Browse Search Feedback Other Links Home Home
The Talk.Origins Archive: Exploring the Creation/Evolution Controversy

Evolution and Philosophy

Predictions and Explanations

Copyright © 1997

Some rights reserved
This work is licensed under a Creative Commons License.




Evolution is sometimes criticised for not being a predictive science, and for not having natural laws. This relates to the issue of whether science should be like physics (see the section on the nature of science), but the two issues raise a more general matter.

It goes to the question whether explanations have to make use of natural laws, and just what are explanations anyway?

One theory about explanation is called the nomological deductive (ND) theory, or less pretentiously, the hypothetical deductive theory. Due to philosophers Karl Popper and GC Hempel [cf Dray 1966, especially the essay by A Donagan], it has the form:

Premises + Universal Law ⇒ Things to be explained

The idea is that if the thing to be explained is a logical, deductive, consequence of the premises and the universal laws, then you have explained it. Once you have a theory of this form, then you can predict that a phenomenon will occur if the initial conditions are right, based on the universal laws of physics, chemistry, etc:

Initial Condition + Universal Law ⇒ Observed phenomenon

There is a version that uses statistical assumptions and permits inductive argument rather than restricting explanation to deductive argument, called the statistical inductive model (SI), but we can safely ignore it here.

The prediction is a deductive consequence of a true theory and proper measurements. Since evolution cannot make predictions of this kind, and in fact any outcome is compatible with the theory, its critics say that evolution is not a complete science (see the section on the tautology of fitness).

However, there are problems with this highly idealised view of scientific explanation, and anyway, I will argue it doesn't affect evolution.

Any set of laws are ideal simplifications. In order to predict where a planet is going to be in 10,000 years, you have to ignore may things, such as the very small bodies, the influence of distant stars and galaxies, friction due to solar wind, and so forth. And it works, to a degree. But that degree is still real. You may only be off a few meters, but you will be off, due to these ignored complications. Physical systems of this kind are stable, in that the initial conditions do not greatly affect the outcome.

Evolution is not like these systems. It is highly sensitive to the initial conditions and the boundary conditions that arise during the course of evolution. You cannot predict with any reasonable degree of accuracy what mutations will arise, which genotypes will recombine, and what other events will perturb the way species develop over time. Moreover, the so-called 'laws' of genetics and other biological rules are not laws. They are exceptional. Literally. For every law, right down to the so-called 'central dogma' of molecular genetics, there is at least one exception.

And yet, we know the properties of many biological processes and systems well enough to predict what they will do in the absence of any other influences. This is proven in the lab daily. So, in this way, we have in biology the extreme end of the continuum of what we have in physics at the other end. The difference is one of degree, not kind. And more and more, physicists are uncovering systems that are similarly unstable and sensitive. You cannot predict in physics what any small number of molecules will do in a flame, or in a large gas volume, for example. And while the weather cannot be predicted at all in fine detail for very long, you can explain last week's weather through the initial conditions and the laws of thermodynamics, etc, after it has happened.

If you take the standard form of biological explanation, it has the same structure as a physical explanation. It just differs in two ways. First, you cannot isolate 'extraneous' influences ahead of time for wild populations. Second, you cannot make a prediction much beyond the immediate short term (hence, nobody can predict the future of evolution of a species). Although a number of experiments have been conducted to test selectionist hypotheses through prediction, such as the studies on finches in the Galápagos Islands by the Grants, mostly, explanations in evolution take the following format:

Initial Conditions at t-n + Properties if biological systems ⇒ Observed phenomenon at t

In other words, they are retrodictions, not predictions. The only formal difference between this and the same form in physics is that the tense is different. This use of the nomological-deductive model in historical cases is called a covering law model [Dray 1957, 1966].

So, physics is not really a different kind of science to evolutionary biology, except in some matters of convenience with experimentation, and the degree of the stability of the systems it sometimes explains, and not always then.

Covering law explanations can be used to retrodict the initial conditions, under certain circumstances. If you know what is now in evidence, and you have laws that generate these outcomes, you can sometimes predict what will be found:

Predicted initial conditions + Universal Laws ⇒ Observed phenomena

For example - you know that certain features of ants are derived (not in the primitive ancestor). You have general laws of evolution that account for the phenomena you observe (actual ants today, and in the fossil record). So, you predict that a certain transitional form will be found. When it is, you have made a bona fide prediction.

What special conditions can this be done under? Well, for a start, if you have a deductive argument if A then B, you cannot immediately infer from the existence or truth of B, that A. It might have been something else. B might have a virtual infinity of possible causes. Before you can make a retrodiction like this, you have to narrow down the field. That is, you have to assume the validity of some theoretical models before you can make the retrodiction/prediction. On the other hand, if you make such a claim, and it pans out, you have certainly strengthened your model.

Finally, note that the ND model is not sophisticated enough to capture everything important about scientific explanations. A good many scientific explanations rest not on laws but propensities, that is, likelihood to behave in a certain way. And many perfectly useful accepted scientific explanations are not deductive, they are inductive. That is, the likely outcome of the initial conditions and the laws is not a rigourous deduction but an induction with all the problems that brings. Still, that's what science does, whether philosophers like it or not (cf Franklin 1997).




Home Browse Search Feedback Other Links The FAQ Must-Read Files Index Evolution Creationism Age of the Earth Flood Geology Catastrophism Debates
Home Page | Browse | Search | Feedback | Links
The FAQ | Must-Read Files | Index | Creationism | Evolution | Age of the Earth | Flood Geology | Catastrophism | Debates