SWIF Philosophy of Mind, 15 March 2001 http://www.swif.uniba.it/lei/mind/forums/allen.htm

Forums Forum 2 Carruthers's reply

Evolving Phenomenal Consciousness
Commentary on Carruthers, Peter. Phenomenal Consciousness. Cambridge: Cambridge University Press, 2000.

Colin Allen

Department of Philosophy
Texas A&M University(USA)

As part of the argument for the superiority of his HOT (Higher Order Thought) theory of phenomenal consciousness, Carruthers offers an answer to the question of how phenomenal consciousness evolved, and he claims that this answer is superior to alternatives based upon other theories. I wish to demur. Evolutionary considerations are the cornerstone of Carruthers' rejection of alternative HOE (Higher Order Experience) accounts of phenomenal consciousness, which he argues do not successfully explain the cognitive role of phenomenal consciousness. They are also a cornerstone of his denial that nonhuman animals, young children, and autistic individuals are phenomenally conscious. The conjectured evolutionary story is too weak for either role.

In Carruthers' story, phenomenal consciousness emerges as an evolutionary "by-product, not directly selected for" (2000:230) of a two-stage selection process: First there was selection for first-order sensory representations (unconscious experiences), then there was selection for a "mind-reading" capacity which required conceptualization of mental states. On Carruthers' view, once the organism's own first-order sensory representations become directly available for conceptualization they are de facto phenomenally conscious. Once experiences become phenomenally conscious, then further adaptive benefits may follow -- particularly, Carruthers thinks, for making appearance-reality distinctions (2000:232; see also Allen & Bekoff 1997).

We have a huge range of phenomenally conscious experiences, from pains and orgasms, to the taste of sour milk and the feeling of breathlessness caused by the thin air and the staggering view from atop a snow-capped mountain. I shall argue that Carruthers' account fails to explain why we are phenomenally conscious in all the ways that we are. In other words, I shall put pressure on the alleged generality of this account, which is in evidence in this passage from chapter 8:

Now the important point for our purposes is that the mind-reading faculty would have needed to have access to a full range of perceptual representations. It would have needed to have access to auditory input in order to play a role in generating interpretations of heard speech, and it would have needed to have access to visual input in order to represent and interpret people's movements and gestures, as well as to generate representations of the form, `A sees that P' or `A sees that [demonstrated object/event]'. Mere conceptual events wouldn't have been good enough. For often what needs to be interpreted is a fine-grained gesture or facial expression, for which we lack any specific concept. It seems reasonable to suppose, then, that our mind-reading faculty would have been set up as one of the down-stream systems drawing on the integrated first-order perceptual representations, which were already available to first-order concepts and indexical thought. .... Once this had occurred, then nothing more needed to happen for people to enjoy phenomenally conscious experiences, on a dispositionalist HOT account. (2000:231)

In this passage, the phrase "access to a full range of perceptual representations" seems to be playing a dual role. First, although only two forms of perception -- hearing and sight -- are mentioned explicitly they seem to be standing service for all the forms of perception which give rise to phenomenally conscious experience, and the full range of perceptual representations should therefore encompass odor, taste, and touch, as well as nociception and other somatic sensations. Second, Carruthers' reference to those elements of perception for which we lack specific concepts indicates another sense in which the range of perceptual representations available for second order thought is supposed to be construed broadly.

I shall have little to say about this second sense of breadth, although I do think there are questions that might be raised about the limits of this account of phenomenal consciousness in the light of such phenomena as chicken-sexing (Biederman & Shiffrar 1987) where the question of how it is accomplished cannot be answered by introspection on the part of the expert chicken-sexers (Harnad 1996). Instead, I shall focus on the first notion of breadth -- the idea that Carruthers has outlined here an account of the evolution of phenomenal consciousness that applies across a wide variety of experiences from different perceptual systems.

Carruthers draws attention to interpretive acts based on speech and gesture, as well as to a more general class of attributive acts that seem to have less to do with communication per se. Because intentional communication between humans takes place predominantly in the modalities of hearing and vision (and perhaps to some extent using touch), Carruthers' focus on these two modalities in the quoted passage seems designed to enhance the plausibility of his thesis that interpretation constitutes a driving force for evolution. But the thesis seems considerably less plausible with respect to other sensory modalities, particularly smell and taste. The way others look to us, sound to us, and the sensations they produce when they touch us are all possible targets of interpretation. But there seems little to interpret in the way other people smell and taste to us. I conclude that the mind-reading faculty has no need for access to smell and taste for interpretive purposes.

What about more general forms of attribution? Well, it seems trivially easy to think of scenarios in which it would be adaptively useful to know what another individual is smelling or tasting. And we also know that natural selection can operate on very small margins, so it is not out of the bounds of possibility that there could have been selection for mind reading with respect to smell and taste. But this just-so story needs fleshing out, especially in light of the fact that it is not a foregone conclusion that these perceptual systems should give rise to phenomenally conscious experience, for there is at least one perceptual system, the vomeronasal system (Monti-Bloch et al. 1998), which responds to pheromones and affects human behavior but with respect to which we utterly lack phenomenal consciousness. Indeed it seems much more straightforward to think of cases where it would be adaptively advantageous (not to mention potentially pleasurable) to know whether one's pheromones have been detected and are generating an intense desire for intercourse in a conspecific, than it is to think of adaptive scenarios for more mundane odors. It is far from clear why we have phenomenally conscious smell and taste but are oblivious to "vomerolfaction" (Cooper & Burghardt 1990).

At best, then, the evolutionary story for phenomenal consciousness with respect to taste and smell is not proven. Neither sense is important to interpretation, and a weak just-so story is all that has been suggested to explain why we should have evolved the ability to attribute gustatory and olfactory states to others. It might be thought that this case can be made stronger by pointing to facts such as that my knowledge that some food type tastes bad to you, might make me less inclined to eat it myself. But I see no reason for thinking that the ensuing adaptive advantage that accrues from lowering the chances of poisoning oneself need to be mediated by conceptual recognition of the gustatory experiences of others, when the very same advantages could be derived simply by learning some non-mental facts about their reduced tendency to ingest this type of food. Of course there could be exaptation for attribution of such perceptual states (although scenarios involving deceptive manipulation of appearances are harder to imagine for taste and smell than for sound and vision), but Carruthers is supposed to be providing us here with an explanation for initial selection of the capacity to attribute gustatory states to others, not its subsequent cooption.

It is not my intention, however, to trade just-so stories. Rather, the point is this: the fact that the vomeronasal system is devoid of phenomenology shows that there is no guaranteed connection between phenomenal consciousness and any given behavior-guiding perceptual system, even among mind-reading creatures such as ourselves. Thus we are entitled to demand from Carruthers a particular and specific explanation for phenomenal consciousness with respect to each of the separate perceptual systems. It is not good enough to say that "mind-reading faculty would have needed to have access to a full range of perceptual representations" for the mind-reading faculty does not in fact even have access to the full range of perceptual systems.

Carruthers' diagrams tend to obscure the separate processing of different perceptual streams. In his figure 8.1 (fig 1a), for instance, the "Percept" box splits its output between two "short-term memory stores, C (conscious) and N (non-conscious)" (2000:228). Clearly the vomeronasal system has no direct line to C, so cannot be regarded as belonging in a general "Percept" box. As far as I know there is no neurological evidence for a single "Percept" box through which all perceptions pass. Hence the diagram should, minimally, be amended to show separate input streams from each perceptual system (even this is likely to be an oversimplification given the complex interactions between different perceptual modalities and conscious experience). There should, then, be separate links from each perceptual system to C, or perhaps even links to separate short-term memory stores, each with separate links to the "Conceptual thinking/HOT" box (fig 1b), and we are owed an explanation for the links between each perceptual system and the parts of the system responsible for consciousness.

Further complexity in the relationship between perception and conscious experience consists in the fact that some modalities interact in ways that affect phenomenal consciousness. For instance conscious experiences of flavor (distinct from taste) are due to a synthesis of gustatory and olfactory inputs, and effects such as the McGurk illusion (McGurk & McDonald 1976) show that conscious experience of speech can be affected by the visual perception of lip movement -- the simple act of closing one's eyes can change which phoneme is heard given identical aural input. Other modalities do not interact in this way; thus, for instance, the phenomenology of vision is not, as far as we know, influenced by olfaction (although flavor perception may be affected by vision). Carruthers might argue that these kinds of phenomena are determined by perceptual systems before the perceptual contents become available to higher order thought and that therefore it is not incumbent upon his theory of consciousness to explain them. But given that olfactory and gustatory neurons project to separate brain structures, and that odor and taste can be distinguished in associative conditioning experiments, there is no clear reason (particularly with respect to mind-reading) why phenomenal consciousness should receive synthetic experiences of flavor rather than concurrent experiences of taste and odor.

I have argued that the evolutionary explanation of phenomenal consciousness delivered by Carruthers is rather weak. This conclusion is yet compatible with his claim that the HOT theory provides a better evolutionary explanation for phenomenal consciousness than other accounts of consciousness, for the truth of this claim depends on the relative strength of those alternative accounts.

In what remains of this commentary, I want to argue that Carruthers gives too short shrift to at least one of the alternatives he considers and rejects on behalf of HOE accounts, specifically that the capacity for conscious discrimination between different experiences might aid learning (2000:215, first item). Carruthers claims that, for example, learning to avoid harmful events is possible without higher order representation of pain states -- all that is required is the capacity to distinguish painful stimuli (i.e. stimuli that trigger nociception) from other stimuli such as tickles. It is "hard to see" the point, he writes, of "discriminating between experiences of pain and experiences of tickling ... in the absence of a capacity for HOT" (2000:215).   He provides no empirical citations to support this claim. But it is in fact borne out to a certain extent by work on spinal nociceptive mechanisms in rats (see Grau 2001). Rats whose spinal cords have been cut in the cervical region show quite sophisticated kinds of associative learning in response to noxious stimuli applied to the hind legs, even though we can be sure that no signals are reaching the brain, and correspondingly certain that there is no phenomenal consciousness associated with these stimuli. The spinal cord distinguishes painful stimuli from other stimuli, and adaptive changes in behavior result. It is true, therefore, that learning can occur in the absence of conscious awareness of pain.

Carruthers is correct to reject simplistic adaptive hypotheses about the function of phenomenally conscious pain. Phenomenally conscious pain is not required either for withdrawal from noxious stimuli, nor for associative conditioning of pain responses. Yet learning is not a unitary phenomenon; certain forms of operant learning seem to require the brain (Grau 2001), as do some forms of classical conditioning (see next paragraph). For pain-related learning, it is by no means clear that these more advanced forms of learning can be sustained on the basis only of "first-order information-bearing states differentially caused by tissue damage in the one case, and stroking or tickling in the other" (2000:215) for while such states may indeed be sufficient to support discriminative avoidance of the noxious stimuli in some circumstances, the varieties of organismic adaptation are far more subtle than Carruthers lets on. One simplistic hypothesis has been replaced with another. Indeed, one of the chief puzzles of conscious pain is its relative independence from tissue damage; whatever role the internal state is playing is much more complicated than mere correlation with tissue damage (see Hardcastle 1997 for an overview). It is also the case that pain has an evaluative component -- we can decide, often in retrospect, whether a certain level of conscious pain was worth the reward or reasonably correlated with the danger, and such evaluations affect subsequent behavior (i.e., consciously-mediated learning results). Here, the point may not be whether there is a capacity to distinguish between experiences of pain and experiences of tickling, as Carruthers puts it, but whether there is the capacity to discriminate among experiences of different intensities. Since these different intensities are variably related to tissue damage, it might very well be adaptive to discriminate between the painful experiences as such so as to allow independent assessment of the severity of the underlying conditions that they purport to represent.

Another intriguing finding relating brain (specifically hippocampal) function to a sophisticated form of learning derives from Clark & Squire's (1998) experiments on "trace conditioning", a form of classical conditioning involving a temporal separation between the conditioned stimulus (CS) and the unconditioned stimulus (US). Clark & Squire used the eye-blink reflex as the response and showed that when presented with an audible tone (CS) that terminated one half or one second before a puff of air (the US), not all normal human subjects learned to blink in response to the tone (prior to the puff), but all those who did learn reported during debriefing that they were explicitly aware of the temporal relation between the two stimuli. In contrast, all subjects can be conditioned to blink in response to the tone when the puff is presented after the onset of the tone but before its termination -- "delay conditioning" -- regardless of their knowledge of the temporal relationship between CS and US. Trace conditioning occurs in rabbits (Clark & Squire 1998), and there are much greater individual differences in success rate for trace conditioning than for delay conditioning in these animals (Thompson et al. 1996). While Carruthers might assert that trace conditioning requires only first order representation of the temporal relation between two events, not between the experiences themselves, we know that the human capacity to represent this event depends on phenomenally conscious awareness of the stimuli; one cannot be explicitly aware of the temporal relation between two stimuli of which one has no phenomenal consciousness. In other words, what is represented seems to be the relationship between the two experiences.

Can such considerations ultimately sustain an HOE account?  Ultimately I'm not sure, and I don't have any particular stake in the outcome. But even if they cannot -- even if higher order thoughts are required to mediate the more sophisticated kinds of learning -- I believe that the considerations above undermine one of the more radical claims that Carruthers takes to be a consequence of his HOT account: namely his denial of phenomenal consciousness to nonhuman animals, young children, and autistic individuals, on the basis of their apparent lack of full mind-reading capabilities. Even if HOTs are required for certain kinds of learning, such learning seems within the range of at least some nonhuman animals whether or not they can "read minds".

Carruthers' denial of phenomenal consciousness to animals runs strongly counter to most people's intuitions on this matter. "It really is something of a scandal," he writes, "that people's intuitions, in this domain, are given any weight at all" (2000:199). At this stage of our understanding of behavior and neuropsychology, a far greater scandal, I believe, would be to circumscribe phenomenal consciousness on the basis of one man's inability to imagine alternative accounts for its evolution.


Allen, C. and Bekoff, M. 1997. Species of Mind. Cambridge, MA: The MIT Press.

Allen, C. 2000. "Where pigs and people fear to tread." In The Smile of a Dolphin, ed. M. Bekoff, pp. 90-91. New York: Discovery Books.

Biederman, I. & Shiffrar, M. M. 1987. "Sexing day-old chicks: A case study and expert systems analysis of a difficult perceptual-learning task." Journal of Experimental Psychology: Learning, Memory, & Cognition 13: 640 - 645.

Clark, R. E., & Squire, L. R. 1998. "Classical conditioning and brain systems: The role of awareness." Science 280: 77-81.

Cooper, W. E. Jr. & Burghardt, G. M. 1990. "Vomerolfaction and vomodor." Journal of Chemical Ecology 16: 103-105.

Grau, J. 2001. "Learning and Memory without a Brain." In The Cognitive Animal, ed. M. Bekoff, C. Allen, G.M. Burghardt, in press. Cambridge, MA: The MIT Press.

Hardcastle, V.G. 1997. "When Pain is Not." Journal of Philosophy XCIV: 381-409.

Harnad, S. 1996. "Experimental Analysis of Naming Behavior Cannot Explain Naming Capacity." Journal of the Experimental Analysis of Behavior 65: 262-264.

McGurk, H. & McDonald, J. 1976. "Hearing lips and seeing voices." Nature 264: 746-748.

Monti-Bloch, L., Jennings-White, C., & Berliner, D. L. 1998. "The human vomeronasal organ: A Review." Annals of the New York Academy of Sciences 855:373-389.

Thompson, L.T., Moyer, J.R., and Disterhoft, J.F. 1996. "Trace Eyeblink Conditioning in Rabbits Demonstrates Heterogeneity of Learning Ability Both Between and Within Age Groups." Neurobiology of Aging 17:619-629.


I am grateful to Gordon Burghardt, Lori Gruen, and Gary Varner for comments.

© 2001 Colin Allen