Scientific explanation lost and found
Creationists want science to reintroduce divine entities and powers into its theories and theorizing. The Creationist's supernatural "alternative explanation" should be given equal time to the natural theory of evolution. Literalist Creationists hold that Biblical revelation should be admitted to explain and justify the supernatural origin of the world, of animal species, of human beings and more. Johnson and ID Creationists want science to incorporate the reality of God and to cite His preternatural divine intelligence as the best explanation of biological complexity. But are these sorts of appeals to supernatural explanation reasonable? In particular, are they reasonable in science?
Many scientists would immediately answer in the negative on the grounds that explanation, supernatural or otherwise, should play no role at all in science. Science, they would claim, does not explain the world, but merely describes it and leaves the explanations to philosophers and theologians. Indeed, Arno Penzius made just such a statement in an interview with Creationist Fred Heeren about his work with Robert Wilson that led to the observation of cosmic background radiation and evidence for the Big Bang. Asked why Wilson had been disposed to accept the steady state theory before those 1965 observations, Penzius offered: "[Wilson], like most physicists, would rather attempt to describe the universe in ways which require no explanation; there's the economy of physics. And since science can't explain anything--it can only describe things--that's perfectly sensible."
Although this is a rather common view among contemporary scientists, it is quite mistaken, and it is important that we see why it is mistaken. First, however, we should note that as a response to the Creationist this would be a weak argument. It provides no good reason to rule out supernaturalism. The supernaturalist could easily agree not to call Creationism an "explanation," but just a straight-forward hypothesis about the world, what things exist in it and what relations hold among them. They are simply offering an alternative description of the world. This gives us a clue about the nature of the scientist's mistake.
The mistake arises in part from ambiguity in the notion of explanation. To see this let us take a moment to examine the concept. At the generic level an explanation of X is something that "makes X plain." That is, explanation brings understanding where before there was confusion or obscurity. But there are many species of explanations that may be distinguished by the nature of the phenomenon to be explained as well as the question that is being asked about it. For example, one may ask someone to explain what is the temperature and light output of the sun. Here we see without difficulty that the explanation one would offer would be just a description--the surface temperature of the sun averages X degrees centigrade and its light output is such and such. The scientist would probably quickly accept this sort of case and then claim that the problematic case is when we ask the person to explain why the sun so shines. But here is the source of the ambiguity that leads to the mistake. There are two different questions that we might have in mind when we ask "why X?"
The first is a question that inquires after the intended purpose, or the ultimate end, of sunshine. We sometimes express this more explicitly asking it in the form of an explain what for question. That is, when we ask "Why does the sun so shine?" we might mean "What is sunshine for?" This is the teleological sense of why-explanation. The term "teleology" comes from the Greek term telos, which means "goal" or "end." In his classification of explanatory types, Aristotle called these "final" explanations. If one considers how to explain, for instance, a sculpture in this sense it would not be appropriate to simply start listing the sculpture's many physical effects (it tips the scales at 2000 kg., it casts an irregular shadow, it makes many observers shrug and say that their 6 year old child could have done better, and so on). Rather the relevant teleological explanation is, say, that the sculpture is meant to express the artist's alienation from contemporary material culture and feelings of irony in having accepted the commission for the sculpture from a major Wall Street trading firm. Teleological explanations are "final" because they refer to ultimate intended goals. It is to this notion of explanation that the positivistic scientist is probably objecting. When science investigates the sun it can tell you that sunshine warms the earth and helps make plants grow, but it cannot say that doing these things is what sunshine is for. If that is what you mean by asking for an explanation of sunshine then let the romantic nature poet give that sort of answer. Science could also discover that sunshine can burn one's retina or cause skin cancer, but it could never discover and offer the explanation that one or another of these effects was sunshine's ultimate purpose.
However, there is a second notion that someone may have in mind in asking why the sun shines in the manner it does--the interrogator may be inquiring about the physical processes that produce the observed light and temperature. This is the genetic sense of why-explanation, and there is no good reason for the scientist to object to this sort of explanation. Indeed, giving accounts of processes that give rise to phenomena is the main thing that scientists do. Furthermore, an explanatory account of this sort is just a special sort of description, so we see again that the distinction between explanation and description that the Positivistic scientist appealed to was mistaken.
There is a simple reason that many contemporary scientists make this mistake; probably without realizing it they are following a philosophical position that was the received view during the first half of the 20th century. The position was advocated under a variety of different names and its specific tenets evolved over time, but we may speak of it under the general name of "Positivism." Positivism's influence continues to be felt in science even though philosophers of science themselves have long since abandoned many of its tenets after continued argument and analysis revealed their conceptual flaws. Positivists held that science should not go beyond what is physically observable, and they explicitly rejected explanation in science because they thought that it was necessarily metaphysical. Their maxim, heard echoed above, was that science describes but does not explain. But Positivists readmitted explanation into science after Carl Hempel, in a series of important articles beginning in 1948, showed how it could be explicated in a way that was not dangerously metaphysical. The contemporary "positivistic" scientist probably absorbed the anti-explanatory view that dominated until the time of Hempel's work and is simply not aware of the more recent developments in philosophy of science.
Hempel developed several explanatory forms to deal with different sorts of scientific cases, but all fall under what was called the "Covering Law Conception" of scientific explanation. The idea is that we may explain X--the explanandum--by showing that it follows from the empirical law (or laws) governing that sort of phenomena together with background information such as the initial conditions of the variables in the law. Let us take his central Deductive-Nomological (D-N) Model to give an example. Suppose one were to ask for an explanation of why a cannonball takes a given number of seconds (say six and a half) to hit the ground after being dropped from a tower? Here the explanandum, E, is the specific duration of the cannonball's to fall. This is explained by giving a logical deduction from the gravitation law (nomological has the Greek stem nomos, which means "law") and the values of its variables for the case at hand. That is, one may derive E from Galileo's gravitation law that governs bodies falling near the earth (or some more general gravitational law) and plug into its equation the figure for the height of the tower from which the cannonball was dropped (the initial condition). We explain the duration of the fall by showing that it follows in this way from the law of gravity. The abstract form of the D-N model looks like this, where the line indicates that what is below is a logical derivation from the premises above.
L1, L2...Lk Laws
C1, C2 ...Ck Initial conditions
----------------- ---------------
E Explanandum (Fact, or feature of an event)
Thus, on Hempel's conception a scientific explanation is a special sort of deductively valid argument, namely one that contains at least one general law in the premises from which one derives the explanandum. Furthermore, Hempel specified that the laws must have empirical content, by which he meant that they had to be testable by observational data. This condition prevented explanation from falling back into metaphysics. Science could indeed explain empirical phenomena by reference to covering laws so long as it was careful to stay within the bounds of empirical testability. Explanation was now acceptable to the Positivists and the anti-explanation tenet was dropped. Actually, Positivism was itself abandoned as a unified philosophical view shortly thereafter when sufficiently many of its other central tenets were also rejected for other reasons. Today almost no philosophers of science still consider themselves Positivists, though there are still scientists who are vaguely "positivistic" in the old, outmoded sense without realizing it.
So, science may indeed offer explanations. However, it is unfortunate that we cannot rest with Hempel's precise D-N model of explanation and proceed immediately with our assessment of supernatural explanation by comparing it to the detailed logical structure of his model. Hempel's work was successful in reintroducing explanation to science, and many of his broad conclusions remain in force, but extensive discussion of the technical details of his particular logical models revealed weaknesses that he was unable to overcome. Other philosophers of science took up the task and have made significant conceptual progress since then.
It turns out that in some ways Hempel's conditions were too weak and in other ways they were too strong. For example, the specific logical form of the D-N model was too lenient and thereby allowed in cases where were not truly explanatory. A variety of proposals have been offered for how to strengthen the requirements. One important version, developed by Philip Kitcher, emphasizes the idea that explanatory understanding may be achieved by unifying our knowledge and thereby reducing the number of "brute facts" we must accept. On this view for a derivation to count as an explanation it must belong to a restricted set of derivations that optimizes unification by minimizing the number of explanatory patterns needed while maximizing the number of conclusions that may be generated. (KEUC431) On the other hand, Hempel's requirement that an explanation be an argument that cites a law may be too strict. Many philosophers now argue that an explanation need not take the form of an argument at all, and that a description may be sufficient. It also may not be necessary that the description include a statement of a law. On the influential account developed by Wesley Salmon, to explain X it may be sufficient to describe the causal process that produced X. Salmon has a detailed theory of causal processes and their interactions that forms the framework for this sort of explanation, but we need not get into its details here. Salmon's main point is that at the most basic level the explanation of something in the world involves something else in the world--the causal processes that led to it. An explanatory account need not include an explicit statement of the causal law, though of course it is understood that the cited causal processes are lawful. Finally, philosophers now agree that Hempel's hope for a theory of explanation that made use of only syntactic and semantic constraints was not possible, and that pragmatic considerations must also enter the picture. Bas van Fraassen has developed this point, showing how explanations of X are fixed in relation to a contrast class--some alternative Y that depends upon the question in which we are interested. Thus, explanation-seeking why-questions are too vague if that take the form of "Why X?" and need to be further specified by contrast, such as by asking "Why X, rather than Y?" Philosophers of science have developed other elements of Hempel's view as well but the three we have mentioned give us enough to proceed with our discussion of supernatural explanations.
A Constructive Theory of Explanation.
Given the above brief history, let me now sketch a simple theory of genetic explanations that we may use that combines the elements we have discussed.
(I) X is explained by what makes it so (in situ).
(II) A person explains X by showing what might make it so (in situ).
The idea of what makes X so is the core of our concept of explaining why X. (Pennock 1995, p. 42) I will call this a "Constructive" theory of explanation, since it holds that why-explanations necessarily involve the processes that make or structure that which is to be explained. Though the locution is not always felicitous, we may say that one explains something by showing what "constructs" it, and sometimes how these processes construct it. I also choose this term as a friendly gesture towards Constructionists, especially moderate ones like the later Latour, but to still maintain a constructive distinction with those forms that endorse extreme social relativism.2 It is an ontological realist view of explanation that recognizes that explanatory accounts have important pragmatic features.
The Constructive theory of explanation is a generic analysis that aims to capture the ways explanations may legitimately vary depending upon the type of explanatory relata--what it means to "make X so" depends upon the type of thing to be explained--and upon the specifics of the why-question we are asking. Given that scientific explanation is a special case of why-explanations, we should expect that its analysis will fall under a generic conception. Scientific explanations are more precise than everyday ones, but they are not radically different in kind, so it will thus be useful to review several mundane examples to illustrate the generic pattern and then see how it applies to the scientific case. Some of the examples that follow could be developed into scientific explanations, but here we just note them as ordinary explanations.
- Why is so-and-so President of the U.S.? Because he was made so by election. (Political practice)
- Why is Johnny angry? Because Billy made him so by calling him a dork. (Psychological explanation)
- Why is the U.S.S. Enterprise sailing into the neutral zone? Because Captain Picard gave the order, and he has the power of command to make it so. (Authoritarian explanations)
- Why are all bachelors unmarried adult males? Because we make this true by the way we use our language. (Linguistic convention)
- Why are triangles three-sided? Because they are made so by the definition of triangle and by the rules of Euclidean geometry. (Formal definitions and relations)
Again we see that although these are all why-explanations, only some of them make even an indirect reference to unification. What is common to them all is that the explanatory accounts cite something that makes, or purportedly makes, the explanandum so.
In science there are several varieties of explanation, and all fit this pattern, and in specific domains the somewhat vague notion of "making it so" may be spelled out in a much more precise manner. The most common variety of Constructive explanation in science is that in which one explains by citing the causal processes that produced the fact. The causal view has its roots in Aristotle's theory of explanation in terms of the "four causes" and has since been endorsed in various forms. Important recent advocates include Michael Scriven (1958) and others, but the view has been developed most thoroughly by Wesley Salmon (1984; 1994). According to his Causal-Mechanical (C-M) model: "To provide an explanation of a particular event is to identify the cause and, in many cases at least, to exhibit the causal relation between this cause and the event-to-be-explained." (Salmon 1984) We need not endorse the full details of these views to recognize the importance of causal explanations in science and the straight-forward sense in which the cause of X is something that constructs or makes X so. (I will take causal explanations as my main example to show how the Constructive account becomes more precise in specific contexts.)
A second sort of scientific explanation is that in which one may explain a fact by reference to its classification, say, explaining a given property of a species by reference to its genus (or other higher order taxon). This fits the Constructive pattern in that we explain why a thing has the characteristics it has by virtue of the kind of thing that it is. For instance, one might explain why whales are viviparous by noting the fact that they are mammals. I know of no detailed treatment of this sort of case that is as well-developed as theories of causal explanation, but aspects of it have been discussed under various other headings. Dray presented the idea of "explanation-by-concept" in historical contexts (Dray 1959, p. 403), and his notion was discussed by Hempel (1965, p. 453-57), who tried to reduce it to his nomic view. More recently Ruben (1990, p. 218-22) endorsed the possibility of "identity explanations" whereby one may explain a fact by citing "another" fact that is identical to the first, but at a different level of description, such as explaining a change in the temperature of a gas in terms of a change of its mean kinetic energy.
A third sort of explanation involves mathematical relations. Being able to account for mathematical explanations had been one of the notable advantages of the Unification over the Causal theory of explanation. These fit naturally within the Constructive notion of making something so by virtue of formal relations. We find this idea entering at the most basic level of mathematical understanding as when we are taught mathematics by first learning that one and one "makes" two. In science we find mathematical explanations appearing in both theoretical and applied contexts. Here I mention a single example. In genetics one explains a one-to-one ratio of males to females in a population in terms of Fisher's sex ratio argument that shows it arises as a mathematical consequence under assumptions of random mating, differential ratios of sons and daughters produced by different parents, and heritability of these offspring production ratios. (Sober 1993, p. 15-7)
I would argue that other forms of why-explanation in science may be couched in one or another of these or in one of the other varieties previously mentioned.
Returning now to the generic Constructive notion, the qualification that an explanation be "in situ" limits the requirements for an explanation. To explain a given item one need not show what would make it so in all situations, but just what made it so in the situation(s) under consideration. This is a pragmatic constraint. An ideal complete explanation of X would include everything that made X so, but this is seldom, if ever, required. We typically take certain background facts for granted in a given context, and need only cite what made the difference, say, in getting X rather than Z, in that situation. To make this point clearer, let me how to think of causal explanation from a Constructive point of view.
As we noted, according to Salmon's C-M model one explains an effect by identifying "the cause." However, the world is a complex causal network and for any given effect there are multiple causal factors that were involved in its production. The Constructive view agrees that causes count as explanations because they "make things so," but does not hold that the ontic relations alone fix "the cause" unless by that we mean the ideal complete explanation, an account of which would include all the factors, such as might be written out in an Ideal Explanatory Text (Railton 1981). Instead, it holds that explanations may be and for the most part are relativized to a situational context and suggests that this is picked out on a pragmatic basis. To be precise about the elements in the basic sort of case I propose that we not think of the causal relation in the standard two-place (C causes E) fashion but rather as a four-place relation that I call the "CaSE" model. This is simply a more fine-grained way of parsing causal relations in a way that recognizes both the ontic elements and the pragmatic ones and thereby licenses certain inferences. I cannot here defend the details of the model or its logic, but the idea is that one factor C in the network is separated out from the others, in the situation ('S'), for pragmatic reasons. Typically this is done by means of contrasting alternatives ('a'), such as noting by the emphasis in the question that one is interested in one aspect of the event rather than some contrasting other (which may be its negation or some specified alternative), or that one is interested in C rather than some possible C' that did not occur (again, its negation or specified alternative), but may be done in other ways.3
So, in a causal explanation of some effect E, we cite its cause C, relative to a "background" situation S, which is fixed (explicitly or tacitly) by our interests. In a loose way we may often think of S as referring to assumed "standard conditions" or perhaps to some discipline-determined background domain. In the most precise manner we may think of it as a ceteris paribus clause. The CaSE model is best exemplified in practice by controlled experimental testing, which I take to be the gold standard of good scientific method; the parameters (S) are fixed in both the experimental and control groups and only the independent variable is allowed to change (C relative to some interesting alternative, a, which is either not-C or some set base line), and the effect upon the dependent variable (E) is observed.4 Because on the Constructive view there is no single ontically privileged way to make the separation, the "same thing" may have different explanations depending upon the way a question is posed, allowing for legitimate explanatory pluralism without falling into unconstrained relativism.5 The story would have to be told differently for non-causal explanations, but similar considerations are involved.
Continuing now to spell out the Constructive view, its notion of being "made so" may involve but does not require necessity. We see this most clearly in explanations of human behavior. Although Johnny was angered by Billy's remark, we don't think that his response had to have been necessitated by the events to be explained by them. Perhaps in an ideal psychological theory scientists would get necessary mental connections, but we may still have good explanations without them. We also see this in statistical explanations. (Salmon takes it as a strength of his view that it can incorporate statistical explanation, recognizing indeterministic causal processes as possible explanations. Ruben appears to want to stick with necessary--"determinative"--relations, but in the end he opens the door a crack to allow the possibility of stochastic explanations, not for a positive reason, but so as not to beg the question of indeterministic causation.) Of course, we sometimes do desire more of an explanation, especially in science. If so, then we may explicitly indicate the higher standard by asking not why X was (or is) so, but why X had to be (or must be) so. In such cases we would expect an adequate explanation to involve necessity, by indicating processes that necessarily made X so.
The Explanatory Virtues
Given the Constructivist analysis of explanation, we may now turn to an exploration of some of the explanatory virtues. According to the Inference to the Best Explanation (IBE) confirmation theorist, we appeal to explanatory virtues to sort out the better or best of competing potential explanations. Here I will discuss just three explanatory virtues and show how Creationism fails in comparison to evolutionary theory for each of them.
(a) Explanatory scope
We begin with one of the most commonly cited explanatory virtue, that related to the scope of the explanation. By this is meant the variety of phenomena explained by the explanatory hypothesis. Whewell discussed this virtue in terms of conscilience--if two or more different sorts of evidence point in the same direction this is a sign that our investigation is on the right track and that we have a good hypothesized explanation. Typically we emphasize the importance of greater variety and express notion this by saying that a given explanation has breadth or wide applicability. For example, the Darwinian evolutionary mechanism has broad scope in that it can explain adaptation of organisms to their environment, the origin of species and features of their bio-geographical distribution, and the tree-structure of biological taxonomy, among other general phenomena. That is, these various patterns of facts are made so by the causal processes of variation, inheritance and natural selection. This is strong evidential support for evolutionary theory; no alternative can account for as wide a variety of phenomena. Other things being equal, we tend to hold that explanations with broad scope are better than those with narrow scope, though there are caveats to this generalization.
Some methodologists use the term "explanatory power" as a synonym for explanatory scope--when they say that something is a powerful explanation they mean that it has broad scope, or broader scope than some alternative, weaker explanation--but others use the term in a more general sense to mean that a given explanation has broad scope and other virtues as well, such as being especially precise, or deep, or simple. This terminological ambiguity may be part of the reason that unification has been confused with other explanatory virtues. I'll use "explanatory power" in the more inclusive sense and will try to sort out unification from scope and other specific virtues.
Explanatory scope is sometimes identified with unification, but these are not the same. Even if we thought that unification is the correct analysis of what it means to be an explanation, a given explanatory hypothesis could explain more or fewer phenomena than some other hypothesis. So, explanatory scope would be a measure of a certain degree of unification. Scope refers to how much gets explained, however the notion of explanation is explicated and it is useful to keep this as a separate virtue even when we reject unification as the correct explication of explanation.
Often we assume that the greater the scope the better the explanation, but it is important to recognize that a particular explanation may have a narrow scope and still be good. For instance, a three hour timer delay for my dishwasher set at noon explains why the machine activates at three o'clock. In general that setting of the timer mechanism explains no (or relatively few) other sorts of events, but in the context it is a perfectly good explanation of the event. Explanations will have broader or narrower scope depending upon the sort of phenomena being considered, so an absolute measure of scope is not by itself an indication of a good explanation. Rather I suggest that scope comes into play for IBE confirmation when we are comparing alternative hypothesized explanations of the same data set; other things being equal we should prefer H1 over H2 if H1 explains more of the data set than H2 does. For example, a chemist may get a series of measurements from a laser absorption analysis of an unknown compound. The hypothesis that it has structure H2 may account for just certain features of the curve, but this should be rejected in favor of structure H1 if the latter accounts for more of the data curve. Creationists claim that their hypothesis of a benevolent deity who specially designed and created organisms explains the organisms' adaptations to their environments, but this hypothesis cannot explain the concurrent existence of maladaptations. It is thus not as good an explanation as Darwin's evolutionary hypothesis which has the resources to explain both.
Finally, we should note that even if H1 has broader scope than H2, the latter may be better supported if their other explanatory virtues are not equal. A possible explanation of the same data set may still be rejected in favor of another with narrower scope if, for example, the way the mechanisms of the latter produce the observed phenomenon can be shown in greater detail.
(b) Explanatory resolution
We tacitly accept that the best explanations are those that account for the phenomena in a manner that is detailed and precise. For example, it is sometimes sufficient to say that the sailor is sick because of a "vitamin deficiency," but we find the explanation better (and thus better confirmed) if we can delineate the biochemical pathways that lead to the observed symptoms, say, from absence of vitamin C to the signs of scurvy. This is so not only in explanations of specific cases but also in general explanations--the current medical explanation of A.I.D.S. is the Human Immunodeficiency Virus, but researchers are not fully satisfied with the explanation since they can so far say only a little about how the virus works.
Surprisingly, methodologists have rarely discussed this property of explanations explicitly and it is not included in standard lists of explanatory virtues. Following the optical instrument metaphor, I will call this the virtue of explanatory resolution. The resolution of a telescope or microscope refers to its ability to resolve or distinguish points nearby one another. The clearness of the image depends upon this property of the instrument, and clarity is also a basic feature of good explanations. We disparage explanations that do not have this property by calling them fuzzy or vague. Let us say that an explanation that provides a great level of detail about the productive processes is one with high resolution, and one with minimal detail has low resolution. I suggest that a hypothesis that accounts for the phenomena with higher resolution is better than one with lower resolution. Indeed, I propose that explanatory resolution is the most basic and the most important of the explanatory virtues, and in general will Trump other explanatory virtues. We find that this is typically so in the judgments of scientists on specific issues.
The centrality of explanatory resolution makes sense given that, as we saw, explaining why X involves showing how X became as it is; that is, showing what made it so. In giving a causal explanation, for example, we expect to see the details of how the causal mechanism can produce the observed pattern of data as precisely as possible. This is more important than having an explanation that unifies but in an imprecise way.
For instance, Lamarckism could be said to unify in a vague manner about the same evolutionary phenomena as Darwinism. It also has only somewhat less explanatory scope. Nevertheless it loses to the latter because it provides no clear account of how the change works. Indeed, even before Darwin's alternative was in place Lamarckism was judged harshly because it could not show in detail a plausible mechanism by which use led to heritable characteristics. On the other hand, Darwin provided a clear process that could produce transmutation through natural selection of heritable variations. Moreover, one of the most persuasive new sorts of evidence in favor of Darwinism was the discovery of DNA, RNA and the molecular mechanisms that underpinned evolutionary theory. Now we have the details of how genetic information is stored, replicated and transmitted. We can show with remarkable precision how the genetic mechanisms produce the heritable properties of an organism, and how variation arises that can then be selected for in an environment, producing organisms that are adapted to their conditions. The primacy of explanatory resolution shows why biologists judged the Darwinian explanation to be far superior to the Lamarckian even though the two are approximately comparable regarding other virtues. Creationism does not even get out of the gate on this explanatory track for their "theory" says and can say nothing at all about how the process of divine special creation is accomplished.
(c) Explanatory focus
Closely related to explanatory resolution is a virtue we may call explanatory focus. We expect that better explanations will be able to account for the distinctive features of the explanandum, that is, why it has this particular feature rather than some other one. As in the virtue of resolution, we are interested in seeing detail, but here the emphasis is on a specific feature in contrast to some specific alternative. We spoke of this in passing earlier in the discussion of the limitation that explanations be in situ, noting that, in situation S, we want to be able to explain why X rather than Z. Lack of this virtue makes for a different sort of vague explanation, one that cannot account for the what is distinctive about what is to be explained. Many explanations in folk psychology lack focus in this way. For example, "stress" has been cited as the explanation of all manner of personal malaise from angry outbursts and nervous breakdowns to chronic fatigue and hives. It may be true that stress is involved in the explanation of these maladies and many others, but medical physiologists still judge it to be a poor explanation because nothing in the "stress hypothesis" can distinguish why the sufferer got one rather than another. Such explanations quickly give way when scientists discover more focused alternatives. Stomach ulcers were commonly explained in this unfocused manner and sufferers were advised, in a similarly unfocused manner, to "reduce stress," but this explanation and therapy are quickly being abandoned now that a bacterium has been discovered that appears to be the specific cause of ulcers. (Tompkins and Falkow 1995) The "unifying" appeal of the stress theory does not carry much relative weight.
This is really what Popper must have had in mind in developing his falsification criterion of demarcation between science and pseudo-science. He explained that his original motivation was to understand why Marxist history and Freudian and Alderian psychology were not scientific theories in the same way that Einstein's theory was. He claimed that it was not "explanatory power" that differentiated them, because these theories explained "practically everything that happened within the fields to which they referred." (Popper 1962, p. 35). What was needed to make something a good scientific theory, Popper thought, was that it could make risky predictions that could potentially falsify the theory. If the foregoing analysis of explanation is correct, then what Popper was really after was not so much a certain type of prediction, but rather the virtue of explanatory focus so that the theory could clearly give an account of why X rather than Z, and so could be rejected if observation proved otherwise. This misunderstanding may account for why Popper originally dismissed evolutionary theory as not being scientific; evolutionary theory is notoriously poor at making predictions because of the complexity and historical contingency of the causal factors in a changing environment. On the other hand the sorts of causal processes it appeals to are certainly able to provide explanatory focus once the selective forces in a given environment are understood. For example, Peter and Rosemary Grant's already classic research on the evolution of Darwin's Finches on Daphne Major documented how changing weather conditions leading to changes in available food sources caused the population to evolve by selectively favored finches with larger beaks were over those with smaller ones. Again, the Creation Hypothesis has no resources to explain why one trait occurs rather than some other except by a generic appeal to God's will, and this equally "explains" both X and Z.
We could continue in this manner down the lists of explanatory virtues with similar results, but rather than multiply examples let us step back and see if we can find the reason for this pattern.
Supernatural explanations
The previous sections have outlined the basics of the theory of scientific explanation and a little of how assessment of explanatory virtues allows us to evaluate the worth of competing explanatory accounts. There are actually a range of other explanatory virtues that have been discussed in the literature, but, like scope, they are subsidiary to resolution and focus. As we have seen, the supernatural Creationist theory seems to fare poorly on these points. But this is not because of any special bias against Creationism, for the conclusion holds for any supernatural theory, simply because of the characteristics of the notion of the supernatural. Let us review three of the main characteristics of the supernatural to see why this is so.
The first and most basic characteristic of supernatural agents and powers, of course, is that they are above and beyond the natural world and its agents and powers. Indeed, this is the very definition of the term. They are not constrained by natural laws. Indeed, on some views it is a supernatural creator that makes the laws in the first place, and those that make the laws have the power to break them. Of course, this is why humans hanker after access to occult powers, since they would supposedly free us from the laws that bind us.
If supernatural agents are constrained at all it may only be by logic. Even God cannot make it so that something is at once P and not-P. When story-tellers need a way to save their protagonist from a misguided pact with the devil, they invariably do it by placing a bet that the devil is unable to win by virtue of the rules of logic. But other than logical impossibilities, there is nothing that we can know that a supernatural agent could not do.
The second characteristic of the supernatural, that we have mentioned before and that follows rather directly from the first, is that it is inherently mysterious to us. As natural beings our knowledge all comes via natural laws and processes. If we could apply natural knowledge to understand supernatural powers, then, by definition, they would not be supernatural. The lawful regularities of our experience do not apply to the supernatural world. If there are other sorts of "laws" that govern that world, then they can be nothing like those that we understand. Occult entities and powers are profoundly mysterious to us.
The same point holds about divine beings--we cannot know what it is that they would or would not do in any given case. God works, they say, in mysterious ways. We cannot have any privy knowledge of God's will and those who have tried to claim it are quickly brought back to earth. When the complex Ptolemaic epicycle theory of the planetary system was explained to Alphonso X, king of Castile, with its equant points, eccentrics, deferents and epicycles--wheels upon wheels within offset wheels--he commented that "if God had consulted him at the creation, the universe should have been on a better and simpler plan." (WHI151) Defending the complexity of his theoretical models, Ptolemy is reported to have said, "You may complain that these models are not simple, but from the point of view of God, who knows what is simple." And of course Ptolemy was right; we cannot say that our notion of simplicity is at all relevant to what God's might be, or if it is even an important property for Him. Scientific models must be judged on natural grounds of evidence, for we have no supernatural ground upon which we can stand for it is necessarily a mystery to us.
A final relevant element of the notion of the supernatural, closely linked to the previous ideas, is that supernatural beings and powers are not controllable by humans. Though our secret desire may be to gain esoteric power through contact with the supernatural, we seem to understand at a deep level that such control would be impossible.
Folk tales and literature consistently tell us that those who would steal the fire of the gods are bound to be burned. Information about the future acquired by supernatural foresight did not allow Oedipus to escape his fate, and only made the inevitable outcome all the more pitiful because his parents' foolhardy attempts to use that information to outwit fate became the very means that sealed it. The power of wishes, however it may be granted and however carefully used, turns upon the wishers who soon learn that they cannot predict or control the negative effects that follow. The sorcerer's apprentice appears to touch supernatural power for one giddy moment before losing control, with usually dire consequences. The very notion of the "Faustian bargain" carries this warning against the temptation of thinking one can control supernatural powers for one's own benefit; pacts with the devil inevitably turn out for the worse, for Satan never does have one's interests at heart. The best protagonists can do in such tales is to return in the end to their prior state, having learned, one hopes, to be content with their natural estate and powers.
Perhaps one may find a few isolated exceptions to this pattern in which humans are unable to control the supernatural, but the pattern certainly holds true of our relation to the divine Creator as Christian Creationists want us to believe in Him. God controls the world and, though we may control ourselves, we cannot control God. Indeed, part of what it means to accept Christ, on the Evangelical view, is to relinquish even the control we have of our selves and to turn our lives over to God's will. Despite this, however, there remains the same temptation to think that we may influence God's supernatural power for our benefit through our actions. What is the difference between the belief that wearing a religious medallion or making the sign of the cross will protect one from evil and the belief that a charm or a knock upon wood will ward off bad luck? Of course one important difference is that many people uphold the efficacy of the former because of their belief in the divinity of Jesus, while most dismiss the latter as silly superstition because they no longer take seriously the idea of the goddess Fortuna. Nevertheless, the two beliefs are the same in the sense that both seem to maintain the hopeful possibility of supernatural control. Does not the belief in petitionary prayer yield to the same subtle temptation? Prayers for healing or discernment, for career advancement, for success in war, or even for peace, that are made "in Christ's name" seem to contain the implicit idea that we may bring God to use His powers towards such ends by our sincere petitions. And do we not find this again in the most basic idea that we may gain forgiveness and eternal life by choosing to accept Jesus as our savior? Evangelicals emphasize that we need only ask for salvation to be given, saying that it is "up to you." All this appears to fit the original impulse that leads us to the supernatural with the hope of benefiting from its powers.
However, we need to recognize that this wishful belief in the possibility of human control of divine and occult powers actually contradicts the idea of the supernatural in a profound manner, for by definition the supernatural is beyond the reach of we mere creatures of the natural world. If the supernatural could be controlled by the natural then it would cease to be super. If we may control the natural world it is only because the world is governed by physical laws that must be "obeyed" even when we are pulling the strings. But in the very idea of the supernatural is the notion that it stands above natural laws and thus outside the possibility of our control. If God were really under our control in any sense then He could certainly not be said to be omnipotent, and probably would not be thought very godly. The whole point of the supernatural is that it may control nature but nature cannot control it; God commands and we must obey, not vice versa.
It seems that the cautionary tales we have noted tacitly recognize the impossibility of natural control of the supernatural and the hubris of believing otherwise. Indeed, even the religious examples of supposed supernatural control by symbol, prayer or action that we mentioned have another interpretation than the one we first considered that respects this view. According to this alternative interpretation our actions cannot control God in any way. Our prayers are not worthy of answer and no good deed or profession of faith can make us worthy of salvation. Redemption is not owed and it cannot be earned. If redemption comes, it comes as a gift. It is out of our control. Admittedly, this is a harsh doctrine and few Christian sects, except perhaps the Calvinists, have faced all its consequences squarely, but this does seem to be the only consistent view. Many try to have it both ways and intimate that, although God cannot be compelled by petitions, actions or even faith, somehow we can be confident that these will indeed make a difference, and that He cannot fail to heed the call of the believers. But even come judgment day, no matter what our actions it still remains God's free choice whether or not to heed our petitions. This certainly seem to be the view that would be held by the Evangelical Creationists. We cannot control God.
These characteristics of the supernatural show why supernatural explanations should never enter into scientific theorizing. Science operates by empirical principles of observational testing; hypotheses must be confirmed or disconfirmed by reference to inter-subjectively accessible empirical data. One supports a hypothesis by showing consequences obtain that would follow if what is hypothesized were to be so in fact. Darwin spent most of the Origin of Species applying this procedure, demonstrating how a wide variety of biological phenomena could have been produced by (and thus explained by) the simple causal processes of the theory. But, as we have seen, supernatural theories can give no guidance about what follows or does not follow from their supernatural components.
The appeal to supernatural forces, whether divine or occult, is always available because we can cite no necessary constraints upon the powers of supernatural agents. This is just the picture of God that Johnson presents. He says that God could create out of nothing or use evolution if He wanted (JDT p. 14, 113); God is "omnipotent" (JDT p. 113). He says God creates in the "furtherance of a purpose" (JDT p. 4), but that God's purposes are "inscrutable" (JDT p. 71) and "mysterious" (JDT p. 67). A god that is all-powerful and whose will is inscrutable may be called upon to explain any event in any situation, and this is one reason for the methodological prohibition against such appeals in science. Because of this feature, supernatural hypotheses remain immune from disconfirmation. Young Earth Creation-Science does include supernatural views at its core that are not testable and it was rightly dismissed as not being scientific because of these in the Arkansas court case, but it at least was candid about a few specific non-supernatural claims that are open to disconfirmation (and indeed that have been disconfirmed), such as that the earth is less than 10,000 years old and that many geological and paleontological features were caused by a universal flood (the Noachian Deluge). So far at least, Johnson has declined to offer any specific positive claims of this sort by which his notion of Creationism could be tested.
Experimentation requires observation and control of the variables. We confirm causal laws by performing controlled experiments in which the purported independent variable is made to vary while all other factors are held constant and we observe the effect on the dependent variable. But by definition we have no control over supernatural entities or forces.
Finally, if we were to allow science to appeal to supernatural powers even though they could not be tested, then the scientist's task would become just too easy. One would always be able to call upon the gods for quick theoretical assistance in any circumstance. Once such supernatural explanations are permitted they could be used in chemistry and physics as easily as Creationists have used them in biology and geology. Indeed, all empirical investigation could cease, for scientists would have a ready-made answer for everything. For example, consider Wayne Frair's alternative creationist explanation of the many general similarities among animals (such as common reactions of humans and rats and monkeys to drugs). These, he says, "can be explained as originating in basic design given by the Creator. Evolution is not needed to account for the similarities." (Frair and Davis 1983, p. 14) In short the "explanation" does not go beyond claiming that this pattern is so because God designed it so. Clearly science must reject this kind of one-size-fits-all explanation. By disqualifying such short-cuts, the Naturalist method also has the virtue of spurring deeper investigation. If one were to find some phenomenon that appeared inexplicable according to some current theory one might be tempted to attribute it to the direct intervention of God, but a methodological principle that rules out appeal to supernatural powers prods one to look further for a natural explanation. Clearly, it is not just because such persistence has proved successful in the past that science should want to encourage this attitude.
The scientists' appeal to supernatural agency in the face of a recalcitrant research problem would be as profoundly unsatisfying as the ancient Greek playwright's reliance upon the deus ex machina to extract his hero from a difficult predicament. Sydney Harris, the preeminent scientific humorist, cleverly made the point in a cartoon that appeared in American Scientist. He pictures two scientists standing before a chalk-board that the first had covered with an intricate series of symbols and equations, but with one gap in the sequence at which is noted "Then a miracle occurs." The second scientist, gesturing towards this notation, states his considered assessment: "I think you should be more explicit here in step two."
Without the binding assumption of uninterruptible natural law there would be absolute chaos in the scientific worldview. Supernatural explanations undermine the discipline that allows science to make progress. It is not that supernatural agents and powers could not explain in principle, it is rather that they can explain all too easily. As such we may think of them as the explanation of last resort, since, like the Greek god in the machine, they can always be hauled down to "save the day" if every other explanation fails. They are the poor person's explanations, that is, the explanations of the intellectually poverty-stricken, since they are available for free. Surely it is not in this sense that the poor in spirit are blessed with seeing God.
Pfft. Radical atheist.