Materialism has been extraordinarily successful at explaining nearly every aspect of the world we observe. Every aspect except consciousness, that is. Even though we have first person experience of the existence and nature of consciousness, it is difficult to determine how the material world is responsible for this phenomena. Some have taken this difficulty to indicate that materialism should be rejected, while others think that it simply demonstrates that consciousness as we know it is simply an illusion. In this paper I will attempt show how consciousness and materialism can be reconciled, through a view I call levels of description, in a way that preserves the reality and casual power of consciousness without introducing any non-physical properties, entities, or causes.
2: Levels of Description
In the simplest possible terms, the levels of description theory claims the following: that the mental system is only one thing, which we describe in various ways, namely the physical, the computational, and the phenomenal. We can consider each of these descriptions equally valid depiction of reality, in the sense that each description has its basis in objective reality, and describes real causation. Even so our descriptions of the mental system do not exist independently of each other; changes in the mental system captured by one level are reflected in others as well. If we accept this explanation of the mental system, there is no need to posit the existence of non-material entities, nor is it necessary to abandon consciousness. A purely physical analogy to this claim might be as follows: the quark description and the neutron-proton-electron description of an atom are equally valid, but they do not exist independently of each other; there is only one atom, but it can be described in several ways. Just as the multiple descriptions of an atom can be considered equally real, and yet not force us to deny materialism, so can the levels of description account of the mind reconcile our experience of consciousness with a materialistic assumptions about reality. An important note to make here is that this theory does not rest on any particular ontological position concerning the structure of reality. It is possible that the descriptions presented here reflect different aspects of the underlying reality, but it is not required for the argument to be coherent. Even if reality was ultimately an undifferentiated Parmenidian oneness, it still makes sense to talk about different the descriptions with which we apply to the mental system. Even if they themselves are purely artificial mental constructs, they are useful mental constructs, and they do reflect both our first person and objective experience of the mental system.
2.1 The Phenomenal
As I mentioned above, I divide the mental system into three types of description, physical computational, and phenomenal. The phenomenal level is simply the description that captures the first person perspective we have concerning our own minds. Our concepts of intentionality, self, and will are parts of this description. The phenomenal description captures the conscious activity of our minds, because by definition conscious activity is something we are aware of, and thus part of the first person perspective we are describing. By its very nature then the phenomenal description is also subjective, since it contains a viewpoint with respect to which these thoughts and perceptions are described. In theory at least we should be able to give some objective criterion as to when a system has a phenomenal description and when it does not. It is because of the way in which the phenomenal is connected with the physical and computational descriptions of our minds, described below (see section 4.2), that it seems necessary that such a criterion must exist, but unfortunately we do not know exactly what it is. The discovery of a way to objectively analyze the phenomenal in this way would solve many of the problems facing philosophers, as well as neuroscientists and AI researchers. Even though we cannot construct the description from the objective descriptions at the moment there is still reason to believe that such a description exists, namely our everyday experience. To deny that such a description of the mental system is possible would be to deny your own conscious experience.
2.2 The Physical
The physical description is much less opaque to objective investigation than the phenomenal description, because it is simply a snapshot of the brain that contains all the information available to scientific inquiry, such as the position and velocity of atoms, the distribution of neurotransmitters, etc. The existence of a physical description of the brain seems to be indisputable. Some would argue however that such a physical description has little or nothing to do with the mental system. We might then provide evidence that changes in the physical description are reflected by changes in the phenomenal level. This paper though is not the place to defend the connection of the physical with the mental, since our goal is to show that materialism can be reconciled with our experience of consciousness and our intuitions as to its casual effectiveness. Since we have made the assumption that everything in existence is purely material, and thus has by a physical description, we must conclude that the mental system, being part of reality, has a physical description too. Secondly, our assumption of materialism forces us to assert that nothing that is part of the mental system is left out of the physical description; there are no extra properties or states of the mental system that are not reflected in some way within the physical description. It is this claim that distinguishes the levels of description theory from transcendental idealism, for in transcendental idealism we would believe that the ultimate reality of the mental system could not be captured by our descriptions, and thus some aspects of the mental system would be omitted from the physical description, but this is not the assumption we are making here. Finally, it is important to remember that while the physical description may capture all the information about the ultimate reality, it is not itself the ultimate reality. We observe reality to behave as though electrons existed, but the notion of an electron is part of our description, not of reality. For a thorough defense of this position see Carnap’s work on the Ramsey sentence and the philosophy of science, in which he shows how physical laws do not pin down exactly the contents of reality, only that there are certain objects in reality that have properties described by the laws.
2.3 The Computational
Finally, we come to the computational description, which is by far the most controversial. A computational description implies the following: that the operation of the system in question can be described as a set of functions that operate on information, and in turn generate information. The information processed by a function can come from other functions or be an external input into the system. Likewise the output of functions may be directed to other functions or become output from the system, or both. The operation of each function can be specified algorithmically, i.e. a finite set of operations and control structures, and likewise determining where information goes and comes from can be calculated in a finite amount of time. Functions may operate in sequence or in parallel, and may take different amounts of time to complete their operations. You might say that this doesn’t sound much like how a computer is supposed to operate; after all doesn’t a computer deal only with numbers, sequentially? This vision of a computer is on a much lower level of the operations than I have described. These operations can be implemented, or simulated, on any computer system and reflect how programs are structured. The next step of course is to demonstrate that we have evidence that such a description can be applied to the mental system. Although I will later demonstrate why the computational description is tightly tied to both the phenomenal and physical description, it will suffice here to show that a computational description can be applied to the physical make-up of the brain. One reason to believe that such a description exists is the behavior of a damaged brain, specifically that damage to certain areas can eliminate or affect only a single capability of the mind, implying in turn that the brain is divided into areas that each perform a specific function. On the other hand, we could start with a computational description of individual neurons, since we know for sure that neurons act in a relatively well behaved way, each being affected by numerous inputs to produce a single output. Then from this neuron-computational description, we could abstract into larger units of operation. In either case, the conclusion is the same, that the operation of the brain can be described computationally.
3: Some Objections
A possible objection to the view being presented here is that we don’t “feel” the physical level controlling our actions, which in turn leads us to “feel” that we have free will in the sense that our actions are not determined in any way by the atomic level. I would hold that this objection rests on the false assumption that somehow our physical minds and our phenomenal minds are two different things, and that they could somehow work in opposite directions. However, under the levels of description theory, they are really both descriptions of the same mental system, and thus the objection is, in a way, nonsensical. Your will to lift your arm can be identified with some physical process occurring in your brain. This is not to affirm some version of identity theory of course, but simply to state that our thoughts reflect the physics of our brain, and that the physics of our brain reflects our thoughts. This may or may not throw into doubt the existence of free will; it depends on how free will is defined, and if one thinks that free will is compatible with materialism, and that is a topic which is outside the scope of this paper.
We might also raise the objection that the levels of description I am claiming exist with respect to the mind are different from the levels of description that we accept exist with regards to non-conscious objects. In other words, it seems obvious how some of the objects in the world have multiple levels of description, but we can perceive easily how these levels can be seen as aspects of one object, while the levels of description that describe our minds are not obviously part of a single object. A classic example of levels of description is an application to the morning/evening star. One description is the “morning star”, which has one set of properties. Another is the “evening star” description, which has its own unique properties. Despite their differences however, we think these descriptions are of the same real object, Venus. However, my claim that the physical and phenomenal descriptions are descriptions of the same mental system may seem different than the claim that the morning and evening star are descriptions of the same physical planet.
The reason that these situations feel different is because we perceive the morning and evening star in the same way, but we do not perceive the phenomenal and physical level in the same way. If we had direct perceptions of our neural activity, I claim that we would not feel that the phenomenal description and physical description were irreconcilable, however since we do not in fact have direct perceptions of such activity, I must argue by analogy. Imagine the following situation: unbeknownst to you I have taken a picture of your arm. Later I present you with the picture. Unless you are unusually disfigured, you will probably not identify the picture as your arm. Normally you have direct phenomenal access to your arm, you feel that it is part of you, but you do not get that feeling from the picture. However when presented with other evidence, such as the testimony of witnesses that in fact the picture was taken of you, you may eventually come to accept that it is your arm and feel differently about it. Likewise when we see our own brains at work, say in an MRI image, it is much like the feeling we get when presented with a picture of our arm, because we almost never perceive our minds in this way (on the physical level). It is easy to deny that the physical activity in the brain is actually one way of viewing the mind. This may account for why the phenomenal description and the physical description seem to be different kinds of thing, and thus by itself this feeling is not sufficient to rule out levels of description as being a valid theory concerning the nature of our minds.
4: Causation and Dependence
Even if we accept that there are various levels of description that can be used to describe the mental system, that still does not tell us how they are related to each other or to the rest of the world.
The image above illustrates the view I will be defending here. This image depicts reality and real causation as well as the various descriptions of it. Just as causality is part of reality so a description of causation is part of each description of the mental system (see section 4.1). The red arrows, illustrating dependence (not causation), show how the descriptions are connected to each other, and the holes in each description illustrate how they fail to capture some of the information that is incorporated into lower levels (see section 4.2).
First let me explain how each of the levels can be seen to describe real causation with respect to the world. It is easy to see how the physical description can be seen to capture casual power, since experiments have shown that electrical stimulation applied to various areas of the brain will cause a person to move their limbs. We also know from first hand experience that the phenomenal level can describe the cause of events in the world, for example under normal conditions a desire to raise your arm results in your arm being raised. Finally, in the case of the computational level we have evidence that there is a system in your brain (probably located in the motor cortex) that responds to input, such as commands from other parts of the mental computational network, by sending signals that result in the arm being raised.
Before I move on to the nature of dependence, it is important to note that the casual power of the mental system is an “all or nothing” endeavor, meaning that if “higher” levels of description can be said to capture the cause of an event in the world than the “lower” ones must describe the cause of that event as well. For example, my desire to raise my arm implies that the computational description and the physical description both contain some other description of this same cause that ultimately results in my arm rising. It is impossible that I might desire for my arm to rise and somehow it rises without the physical description also describing the cause of it rising. This is because it is really the mental system that interacts casually with the rest of the world; when we describe the system of course that description includes casual effects, but the real casual power is in the system itself, not in our descriptions. When the mental system causes an event to occur, different descriptions naturally reflect that activity. The reason that I mentioned that a causation effect in a “higher” level of description implies that we will find such effects in “lower” levels is not because the higher level is causing the lower levels to act, but because all the activity of a higher level of description has its basis in a an aspect of the mental system that the a lower level of description also describes, but not all activity in the mental system that is captured by a lower level of description is also captured by a higher level of description; for example there are physical events in our brain that do not become part of the phenomenal description, such as the movement of blood within the brain.
This also answers the epiphenomenalism objection, which is the assertion that under this theory the higher levels (i.e. the phenomenal) no longer have real casual power. This is to say that since the physical level provides a cause for every action, there is no room for the phenomenal level to be the cause of anything. Such an objection is founded on a misunderstanding of what a level of description is. Cause and effect are properties of the world, but the levels of description are not part of the world, except in the sense that we are thinking about them. Thus it is a mistake to claim that any description of reality has casual power; like a painting, a description reflects what is in the world, and like a painting it has no effect on the reality that it depicts. Thus the mistake made by this objection is not to reject the phenomenal as having real causal power but to accept to physical description as having this power when it does not. It is best to view the situation with respect to causation in the following way: the mental system is a real cause of events in the world, and vice versa. Since the physical description is complete (from our assumption that materialism is correct), then it must contain a description of every effect the mental system has on the rest of the world. The phenomenal description, which does not necessarily contain all the information about the mental system may or may not contain a description of individual casual effects of the mental system, for example unconscious thoughts are part of the mental system, and thus sometimes they may be the cause of events in the external world, but they are not part of the phenomenal description. On the other hand, our will to do something is also part of the mental system, and in this case it is part of the phenomenal description, meaning that this case of mental causation is reflected by the phenomenal description. It is not our consciousness, which is part of the mental system, that is epiphenomenal then, but our descriptions, because they only can describe the real causation not be a source of it themselves; the descriptions are ways we view the mental system, not part of it. If this situation were really epiphenomenal we would expect to find cases where the phenomenal description indicated that there would be a causal effect on external world, but the physical description somehow lacked this cause and thus no action was taken. However since the physical description is complete with respect to the information that is part of the mental system, any possible causation initiated by the mental system must be reflected by the physical description, and thus a situation where the phenomenal indicates causation without the physical description also capturing that causation is impossible.
Already I have indicated that the different levels of description are connected, and that their connection is not casual. For lack of a better word we can say that one level depends on another for its existence. The claim that description A depends on description B is simply shorthand for the following claim: the state and properties of description B give us (epistemic) reason to think that description A applies to the system. (1) It easy to see that phenomenal description depends on the physical (indirectly, through its dependence on the computational, as I will later argue). We know that head wounds, i.e. drastic changes to the physical description, are accompanied by changes in the phenomenal description of the system (through the testimony of those wounded). From this we can conclude that changes in the physical description of the system imply that we will find changes in the phenomenal description, i.e. the phenomenal depends on the physical.
From the division of the brain into various functions and channels of communication we can see that the computational description that depends on the physical description. The more pressing question we are now faced with is if the phenomenal depends on the computational or on the physical description alone. The way the brain responds to damage is a strong indication that the phenomenal description depends on the computational description and not directly on the physical description. Let us assume that there is a structure in the brain that controls conscious vision (in reality of course conscious vision is more complicated than this). Let us consider two cases of damage to this structure, in one case half the tissue in this structure is removed while in the other case of damage removed the other half, so that between these two patients there is no tissue in common. Despite this, because of the redundancy in the construction of the brain, the function of this structure is unimpaired (with respect to the entire computational description), and as a result both patients have conscious vision in phenomenally the same way they did before the damage. However patients with enough damage to impair the operation of the structure will not have conscious vision in the same way they used to. From this we can conclude that the phenomenal description depends solely on the computational description, because the correct functioning of the structures in the brain is the only requirement for normal phenomenal experience, the physical description is only related to the phenomenal indirectly though the computational description’s dependence on the physical. This does not however show that the computational description is independent of the physical description for the following reason: even though the overall computational description remained the same in both patients overall the description of computation with respect to the damaged structure has changed, for it now has less tolerance for future damage.
Now that I have convinced you that these various descriptions depend on each other, I need to mention that it is an incomplete dependence in the following sense: a given description does not depend on all the features of the “lower” description. Some of the physical activity within the brain does not contribute meaningfully to the computational description, for example the exact chemistry of a glial cell (a cell that holds a neuron in place) does not contribute meaningfully to computational activity. This is not to say that the non-contributing parts of the physical description could never influence the computational description, for example if the glial cell dies the neighboring neurons would be affected, simply that it does not contribute meaningfully to current computational description. This may sound a little vague, but when applied to the relation between the computational description and the phenomenal description we can easily see it in action. The unconscious is the most noticeable example of computation that the phenomenal description does not depend on. For example, consider the case of a hypnotic command. In such a situation, the phenomenal description of the subject does not include the command (otherwise they would be aware of it, by definition), but later, if the command is triggered it definitely has an influence on their phenomenal description, since they are aware of their own behavior and attempt to formulate explanations for it. This also answers the question “is there something that it is like to be in an unconscious state?” negatively, since there is only something it is “to be like” when there is a phenomenal description involved.
4.3 Multiple Realization and Mental Laws
This dependence does not rule out multiple realizations of the same description either. It is hard to give examples of multiple realizations of a phenomenal description, since this would imply that there was the same consciousness in two different places, but we encounter multiple realizations of computational levels of description constantly. Consider that there is a computational (information processing) description of the addition of two numbers. Any calculator we encounter is able to carry out this process, and thus has the same computational description (at least partly, since not all calculators support the same operations). In this way, a single description, the computational description of addition, is realized in many physical objects.
Although multiple realization of a complete phenomenal description is unlikely, simply the possibility of multiple realizations complicates any discussion concerning possible laws of mental activity. Let us say that we are given a complete phenomenal description, X. Ideally mental laws would tell us which phenomenal description X would evolve into over time. However changes in the mental system are not changes in one level of description, but changes in the system as a whole, so even though we can talk about changes in the phenomenal description, prediction of future states must be based on the complete mental system, not simply a partial description of it. Unlike other levels of description, the physical description contains all the information that is part of the mental system (once again this follows from our materialism assumption), and because in theory we can predict the behavior of physical systems over time we might suppose that given physical system x, which has a phenomenal description X we could predict that it would change into system y, which has the phenomenal description Y. Unfortunately we started off with a phenomenal description, not a complete mental system, and because of the possibility of multiple realizations of that phenomenal description we must accept that the system we have identified with X may really have any one of they physical descriptions x1, x2, … xn, which in turn we can predict will change into a system with physical description y1, y2, … yn. It is possible, even likely, that there will not be a single phenomenal description that will apply to all of these systems, and thus the best we will be able to say is that X will become Y1, Y2, … or Yk (k ≤ n). Thus we can conclude that with respect to the phenomenal description there are no laws of the kind we find in a deterministic physics. However the mental system as a whole does obey the kind of laws found in physics, because the physical description obeys them, and the other levels of description are dependant on the physical description.
I have made the claim that the phenomenal description at one point in time can be legitimately said to describe causes that will result in future phenomenal descriptions, so one thought can legitimately be said to be the cause of another thought. Note that we can’t claim that the phenomenal description contains real causes, given that earlier it was argued that causation only exists with respect to the mental system and not within a description. Still we want to claim that the phenomenal can legitimately be seen as the cause of the phenomenal, by showing that the phenomenal describes cause and effect within the mental system that determines which phenomenal description will exist in the future. Some might object to this however, stating that the phenomenal level can not describe “real” causation because casual relations between events must instantiate a casual law. A phenomenal event X might result in various phenomenal states, Y1, Y2, … Yk, as I have shown above, which may throw into question the existence of a casual law of the kind desired. However to make this objection implies that one must believe that the physical description can legitimately be said to describe the cause of future physical descriptions since the physical world obeys casual laws. Let us then consider two physical descriptions, p1 and p2, such that p1 is said to be the cause of p2. Given each of these physical descriptions there is also a computational description c1 and c2. Because whenever we have p1 we must (because of dependence) c1, and likewise for p2 and c2 we must conclude that there is a single computational description, c2, that will result from c1, and thus c1 is legitimately the cause of c2. In the same way, we can move from the computational description to the phenomenal description. Thus if we accept that the physical description of the world reflects casual relationships with respect to future physical descriptions so must we accept that the phenomenal description reflects these casual relationships with respect to future phenomenal descriptions. Of course a different account of causation may yield different conclusions, but I will leave that to the professionals.
Under this account then it has been demonstrated why we have reason to believe in the casual effectiveness of consciousness, both with respect to the world and with future states of consciousness. It has also been shown how we can reconcile the existence of consciousness with a purely material universe, in which consciousness itself is not a basic building block. With these problems solved, then it would seem that we no longer need to reject materialism because of an incompatibility with our experience of consciousness. Of course this doesn’t provide us a reason to believe materialism to be valid, but at least we don’t have reason to reject it out of hand.
1: Although it is possible that this dependence of descriptions reflects an ontological dependence of aspects, it is not necessary to agree with any ontological description of reality for the purposes of this argument, as mentioned earlier.
Carnap, Rudolf. An Introduction to the Philosophy of Science, 1966 (amazon)
Davidson, Donald. “Mental Events” in Philosophy of Mind: Classical and Contemporary Readings, 2002 (amazon)
Dennett, Daniel C. Consciousness Explained, 1991 (amazon)
Edelman, Gerald M. and Tononi Giulio. A Universe of Consciousness, 2000 (amazon)
Fodor, Jerry A. “Special Sciences (or: The Disunity of Science as a Working Hypothesis)” in Philosophy of Mind: Classical and Contemporary Readings, 2002
Kim, Jaegwon. “Multiple Realization and the Metaphysics of Reduction” in Philosophy of Mind: Classical and Contemporary Readings, 2002
Kim, Jaegwon. “The Many Problems of Mental Causation” in Philosophy of Mind: Classical and Contemporary Readings, 2002
Searle, John R. Mind: A Brief Introduction, 2004 (amazon)
Smith, David W. Mind World: Essays in Phenomenology and Ontology, 2004 (amazon)