Epiphenomenalism and the Meta-Problem of Consciousness

                                William S. Robinson Iowa State University

Abstract: Two questions comprising the Meta-Problem of consciousness are distinguished and addressed: Why do we think consciousness is problematic? and How is it possible to talk about our experiences and their qualities? The proposed answers are compatible with a wide variety of views. A particular concern in this commentary will be to show that they fit very naturally with epiphenomenalism. Chalmers’s ‘coincidence argument’ and the question of how we know we are not zombies are also addressed.

David Chalmers (2018, to which parenthetical numbers refer) has provided a thorough map of the dialectical territory surrounding two important questions – questions corresponding to his first (6) and second (7-8) ‘approximations’ to the meta-problem of consciousness (M-PC). In part 1 of this commentary I elaborate upon and support an answer to the first question – an answer that Chalmers mentions, but only very briefly. In part 2 I outline an answer to the second question, and explain its relation to epiphenomenalism and rival views. Epiphenomenalism naturally raises Chalmers’s ‘coincidence’ problem, and part 3 briefly sketches a line of response to it.

1. Why do we think consciousness is problematic?

Issues that philosophers raise using phrases such as ‘problems about consciousness’ do not first occur to people under the abstract term ‘consciousness’. For non-philosophers, consciousness is mostly something one loses by falling asleep, receiving a blow to the head, or consuming various drugs, and is something one may regain after some passage of time. It is common knowledge that all of those causes involve changes in the states of brains.

            The beginning of puzzlement about consciousness lies, instead, in some realizations about sensory qualities. In innocent childhood, we just see colored things, savor tasty food, smell fragrant flowers, hear noisy things, and so on. The qualities are in the things we perceive.[1] Eventually, however, we have experiences that lead to difficulties with this picture. Common examples include afterimages caused by a photographic flash or a mirror reflection of the sun into our eyes. Or we may close our eyes while tanning our front side, and notice an orange field. We may notice that a slightly bitter food tastes very bitter if we’ve just eaten something sweet, or that if we wash our hands when they are cold, water from the ‘cold’ tap seems unusually warm. I will refer to these cases, and others like them, as ‘discrepancy cases’.

            Discrepancy cases tend to show that experiences depend on us. The orange we experience while having an afterimage or while tanning is not a color that we attribute to any object. We do not believe that a food has come to have a high degree of bitterness because we just ate something sweet, or that the tap water contains any more warmth than usual. But if these qualities are not in the sun, or our eyelids, or the food or the water, what are they in? What about dreams, which are not at all like hearing stories, but instead seem to be visual phenomena? What can the visual qualities in dreams be qualities of?

            Well, if they are not in anything outside us, it would be natural to think that the bearers of visual qualities (and extra warmth or bitterness) are somehow in us. But the ideas that the flesh of our eyelids turns orange just because we face a bright light, or that our tongue is bitter when we have an extra-bitter taste, or that our hands are warmer when we feel water as warmer than normal, seem easily dismissed.

            If the qualities in discrepancy cases are not in something outside our bodies, and are not in a body part of ours, then it seems that one of the following must be true. Each alternative leads to problems, and I have given brief indications of some of them.

(a)        The bearers of qualities in discrepancy cases are items that are not part of the physical world.

This answer leads to the problem of having to explain the relation between physical and nonphysical entities. Causes and effects usually have some nature in common, e.g., mass in the case of momentum transfer. But being nonphysical seems to exclude such commonality, and that is easily understood as a problem.

(b)       The qualities that are said to be ‘in’ something are in fact never in anything that exists in our world. (They are in items that exist in some possible worlds that we can think about, but they are not in anything in this world.) 

This answer leads to the problem of imagining non-existence of this kind. We can easily imagine familiar things being arranged in ways they never are, e.g., horse-like animals with a horn, but sensory qualities do not seem to be arrangements of anything. (Orange is, indeed, a mixture of red and yellow, and hues can be instantiated with various saturations, but unique hues are not complex in the same way – nor in any obvious way.) It is easily perceived to be a problem how we could know what color words, e.g., mean if there are nowhere any actual examples of colored items.

(c)        Qualities in discrepancy cases are physical properties of physical things in our world, such as reflectance profiles (vision), molecular structures (taste and smell), compression waves (hearing) and so on. We are very good at detecting when such qualities are present and when they are not, but in abnormal cases we sometimes misrepresent them as present when they are not instantiated anywhere in our vicinity.

This answer leads to the problem that the qualities in discrepancy cases do not seem to be complex in the way that the indicated physical properties are. There is nothing in the odor of vinegar that remotely suggests three kinds of parts in a network of relations. There is nothing in pain that suggests the rapid series of events (i.e., neural firings) that either realize or cause it. (Sensory qualities may, of course, be said to be ways in which these physical properties appear, or are represented. But those claims would just return us to the same three alternatives when we ask what a case of appearing vinegary or painful actually is.)

            Evidently, none of the problems I have raised for the three alternatives is a knockdown argument against them. That was not the point of identifying the alternatives and their problems. The point is rather this: The problems in each case are very easy to see. Most readers of this commentary will have long been familiar with them in one formulation or another, but that is only my reason for being brief. The point I stress is that the problems I have mentioned can be seen by almost anyone, and extensive debate about solutions proves that they are difficult problems. The account of this section is thus a strong candidate for an answer to the question “Why do we think consciousness is problematic?”

            The problem of consciousness as I have just explained it can be summarized as the problem of how to locate sensory qualities in our account of what there is. Chalmers briefly identifies this problem when he remarks that ‘Phenomenal states seem problematic in large part because they seem to have a specific qualitative nature that is hard to explain in physical terms . . . . Ultimately, we need to explain why these qualitative properties seem to populate our minds . . . .’ (22). Unfortunately, this fundamental fact is presented only as an obstacle to accepting an ‘independent roles’ explanation of our problem intuitions. In contrast, I am offering the difficulty of locating sensory qualities as a sufficient and most plausible explanation of why we think that consciousness is problematic.

            It may be objected that the task set by Chalmers’s paper is to give an explanation of why we think consciousness is problematic in topic-neutral terms, and that the foregoing account fails because it invokes sensory qualities. However, unlike the philosophers’ term of art, ‘qualia’, which Chalmers explicitly excludes from acceptable terms (16), ‘sensory qualities’ is topic-neutral. That is because (i) it is used here only as a general term for colors, flavors, scents, and qualities involved in other sensory modalities, and (ii) terms for those qualities are topic-neutral terms. “Orange”, for example, does not have built into it that its instances are nonphysical entities. Views that regard sensory qualities as properties that are instantiated in nonphysical experiences require support by arguments, not mere consultation of dictionaries.

2.  How is it possible to talk about our experiences?

The reasoning in part I is reasoning about our experiences and sensory qualities that are somehow involved with experiences. It therefore presupposes that we can talk about such items. The second part of the M-PC, as I see it, is to explain how such talk is possible.

            Words for colors, flavors, scents and other qualities differ in different languages and evidently must be learned. So, we cannot think of experiences that involve some particular quality, e.g. orange, as directly latching onto our vocal apparatus and causing it to produce an utterance of ‘orange’.

            To have a remotely plausible picture, we must instead suppose, first, that an experience that would correctly be said to be an experience as of quality q involves a brain event (or possibly a small set of closely related events) of some distinctive kind in the language learner. Let us call such a brain event ‘BE(q)’ – ‘BE’ for brain event, and ‘q’ for the quality involved. Physicalists will think of BE(q) as identical with an experience of q (though they may disagree about whether an experience instantiates q or merely represents q). Epiphenomenalists will think that BE(q) causes a non-physical q experience, and that the latter has no effects of its own. Interactionists will think of BE(q) as an effect of a non-physical q experience that was brought about by some brain event(s) that occurred earlier than the q experience. The ‘BE(q)’ notation is to be understood as neutral among these accounts. It is a physical brain event that occurs when and only when a q experience occurs, but no account of why there is this covariance is built into its meaning.

            To explain the ability to talk about our experiences we must further suppose that that there is a positive correlation between the presence of q objects and adult speakers’ utterances of ‘q’ – for example, a positive correlation between presence of oranges, or certain crayons, or uniforms of Netherlands’ national teams, etc., and adult speakers’ utterances of ‘orange’ in preference to other color words. (Here and throughout I take English as the example language.) We must further suppose that such correlations are reproduced in the brain, in the form of positive correlations between BE(q) events and other brain events that are distinctive of occasions on which the learner hears the corresponding words for qualities.

            Finally, we must suppose that the brain is able to track such correlations, and alter synaptic strengths so that when we are about to produce a report of what is salient in our experience when, say, BE(o) occurs, the lexical system supplies ‘orange’ in preference to other predicates.

            We would, of course, like to have more details than we now have about how the association between BE(q)s and effects of adults’ words upon language learners is set up. But if we agree to the sketch of learning just given, we are in a position to see that epiphenomenalism is in as good a position as its two main rivals to account for the reliability of our experiential reports. That is because the account of learning will be almost the same on all three accounts.

            Let us begin with physicalism. This view presumably accepts that there is a set of physical processes that changes synaptic strengths such that when, say, BE(o) occurs and an experiential report is to be made, neurons in the lexical system are activated in such a way that ‘orange’ tends to occur as the predicate in the report. These physical processes will follow laws that explain the relation between timing of neural activations and changes in synaptic strengths. The physical properties of neurons and synapses, physical laws, and the positive correlations described earlier will explain the reliability of the connection between presence of, say, orange objects and utterances of ‘orange’ when experiential reports are made. The alleged identity of an experience and BE(o) (whether or not that is conceived of in representationalist terms) will then imply reliability of connection between experiences of orange and utterances of ‘orange’ when experiential reports are made.

            Epiphenomenalism denies efficacy to non-physical experiences, but it does not deny efficacy to any physical thing or event. It thus has exactly the same physical resources available to it as physicalism has. Almost all of its account of reliability of the connection between reports of experiences of orange and presence of orange objects will thus be word for word the same as the physicalist account. The only difference will come at the end: Where physicalism posits identity of BE(q)s and experiences, epiphenomenalists say that BE(q)s cause experiences. But this latter view gives exactly the same degree of reliability for experiential reports as does physicalism. Neither the identity claim of physicalism nor the causal claim of epiphenomenalism enters into the explanation of the setting up of the synaptic connections between BE(q)s and the lexical system – that is done through physical properties of neurons and synapses, and laws of physical nature. So, the difference in the two views’ accounts of what experiences are will make no difference to their accounts of how our experiential reports can be reliable.

            A parallel point holds for interactionism. That view holds that BE(q)s are caused by non-physical experiences. But if a word for a sensory quality can be learned, there will have to be a positive correlation between presence of q objects and adults’ utterances of ‘q’ in preference other quality words. If experiential reports can be reliable according to interactionism, that will be because presence of q objects causes experiences of q, which cause BE(q) events, and the latter are connected to the lexical system in such a way that the predicate ‘q’ tends to be produced if an experiential report is made. But now notice that the explanation of the reliability of experiential reports will still be able to be given in the same way as in physicalism. That is, the reliability will depend on the brain’s ability to turn positive correlations between presence of q objects, adults’ utterances of ‘q’ and BE(q)s into strong synaptic connections between BE(q)s and neural producers of the appropriate predicates in the lexical system. The only difference in the accounts will be that interactionism holds that the relation between experiences and BE(q)s is one of cause to effect, rather than effect to cause, as in epiphenomenalism, or identity (with or without representationalism) as in physicalism. But that difference makes no difference to the degree of reliability available in the accounts.

            There are, of course, other differences among these views. Epiphenomenalism struggles to give an account of how physical events cause non-physical experiences. Epiphenomenalism can nonetheless be claimed to be superior to interactionism on grounds of parsimony: Interactionism must also allow that physical processes initiated by impingements on our sense organs cause non-physical experiences, but then, in addition, is must allow physical events to be caused by non-physical experiences. Epiphenomenalism cannot claim to be more parsimonious than physicalism, but it can claim to avoid having to explain how a congeries of neural firings can either constitute the orange of an experience of orange, or represent presence of orange when all there is that is relevant and conceded by physicalism to be in our world is reflectance profiles (complexes of reflectance percentages at different wavelengths) and patterns of neural activations.

            The aim of this section is to explain how we can talk about experiences. I have given part of the answer, but my explanation of reliability of experiential reports assumes that we can at least talk about experiences as such. So, the next question is to explain how our talk can get to be about experiences at all. How can we learn to properly wield the rubric ‘I have a ________ experience’?

            The materials required for a response to this objection are already present in part 1. Very plausibly, we learn ‘q experience’, ‘seems to be q’, ‘looks q’ and the like from discrepancy cases. We are inclined to say the water is decidedly warm, but it is coming from the cold tap, and we do not believe anyone has surreptitiously installed a heater. We might say ‘It’s warm, but I don’t see how it could be’. It is not surprising that when a few such situations have occurred (involving a few different qualities) we should invent a shorthand for ‘one of those situations where something is so but I don’t see how it could be so’. It is thus not surprising that verbs like ‘seem’, ‘appear’, or ‘look’ should be in our language, nor that terms like ‘appearances’, ‘ways things look/seem’ and so on should occur.[2]

            Since we do learn words like ‘experience’, ‘looks’, and ‘seems’, physicalists must presume that there is a physical account of that learning. Epiphenomenalists do not deny the efficacy of any physical event or property, so they can give the same account. They also have a plausible reason why talk of experiences is about sensory qualities such as orange or warmth rather than about their complex physical causes. To wit, it is the sensory qualities, but not their causes, that are instantiated in our consciousness when we are (also) conscious of using the words in which we report what we experience.[3]

            Epiphenomenalism is also a view that meets Chalmers’s ‘meta-problem challenge for theories of consciousness’ (36). For according to it, the mechanism that produces judgments about our experiences includes brain events that are the causes of our experiences.

3. Coincidence and Acquaintance

There are well-known, often repeated objections to epiphenomenalism, and it cannot be expected that they should be addressed in this commentary.[4] But there is one point specifically identified by Chalmers (47) to which it is feasible and desirable to respond here. This is the coincidence argument: Briefly, if the explanation of our intuitions that consciousness is problematic is independent of [i.e., in our case, not causally contributed to by] consciousness, then the correctness of those intuitions is a coincidence – which is very hard to accept. More fully, the argument (47) is this:

1.         There is an explanation of our phenomenal intuitions that is independent of consciousness.

2.         If there is an explanation of our phenomenal intuitions that is independent of consciousness, and our phenomenal intuitions are correct, their correctness is a coincidence

3.         The correctness of our intuitions is not a coincidence.


4.         Our phenomenal intuitions are not correct.

‘Coincidence’, however, is an ambiguous term. In its ‘between-worlds’ meaning, it is plausibly a coincidence that our world contains both its laws of gravitation and electromagnetism. Maybe Spinoza was right, and there is no real contingency anywhere, but it is plausible that there are other possible worlds in which those two laws come apart.

            In this sense of ‘coincidence’ affirmation of premise 3 of the coincidence argument implies that there is no possible world in which the physical laws of this world obtain but the psycho-physical laws of this world do not. But many (Chalmers famously among them) have found it plausible that such a world is possible. So, affirming premise 3 in the between-worlds sense amounts to simply begging the question against a plausible view.

            Premise 3 is, however, extremely plausible and non-question-begging if ‘coincidence’ is read in its ‘within-a-world’ meaning. For then, premise 3 says that it is not just dumb luck that our phenomenal intuitions are correct. However, on this reading of ‘coincidence’, the epiphenomenalist view outlined in this commentary implies that premise 2 of the coincidence argument is false. Such a view satisfies the antecedent of premise 2, i.e., it explains our phenomenal intuitions independently of any causal contribution by consciousness, and counts them as correct, but it implies that the consequent of premise 2 is false: It cannot happen that BE(q)s cause q experiences and the complex set of causal and correlational relations described in §2 holds, and yet our phenomenal intuitions are mistaken.

            The coincidence argument thus trades on an equivocation on ‘coincidence’ and we should not be daunted by it.

            A question may now be raised as to how we know we are in a world in which the relations proposed in epiphenomenalism and §2 obtain. ‘How do you know you are conscious? How do you know you are not a zombie?’ Epiphenomenalists’ best answer is that we do not know we are conscious by any argument. We know it because we are in fact conscious and the laws of nature, mechanisms of reporting, and our linguistic history guarantee the reliability of our remarks about consciousness and sensory qualities. (For elaboration, see Robinson 2006; 2013.)

            It is instructive to note that interactionists cannot have a better answer. That is because one could surely not have a better reason to accept ‘This experience causes my judgment about it’ than one had to accept ‘This experience exists’. Thus, the causal claim could not be a premise in a proof that one has experiences. But that causal claim is the only point of disagreement between epiphenomenalism and interactionism. So, epiphenomenalism does not deny any premise that interactionists could use to try to prove to themselves that they are not zombies. Similarly, physicalists who do not outright deny experiences are surely more certain that they have experiences than that their experiences are either only representations or are identical with brain events. So, their point of difference with epiphenomenalists cannot put them in a better position to know that they have experiences.

            The last two paragraphs seem likely to remind readers of Chalmers’s remarks about ‘immediate knowledge’ (25), and of his use of the term ‘acquaintance’ (25 and 39). Commitments of this latter term are, however, not entirely settled. (See, e.g., Chalmers, 2010, pp. 286-291.) It thus seems desirable to clarify that the epiphenomenalist account I have outlined here makes no use of a relation of acquaintance, if that relation is conceived of as something additional to (a) instantiation of phenomenal qualities and (b) having such instances as episodes in one’s conscious life. The account in §2 involves many causal and correlational relations, but it neither has nor needs a place for a special relation of acquaintance.

            There is, evidently, much more to be said regarding everything in these comments. I hope and believe, however, that I have given readers a sketch of an epiphenomenalism-compatible view about the M-PC that will lead them to further explore it.

Acknowledgement: Thanks to Chris Schriner for comments on an earlier version.


Chalmers, D. J. (2006) Perception and the Fall from Eden, in Gendler, T. S. & Hawthorne, J. (eds.) Perceptual Experience, Oxford: Oxford University Press. Reprinted in Chalmers (2010), pp. 381-454.

Chalmers, D. J. (2010) The Character of Consciousness, Oxford: Oxford University Press.

Chalmers, D. J. (2018) The Meta-Problem of Consciousness,  Journal of Consciousness Studies, 25(9-10):6-61.

Robinson, W. S. (1982) Causation, Sensations and Knowledge, Mind, 91:524-540.

Robinson, W. S. (2004) Understanding Phenomenal Consciousness, Cambridge: Cambridge University Press.

Robinson, W. S. (2006) Knowing Epiphenomena, Journal of Consciousness Studies, 13:85-100.

Robinson, W. S. (2007) Evolution and Epiphenomenalism, Journal of Consciousness Studies, 14(11):27-42.

Robinson, W. S. (2012) Phenomenal Realist Physicalism Implies Coherency of Epiphenomenalist Meaning, Journal of Consciousness Studies, 19(3-4):145-163.

Robinson, W. S. (2013) Experiencing Is Not Observing: A Response to Dwayne Moore on Epiphenomenalism and Self-Stultification, The Review of Philosophy and Psychology, 4(2):185-192.

Robinson, W. S. (2016) Hidden Nature Physicalism, The Review of Philosophy and Psychology, 7(1):71-89.

Robinson, W. S. (2018) Russellian Monism and Epiphenomenalism, Pacific Philosophical Quarterly, 99:100-117.

Robinson, W. S. (2019) Epiphenomenal Mind: An Integrated Outlook on Sensations, Beliefs, and Pleasure

[1] This naïve attitude has been described in Chalmers (2006).

[2] For more detail, see Robinson (2019).

[3] See Robinson (2006, 2012) for elaboration.

[4] I have addressed them in several places, including Robinson (1982, 2006, 2007, 2013, 2019). I have also responded to a Chalmersian view concerning constitution of phenomenal beliefs, in Robinson (2006).

Comments are closed.

%d bloggers like this: