Jack Maze pointed out on more than one occasion that biology, unlike classical, mechanistic physics and chemistry, is highly variable and characterized by the organization of individual life forms and groups of related individuals. Cy Finnegan added that biology is rich in data, but poor in theory. In his History and Philosophy of Biology course, Cy surveyed the ideas of quite a few philosophers and historians of science, but he focused with special intensity on the structure of causal scientific explanations. In order to understand the differences between theory and not-theory, especially in biology, the consideration of the structure of a valid explanation is important in keeping our thoughts clear.
Science bases explanations on things called "laws," or the regular (partly predictable) behaviors and interactions of the forces, energies and substances in the universe. Laws outline the constraints that shape natural phenomena and provide the drives that bring them into existence. These laws offer partial prediction in that they generate a pattern - a series of events that can be related to a single causal source, a law of nature, though the specifics of those events may not be fully knowable until they occur. Regularities of nature are called laws because they seem never to be violated, as far as we can tell, but our understanding of them has often been (and continues to be) crude, incoherent or incorrect. Scientists can't go beyond the laws of nature in explaining events because these form the "ultimate" sources of actions and behaviors in our causal explanations of objective reality. Religions, having no need of evidence, can "explain" whatever they like beyond the laws of nature, according to their own sets of rules. It is entirely possible that the laws of nature were created and set into motion by some sort of distant deity. This idea of deism seems to have grown most strongly from the Enlightenment, when some humans decided their intellect was sufficient to understand nature, though it may all have been kicked into motion by a distant and rather disinterested God. But it is not possible to scientifically test (or even approach) any ideas about god-like entities or their activities. There can be new understandings of the laws of nature and how they might work, but we can't go beyond them - there is a hard barrier here for science. Science and religion need not be in conflict, but the constraints of science, at least, keep them clearly separated.
The philosopher Carl Hempel (Philosophy of Natural Science, 1966, Prentice-Hall, Inc.) first gets the reader's attention by telling the story of Semmelweis and childbed fever. Semmelweis could clearly see that there was a strong correlation between the women in the First Maternity Division and an abnormally high incidence of fever, but he needed a cause, an explanation, information that would help improve the situation. After testing and rejecting a frustrating series of possibilities, his colleague dies from a disease with the same symptoms after being accidentally stabbed. He thinks it must be something about touching corpses, so he comes up with giving hand washing a try, and it works! Semmelweis still knew nothing about disease-causing bacteria, and too many colleagues didn't appreciate this insistence on hand-washing, but his discovery helped lay down a foundation that could be built upon by later microbiologists and epidemiologists - science as ever-growing sets of hypothetical knowledge developed from the interpretation of new information, or new ways of looking at existing information, providing the foundation for technological applications.
Hempel went on from his introductory example. His idea was to develop a logical framework that can be used to examine scientific theories and hypotheses by making explicit how events come to be caused, given specified initial conditions and a relevant natural law. This explanatory framework, called Hempel's "Covering Law Model," labels the initial conditions plus the appropriate law(s) as the explanans, or the things that, when brought together, cause the explanandum (the event) to occur. The links or bridges between the explanans themselves, and between the explanans and the explanandum form a logical, causal relationship. If I drop a rock from a short distance above the ground (the initial conditions), gravity (the law) will cause the rock to spontaneously move toward the Earth's surface (the event).
Semmelweis' Covering Law Model approach to scientific testing was quite general - if medical students wash their hands between dissecting a corpse and delivering a baby (conditions), the toxic substance on a corpse that causes childbed fever (law) will be removed (these are the explanans) and the death rate of women will go down (the event, the explanandum). My friend Ruth Deery suggested other examples: the egg shells of birds of prey become thin (the explanandum, or event) when these two parts of the explanans occur: DDT accumulates in the tissues of the birds (initial condition) and interferes with normal egg shell development (laws or regularities of chemistry and bird physiology). Similarly, as greenhouse gases increase in the atmosphere (initial condition), they trap heat escaping from the Earth (general laws of chemistry and physics), causing events such as melting ice caps, followed by a rise in sea levels and increased coastal flooding, not to mention droughts and bigger, nastier, more energetic storms. Other events may be caused that may not be specifically predictable, such as the exact timing, location and track of the next ferocious hurricane. Many events are difficult to fully predict by their intrinsic nature, and are more easily explained than fully predicted, but they can be anticipated in a general sense.
Hempel's Covering Law Model provides a structure for looking at hypotheses and theories with relative clarity, demanding explicit (or as explicit as possible) identification of the conditions, laws and event(s). By distinguishing between that which explains (the laws and conditions of the explanans) and the explanandum, or the event to be explained, clarity is greatly improved and circular arguments, or tautologies, can be avoided. The concept of adaptation is a classic example of something explaining itself. All organisms are adapted. Of course they are. Non-adapted organisms never existed because basic adaptation is one of the initial conditions life forms must possess in order to be alive. But adaptation is sometimes also cast in the role of the event, the explanandum, the outcome of evolution. Adapted life forms experience random mutations, becoming more highly adapted after the "law" of natural selection kills off the less adapted, or something like that. As the developmental biologist Conrad Waddington pointed out some decades ago (add ref.), the best-adapted organisms are bacteria, so there must be something else going on here for there to be more complex, multicellular forms. Adaptation can only be invoked once, either in the explanans or the explanandum. I think it is properly part of the initial conditions of life and belongs with the explanans, when evolution is the thing to be explained as the explanandum. Adaptation sometimes becomes a circular argument, which is by definition "true," though not very helpful. If adaptation is the event, the explanandum, then it collapses back into an explanan - adaptation causes/explains adaptation. If adaptation is the law of nature that causes evolution, then Waddington was right - there should be no such things as elephants. Maybe the answer to why a plant takes a particular form or has a certain characteristic isn't: "Because it's adaptive." Maybe the answer is: "Why not?"
Cy Finnegan spent some time describing another important component of the Covering Law Model, the "bridge principles." These are the logical connections between the conditions and laws of the explanans, and also the links between the explanans and explanandum. These bridges are necessary parts of the structure of theories because they insure that the causal explanation is actually relevant to the event being explained. Is adaptation an initial condition? Is it the event to be explained? Is it both? That won't work. Darwinian (invoking natural selection, a variation-decreasing force) and neo-Darwinian (incorporating genetics) theory seems to lack coherence when I try to fit the pieces into a Covering Law structure. How can a variation-decreasing force be the same force that spurs variation, not to mention the self-organization and increasing complexity through time (including the move from the unicellular to the multicellular) that are characteristic of life and its evolution? Jack Maze noted, "Interestingly enough, it is the lack of explanatory cohesion in Darwinian and neo-Darwinian theory that some creationists exploit in their attacks on evolution." I can't argue with that.
Theories and the events they explain must form a partly coherent whole. Even if there are vague or incomplete areas, the connecting bridges must have some clarity, be logically possible and somehow empirically testable. Though testing is not always direct, it must still be linked to the question, to the explanation offered through the theory. The dispute isn't between neo-Darwinian theory and "creationism" or "intelligent design." Ignoring creationism, there is more than one reasonably scientific theory or causal explanation for evolution, and some make far better candidates than others in terms of accurate descriptions, logical connections and relevant explanatory power. Scientific explanations should be more than just assertions. The constraints of the Covering Law Model are demanding and even rather rigid, but there is room for an explanation to change and be re-formed within its constraints. "Creationism" or "intelligent design" doesn't play by the rules of science, so it gets shut out of the ballpark. Faith by its nature is never inaccurate or wrong, and it can't be interchanged with skeptical, self-critical rationality. Bacterial flagella, though they may look like something developed in a modern robotics lab, do not provide any actual evidence of a tinkering deity. Deities don't lend themselves to evidence or empiricism, positive or negative. They inhabit a legitimate metaphysical realm, but it is different from the place inhabited by science; there is not much overlap.
Evolution is not a fact; it is a theory. Fossils are facts. Living organisms are facts. The vast body of evidence for relationships among existing organisms, and among organisms past and present, based on the characteristics they possess and share, is factual. Evolutionary theories offer explanations for the assumptions of relatedness that are based on evidence from organisms, both living and extinct. But theories are not opinions, though opinions may play a role. Theories can be dissected and presented in the form of the explanans and explanandum of the Covering Law Model, but opinions, all by themselves, cannot. One can't substitute divine intervention for the laws in the "Covering Law Model." There may be some deity underlying the laws of nature, but science can't go there; it's cheating. Jack and Cy further noted, "Religion masquerading as science, ironically, has no lasting effect on science, but denigrates religion by replacing the divine with the mundane..." I don't much worry about the second part of their statement, about the fate of divinity, but I sure hope they're right about the "no lasting effect" on science part.
It's interesting to think about how our understanding of the laws of nature keeps changing. As Jack might say, we try to touch the truth, but then it moves. The example of the dropped rock works for just about any version of the law of gravity that explains why the rock falls - Newton's, Einstein's or perhaps a more recent version. We could try to think of experiments to distinguish between the different descriptions of gravity - as an attractive force between two masses or as masses distorting the fabric of space, or other possibilities. The rock still falls to the ground, but the understanding of why this event would occur has changed more than once and will probably change again. There is a certain level of prediction here: given the initial conditions, the event of the falling rock can be predicted. In many cases, though, the full details of the event may not be predictable. This unpredictability may be a property of nature, a property tied to the ideas of entropy and emergence. Ruth Deery said, "The example that sticks in my head is that there is no way the taste of salt could be predicted before sodium and chlorine atoms are joined into a sodium chloride molecule." Here is an interesting characteristic of many aspects of nature - table salt has a set of properties that neither one of its components has. A wave or convection current in water has properties beyond those of the water molecules. A convection current can be neither predicted nor described exclusively from the molecular properties of water - the current expresses a different level of organization. It is more than the sum of its atomic or molecular parts; it demonstrates some form of emergence. Explaining convection calls for other regularities or laws of nature beyond the water molecules.
Another philosopher of science, Israel Scheffler ("Explanation, prediction and abstraction." in: Danto and Morganbesser, eds., Philosophy of Science. 1960. Meridian Books), made an important point about predictability when commenting on the Covering Law Model. He argued that although the model could be used to make predictions in many cases, explanation was its most important function. It is not necessary to always be able to predict, except in a general sense; it is necessary to be able to explain, or at least try to, as explicitly as possible. The Covering Law Model offers a structured and clear way to think about theories and hypotheses, to find their strengths and flaws without requiring that they be fully predictive.
To say that prediction is not an important aspect of science runs counter to some versions of the "scientific method," as it is often taught: make predictions as part of the hypothesis, collect data (often after manipulation of a single variable such as the effects of fertilizer or sunlight on plants), test the hypothesis, and reach conclusion as to the success of the predictions. But some aspects of nature are not very predictable. It's unreasonable to think we could predict all the details of a new species that has not yet evolved. This doesn't mean that phenomena such as evolutionary events are scientifically unapproachable. The idea that all natural events should be fully predictable is related to the idea of the mechanistic universe. Various references to a "watchmaker" (God) and sometimes to a "blind watchmaker" (the natural selection of Darwinian or neo-Darwinian theory) have been common throughout the brief history of western science, but it's not an accurate description. A less misleading analogy might be to think of the universe in terms of emergent properties - collections of atoms or molecules that become something more when energy flows through them, becoming no longer random, but self-organizing into convection currents, tornados and other structures, including developing, living things.
Jack Maze and Cy Finnegan offered additional clarification on predictability: "To those who may recoil in horror from the statement that prediction is not an important aspect of science, we would offer the following. Prediction can be used in two ways. One, as emphasized above, is a general statement that incorporates natural law in describing what may happen in the future. The other is a part of hypothesis testing. Hypothesis testing has the form of "if-then" statements: if a certain condition exists, then a certain observation is to be expected, is predicted. This is a different use of prediction and is often the one inferred when doing science." A prediction can be made with fair certainty that gravity will cause the rock to fall to the ground. Given the right conditions, an energized weather system can be predicted to form a hurricane. But once the hurricane forms, its specific behavior can be hard to predict.