Theory reduction and instrumentalism

I said that we can explain not only particular events but also general truths and that Hempel's theory takes account of this. There is a variety of general truths we might want to explain, among them some laws.

Purple and white peas, when crossed, give rise to purple and white peas

is an observational law. This too can be derived logically from MG, along with the two laws and correspondence rules, which tell us that P is dominant over W. We can deduce laws in the mixed language, such as the law that

Homozygous red peas crossed with white peas will have only pink offspring.

Finally, of course, we can deduce theoretical laws from MG, such as that

Two homozygous genotypes of distinct alleles will produce only heterozygous genotypes in the first generation.

The fact that we can explain generalizations on this model is of very great significance for the received view: it allows us to tell a story about how science can develop. One of the striking features of the history of science is the way in which earlier theories get super­seded by later ones. Sometimes, of course, we just discover the old theory was wrong. It makes false predictions. But sometimes we dis­cover that we can keep much or all of the old theory while the new theory develops, because the new theory explains the old theory.

Something like this happened in the history of genetic theory.

mation. It was possible, therefore, to explain why all of the first eleven of Mendel's claims were true. They were true because small portions of the chromosome obeyed these eleven principles. For example, every cell of the organism, except the gametes, had two chromosomes, one from each parent, and that was why there were two alleles at each chromosomal locus in each cell. Thus, according to the DN model, these eleven claims of Mendel's theory could be explained by the chromosome theory, because they could be derived logically from it.

But some genes failed to obey the laws of independent assort­ment. They were not inherited independently. This was because, if two genes were on the same chromosome, when the chromosomes came to be divided between the gametes, the genes were bound together and so were inherited together. If the genes for stem tex­ture and flower color had been like this, for example, then, as we saw, Mendel might have got only two out of the four theoretically possible kinds of offspring.

Nevertheless, where genes were on different chromosomes, they did obey Mendel's second law of independent assortment. So, as it turned out, the second law, which was one of the most significant parts of Mendel's theory, was in fact true only in special cases: the cases where the genes were on different chromosomes. Not only was the law true of pairs of genes on different chromosomes, the discovery of the chromosome allowed one to see immediately why the law was not obeyed by genes on the same chromosome and to predict what would happen in that case.

Thus, when chromosomes were discovered, genetic theory was able to build on Mendel's theory. On the received view of theories, we can see how science can be progressive.

The process of showing that an old theory can be derived from a new one as a special case is called theory reduction. We can derive the old theory from a new one, using the special conditions under which the old theory works as the antecedent conditions of the explanation. On the DN model the successes of the old theory are, thereby, explained. Thus, in the case of genetics, the fact that Mendel's theory was superseded by the chromosome theory didn't mean that all the explanations it had made possible had to be given up. The old explanations were still adequate in all those cases that didn't depend on the second law, just because all the other laws were still true. And even those Mendelian explanations that presup­posed the second law could easily be salvaged in any case that involved only pairs of genes on different chromosomes.

This view of theoretical progress also accounts for an important fact about the many so-called crucial experiments in scientific his­tory—those experiments that play a decisive role in the changeover from one theory to another. On this view of scientific progress, such crucial experiments play the role of showing where an old theory breaks down. But, because the old theory usually works in many cases, the circumstances of the crucial experiment are important in defining the antecedent conditions under which the old theory do>es work. The crucial experiment contributes to the progress of science not simply by getting us to jettison the old theory, but by showing something about its limitations. Thus the experiments that demon­strated that not all genes were independent showed that Mendel's theory was limited in its application, a fact that the chromosome theory was able to explain.

The received view of explanation and of theory reduction was realist.

This view of theories is called “instrumentalism.” Instrumentalism, then, holds that theories are just instruments that allow us to predict phenomena. But instrumentalism, though it is quite consistent with the fact that scientific theories have led to a great increase in our capacity to predict (and thus sometimes con­trol) what happens in our world, certainly doesn't seem to capture what most scientists think they are doing. After all, according to the instrumentalist view, Mendel's theory wasn't really about unobserv­able entities called genes at all. The only part of Mendel's theory that matters for the instrumentalist is the observation language. Indeed, any theory that made exactly the same predictions as MG in the observation language would be just as good. That's because the instrumentalist gives up the empirical condition of adequacy for explanations.

One of the major arguments for instrumentalism is epistemolog­ical. Instrumentalists, like logical positivists, are radical empiricists. They want to say that beliefs are justified only if they have empiri­cal support, only if there are observations that lead you to believe them. We can see why this might lead you to think that you ought not to believe in unobservable entities.

Consider any theory, such as MG, that refers to unobservable things.

This epistemological argument for instrumentalism amounts to a challenge: the instrumentalist wants us to show why we should care about matters to which no possible evidence is relevant. That most of us do care is obvious enough. It is one thing to suggest that we can only use terms that connect with things we can observe, which is what the received view says, and another to say, with the instru­mentalists, that we have no reason to believe that there are things we could not, under any circumstances, observe. Indeed—we can respond to the instrumentalist's challenge—surely, whether or not we can observe a thing is just a fact about us. And why should the furniture of universe depend on us? The issue here is essentially the one that came up in the private-language argument and in the argu­ment for verificationism. Wittgenstein said that we must be able to check that we are referring to a thing properly if it is to exist. The Verificationist says we must be able to know about something if it is to exist. And the instrumentalist says we must be able to observe a thing if it is to exist. All these views are to some extent idealist: they hold, in opposition to realism, that the existence of an object depends in some way on our ideas, on its relationship with our minds.

Not only is instrumentalism idealist, its consequences are in other ways counterintuitive.

4.8