What Should Be the Roles of Conscious States and Brain States in Theories of Mental Activity?

The behavioural revolution–its enduring legacy

Simply put, consciousness was vigorously rejected as a proper subject for science. As John B. Watson put it, “This suggested elimination of states of consciousness as proper objects of investigation in themselves will remove the barrier from psychology which exists between it and the other sciences,” (Watson, 1913/1994, p. 253). The reason? “Behaviourism claims that ’consciousness’ is…merely a word for the ’soul’ of more ancient times. The old psychology is thus dominated by a kind of subtle religious philosophy” (Watson, 1924, p. 3). Furthermore, for a mechanical metaphor, a dominant machine of the time was selected: a Bell Telephone Switchboard. Stimulus “connecting” could produce a muscular “ring,” and so psychological theory could simply be Stimulus-Response Behaviour theory.

With this core commitment, B.F. Skinner (1974) in his most systematic – and enduring – treatment of Radical Behaviourism, called for the rejection of the person as agent, identified with an introspected and nonphysical world. Behaviourists focused their experimental inquiry on rats and pigeons unequipped to report anything behaviourism had rejected. Just rid the science of what Skinner called “the inner man” – and there would be no reason to fear an ontological nonmaterialism or free will in the sense of indeterminism.

That was the enduring confusion. Theoretical assertions are subject to empirical evaluation. Within the science, as we know it, metaphysical assertions are not (Dulany, 2003).

Mental contents

Most significantly, the mentalistic metatheory is set apart in holding that conscious states are the sole carriers of symbolic representation – permitting its explanatory power. In consciousness, we represent a present in perception, a past in forms of remembrance, and a future in expectations, intentions, hopes or fears. With higher order representations, we may even represent our own conscious states and mental episodes. Therefore, as we learn what consciousness explains, we enrich the explanation of consciousness.

Symbols are functionally specified. They may (a) activate other symbols, e.g., “’café’ activates ’latte,’” and (b) serve as subjects or predicates of propositional contents, as in “A latte is on my desk.” They may (c) participate in the special proposition “’This’ represents ’that,’” e.g., “This ”’latte’ refers to content of ’that cup,’” and (d) in the intention – controlling actions that warrant the preceding proposition: I grasp that cup and sip.

Symbolic contents may be identity symbols, attentional identifications of things as such, or literal symbols that precede and surround that attentional identification. We attentionally hear what a solo instrument plays from, how it sounds and the surrounding accompaniment.

Symbolic contents may be propositional, with subject values, such as the relation of what named to what considered or scored correct, and with predicate values, such as the likelihood of a certain outcome given the occurrence of what is named in the subject. They may also be nonpropositional such as a perception of a trumpet or oboe.

Modes are the familiar carriers of conscious contents and vary quantitatively in many experiments. They may be propositional modes such as “perceive, or believe, or intend that ——,” each carrying a propositional content. Alternatively, they may be nonpropositional modes such as a “feeling or only sense of ——” for nonpropositional content.

For mental mode and content, we can speak of a sense of agency, of possession that may vary in strength and frequency in normalcy, or be diminished in neurosis, and even missing or erroneously attributed in a psychotic syndrome.

Mental episodes

Mental activity consists of mental episodes – conscious contents interrelated by nonconscious mental operations. In deliberative mental episodes, propositional contents are interrelated by deliberative operations; they are decisions or inferences. In evocative mental episodes, nonpropositional contents are interrelated by associative – activational operations. We can think of these operations as simply the relations among these conscious states,

sometimes represented by a mathematical function within a model,

– or as the neural processes interrelating the mental states’ coordinate brain states. On this analysis of the commonly but too loosely termed “explicit” or “implicit” (learning, memory, or thinking), these processes are not distinguished by being “conscious” or “unconscious,” but by deliberative and evocative mental episodes (Dulany, 1997).

Forms of mental episodes are interrelated. For someone learning to drive, for example, “Red means Stop” can with associative repetition become the evocative episode, “’Red’ activates ’Stop.’” And later, one can readily represent that evocative episode in higher order awareness with the proposition, “Red means Stop.”

The domain for mental episodes lies between the final output of sensory transduction and the final input to motor transductions.

Higher order awareness

Consciousness may carry symbols representing a past, or even future, conscious state or mental episode (“metacognitively”) by imperfect operations of memory, prediction, or inference. But contrary to Higher Order Thought theories, (e.g., Gennaro, 2004), no mental state on this metatheory must be graced by a higher order state to acquire a “property” of consciousness. That mental state is a conscious state.

The nonsymbolic

Nonconscious memories are nonsymbolic neural networks – consistent with that part of connectionism. What we “know,” “believe,” or “intend,” despite a loose vernacular, does not have the same functional specification in inactive memory as in consciousness. Memories in those networks are constructed by experience, not “stored,” and remembering is a process of activation and construction, not “search, identification, and retrieval.”

Mechanisms

One example is the search for the neural mechanisms underlying symptoms in schizophrenia (e.g., Wibble, Preus and Hahimoto, 2009). Cacioppo and Decety (2009) called for a general programme of identifying the mechanisms underlying more general psychological processes, with some focus on use of fMRI. And, the P300 in ERP has long been identified with the general process of “context updating” (Donchin and Coles, 1988).

Classes of states

Also, well known is fMRI evidence revealing brain areas underlying classes of conscious states, e.g., the fusiform face area for facial recognition (Kanwisher, McDermott and Chun, 1997), and a parahippocampal area for place recognition (Epstein and Kanwisher, 1998).

The challenge of specific states

To some degree, “pattern analysers” facilitate distinguishing more specific states, at least when a controlling stimulus provides the specific state to be identified, for example, vertical or tilted grating, leftward or rightward motion or blue jay or sparrow in Kamitani and Tong (2005). Then, further tests were directed at distinguishing which of competing stimuli were attended – with orientation predicted by activation of sensory areas, V1 – V4 and objects by activation of higher areas.

But could there be a “dictionary” of imaging outputs, something that calls for numerous identifications? Discriminating the few with specific training is not equivalent to identifying the many in many contexts over many persons, with control for wandering thoughts. We can also report abstract concepts – “justice” or “energy” – but are they identifiable with neuroimaging? I think, too, that propositional forms, those central to deliberative thinking, present a particular problem – with the extraordinarily large combinations of subjects and predicates and degrees of belief or asserted likelihood, as well as evaluations and the sense of ownership of those thoughts.

General methodological challenges

While recognising significant contributions, Poldrack (2006, 2008) also points to specific limitations, for example, the problem of “reverse inference” from fMRI imaging to localisation of function when the localisation has been identified with various psychological functions. He elaborates especially interesting challenges from psychiatry and advertising (Poldrack, 2009).

In addition, Vul, Harris, Winkiellman and Pashler (2009) identify a large number of studies in which there were unrealistically high correlations between fMRI indices and various personality, emotion and social cognition measures, unreasonably high, given the modest reliability of both measures. They identified many studies in which voxels were selected for computation of their mean values only if the correlations of their own indices – amount of deoxygenated haemoglobin in the blood (the BOLD signal) – with the personality measure met some correlational criterion. This selection of voxel signals artificially reduced unreliability of the fMRI measures, thereby inflating the overall correlations – a statistical violation. Vul and Kanwisher (2010) identify more cases and elaborate the problem, including an especially revealing concluding section entitled “Why the nonindependence error is prevalent in fMRI.“

Unconscious volition?

A major aim in this literature has been the rejection of control of actions by conscious volition – intention, by demonstrating that the “will” is only an epiphenomenon – its control being an illusion.

In a famous series of experiments, Libet (1985) instructed subjects to voluntarily flex their wrist, as many as 40 times – with three critical measures: EMG of the hand movement, a postresponse pointing to a clock to report the time of an “urge to move” (about 200 ms earlier), and a “readiness potential” with EEG that was a negative shift in electric potentiality somewhere in the premotor or motor cortex (about 550 ms earlier). Averaged over trials and subjects, the “readiness potential” precedes the reported “urge” to flex – the intention – will – volition – by 350 ms or so: the “Libet lag.” That “readiness potential” has commonly, though surprisingly, been interpreted as the valid measure of the causally controlling intention – will – volition, permitting the reported urge to be described as merely a noncausal epiphenomenon. That interpretation has been both vigorously endorsed and challenged in various published symposia and books (e.g., Pockett, Banks and Gallagher, 2006).

In particular, a prior “readiness potential” could reflect what its term says – the activation that follows multiple responses and a general decision process, that is, “readiness” to form that more specific intentional urge to respond again. The reported time of intentional urge does precede the response time, and subjects were also able to voluntarily withhold a response. Brass and Haggard (2008) proposed a more complex decision process, reviewing fMRI evidence for “what,” “when” and “whether” states, actually conveyed in this way by the experimental tasks, with variously reported loci, from presupplementary motor area to prefrontal and parietal. These are images that can be interpreted as covering “readiness potential,” decision and conscious intention.

Is Libet’s legacy part of the behaviourist’s legacy? Pockett (2004, p. 624) writes that with “Libet’s main legacy…[we could] perhaps finally solve, the metaphysical debate about free will and determinism.” This, I believe, is the conceptual error that has motivated much of this literature. A theory that actions may be controlled by conscious intentions in no way entails free will in the sense of indeterminism. Intentions have their own causal antecedents and may together be represented in theoretical assertions, empirically investigated on the methodologically useful assumption of determinism – as in a literature extensively developed and reviewed in Ajzen, I. (in press).

A resolution in coordinate models? On the general assumption that conscious states are coordinated with neural states, a theory that conscious intention causally controls observable actions is coordinated with a theory that their underlying neural states causally control neural response processes.

Unconscious thinking – problem solving?

The belief is in the culture, even in anecdotal “examples” provided in undergraduate science classes – Poincaré’s creation of the Fuschian function and Kekulé’s creation of the benzene ring structure. But Ericsson and Simon (1993, p. 162) write, “A critical reading of Poincaré does not provide any evidence for unconscious processing, nor for the belief that Poincaré himself favoured that interpretation.” And the scholarship of Wotiz and Rodofsky (1984, p. 720) debunks the Kekulé story, quoting his own regret for the false claim he made after dinner and alcohol: “Today I have the feeling it would be better if one burned the whole rubbish and did not allow anything to be printed.” As the authors conclude, however, “…we have our doubts that the truth will get in the way of a good story,” (p. 723).

Indeed, there have been vigorous experimental attempts to show an “incubation effect” in problem solving, in solution of anagrams to complex puzzles, defined by this result: If Presentation of Problem is followed by Time Off (with distraction and delay), then performance on the Delayed Test is superior to Presentation followed by Immediate Test. Over the years, the reviewers agree: The “incubation effect” is inconsistently found, and the design is subject to artifacts that can prevent distraction from ruling out conscious problem solving, e.g., “physical refreshment, fruitful forgetting, losing commitment to an ineffective approach, and noticing clues in the environment. Extended unconscious thinking does not occur” (Perkins, 1981, p. 81). And Mandler (1994, p. 220) concludes, “there is no direct evidence that complex unconscious ’work’ (new elaboration and creations of mental contents) contributes to the incubation effects.”

Unconscious thinking – decision?

With the more recent work, (e.g., Dijkterhuis and Norgren, 2006), there has been a revival of the general claim within a similar paradigm, but with theory focused on decision processes. In one set of experiments, “conscious thought” was identified with the period following a presentation of alternative products – e.g., cars, apartments – and before instructed product evaluation and choice. “Unconscious thought” was then identified with a period of distraction after presentation and before evaluation and choice. The findings emphasised were better choice and postchoice evaluation with “distraction” than with “conscious thought.” In addressing the general metatheoretical matter, Dijksterhuis and Aarts (2010, p. 475) suggest that “One way to approach this issue is to propose that, in principle, the operation of higher cognitive processes does not care much about the conscious state of the individual.” That legacy?

This series of experiments is being met by a burst of experimental critiques – e.g., Lassiter et al., 2009; Rey et al., 2009; Waroquier et al., 2009. In essence, these experiments consist of failures to replicate and the identification of experimental biases. For example, subjects could form their favourable evaluations of familiar attributes on a first presentation – and so could, on ordinary “experimental demand,” revise those evaluations downward. As Lassiter et al. (2009, p. 671) refer to their findings, “…such judgments are ultimately a product of conscious rather than unconscious thinking.”

Unconscious implicit learning?

This is one of the most vigorously investigated and critiqued claims for unconscious symbolic representation. It would consist of unconscious abstraction and representation of guiding and symbolic rules (Reber, 1967) – a claim with many attempts to support and many critical analyses, from Dulany et al. (1984/2009) onward, for example, Shanks and St. John (1994), Perruchet and Vinter (2002) and Pothos (2007). I must only list some of the methodological – conceptual biases: (a) “Post experimental” assessment of awareness of sets of rules beyond memory limits. (b) Neglect of correlated and explanatory conscious contents, such as features and exemplars. (c) What is not conscious structural knowledge said to be unconscious knowledg – despite explanation by conscious judgment knowledge. (c) Failure to recognise implicit learning in the sense of establishment of associative – activational (evocative) relations apart from rules – as represented in theory (Dulany, 1997).

Prosopagnosia

With damage to a region of the fusiform gyrus, the patient may be unable to recognise faces consciously, yet nevertheless respond systematically in various ways to the person’s face. For example, Tranel and Damasio (1985) found greater GSR to familiar than unfamiliar faces. Furthermore, DeHaan et al., (1992) even found that classifying a name as politician or nonpolitician was slowed when the face was accompanied by a face from the opposite category. On the standard explanation, (e.g., Young, 1994), this is unconscious facial recognition, a dissociation between consciousness and a facial recognition system.

The key to a mentalistic explanation is recognition of a role for literal awareness dissociated from identity awareness. Identity awareness is blocked when activation fails to reach temporal areas, but literal awareness of face can activate other neural networks, producing a GSR to a familiar face, as in Tranel and Damasio (1985), and activating various occupational associations incompatible with the erroneous labels, as in DeHaan et al. (1992). Is the mentalistic explanation more tenable and promising? Literal facial forms are definitely represented in the patient’s awareness, activating other networks, as shown in their normal ability to match facial photos, familiar or not (DeHaan et al., 1987). With Farah (1994), this agrees that the recognition system is damaged, but the remaining parts produce these effects.