In this paper I will examine a response of Paul M. Churchland to Bas C. van Fraassen's account of empirical adequacy and its relation to observability in his book The Scientific Image. I wish to show that while not all of Churchland's arguments are conclusive against various aspects of van Fraassen's position, they are compelling and persuasive with regards to the more controversial problems. The three main topics I will focus on are what should count as observable, what things we can and cannot observe, and what pragmatic and epistemic consequences the answers to the previous questions can reasonably be taken to have. |
Paul M. Churchland in his essay "The Ontological Status of Observables," written in response to Bas C. van Fraassen's The Scientific Image, bases one main distinction between his own scientific realism on the one hand and van Fraassen's constructive empiricism on the other hand in the different ways the two schools of thought treat the ontological status of observable and unobservable objects. A realist wishes to extend full ontological existence to all entities postulated by our best current scientific theories, whether they are observable or unobservable, with the term 'observable' being a highly technical term whose correct extension is the topic of discussion in both Churchland's paper as well as in the current paper. Constructive empiricism of van Fraassen's flavor says that we ask too much of our best scientific theory if we ask it to be true about all things the theory treats with. Rather, we should be content with simply truth about observable objects, e.g., this chair I'm sitting on, this book in front of me, this cat sitting beside me. One need not ask any more of one's scientific theory than that it say true things about observable objects, and that beyond that there is no reason to believe that a theory which is empirically adequate with regard to observables has to also be true with regard to unobservables. This is precisely the bone of contention which Churchland wishes to address.
Van Fraassen appears to use the term 'empirical adequacy' in two ways, as needed. In the first, a theory is empirically adequate if everything that it says about observable objects is true. In the second, a theory is empirically adequate if everything that it predicts about observables is true. The difference can be subsumed by taking 'says' to include both present descriptions as well as descriptions of future states, as this is merely what prediction is: describing a future state of the world and then waiting to see if our description is correct. An accurate prediction is one that, in time, becomes an accurate description. Churchland attributes simple descriptive (not predictive) accuracy to van Fraassen's idea of empirical adequacy when he says that "van Fraassen asserts that descriptive excellence at the observational level is the only genuine measure of a theory's truth" (35). At a later point, Churchland says that "roughly, a theory is empirically adequate if and only if everything it says about observable things is true" (37).
If it is reasonable to take questions of empirical adequacy vs. truth as what distinguishes a scientific realist from a constructive empiricist, then the question that Churchland pursues in his paper is one of crucial importance. Specifically, if a constructive empiricist (or indeed, any empiricist) is to draw a hard line between facts about observables and facts about unobservables, he needs to be able to give a principled distinction between the two: Some genuine ontological characteristic that can underwrite the epistemological differences on which an empiricist needs to rely. As Churchland shows in his essay, giving an argument for a principled distinction is extremely difficult to do.
Before considering the strengths and weaknesses of Churchland's arguments against van Fraassen's position on the observable/unobservable distinction, it is important to try to get a clear picture of what van Fraassen means by "observable." This is not an easy task, for at different points he says different (and sometimes seemingly contradictory things). Van Fraassen at one point gives a general principle for observability (one which he notes is not meant to be a definition):
X is observable if there are circumstances which are such that, if X is present to us under those circumstances, then we observe it (16).
Thus, "seeing with the unaided eye is a clear case of observation" (16) [1] and "a flying horse is observable...and the number seventeen is not" (15). Similarly, the moons of Jupiter are observable because they "can be seen through a telescope: but they can also be seen without a telescope if you are close enough" (16). It is worth noting here that it is not only objects (such as trees, people, cats) that are observable, but also things like motions, which are not objects but events. In a discussion of planetary motions, van Fraassen says that "for [Ptolemy], there was no distinction between true and apparent motion: the true motion is exactly what is seen in the heavens" (45, emphasis mine), even though "what that motion is, of course, may not be evident at once: It takes thought to realize that a planet's motion really does look like a circular motion around a moving center" (45). The introduction of the Newtonian theory of planetary motions forces us to distinguish between actual motion and apparent motion; but the apparent motion is still something that is observable by us. Furthermore, as van Fraassen points out, there is an "intertranslatability of statements about objects, events, and quantities" (58). If we can talk of both "there is a molecule in this place" and "the event of there-being-a-molecule occurs in this place" (58), then what is an observation-statement about an object can be translated into a corresponding observation-statement about events or quantities.
Embedded in the above discussion of what is and is not observable are three different levels of observability that must be distinguished from each other. Churchland gives the levels as follows (40):
(1) things observed by some human (with unaided senses)
(2) things thus observable by humans, but not in fact observed, and
(3) things not observable by humans at all.
Another, perhaps more perspicuous way that these three distinctions between observables can be put is this:
(1*) Things that actually occur and are observed.
(2*) Those of type (1*) and those that actually occur and aren't observed, but are, in principle observable.
(3*) Those of type (2*) and things that don't occur but would be observable if they occurred.
According to Churchland, an empiricist like van Fraassen would want to accept objects and events of type (2), but not (3), and he wants to show that van Fraassen could not give a good reason for not accepting (3). As Churchland says, "admittedly, for any distant entity, one can in principle always change the relative spatial position of one's sensory apparatus so that the entity is observed: one can go to it. But equally, for any microscopic entity, one can in principle always change the relative spatial size or configuration of one's sensory apparatus so that the entity is observed" (40). Thus, one would think that such contingent features about our physical make-up, such as size, spatial location, etc., are not significant contributing factors to what counts as observable. However, this is precisely the type of observability that van Fraassen wishes to deny: He wishes to be able to consider the moons of Jupiter as observable because, in principle, we could travel to them and perceive them with our unaided human senses. But, also in principle, he argues, we cannot do the same thing with viruses, and other minutiae; there is something inherently different about our changing our size that causes us to be able to call things seen by changes of spatial position observable but things seen through changes of size are not.
Much here turns on the use of the phrase "in principle," used in my formulation above as well as in both Churchland and van Fraassen's rhetoric. Whether this distinction can be motivated is questionable. Churchland argues that there is no sufficient distinction between the second and third levels. I think the point can be pressed even further: The second level is not a viable option at all. It will either collapse into the first or collapse into the third, and this turns on the use of the phrase "in principle" in the second level. In trying to widen his scope to encompass objects like the moons of Jupiter, van Fraassen appears to want to be able to accept counterfactuals of a certain type: Those of the sort "What I would have observed if I was there where (when) it happened (or existed)," e.g., ones that involve problems with space and time, not size or other such characteristics. Van Fraassen says that things like the moons of Jupiter are in principle observable, crucially because if we were at their location, we would be able to see them, without changing our human senses. What van Fraassen does not want to admit as observable are things like viruses (which we ourselves are too large to see), or neutrinos (which we are just simply not able to sense). Thus counterfactuals of the type "If we were at location x, we would be able to see y" (even though we are never at location x and we never see y) are acceptable, but those of the type "If x were to happen, we would be able to see y" (even though x never happens and we never see y) are not acceptable.
When the different types of counterfactuals are presented in this fashion, it looks, naively, as if there might be a way to give a principled distinction between the two, to vindicate van Fraassen's desire for the second level without the third level. What someone might try to say is this: "I want to admit things like the moons of Jupiter into my observable ontology, and thus worry about whether my best scientific theories are adequate with regards to these objects, because, even though I can't actually see them with unaided sensory apparatus, there is no limitation, in principle, stopping me from traveling there and seeing them. However, this does not mean that I have to worry at all about things that I do not observe nor ever could (either because they are too small or because they will never happen)." It is, in a sense, a matter of the "nearness" and "farness" of the type of counterfactual. Those of the second type require possible worlds that differ very little from the actual world, and thus are in some sense "closer," while those of the latter type require more differences, and thus are "farther." These "far" counterfactuals are not themselves impossible, but, in some sense of the probable, they could perhaps be considered less probable or likely.
However, this difference in types of counterfactuals ends up being just a matter of phrasing. The two types of counterfactuals can be represented in a different way that clearly displays their similarity. More than this, though, it can be shown that if empirical adequacy is concerned with predictive excellence as well as descriptive excellence, some counterfactual statements of the third level must be admitted.
To show the similarity of the two types of counterfactuals, consider these rephrasings:
* What I would have observed if I was there where it happened.
* What I would have observed if it had happened where I was.
To make the similarity even more apparent, "where it happened" is identical to "where I was" in both cases. A theory that accepts counterfactuals of the first type but not of the second is unable to be empirically adequate in terms of prediction. Here is a straightforward example: You are sitting across the table from me, and I have in my hand a light rubber ball. What would happen if I threw the ball at you?
One would think that our scientific theories should be such that if I could accurately describe this event once it had happened, I could predict precisely what would happen even if I never threw the ball! Yet, since this counterfactual is a counterfactual from level three, van Fraassen's empiricism does not cover these circumstances. To say that we do not wish our theories to be adequate concerning counterfactual situations, because they never happen, seems to me to be wrongheaded. Thus, while there appears to be no epistemic reason why we shouldn't accept level three, there appears to be a pragmatic reason why we should accept them.
One response that a van Fraassian could make is this: Discussions of future states are not counterfactuals. [2] A counterfactual, by definition is something which is "counter to fact." Since future events have not yet occurred, there is no fact to run counter to. (I am for the moment foregoing assuming a non-deterministic view of the world. If the future state of the world is completely determined by the state of the world as it currently is, then this response fails immediately because we can construct things that are counter to (future) fact). In return, my response would be this: This takes too narrow a view of what can count as a counterfactual. The situation described above, where it seems to be the case that I should be able to predict what will happen if I were to throw a ball at you (in the future at some point), even if I never throw the ball. The difference between counterfactuals like "If I were to throw the ball at you, then..." and "If the world were flat, then..." is that in the second case, we know that the antecedent is false, whereas in the first case, we don't know. In order to make the cases exactly analogous, one would have to amend the antecedent of the first to "If I were to have thrown the ball, then .". This emendation, I believe, is uncontroversially a counterfactual. I believe this shows how it is possible to create future counterfactuals, even though there is no fact yet to run counter to. Therefore, the problem of requiring theories to be empirically adequate in terms of prediction while simultaneously refusing to admit counterfactuals of any type remains.
Let us suppose, for a moment, that empiricists are not to be concerned by this, and suppose also that empirical adequacy does not (perhaps cannot) include predictive excellence. If it is descriptive excellence, though, we are left in the funny situation, at level two, in that we could never know that our theory was empirically adequate, because counterfactuals regarding location are still precisely that: Counterfactuals. Van Fraassen seems to want empirical adequacy of a theory to cover what things would be like on the moons of Jupiter if we were there to observe them, but we will never be able to know if such a theory is empirically adequate [3]; we have no epistemological access to events which never occur nor objects which never exist. Therefore, if we wish our theories to be adequate concerning these things as well as actually occurring events and existing objects, we will never be able to determine when or if our theories are adequate. We seem to be falling back to the position of level one, where observable gets conflated with observed. An empiricist cannot reasonably hold on to a definition of observability of type two: Either one must accept counterfactuals of all types (in which case they would subscribe to type three) or one must forego counterfactuals of all types, including all predictions, since a prediction is the statement about a possible situation that does not become actual until an experiment is made (in which case they must subscribe to type one).
Though there certainly are passages in van Fraassen that lead one to believe that he would be amenable, in terms of delineating observability, to counterfactuals of the type expressed by level two, (e.g., what is not observed now but would be observed were we in the correct spatial location), there are other places where it is not clear that van Fraassen would accept counterfactuals of this type, in terms of for what we want our theory to be empirically adequate. More precisely, while he has explicitly stated that the moons of Jupiter are, in principle, observable, because we would be able to see them if we were there, it is not clear whether this observability is sufficient to warrant a desire for empirical adequacy concerning any observations about the moons. However, one would naturally think, a theory should be empirically adequate for all objects and phenomena that are observable, including objects and phenomena that we never come in contact with, the following quote adds some puzzlement to the discussion:
We have here a counterfactual: if two bodies have different masses and if they were brought near a third body in turn, they would exhibit different acceleration. (Remember that empirical adequacy concerns actual phenomena: what does happen, and not, what would happen under different circumstances) (60).
Van Fraassen's position seems entirely untenable: First he wishes to say that things which we do not see but would see if we traveled to them are observable, but then he wishes to say that empirical adequacy should not concern itself with these observations, because they will never be made. But this a fortiori contradicts the general idea of empirical adequacy that van Fraassen himself subscribes to earlier in his book: "truth about observables." Here, the moons of Jupiter are observable, but since they are so only under counterfactual situations, we need not ask our theories to be empirically adequate with regards to them.
Though Churchland does not present it this way, the main thrust of his arguments against what he calls van Fraassen's selective skepticism is that van Fraassen is taking contingent physical facts, facts about what sensory apparatus we humans actually have rather than apparatus that we humans could have had, as contributing factors to the necessary epistemological status of objects. The world is such that we in fact are able to change our spatial position, but we happen not to be able to change our spatial size, and thus things that can be observed through changes of the first are to be counted as "observable," but things that could be observed through changes of the second are not to be counted as such. One of Churchland's responses is to propose a thought experiment, concerning an arboreal scientist, Douglas van Fiirrsen, who has greater sensitivity of sensory apparatuses but is rooted in place and so cannot move. Van Fiirrsen, says Churchland, "urges an antirealist skepticism concerning the spatially very distant entities postulated by his fellow trees" (39-40). This is just one illustration of the problem of trying to take contingent features of the observer as a guide to differences in the observed. Van Fraassen's objection to this thought experiment, that we should be concerned only with how we are not how we could be, is addressed in the final section of this paper.
Elliott Sober in his article "Epistemology for Empiricists" takes a different tack. Rather than concentrating on the problematic idea of taking contingent features about our physical makeup as contributing to some epistemological characteristic that objects or events necessarily have, he say that "nonetheless it is entirely unclear why this difference between moons and viruses should matter to us epistemologically.the mere fact that Jupiter's moons are observable while individual viruses are not is epistemically irrelevant" (Sober 41). His basis for claiming this is so comes from the principle of actualism, which
says that we should form our judgments about hypotheses based on the evidence we actually possess; possible but nonactual evidence does not count.It is not the distinction between observable and unobservable that is fundamental, but the difference between observed and unobserved. Empiricists should attempt to show how actual observation provides more certainty than other supposed routes to knowledge (Sober 43).
The principle of actualism, thus can be taken to support the claim that there is no important sense of observable in the second type. Either we can accept counterfactuals and worry about empirical adequacy about things that we will never see but could, perhaps, see, or, in keeping with the principle, we can eschew everything but actual evidence gained from observation, and let that guide us in our knowledge.
Churchland also argues that since "[o]ur history contains real examples of mistaken ontological commitments in both [observable and unobservable] domains" (36), we should regard observable entities in the same skeptical light as unobservables. Furthermore
We are too often misled, I think, by our casual use of observes as a success verb: we tend to forget that, at any stage of our history, the ontology presupposed by our observational judgments remains essentially speculative and wholly revisable, however entrenched and familiar it may have become (37).
While I think that this is a correct point, I do not think van Fraassen would agree with it. What does it mean to say that our observable ontologies are "essentially speculative and wholly revisable"? If at some point in history we can "see" witches and at a later point we can longer "see" them, what is going on?
One thing that is clear is that our actual perceptual ability to see the objects that we call witches, whether or not we call them witches, cannot have changed. Say I am in the presence of the Wicked Witch of the West. Ceteris paribus, I can see her. Thus, one could describe the situation as that of one where I see a witch. These types of observational claims, (e.g. "I see the Wicked Witch of the West") are as liable to error as claims about unobservable things. We could simply be mistaken, and only think that we are seeing the Wicked Witch, when in fact it is just an illusion. But this is not the type of instance that Churchland is speaking of when he talks about seeing as a success verb. 'Seeing' is not always a success verb; there are certainly occasions where I may think I am seeing something, may describe myself as seeing something, but in fact actually do not.
So I could be mistaken as to whether I am actually seeing the Wicked Witch. However, two further questions need to be asked: Can I see that she is a witch? and Can I be mistaken when I see that she is a witch? Seeing in this propositional sense does appear to be a success verb; we either do or not see that X is p, and this we cannot be mistaken about. (We cannot see, e.g. that the cat is grey unless the cat actually is grey. See Dretske (1967) ch. 3)
To say that at one time we can see witches and at other times we can't is incorrect; if something is observable then de facto it will always be observable. However, Churchland is correct that at one point in history, we were able to see that certain people were witches, but that today we no longer can. Does it make sense to say that this constitutes a revision of our observable ontology? As van Fraassen points out in his discussion of the observable/unobservable distinction, "it is also important here not to confuse observing (an entity, such as a thing, event, or process) and observing that (something or other is the case)" (15). The first type of observing is acceptable; the second is irrelevant for van Fraassen's purpose. Therefore, I think van Fraassen's response would be that we see what we see (and while we are sometimes mistaken about what we see, this is non-threatening), and that this domain does not change if our perceptual apparatus does not change. It is in some different sense that we "revise" our observable ontology, because propositional seeing, seeing that something or other is the case, relies on theory; without a theory of witches, we would not be able to say that we see that something is a witch. If we revise our ontologies in the way that Churchland suggests, van Fraassen would argue, that is only because we have revised our theories. The fact that we revise our ontologies based on our theories does not affect whether or not we do or should revise our ontologies based on observation.
The only way that such revision could carry any important weight is if our being wrong about seeing that something is a witch in today's world rests upon some fundamental physical essence of which a person can be right about. We are comfortable revising our views of theoretical entities to match observation, and we do so regularly. Less regularly we revise our view of observable entities to match theory: Every day we observe tables, trees, books, cats, and we also observe that they are tables, trees, books, cats. These categories of "natural kind objects" are extremely theory laden. It is theory that allows us to see that x is P. It is theory that gives us the ability to see that things, rather than just see things. To say that we see that something is a tree, rather than to just say that we see a tree, presupposes that there is some category of objects "tree." If our theories failed to distinguish between, say, four-legged chairs and three-legged stools, there would be no way that we could see that something was a three-legged stool. Yet as soon as that concept is introduced to our ontologies, then we will, when we are sufficiently familiar with the new concept, be able to see that things fall within the bounds of the category. As our beliefs and knowledge about the world changes, so will what we are able to see, in this propositional sense of seeing.
The presence of these natural kind terms (and by this I mean category labels that apply to both 'natural' objects as well as 'artificial', e.g., to both cats and tigers and chairs and trees) is a reflection of some kind of metaphysical structure to the world. [4] Such metaphysical underpinnings would be abhorrent to empiricists like van Fraassen and thus that is one reason why he wishes to eschew seeing that when defining what is observable and what is not. [5] It is not so much that what we can observe (in the sense of "what we are (physically) able to observe") is revisable, but rather that what we can observe that is revisable, because such revision is based upon theories, and our theories change with time.
There is good reason, though, to think that restricting from our ontologies things we see that is misguided. The first is in the case of events. Van Fraassen at times appears to want things like events to count as observable, but events are generally cases of seeing that. I do not "see the motion" of a ball as it flies through the air, because there is no object that is "the motion." Rather I see "that the ball is moving" or "that the ball has changed in velocity." Consider for a moment a more concrete example: Copernicus' grand idea that all the planets orbit the sun, rather than the sun and the other planets orbit the earth. Is this heliocentricity of the solar system observable? I think two different answers could be supported given a van Fraassien idea of observation.
Answer #1: No. Our evidence for the heliocentricity of the earth primarily comes from mathematical equations (such as Newton's laws, which were developed in congruence with a heliocentric notion). One could say that our belief in heliocentricity is indirectly supported by observation in two ways. The first is Newton's laws predict both the motions of the planets and thus the heliocentricity, and these laws are empirically adequate with respect to the motions. The second is the observations we make of parallaxes and the retrograde motion of Mars. These two observations themselves are not directly observations of the planets orbiting the sun, but they are indirect observational evidence that could raise the likelihood of the solar system being heliocentric rather than geocentric. The only way that it appears correct to say "We see that the sun is the center of our solar system" is by using a sense of "see" that does not connote actual physical sensory perception. Rather, it is being used as a synonym for "realize" (Dretske, 80). Thus, one could say that we do not observe the heliocentricity of the solar system (or, we do not observe that the sun is the center of the solar system).
Answer #2: Yes. It is harder to support a favorable answer to the question, but one possibility presents itself immediately and thus must be dealt with. It seems natural to compare seeing the heliocentricity of the solar system with seeing the moons of Jupiter. Surely, if we move out far enough, we'll be able to see the planets orbiting the sun. However, the case is not exactly analogous. We only have to move closer to Jupiter to see its moons; we would have to move drastically farther out, though, to see the planets orbiting the sun. In fact, if we moved out far enough to be able to see all the planets from Mercury to Pluto in order to ascertain their orbits, we would no longer be able to even see the planets. From beyond Pluto, without a telescope or other visual enhancer, the planets are too dim to be visible. Thus, this argument for the observability of the heliocentricity of the sun is tenuous, for it seems to be that we need to both be close to see the planets and yet far away to see their orbits. This, unlike the other examples of spatial location change we have seen, and unlike change in size, is prima facie impossible. We cannot be in two places at once, and yet this appears to be what is necessary.
It seems that the question must be answered negatively, then. Yet this is not a position that sits well, I believe; there are probably few scientists who doubt that the planets do indeed orbit the sun, that would be willing to say that "since we cannot observe this, we must be forced into skepticism concerning its truth." There is other, non-observational evidence that we have for its truth.
This example illustrates an important point that is at the root of all this: The only things for which we should be seriously skeptical about are things for which we have no evidence, direct observational or other, for their existence. The realist/empiricist division is based upon whether or not we have reason to believe in so-called "unobservable" things like the mitochondria living within our cells. A realist would say that we do, because we appear to have evidence (through microscopes, etc.) for their existence, while an empiricist would say no, because we have not yet observed them in van Fraassen's preferred definition of observation. What both the realist and the empiricist are agreed on, though, is that neither of us has reason to believe that within the mitochondria are the further creatures, farandolae, of Madeleine L'Engle's fancy. Though our only evidence for mitochondria is gained through microscopes (which for van Fraassen would not counted as observational evidence) it seems as if this does give us both reason to believe in the existence of mitochondria (and thus ask of our best scientific theories that they give us true descriptions and predictions concerning mitochondria) and reason not to believe in the existence of farandolae, because these cannot be observed even with the aid of electron microscopes (nor have we any other reason, from theory or not, to believe in their existence). To say that mitochondria are on the same epistemological level, concerning evidence for our belief in their existence, as farandolae seems fundamentally misguided.
To make this point explicitly clear, let us consider three things: Cells, mitochondria, and farandolae. The first we can see with the naked eye, and thus have excellent reason to think exist. [6] The second we can see through microscopes, and thus have good reason to think exist. The third we cannot see at all, and thus have no reason to think exist. According to van Fraassen's definition of empirical adequacy, we should only expect from our theories truths about cells, and not mitochondria and farandolae. Furthermore, on his view, our seeing mitochondria through a microscope is not enough to warrant the truth of "Mitochondria exist." But if there can be given no principled reason for van Fraassen's choice of what is observable and what is unobservable, there can be no such thing as "empirical adequacy," which is defined in terms of "truths about observables." Putting mitochondria on the same side of the fence as farandolae instead of cells is non-intuitive and not supported by van Fraassen's observable/unobservable distinction.
In passing in the previous two sections we have been given reason to view the observability of an object not as an epistemologically-significant feature of the object. I think that Churchland was conscious of these possibilities in his essay, but the further argument that he presents against van Fraassen's claim that virtues like simplicity, coherence, and explanatory power are only pragmatic virtues and not epistemic virtues is not very persuasive. When the epistemic choice is between two radically different ways to view the same empirical facts, we cannot appeal to those very same facts to ground our choice. "In such a case," says Churchland, "the choice must be made on the comparative global virtues of the two global alternatives" (41). However, the fact that we must rely on these superempirical virtues to make our choice is not proof that they must be epistemic virtues and not pragmatic. If we appeal to simplicity, coherence, etc., in picking which global world view we are to view our empirical data in, we have no guarantee that these virtues will lead us to picking out the right global theory (or, in the realist vocabulary, the true global theory). Churchland is correct when he compares questions of which global theory to adopt to Carnap's idea of internal/external questions. Carnap (1947) explicitly says that questions of which framework to adopt are questions of what is pragmatically useful, not what is epistemically correct. The choice of frameworks is
rather a practical question, a matter of a practical decision concerning the structure of our language. We have to make the choice whether or not to accept and use the forms of expression in the framework in question (207).
Both Churchland and van Fraassen would balk at this specific wording of the problem, but that is only because of Carnap's framing of this as a linguistic question. (The problems that stem from trying to construe empiricism as a linguistic thesis have been amply shown in other places. As van Fraassen says, "no concept which is essentially language-dependent has any philosophical importance at all" (56)). But the general claim is still tenable for both: The choice of which scientific glasses we adopt through which to view our empirical data is not something that can be guided by the empirical facts themselves, when both theories are empirically equivalent. We must be guided by other pragmatic virtues, but this does not show that explanatory power, simplicity, and the other similar things that we appeal to in making our decision should be construed as epistemic virtues.
Churchland concludes his essay by considering two thought experiments. The first asks us to imagine people who have no biological sensory organs that allow them to 'observe' what we humans can, but "on top of [their] skulls a microcomputer fitted out with a variety of environmentally sensitive transducers" (42). These computers are connected to the brain so that their input from the outside world is passed on in such a way that the people are provided with "much the same information that our perceptual judgments provide us" (43). Given the definition of observability put forward by van Fraassen, there is nothing that is observable for this society, so "there is no question, therefore, of their evaluating any theory by reference to its 'empirical adequacy'" (43). Churchland uses this as an objection to van Fraassen's delineation of observability because intuitively, it seems correct that this society could indeed actually carry out science in much the same way that we do. But this misses the point: The problem is not that observability is relativized to human, rendering it a useless concept when discussing these thought experimental humanoids. These are not humans, they are something else, and could easily have a differently-relativized definition of observable, in the same way that observable for, say, a honey bee, is different from what is observable for us. The problem is not that van Fraassen has relativized observability to human-observability, it's where he's drawn the line that's problematic.
The second thought experiment concerns people who are able to see, not less than we do, but more, including the "microworld of viruses, DNA strands, and large protein molecules" (43). These people are born with electron microscopes in the place of their left eyes; further, we are to suppose that these electron microscopes are biological in composition, as human eyes are. Churchland's conclusion is that there is no good reason why we cannot count the information gathered by either the first or the second groups of humanoids as observable evidence. Thus, if we wish to consider as observable evidence what the second group sees through their biological electron microscopes, there seems to be no reason why we could not now count what we see through our non-biological electron microscopes as observable evidence.
Van Fraassen's response is that producing thought experiments like this is not productive because the entities postulated are not human, and even if humans at some point will evolve the characteristics present in these humanoids, then what is observable for us would not equal what is observable for these future humans. Thus, considering the perceptual apparatus of these non-human creatures has no bearing on what is observable for humans (1985, 284). This is a reasonable critique of the thought experiment, but does not consider a reasonable extension that can be made of it, to which I alluded in part in footnote 1. Before discussing a pro-Churchland response, I would like to first consider another quote concerning observability from van Fraassen:
But science itself designates certain areas in this picture [of the world] as observable (57).
What does it mean for "science itself" to "designate" something? First of all, I would say that if this is to mean that it is science that demarcates what is observable and what is not, then unless science has given us a reason to think that our unaided perceptions have any greater access to true beliefs about the world, then we are not warranted in assuming that we are on firmer epistemological ground with evidence gained from these senses (which van Fraassen is committed to, otherwise he would have no reason to say that knowledge gained through sensory perception contributes greater evidence to the existence of objects than knowledge gained through other sources, such as mathematical equations). Secondly, in giving science the authority to designate what is and is not observable, it is reasonable to think that as science changes, so will what it designates. This is the argument sketchily advanced in the earlier footnote: If science allows us (as it does right now) to perform surgeries on eyes so that they can see better than they could before (thus presumably see, among other things, more minute items), why should we think that at some point in the future science won't allow us to perform surgeries implementing electron microscopes in the fashion that Churchland discusses? To say that humans with surgically attached electron microscopes are no longer humans would be to say that humans who have had any eye corrective surgery are no longer humans. But if science allows the possibility of biological electron microscopes, that means that what could be observed through those new eyes would by necessity have to count as observable now: For what is observable does not change in extension. If science will one day allow us to add electron microscopes as parts of our physical bodies, then what things are observable through them both then and now must count as observable both then and now. Because the extension of observable does not change, if something is observable, it is observable now, even if it is not observed. [7] To say otherwise would be inconsistent. Again, it is unclear whether van Fraassen's position is even tenable.
This, in conclusion, brings up a question of pragmatics for van Fraassen and empiricists of his vein: If we aren't going to include things that are visible only through microscopes (even non-electron ones), we're severely limiting what we can say. While the ontological status of things such as quarks is legitimately controversial, as we have little concrete evidence for them at all, there are other things that would count for van Fraassen as unobservable whose ontological status is not so controversial. Do we want to say that cell nuclei, for example, are theoretical entities in whose existence we have no real reason to believe? Given van Fraassen's use of "observable" in his definition of empirical adequacy, he cannot say that cell nuclei are observable, for we cannot see them without the aid of microscopes. Thus, cell nuclei are on the same ontological level as viruses, namely, a level about which we must be skeptical.
Van Fraassen himself notes two examples of things that we would not have thought would be observable (in his sense of the word) that turned out to actually be:
Certain crystals, modern science tells us, are single molecules; these crystals are large enough to be seen.so, some molecules are observable. The second was mentioned to me by David Lewis: astronauts reported seeing flashes, and NASA scientists came to the conclusion that what they saw were high-energy electrons (58).
Given that we have a single effective and cohesive theory that postulates the existence of both large and small molecules, high- and low-energy electrons, without distinguishing between them in any significant fashion, why should we count some molecules observable when others are not? Why should we call high-energy electrons observable, and thus accord them positive ontological status, and not low-energy ones? To ask why it is not appropriate to say "I can see big Xs but not little ones, that's why I believe in the existence of big ones and not little ones" requires the question "Why do you believe that the big Xs and the little Xs are actually the same object?" If the answer is that our best scientific theory tells us they are the same and if this theory gives us no reason to distinguish the ontological status of the two types, then it seems incorrect that we should feel the need to appeal to contingent facts about our physical make-up, and say that these give evidence for a significant difference in ontological status. Indeed, if it is true that science is what designates what is observable, it is wrong for us to appeal to these contingent facts.
It seems that there is an epistemological difference between things we can see only through an electron microscope and things that we cannot see at all with our current technology. It is those theoretical entities which do not impinge on our physical sense at all that are on a shaky epistemological foundation; it is concerning those entities that we should be more cautious in postulating their existence.
An empiricism of van Fraassen's type seems, in the end, to be horribly hamstrung by the limitations put upon what is observable. If we are to gather our evidence from and base our knowledge upon only things that are observable in van Fraassen's sense, then much of what is taken as knowledge by the scientific, and non-scientific, communities must be given up as unsupported. We can no longer know that we are in a heliocentric solar system. Cell nuclei, molecular structure, all these must be passed over as mythologies. Any evidence that we could appeal to for support will not be observable in van Fraassen's strict sense. But how many people would truly be comfortable in saying that these things are no more than theory, and should not be counted as true? If van Fraassen's empiricism is to be a normative empiricism - a claim about what science and scientists should do - he has not given adequate reasons why we should accept this normative claim. If it is meant to be descriptive, it is clearly false: Scientists do not all function in this restrictive fashion.
[1] This brings up an interesting question: If our sensory apparatus must be unaided simpliciter for the information we gain from them to count as evidence, then we should not just deny the use of microscopes to enhance our vision, but eyeglasses as well. By van Fraassen's principles, nothing that I see do I actually observe, as either I am wearing glasses (thus my perception is no longer unaided) or my vision is so physically impaired that I cannot actually see anything. In fact, with my unaided visual perceptions, I have very good evidence that the world contains no definite boundaries between objects, that books do not actually have words written on them until the pages are less than three inches from my nose, that the world is tinted either blue or red depending on which eye I have open and which closed. Why would I believe this? Why, I've got good visual evidence for it! I certainly can't trust the things that I can observe with my glasses on. At most, I should be quite skeptical about their existence. Furthermore, our best scientific theories allow us to perform surgeries such that my eyes could be corrected to see better without my glasses than I currently do with my glasses. What is to stop us from thinking that it may be possible in the future for us to have the technology and skill to be able to surgically graft on eyes that function like electron microscopes? This question will be relevant further on in the paper, in discussion of a counterfactual proposal of Churchland's.
One response to the first objection would be this: It is not any particular observer that defines what is observable for the human category, it is rather an ideal observer. This is not very satisfactory because it is not clear what an ideal observer would be. It is some human with 20/20 vision? Someone with 20/10, or better? Or is it perhaps someone who has had the electron microscopes grafted on in place of an eye, wired such that it connects directly to the brain? This is a problematic question, and I cannot answer it here.
[2] I'm indebted to Zach Weber here for pointing out this objection.
[3] This epistemological point may or may not be an objection.
[4] What that structure is precisely is in dispute, but there are a number of options. See Sidelle (1998).
[5] See earlier quote from van Fraassen about the need to differentiate between the different types of seeing.
[6] Not many, but there are some single-celled amoebas that are visible without microscopes. It will be sufficient to concentrate just on these for the purposes of illustrating my point here.
[7] Observability is not time-indexed. The moons of Jupiter, on van Fraassen's view, were always observable, even before Galileo constructed his telescope.
© 2002 Sara L Friedemann