The Massive Multiple Realization of Psychology Properties In my talk today, I want to cover some excerpts from a series of papers I have coauthored with Carl Gillett. The central thesis of our work is what we call the Massive Multiple Realization of Psychological Properties. (Massive Multiple Realization) Many human psychological properties are multiply realized at many neurobiological levels. Something like this view is, of course, the orthodox view in the philosophy of mind, but it is an orthodoxy that has recently been challenged by Shagrir, (1998), Bechtel & Mundale, (1999), Bickle, (2003), Polger, (2004), and Shapiro, (2004). What we think unifies these criticisms of orthodoxy is a drive to raise the standards for scientific accuracy and metaphysical clarity. Shagrir, Polger, and Shapiro, all press for greater clarity on the metaphysics. Bickle and Bechtel and Mundale propose to look much more closely at the results of actual neuroscientific research. The general tenor of our reply to these criticisms is to re-raise the standards. We think that if one really develops a precise metaphysical theory of realization and multiple realization and if one really looks closely at actual neuroscientific practice, one finds good reason to believe in (MMR). Neuroscientists, like all biologists, hold two fundamental beliefs about nervous systems. First, they believe that nervous systems can be studied at any number of distinct, but interdependent, levels of organization in which entities at one level are explained by the qualitatively different entities at one, or more, lower levels that are taken to compose them. 2 Neuroscientists study structures as large as communities of interacting organisms and as small as individual proteins. There are thus a number of neurobiological levels. The second fundamental belief shared by neuroscientists is that nervous systems display individual variation. Subsequent research has shown that Darwin surely understated the case, especially in the case of the subject-matter of the neurosciences, when he observed that not all individuals of a species are cast from the same mold. Organisms obviously vary in their genetic make up, but given distinct histories of interaction with their environments even genetically identical individuals will diverge in their phenotypic details. In truth, no two organisms are exactly alike, molecule for molecule, or cell for cell, or organ for organ – especially when the molecules, cells and organs in question are those studied by the neurosciences. Combining these two fundamental beliefs, we may say that, as far as can currently be determined, individual variation appears at every level of neurobiological organization. As a result, since component entities such as realizer properties vary at particular levels, we contend that we have overwhelming scientific evidence for what we will term the ‘Massive Multiple Realization’ (MMR) hypothesis about psychological properties: (Massive Multiple Realization) Many human psychological properties are multiply realized at many neurobiological levels. Putting the thesis in other words, Massive Multiple Realization is the claim that for many human psychological properties the instances of these properties are realized by different lower level properties at levels of entities studied in neuroscience. There are, of course, a number of stronger, but related versions of this hypothesis. For example, one can change the two “many” quantifiers to “most” or “all” and not limit the hypothesis merely to human psychological 3 processes. For the space of the present discussion, however, I shall examine only this weaker hypothesis. My aim in this paper will not be to provide so much in the way of a defense of this hypothesis. Instead I will work on two preliminaries. The first of these preliminaries is to articulate the kind of realization and multiple realization we think is in play in this thesis. This is where Carl and I propose to take the metaphysics of realization and multiple realization really seriously. The second of these preliminary goals is to show how this apparatus applies in an extended case study. Although Carl and I believe that we can offer additional justification in support of (MMR), providing such justification will not be possible in the space of this talk. A single case study will have to suffice. Maybe another way of introducing my talk is to say that it is not so much a defense of a thesis as it is an introduction to one of my philosophical projects, namely, a more metaphysically and scientifically sophisticated theory of realization and multiple realization in the sciences. So, with that brief roadmap, let me begin explaining our take on realization. 1. Realization The term “realization” is used in many ways in different philosophical projects in different areas of philosophy. The sense that we have in mind appears to be what John Searle was describing in one of the “oldies, but goodies,” Minds, Brains, and Science: it seems to me that there is no difficulty in accounting for the relations of the mind to the brain in terms of the brain’s functioning to cause mental states. Just as the liquidity of the water is caused by the behavior of elements at the micro-level, and yet at the same time it is a feature realized in the system of microelements, so in exactly that sense of “caused by” and “realized in” mental phenomena are caused by processes going on in the brain at the neuronal or modular level, and at the same time they are realized in the very system that consists of neurons. And just as we need the micro/macro distinction for any physical system, so for the same reasons we need the micro/ macro distinction for the brain. And 4 though we can say of a system of particles that it is 10°C or it is solid or it is liquid, we cannot say of any given particle that this particle is solid, this particle is liquid, this particle is 10°C. I can't for example reach into this glass of water, pull out a molecule and say: “This one's wet.” (Searle, 1984, p. 22). So, take Searle’s example of the temperature of a sample of gas being 10°C. For simplicity of exposition, let us suppose that our sample consists of pure hydrogen, that is, molecules of hydrogen made up of exactly two atoms of hydrogen. In this case there is a “macro-level” individual (the sample) bearing a “macro-level” property, namely, having a temperature of 10°C. The sample of gas has its property in virtue of a collection of “micro-level” individuals (the hydrogen molecules) and their “micro-level” properties (their masses and velocities). So, it is because the molecules of the gas have the masses and velocities they do that the sample of the gas is 10°C. And (almost), as Searle says, even though the sample has a temperature of 10°C, no molecule has a temperature of 10°C. This skips some important technical details and the example ultimately does not work, but I want to stay with it because it has a simple structure that will make it easier to convey the formal apparatus of our theory of realization. We can state the foregoing a bit more formally by saying that the gas sample’s property, G, of being 10°C is realized by the masses m1, m2, …, mm, and velocities v1, v2, …, vm of the gas’s constituent molecules s1, s2, …, sm. Letting F1-Fn = {m1, m2, …, mm, v1, v2, …, vm} we can say that G is realized by F1-Fn. Notice that, in focusing on the masses and velocities of the molecules of hydrogen, we are picking out the properties of the molecules that are relevant to the temperature of the gas. We are ignoring other properties of the molecules, such as the fact that they are diatomic, that they contain two protons and two electrons each, that they are electrically neutral, that their covalent bond is due to the filling of a 1s orbital, and so forth. All that matters for the temperature of a gas are the masses and velocities of the constituent particles. 5 The foregoing suggests the following sketch of an account of realization, what we call the “Dimensioned view”: (Dimensioned Realization) Property/relation instance(s) F1-Fn realize an instance of a property G, in an individual s under condition $, if and only if, under conditions $, s has G in virtue of s’s constituents having F1-Fn, but not vice versa. This schema is again a simplification of the Dimensioned view, but I think it will do for present purposes, since as it stands, it captures a number of important features of realization that may well be only implicit in Searle’s examples. First, there are many cases in which the realization of G by F1-Fn will be relative to, or will presuppose, certain background conditions, such as standard temperature and pressure, a particular range of gravitational fields, and so forth. We capture this in our schema by reference to the background conditions $. Second, for present purposes, I will treat the “in virtue of” relation as the bottom of the expository heap. One may go deeper in more metaphysical studies, but this seems to be deep enough to capture the surface metaphysics of science. That said, I do wish to note that by “in virtue of,” I will, however, note that I mean a logical or mathematical determination relation. So, for example, the realization relation we are concerned with is not like the relation between the determinable red and determinate shade of red, scarlet. In the sense that concerns us, determinables are not realized by determinates. Although I will not go into it, this is the point were Searle’s example will break down for us. Third, the “in virtue of relation” is also a kind of non-causal determination relation. The determination relation we are describing is unlike causation in a number of respects. A) Causation is a kind of “horizontal” determination relation that is temporally extended, relates wholly distinct entities and often involves the transfer of energy and/or the mediation of force. So, when one billiard ball collides with another, the first moves toward the second, then collides 6 with it, and both move apart. This process is extended in time. By contrast, it takes no time for the masses and velocities of the hydrogen gas to realize the temperature of the gas. The instantaneous masses and velocities determine the temperature. B) When two billiard balls collide, the first ball is distinct from the second. By contrast, the molecules of the sample of gas are literally parts of the whole sample. So, causation, but not realization, involves wholly distinct entities. C) As we can see in the collision of the billiard balls, causation involves transfer of energy from one billiard ball to another. But, there is no transfer of energy when the molecules of the gas realize the temperature of the gas. Third, there is Searle’s observation that it can happen that a higher level property, such as having a particular temperature or solidity, is qualitatively different than are the properties that realize it. Searle noted that, while a sample of gas may be 10°C, no molecule of the gas is. (This is correct in a rough and ready way, since after all it is possible for a single hydrogen molecule to have the right mass and velocity to bear the temperature of 10°C.) Where the gas has a temperature, the individual molecules have qualitatively different properties, such as mass and velocity. Fourth, it is often the joint action of the lower level properties that brings about the higher level property. So, it is all the masses and all the velocities that together bring it about that the sample of gas has the temperature it does. These are features of the realization relation that we will return to in our case study. 2. Multiple Realization A temperature of 10°C can be realized in many ways. The same ensemble of molecules we mentioned earlier can have their velocities changed while preserving their average velocity. Or, a different set of molecules with different masses and different velocities could also realize 7 the same mean kinetic energy. Let me formalize this a bit more as preparation for our definition of multiple realization. So, again, G is again the “macro-level” property of being 10°C. As we noted above, one way to realize G is to have molecules s1, s2, …, sm bearing properties F1-Fn = {m1, m2, …, mm, v1, v2, …, vm} under conditions $. Another way involves a different set of molecules s*1, s*2, …, s*p bearing properties F*1-F*m = {m*1, m*2, …, m*p, v*1, v*2, …, v*p} under conditions $*. We, thus, get the gist of multiple realization if F1-Fn ≠ F*1-F*m. The foregoing, we think, captures most of the familiar ideas about the kind of multiple realization we are looking at. There is, however, one more absolutely essential refinement. For multiple realization of G, F1-Fn and F*1-F*m must be on the same level. Elaborating on our temperature example will show why. Suppose, as before, that we take one realization of G to be the molecules s1, s2, …, sm bearing properties F1-Fn. Now, think about the two atoms of hydrogen that make up each molecule of hydrogen. Instead of molecules s1, s2, …, sm we might enumerate the atoms s’1, s’’1, s’2, s’’2, …, s’m, s’’m. Similarly, instead of the molecular properties {m1, m2, …, mm, v1, v2, …, vm} we can think of the atomic properties, {m’1, m’’1, m’2, m’’2, …, m’m, m’’m, v’1, v’1, v’2, v’2,…, v’m, v’’m}. Similarly, we can denote this last set by F*1-F*m. Now, suppose, merely for the sake of expository simplicity, that the atoms also realize the temperature of the gas. This apparently allows one to have the temperature of a gas realized both by the molecules of the hydrogen and by the atoms of hydrogen, so that we get two realizations of the temperature, hence multiple realization of the temperature. But, this is not the kind of example that is typically counted as multiple realization. So, we need to be sure to exclude this kind of case. We can now extrapolate from the foregoing example to generate a schema for the kind of multiple realization we have in mind: 8 (Multiple Realization) Instances of a property G are multiply realized if and only if, (i) under condition $, an individual s has an instance of property G in virtue of instances of properties/relations F1-Fn to s’s constituents, but not vice versa; (ii) under condition $* (which may or may not be identical to $), an individual s* (which may or may not be identical to s) has an instance of a property G in virtue of the instances of properties/relations F*1-F*m of s*’s constituents, but not vice versa; (iii) F1-Fn ≠ F*1F*m and (iv), under conditions $ and $*, F1-Fn of s and F*1-F*m of s* are at the same scientific level of properties. Where I think that conditions (i)-(iii) are pretty familiar, I think it is important to emphasize the importance of the less widely appreciated condition (iv). Because of condition (iv), Carl and I often describe our theory as the theory of indexed realization and multiple realization. Even though it is common practice to speak merely of “multiple realization,” we think that it is important to treat this only as a kind of typically harmless shorthand. In all strictness and explicitness, we think that realization and multiple realization make an implicit reference to one or another level. One motivation for this strictness and explicitness noted above is simply doing justice to what is commonly meant by multiple realization. There are, however, two other reasons for noting this relativization. First, it forestalls certain objections to particular examples of multiple realization, and, second, it helps us understand at least part of the philosophical dialectic in debates about multiple or unique realization. Both of these I want to explain by some allusions to biochemistry. In my 2007 paper in Synthese, I claimed that all human psychological properties are multiply realized by distinct sequences of amino acids. Here is the quick and dirty version of an argument for this claim. Every human psychological property are realized, in part, neurons, neurons are realized, in part, by proteins, and proteins are realized, in part, by amino acids. In addition, our best scientific evidence suggests that there are different combinations of amino acids that realize the same psychological property. There is plenty of room here for arguments 9 about the scientific details and the state of empirical knowledge in this area, but there is also a simple, armchair philosophical objection as well. I want to address that briefly here. One might say that any notion of multiple realization that says that psychological properties are multiple realized by biochemical properties is ipso facto wrong. Any such theory simply trivializes the notion of multiple realization. If that’s what you, Aizawa, mean by the multiple realization of human psychological properties, then of course there is that kind of multiple realization. That’s just trivial. Is there anyone who would ever have doubted this? The answer here appeals to our condition (iv). The sense that the multiple realization of psychological properties by biochemical properties is obvious and trivial is captured by the indexed theory of multiple realization, since it allows us to say that in fact psychological properties are multiply realized by biochemical properties, but it also allows for the existence of other interesting questions, such as whether or not psychological properties are multiply or univocally realized by some other, higher level of neuroscientific structure. The long-standing, philosophically interesting question is not whether psychological properties are multiply realized at some level. They obviously are. The long-standing philosophically interesting question is whether psychological properties are multiply realized at all levels. If there were some level at which psychological properties were uniquely realized, then there would be some level at which psychological properties might be type-type reduce. If psychological properties were uniquely realized in, say, some neuronal level properties, then there would be a point at which an argument for type-type reduction might begin to get some traction. So, the indexing of multiple realization provides us with a way of meeting the charge of trivialization. The short of it is that the indexing allows us to recognize the truth of some (trivial) multiple realization claims, while allowing us to explore another long-standing philosophical 10 question. The second reason for recognizing indexing actually grades into the first. Recognizing indexing provides us with a way of more clearly articulating certain replies to multiple realization arguments. Part of what is going on in certain replies is to concede multiple realization at one level, but to hold out hope for univocal realization at another. Let me run through just one illustration. This appears to be at least part of what is going on in this wellknown observation by Jaegwon Kim, [T]he fact that two brains are physico-chemically different does not entail that the two brains cannot be in the “same physico-chemical state.” ... To argue that the human brain and the canine brain cannot be in the same brain state because of their different physicochemical structure is like arguing that there can be no microphysical state underlying temperature because all kinds of objects with extremely diverse microphysical compositions can have the same temperature; or that water-solubility cannot have a microstructural “correlate” because both salt and sugar which differ a great deal from each other in atomic composition are soluble in water. If the human brain and the reptilian brain can be in the same “temperature state,” why can they not be in the same “brain state,” where this state is characterized in physico-chemical terms? (1972, pp. 189-190). There are at least two things going on in this passage. When Kim says that, “all kinds of objects with extremely diverse microphysical compositions can have the same temperature,” part of what he may mean is a point I articulated earlier. In spelling out what realizes the temperature of a sample of hydrogen gas, only certain properties matter. The masses and the velocities of the hydrogen molecules matter, but the fact that they are diatomic, that they contain two protons and two electrons each, and so forth does not matter. Another thing that could be going on is an issue of levels. Kim might mean that two samples of a gas can have the same temperature and the same mean kinetic energy, but then all sorts of different combinations of masses and velocities for the atoms can have the same mean kinetic energy. So, the picture could be that there is one high level property, temperature, then one middle level property, mean kinetic energy, and, finally, several lower level ensembles of masses and velocities. The level theory of 11 realization accommodates this idea. Stated quite generally, in point of logic, one can have a property at level n that is univocally realized at level n-1, then multiple realized at level n-2, then uniquely realized at level n-3, and so forth. So, leveled realization can help articulate part of what Kim might have been up to. But, it assuredly articulates some of the dialectic I’ve encountered in claiming that psychological properties are multiply realized by distinct amino acid sequences. In replying to my argument in Synthese, Bickle has not denied that there are multiple chains of amino acids underlying psychological properties. Instead, he has sought a higher level at which there might be univocal realizations. Another way to frame my argument is to say that psychological properties are multiply realized by distinct primary structures of amino acids, i.e., the sequences of amino acids in chains. Bickle has claimed (roughly) that these amino acids fold up into more complicated tertiary structures and these higher level tertiary structures provide a unique realization of psychological properties. I do not think that this is correct, but my present point is merely that one can make sense of Bickle’s move in terms of higher level structures that might provide for a unique realization. So, to summarize what I have said about multiple realization, the idea is basically that you get multiple realization when you can make a higher level property out of different combinations of relevant lower level properties at the same level. The indexing of multiple realizations to levels is valuable for three reasons. First, it allows us to express what philosophers commonly mean by multiple realization. Second, it allows us to express what is correct, but not the entire story behind, claiming that psychological properties are multiply realized by distinct sequences of amino acids. Third, it allows us to articulate more clearly some of the dialectic of arguments about multiple realization. 12 3. A Scientific Case Study To this point, I have been presenting the theories of realization and multiple realization with an eye toward a clear exposition of the view. With an eye on clarity, I have cut a few corners, but I think the general picture comes through. The idea has been to be true to the metaphysics of the target conceptions. Now, however, I want to discuss the second dimension of the Gillett-Aizawa project, namely, how these theories of realization and multiple realization make sense of actual scientific practice. To do this, I will be working part way through a case study, the realization and multiple realization of color vision in humans. I will show how realization works at a variety of levels and how it is at least plausible to conclude that there is multiple realization at these levels. At the outset, I should say that very little scientific work makes explicit use of the term ‘realization’ in anything like the way in which it has been used by defenders of the received view in the philosophy of mind. Thus, like so much philosophy of science, investigation of the realization of higher level properties by lower level properties constitutes a kind of examination and reconstruction of scientific practice, theorizing etc. The case study to be presented here involves human color vision and in particular the contribution made by the eye. If there is multiple realization of the color-relevant properties in the eye at many levels, then there will be multiple realization of color vision at many levels. So, consider, first, the relevant individuals involved in the hierarchy shown in Figure 1. 1A shows the entire eye containing the retina. 1B shows the retina and the primary types of nerve cells to be found within it. 1C shows the outer segment of a cone cell. Cones are the photoreceptors that are involved in color vision; the outer segment is the portion of the cone in which light is converted into a biochemical signal. 1D shows two photopigment molecules embedded in the 13 membrane of the outer segment. 1E is a schematic drawing showing the individual amino acids that constitute the opsin portion of the photopigment molecule. The seven cylinders represent the portion of the opsin that spans the outer segment of the membrane. To see how our realization and multiple realization schemata work, we will begin with the realization of properties of the green cone photopigments. Human color vision is classically taken to involve three distinct cone photopigments, sometimes called the red, green, and blue photopigments. The green photopigment is made up of a protein component called an opsin and a chromophore (11-cis-retinal, a derivative of vitamin A). The green photopigment has peak light sensitivity of about 530 nm (λmax = 530 nm) and the sensitivity curve has roughly a bell shape. (See Figure 2.) This sensitivity curve determines the probability of capturing a photon of a given wavelength. The green cone opsin is a chain of 364 amino acids. So what looks to be going on here is that the “micro-level” or lower level properties and relations of the 364 individual amino acids in the chain, in conjunction with the properties of the 11-cis-retinal, realize the “macro-level” or higher level properties of the green cone pigment. Our theory of realization allows us to describe this. In our hydrogen gas example, we had only one type of micro-individual, a hydrogen molecule. In this case, however, we have one or more instances of 20 different amino acids to contend with along with the 11-cis-retinal. In the hydrogen gas example, only two properties of each molecule were relevant to the realization of the temperature of the gas, namely, the molecule’s mass and velocity. In the green photopigment case, however, we have many more properties and relations that are relevant to the peak and the shape of the photopigment’s light sensitivity curve. The 11-cis-retinal, for example, has the property of capturing photons in a portion of the visible spectrum of electromagnetic radiation. The individual amino acids, 14 however, have properties such as polarity, size, and charge, along with peptide bonds between them. This makes it a bit messy to come up with a nice notation such as we had for the masses m1, m2… mm, and velocities v1, v2, …, vm in our hydrogen gas example, but so I will provide only a rough account. So, it might be something like P’1, P’’1, P’’’1, P’2, P’’2, P’’’2, P’’’’2 , …, P’365, where the subscripts indicate which of the 365 different individuals bears the properties/relations and where the superscripts indicate one or another of the properties/relation. In our formal schema, however, we just wash out all this additional notational apparatus and enumerate all the properties and relations as F1-Fn. So, plugging this into our schema, we would say that property/relation instances F1-Fn realize an instance of the property G of having a λmax = 530 nm in an individual green photopigment molecule s under normal physiological conditions of the retina $, if and only if the photopigment molecule has G in virtue of the molecule’s having F1-Fn under conditions $, but not vice versa. And, of course, we could tell a comparable story about the shape of the molecule’s spectral sensitivity curve. What makes the biochemical level so apt for multiple realization is that one expects, on general biological principles, that there should be various genetic mutations that give rise to differences in amino acid sequences, but which give rise to proteins with the same property. That is, one of the cornerstones of evolutionary biology is that there is a pool of variation upon which natural selection can act and that many of these variations are selectively neutral. So, in Aizawa, (2007), this was the empirical basis upon which I argued for the multiple realization of psychological properties. What makes this example of the green photopigment especially interesting is that in the last ten years, neuroscientists have actually measured the absorption spectra of molecularly distinct photopigments and found them to have the same λmax. That is, there are genetic mutations that have given rise to chemically distinct green photopigments that 15 have the same λmax. These are the gray and the green curves in the upper left panel of Figure 3. And, as one might expect, the same holds for the red photopigment. So, we now have two chemically distinct green photopigment molecules that have different sets of lower level properties, but the same λmax. Put in our formalism we can say that two instances of the property G of having λmax = 530 nm are multiply realized since (i) under normal physiological conditions $, one molecule has λmax = 530 nm in virtue of the instances of properties/relations F1-Fn belonging to that molecule’s constituents, but not vice versa and (ii) under those same physiological conditions $ another molecule has λmax = 530 nm in virtue of the instances of properties/relations F*1-F*m of the other molecules constituents, but not vice versa; (iii) F1-Fn ≠ F*1-F*m and (iv), under conditions $ and $*, F1-Fn and F*1-F*m are at the same scientific level of properties. With our first step in our case study, I want to remind everyone of the features of realization to which we have drawn attention. First, there is a non-causal, non-logical determination relation between the properties/relations of the amino acids and 11-cis-retinal and the properties of having λmax = 530 nm. Second, the lower level properties are qualitatively different than the higher level properties. No individual amino acid or the chromophore (11-cisretinal) has a λmax = 530 nm, a property that the entire protein does. Instead, the relevant properties of the amino acids are their size, charge, polarity, bonding arrangement, and so forth. Third, it is the joint action of the properties of the amino acids and the chromophore that together realize higher level property λmax = 530 nm. This concludes our application of our schemata for realization and multiple realization to the biochemical level. Now I will move up to a slightly higher level to the realization of an entire nerve cell, green cone. At this level, I will offer a bit 16 less about the realization and multiple realization schemata and a bit more about the scientific details. To explain this case, let me begin with the individuals involved. At the lower level, the relevant individuals include water molecules, ions (such as Ka+, Na+, and Ca++), phospholipid molecules, proteins, etc., and, at the higher level, the individual under consideration is a human green cone. (See Figures 4 and 5.) Phospholipid molecules have both a hydrophilic and a hydrophobic region. Given this configuration, they spontaneously form a bilayer structure in which the hydrophilic regions face outward to an external aqueous environment in either the extracellular space or the cytoplasm, while the hydrophobic tails of the molecules cluster together inside the bilayer. This phospholipid bilayer constitutes the cell membrane, illustrated in the right half of Figure 5. Proteins, for their part, also have hydrophobic and hydrophilic portions which help embed them in the cell membrane. (See again the right half of Figure 5.) Human cone opsins, for example, have an evolutionarily well-conserved set of seven transmembrance amino acid sequences. (See Figure 6.) Ion channels, also, have amino acid sequences that enable them to span the cell membrane and provide bindings sites on one or another side to regulate the flow of ions through the channel. Cytoskeletal proteins, also partially embedded in the cell membrane, shape a cell into exotic configurations such as those of the rods or cones. Now consider how this apparatus transforms the capture of a photon into a neuronal signal. Photopigment molecules are embedded in the cell membrane in the outer segment of the cone. (Recall Figure 5.) Upon absorption of a photon, a single photopigment molecule will change conformation and release into the cytoplasm a molecule of all-trans-retinal leaving an activated opsin molecule in the membrane. (See Figure 7.) One activated an opsin molecule 17 binds to a single G protein molecule located on the inner surface of the cell membrane. This G protein molecule, in turn, activates a molecule of an enzyme, cGMP phosphodiesterase, which breaks down cGMP. When intracellular cGMP concentrations subsequently decrease, cGMP is removed from a cGMP-gated Na+ channel, leading to the closure of the channel. Closing of the channel, blocks the influx of Na+ into the cell. In concert, vast numbers of photopigment molecules, G protein molecules, ion channels, and Na+ ions go through this process leading to the hyperpolarization of the cell. This hyperpolarization propagates from the outer segment to the synaptic contact of the cone, where it reduces the rate of release of the neurotransmitter glutamate. This reduction in neurotransmitter release is the cone’s signal that the cell has been illuminated. The foregoing lower level processes may be summarized schematically as follows: Photon capture → all-trans-retinal release → G protein activation → cGMP phosphodiesterase activation → cGMP decrease → cGMP released from ion channels ion channel closure → cone hyperpolarization → decreased glutamate release. Obviously a large number of these molecular processes occur together and these lower level processes implement the cellular process of signaling the presence of light by release of glutamate. Consequently, we can thus also see that the cone’s property of releasing a neurotransmitter in the presence of light is evidently realized by the properties and relations of the molecular individuals within the cell. So, that is a fair bit of the science of the biochemistry of phototransduction. Let me now plug this into our realization and multiple realization schemata. Our higher level property G will be the property of reducing the release of a neurotransmitter in response to the absorption of light of a particular band of frequencies. Or, oversimplifying somewhat, we might just say that G is 18 the property of transduction. The lower level properties F1-Fn will be the multitude of properties of the many individual green photopigment molecules, the many G protein molecules, the many cGMP phosphodiesterase molecules, the many cGMP molecules, the many Na+ ion channels, the many Na+ ions, the many synaptic vesicles, etc. One property of a green photopigment molecule that is relevant to vision is its λmax, another relevant property of the green cone opsin is its ability to bind and activate a G protein molecule, a relevant property of the G protein molecule is its capacity to activate a molecule of cGMP phosphodiesterase, the relevant property of cGMP is to bind and unbind from the Na+ channel, and so forth. These properties together can be placed into our schema for realization. This is a bit quick and imprecise, but I hope the basic picture of how this counts as realization is clear. Now turn to multiple realization. What we would need for this would be to have two cones that have the higher level property G of changing its release of neurotransmitter in response to the absorption of light, but which differ in their lower level realizer properties. To this end, we cannot rely on cGMP or Na+ ions. cGMP is cGMP and Na+ ions are Na+ ions. They do not vary in the properties they contribute to the realization of G. Also, since one might think that the identifying properties of a medium wavelength cone come from the properties of its constituent green photopigment, I want to argue for the multiple realization of the cone based the properties of other proteins in the biochemical cascade. So, think for a minute of just, say, the G protein molecule. Among its relevant properties is the fact that it is activated by the green cone opsin. But, as with the green cone opsin, there will likely be different amino acid sequences for the G protein molecule, hence different G proteins that differ slightly in their activation by the green cone opsin. Note that sometimes different amino acid sequences give rise to the same higher level property and at other times different amino acid sequences give rise to different 19 higher level properties. When different amino acid sequences give rise to the same higher level property, we can point to such cases as instances of multiple realization at the protein level. When different amino acid sequences give rise to different higher level properties at the protein level, we can use those higher level properties at the protein level as a basis for a case for multiple realization at the cellular level. So, the idea in short is this. We get one set of realizing properties and relations F1-Fn using the properties of one G protein and another set of realizing properties and relations F*1-F*n using the properties of another G protein. As a caveat on this, I should say that I do not know of experimental work specifically verifying my claim that G proteins vary in their amino acid sequences, hence in their higher level protein properties, but such studies may exist. In this case, I base my estimate of the likelihood of different amino acid sequences in G proteins, hence different properties of G proteins, on general evolutionary considerations. The last case I want to consider is a cell-to-tissue case which concerns how a certain property of the retina is realized and multiply realized by cellular level properties. In this example, the lower level individuals are the particular photoreceptor cells, the amacrine cells, bipolar cells, horizontal cells, and retinal ganglion cells. (See Figure 8.) The higher level individual is a retina. The lower level realizing properties and relations are properties of the cones, amacrine, bipolar, horizontal, and retinal ganglion cells. These lower level properties and relations include releasing glutamate in the presence of light within a given band of frequencies, binding certain neurotransmitters, having certain electrochemical synapses, and certain patterns of connectivity. The higher level property of the retina is its signaling a pattern of color in the visual field. 20 At this level, for the sake of simplicity, I will only focus on the contribution of the photoreceptors to the realization and multiple realization base of human color processing in the retina. As you know, human color vision is generally treated as trichromatic. The three types of cones in the normal human eye are sometimes referred to as red, green, and blue. That is, each type of cone changes its glutamate release in response to a different band of frequencies of incident light. (See Figure 9.) Each type of cone releases glutamate as it does in virtue of containing a chemically distinct photopigment. That is, each photopigment consists of a protein component, an opsin, covalently bonded to an 11-cis-retinal component and it is the opsin components that vary from cone to cone. The amacrine, bipolar, horizontal, and ganglion cells, of course, contribute to the retina as well, but again for simplicity we will set their role to one side. This cell-to-tissue example again supports the existence of the features of the realization relation we described in our previous cases. First, the lower level properties and relations of the cells stand in a synchronous, non-causal, non-logical determination relation to the higher level property of signaling patterns of color in the visual field. And there is no transmission of energy or mediation of force between the lower level properties and relations of the cells to the higher level property of the retina, where the relevant properties, and the individuals that have them, are not wholly distinct. Third, the relata in this case of realization are again of qualitatively distinct kinds. The relevant determining properties and relations of the cells include their capacity to release glutamate in response to illumination, releasing certain neurotransmitters, binding certain neurotransmitters, having certain electrochemical synapses, having certain patterns of connectivity, etc., whereas, in contrast, the determined property of the retina is its signaling patterns of color in the visual field. Fourth, in the case of the retina, once again many properties 21 and relations of individual cells together non-causally result in the retina’s property of signaling a pattern of color in the visual field. What reason is there to think that we have multiple realization of the color processing capacity of the retina? There are many. The simplest, however, seems to come from the recent discovery of large individual variations in the cone mosaic, the layout of the cones in the fovea of the retina. One of the things that matters to color vision is the presence of red, green, and blue cones. Yet, the red-green cone ratio varies substantially. Figures 10 and 11 are illustrations of the differences between two subjects, JW and AN, reported by Williams and Roorda, (1999). The following provides quantitative data on these differences. Subject Red cone (%) Green cone (%) Blue cone (%) Red: Green cone ratio JW 75.8 20.0 2.1 3.79 AN 50.6 44.2 5.6 1.15 Although all three types of cones (R, G, & B) matter for normal color vision and although the ratios of these types of cones matters for color vision, the ratios can still vary by a sizeable amount. So, it looks as though there are various combinations of instances of the properties and relations of the red, green, and blue cones that can enable normal color vision. Another way to get multiple realization of a color processing retina is through dichromacy. Dichromacy is a genetic disorder in which one type of cone is not present, typically either the red or green cones. These are the most familiar types of color blindness. Yet, color blind individuals are not completely unable to see colors. Instead, there are certain color discriminations they cannot make. They make fewer color discriminations than do color normal individuals. So, their retinas are still color processors. There are, however, three ways of being a dichromatic retina, each corresponding to the loss of a distinct photopigment. The retina of a 22 ‘protanope’ lacks red cones, the retina of a ‘deuteranope’ lacks green cones, and the retina of a ‘tritanope’ lacks blue cones. Each form of dichromacy corresponds to a distinct realization of an instance of the property of signaling patterns of color in the visual field. So, we have a second type of multiple realization in the human retina. A third type of multiple realization involves the less dramatic, and perhaps somewhat less familiar, cases of what are termed ‘anomalous trichromats’. These individuals possess three distinct types of cones, but the sensitivities of the cones are not those of normal cones. So, for example, in one of the most common forms of red-green colorblindness, the red cones and the green cones release glutamate in response to relatively similar bands of electromagnetic radiation, hence there is not enough difference in the properties of the red and green cones to implement a higher level process that can signal differences between certain patterns of color. This gives rise to color-sensitivity “blindspots”. Anomalous trichromacy is, of course, a kind of color processing, and its subjects do realize instances of the property of signaling patterns of color in the visual field, but in addition it grades off into normal color vision. There is no sharp dividing line between anomalous trichromacy that is normal and anomalous trichromacy that is abnormal. To take but one well-known example in the biochemistry of vision literature, the human red cone appears to be ‘polymorphic’. That is, it comes in two forms. One form of the red cone has the photopigment with an amino acid sequence with a serine amino acid at position 180, where the other photopigment has alanine at that position. Both forms are quite common in the human population. They correspond to the two red curves in all the panels of Figure 3. So, one has the multiple realization of normal human color vision by having some individuals with cones containing one red photopigment and other individuals having cones containing the other. 1 1 See Sharpe, et al. (1998). 23 As a fourth and final example of multiple realization of color processing in the retina, I want to draw on some recent work on the biochemistry of opsins. In a classic paper, Nathans, Thomas, and Hogness (1986) identified the gene sequences, hence the amino acid sequences, of the three opsin components of the photopigments. In addition, they found that normal humans vary in the number of genes coding for the green pigment. In other words, there are multiple loci each coding for a green photopigment. Subsequent research has also found that normal humans vary in the number of genes coding for the red pigment.2 This suggests the possibility that a given individual can possess distinct versions of the gene for the green and red photopigments at the different loci, hence can possess distinct green and red photopigments, and distinct green and red cones. Further, it is hypothesized that part of the reason for individual variation in color sensitivity within humans is due to differences in the number of different kinds of cones. Some individuals might thus have, say, only one type of green cone and one type of red cone, where other individuals might have, say, two different green cones and seven different types of red cone. So, let me summarize how I have applied our schemata for realization and multiple realization to my scientific examples. First, I argued that the properties and relations of amino acids and 11-cis-retinal realized the light absorbing property of the human green photopigment and that this could be multiply realized by distinct amino acid chains that have the same absorption spectrum. Second, I argued that the properties and relations of photopigment molecules, phospholipid molecules, water molecules, potassium ions, sodium ions, G protein molecules, cGMP molecules, etc. realize a cone’s property of transducing light. I argued that the cone’s property of transducing light was multiply realized by G proteins with distinct activation 2 Neitz and Neitz, (1995). 24 properties. Third, I argued that color processing in the human retina is realized by properties of the cones, amacrine cells, bipolar cells, horizontal cells, and retinal ganglion cells. Color processing is multiply realized by variations in the cone mosaic, dichromacy, anomalous trichromacy, and the apparent variety in the number of distinct cone opsins normal humans carry. 4. Conclusion The long term goal of my work with Carl is to provide a clear and correct theory of a particular kind of realization relation and a corresponding notion of multiple realization, then to use that in combination with recent neuroscientific work to justify the Massive Multiple Realization Hypothesis that many human psychological properties are multiply realized at many neurobiological levels. In this talk, I tried to provide an introduction to this project. I presented a bare bones sketch of our views on realization and multiple realization, then showed how it provides an account of the realization and multiple realization relations that neuroscientists have found to underlie some psychological processes in human vision. This is far from a completed project, so I would be very interested in comments and question on it. 25 Figure 1. A) The Eye. B) A Cross-section of the Retina Showing the Principal Cell Types (Including the Rods and Cones). C) The Outer Segment of a Cone. D) Photopigments Embedded in the Membrane of the Cone Outer Segment. E) The Amino Acid Chain of a Cone Photopigment. 26 Figure 2. Green Cone Photopigment Sensitivity Curve. (Modified from Blake and Sekuler 2002, Figure 2.23, p. 74) 27 Figure 3. Absorption Spectra of the Normal Green Human Photopigment (Green Curve), the Two Common Red Photopigments (Red Curves), and Various Red-Green Hybrid Pigments (Gray Lines). The labels, such as R2G3, name the hydrids. (Merbs and Nathans 1992, Figure 1) Figure 4. Human rods and cones. (From Blake and Sekuler 2002, Figure 2.29, p. 69). 28 Figure 5. Photopigment Molecule Embedded in the Cell Membrane and Phospholipid Molecules of the Membrane Constituting a Cone. (Modified from Sharpe, Stockman, Jägle, and Nathans 1999, Figure 1.2, p. 6.) Figure 6. Schematic of an opsin embedded in the cell membrane. The seven cylinders represent portions of the opsin spanning the cell membrane. (Based on Sharpe, Stockman, Jagle, and Nathans 1999, Figure 1.17A, p. 43 29 Figure 7. Phototransduction Path (in a Rod). Figure 8. Cells Constituting the Human Retina. 1) Rods, 2) Cones , 3) Horizontal Cells, 4) Bipolar Cells, 5) Amacrine Cells, & 6) Retinal Ganglion Cells. (Modified from Wässle 2004, Figure 1, p. 2.) 30 Figure 9. Absorption Spectra of the S-, M-, and L-cones. (From Blake and Sekuler 2002. Figure 2.23, p. 74).