the processing level) would also account for the dissociations mentioned in Shevrin's commentary. Conversely, from a third-person perspective, the integration and contextualisation produced by focal-attentive processing would account for the "quality fixing" he describes (the integration of quality and content of experience, observed from a first-person perspective).

[18.] Unlike Dretske, I would argue that this applies equally to external or bodily events as-experienced. As this is peripheral to Dretske's main theme, I will not develop this point here. But see Velmans (1990a), for a detailed discussion.

[19.] The term "consciousness" like any other, is "theory-laden." And, as our understanding of how it relates to "mind" and "brain" increases, our understanding of "consciousness" will deepen. But that does not mean there is nothing to which the term refers.

[20.] Wagstaff makes the point that the accuracy of first-person reports cannot be guaranteed, given that subjective reactions to consciousness necessarily follow the events themselves. This, again, applies equally to third-person reports.

[21.] As Corteen points out, it might be possible to translate everyday consciously experienced decisions (about shopping in supermarkets) into their preconscious, information processing antecedents, but for practical purposes, what would be gained?

[22.] Shevrin thinks this produces a paradox from my point of view, but it fits in naturally. The subject's unconscious can be explored, either from a first- or third-person perspective. What emerges from each perspective depends both on the focus of interest and on what is accessible. In so far as the accounts (obtained from either perspective) are accurate, they are complementary.

[23.] If consciousness follows the processing to which it relates, and does not enter into human information processing, this reductionist position would in any case have to be abandoned. Another obvious problem for Lloyd is the extensive evidence for causally active unconscious cognitive states (see TA; Dixon, 1981; Kihlstrom, 1987).

[24.] Given the widely held presupposition that the third-person perspective is privileged, this point may be less obvious. I simply mean that in most everyday circumstances, it makes perfect sense to explain what one experiences in terms of previous experiences. One experience of the world might cause a thought, one thought might cause another, or an emotion, and so on.

[25.] Mixed perspective explanations can be given either of brain processing and overt action, or of conscious awareness itself (along with its contents). Consequently, I disagree with Navon, that theories of mind function cannot subsume "an object of explanation, like the awareness of persons" - and that first-person accounts in any case attribute causal roles to the contents of consciousness, rather than to "consciousness per se." It depends on what one is trying to explain. Sometimes, the existence of consciousness, or the formation of one or other of its contents is the object of explanation. At other times a difference in functioning (with or without consciousness, or one of its contents) is the object of explanation. In such instances, either consciousness as such, or one of its contents, could form part of the explanation provided that one recognizes that these are all mixed perspective explanations. Whether it be cause or effect, the subject's consciousness is not reducible to anything the experimenter can observe.

[26.] Whether such correlates could exist in the brain without accompanying consciousness is partly an empirical matter and partly a matter of definition. Even if the correlates could not occur without consciousness by definition, this would not establish a causal role for "awareness" in neural functioning. From a third-person perspective, causal sequences would remain explainable in purely neural terms. Consequently, I do not agree with Dagenbach, that unless the form of processing to which consciousness most closely relates can be shown to occur without consciousness, then consciousness plays a causal role in processing.

[27.] What neural form such integrated correlates might take, remains an open question. As Koch & Crick point out, it seems likely that the integrated representations involve some form of co-ordinated activity in disparate regions of the brain, which is somehow distinguished from other concurrent activity. Whether this is achieved by the phase-locked synchronous firing of the neurons involved (Koch & Crick), by the contour of statistical deviations from average neuronal firing patterns (John, 1976), by a form of neural holography (Pribram, 1971, 1982; O'Keefe, 1985), or by some entirely different mechanism, can only be decided by empirical research.

[28.] Whether first- and third-person accounts provide descriptions of the same thing is a separate matter (see Economos). Note that the information encoded in conscious experience and correlated brain states is usually about inner or outer events other than the brain states themselves. Accordingly, first-person accounts are usually descriptions of inner or outer events as-experienced (not descriptions of the correlated brain states). The extent to which a third-person account can be about the same thing, depends on the extent to which the inner or outer events are accessible (either observationally or inferentially) to an external observer. Conversely, the extent to which first-person accounts can be about the same thing as (third-person) descriptions of brain states and processes depends on the extent to which those states and processes are accessible (through experience or inference) to both the subject and the experimenter. There are symmetries and asymmetries of access between experimenter and subject, depending on what is being accessed (see Velmans, 1992). However, the experimenter is not necessarily privileged (Economos mistakes my intention on this point).

[29.] This position also contrasts sharply with that of Koch & Crick, who simply list which TA conclusions they agree with, and which are "incorrect" or "irrelevant" (i.e. different from their own), without supporting argument. Without further ado, they go on to develop their own reductionist position, thereby begging all the questions relating to this issue.