Scientific Realism and Naturalism

Varieties of Naturalism

[ Home ] [ Pragmatism ] [ Humanism ] [ Atheism ] [ Pantheism ] [ More Links ]

Scientific Realism

Scientific Realism is essential to philosophical naturalism. Philosophical naturalism is based on science's ability to provide knowledge about nature. Science attempts to know nature's patterns and the underlying entities responsible for those patterns. Unlike positivism, which holds that science should only attempt to know nature's patterns, scientific realism holds that science should also try to know, and partially succeeds in knowing, about hidden entities behind these patterns. Scientific realism is the view that the entities postulated by highly confirmed scientific hypotheses really do exist, and have the properties more or less as described by these hypotheses. You should visit "Scientific Method" to learn about the terminology that is used in this section.

Scientific realism is always tempered by fallibilism, the reminder that scientific knowledge is imperfect and could be modified by future science. But this proclamation that science is fallible and revisable can inspire scientific anti-realism.

Objection One by the Scientific Anti-realist

The scientific anti-realist can argue as follows: If science is always fallible, then there never is a reason to believe that it discovers truths. Unless it discovers truths, science cannot reasonably claim that the things that its theories try to describe actually do exist. So we should not think that the things that science talks about really do exist. This argument is called the "Pessimistic Induction": all past scientific theories have been shown to be false, so all present and future theories will likely be false as well.

What is at stake here? If this argument cannot be refuted, its victory leads to complete skepticism about science and refutes philosophical naturalism.

Reply by scientific realism: The scientific realist should reply that perfect knowledge of truths about things is hardly required for reasonably believing that they exist. After all, we have practical reliable knowledge about matters like refining iron from ore, and it is quite reasonable to believe that iron exists. Science is just the extension of practical reliable knowledge.

This reply to the first objection by scientific anti-realism is the best place to begin explaining why scientific realism is reasonable. There are several good reasons for accepting scientific realism. Joined together, they provide the justification for scientific realism and hence for philosophical naturalism as well.

1. It is common sense to believe in hidden causes for observable patterns. Ordinary common sense consists of reliable practical knowledge about not just the easily observable patterns of nature but also about many of the things responsible for those patterns. We are very good at investigating the hidden causes of events in order to reveal them: what caused the window to break, what caused the loud noise outside, what caused the dog to bark, etc. Ordinary intelligence assumes hidden causes for observable events and applies methods of investigation to reveal these causes by bringing them into our experience. We are naturally curious about what causes interesting events, and we are especially interested about the causes of patterns of events. Human intelligence is very good at detecting and focusing on natural patterns, and we can explain a single event by noticing that this event should be expected since it is part of a pattern. But intelligence usually goes farther, not stopping at patterns, and seeks the hidden causes of patterns. What animals keep causing those patterns of tracks? Which tree produces those consistently tasty nuts? Which clouds always cause the worst rainstorms? The search for hidden causes of patterns requires the application of abductive inference.

2. Intelligence frequently discovers the hidden causes of natural patterns. Intelligence discovers hidden causes so frequently that the human brain is now highly evolved with fairly efficient curiosity and inquiry techniques that children instinctively use for survival and play, and adults refine these techniques into science. The abductive belief in hidden causes is so useful for practical reliable knowledge that intelligence cannot function without such instinctive practical belief. This practical belief in hidden causes explains why science also requires the application of abductive inference and why we provisionally accept the conclusions of confirmed abductive inferences. Making a hypothesis about a hidden cause is an intelligent effort to understand nature, and it is intelligent to provisionally accept the existence of a hidden cause that is postulated by a confirmed hypothesis. Science is merely the extension of investigative methods of ordinary intelligence, and scientific realism is therefore demanded by intelligence.

The limitation of this argument for scientific realism is that ordinary intelligence deals with the practical, everyday world of directly observable objects. We know what it is like to verify beliefs in hidden causes, since we frequently reveal these causes in direct experience after investigation. Some sciences only deal with directly observable objects -- their hypotheses and theories only postulate Type I entities -- but most sciences also postulate causes that are not directly observable: the Type II, III, and IV entities. Should we also be realists about instrumentally observable entities, and about non-observable entities too? Some scientific anti-realists say no.

Objection Two by the Scientific Anti-realist

The scientific anti-realist can argue as follows: Sciences about observable natural patterns and responsible entities that could be directly observed are legitimate sciences, but scientific realism should stop there. We should only be realists about what can be positively verified in our direct experience. The sciences should not try to describe Type II, III, or IV entities, and we should not think that such things really do exist.

What is at stake here? If this argument cannot be refuted, its victory leads to the type of extreme empiricism called scientific positivism or constructive empiricism: only natural observable patterns and the directly observable entities causing those patterns really exist. This scientific positivism is a kind of philosophical naturalism, but a very limited kind.

Reply: The scientific realist should reply that even if empiricism's demand for observable proof is taken so seriously, it is possible to reasonably believe in at least some Type II and Type III entities. First, instrumentally observable entities can be trusted if the instruments can be trusted, and we can reasonably trust many scientific instruments by methods of direct observation alone. Second, if it is reasonable to believe in many Type II entities, it may be reasonable to believe in Type III entities, since scientific progress frequently converts unobservable entities into instrumentally observable entities, increasing our confidence in postulating Type III entities.

This reply to objection two has two stages: First, defending instrumentally observable entities; and second, defending unobservable entities that potentially could become instrumentally observable entities.

First, why should we trust telescopes and microscopes? We come to trust them in the same way we trust our senses: by comparing the information we get from them as we use them under different conditions. We can learn that we have good eyesight by checking whether we can perceive things from farther away as well as we can when we perceive them nearer to us. We can learn whether a pair of glasses improves our eyesight, and we trust a reliable pair of glasses to show us real objects in front of us (should we think that putting on glasses suddenly presents us with a completely false version of reality?) Similarly, we can learn that a telescope accurately depicts a distant tree by comparing that observation with own perception of that tree using our eyes up close. Might the telescope change the way it works when we turn from looking at distant trees to looking at the distant moon to see craters? Skeptics about Galileo's observations of moon craters, sunspots, and moons of Jupiter suggested that a telescope might not work properly when aimed upwards. But direct inspection of the telescope and its parts shows that a well-constructed telescope does not change its functioning if it is pointed upwards into the sky. Some anti-realists have argued that trusting an instrument necessarily requires acceptance of a complete theory about why the instrument works, but such a theory would require postulating non-observables in the first place, begging the question in favor of scientific realism. The reply by the scientific realist is that it is valuable, but not necessary, to have a theory about an instrument's functioning. Summing up, we should be scientific realists about instrumentally observable entities because:

3. An instrument such as a telescope or a microscope can be tested and trusted by ordinary methods, already described above, that only involve direct observation. Of course, this reply only is relevant to instruments like telescopes and microscopes that detect very distant or very small objects, however (along with other instruments that amplify the senses, such as microphones for making sounds louder). This reply does not apply to other kinds of scientific instruments that detect things like electrical current or air pressure or chemical acidity. More arguments, explained below, will describe how scientific realism deals with these kinds of instruments.

Second, why should we postulate unobservable entities even though there currently is no way to even instrumentally observe them? Our experience with inventing and using sense-amplifying instruments demonstrates that nature consists of things much smaller and much larger and much farther away than our human senses can detect. And since the power of our instruments has grown over time, revealing more and more of nature to us, it is reasonable to believe that much of nature has not yet been observed, directly or instrumentally. Why shouldn't science be permitted to make hypotheses about possible entities in that yet-to-be-observed part of nature? Furthermore, many scientific hypotheses about Type III entities, confirmed by experiment, have later been observed by scientific instruments invented after the original hypothesis. Examples: Atoms were proposed before they could be observed by the scanning tunneling microscope; genes were proposed as the transmitters of biological information before DNA was observed by microscope; planets around other stars were proposed before they were observed by more powerful telescopes. Since many Type III entities proposed by confirmed hypotheses have later been verified by improved instruments, it is reasonable for science to propose hypotheses about Type III entities and believe that these entities exist when these hypotheses are highly confirmed by experiment. Recall the "Pessimistic Induction" from Objection One by the Scientific Anti-realist? The scientific realist can in turn propose the "Optimistic Induction":

4. Since many successful hypotheses about Type III entities have later been verified, it is reasonable to conclude that many more hypotheses proposed in the future about Type III entities will also someday be verified by more powerful instruments. Of course, we cannot yet know which current hypotheses about Type III entities will be verified in the future, but at least it is reasonable for science to try to postulate them now. As for Type IV entities, the Optimistic Induction argument does not help. There are no cases where confirmed Type IV entities have later been verified by improved instruments. This lack of cases is not surprising. By definition, we do not have any conception of what it could possibly be like to instrumentally observe Type IV entities (what would seeing gravity, or a black hole, be like?). So we would not try to invent instruments to observe them -- how could we confirm that we succeeded without any conception of what to look for?

The scientific realist does want to defend the reasonableness of postulating Type IV entities, but that requires a separate defense against scientific anti-realism. Before that issue is explored, the scientific anti-realist has another argument against Type III and IV entities.

Objection Three by the Scientific Anti-realist

The scientific anti-realist can argue as follows: Sciences may postulate and confirm hypotheses about Type I and Type II entities, but scientific realism should stop there. Consider how abduction works: we can never prove a hypothesis true using abduction, since there are potentially many more hypotheses, not yet thought by anyone, which could explain the same predicted patterns equally well. In fact, for any highly confirmed hypothesis H about a Type III or Type IV entity, there are so many potential hypotheses with the same empirical adequacy (able to explain the same natural patterns) that the marginal increase in belief gained by H should be zero or very close to zero -- there is no sufficient reason to believe in that entity's existence. Even if the sciences should be permitted to postulate such entities (maybe to increase logical coherence or predictive power), we should not think that the Type III or Type IV entities postulated by science really do exist.

What is at stake here? If this argument cannot be refuted, its victory leads to the type of rational empiricism called instrumentalism: only natural observable patterns and the directly or indirectly observable entities causing those patterns really exist. Instrumentalism is a kind of philosophical naturalism, but a very limited kind.

Reply: The scientific realist should reply that even though there are many potential hypotheses with the same empirical adequacy as any well-confirmed hypothesis, the real scientific value of a hypothesis, and hence its long-term credibility, lies in its pragmatic value: how well it meets the additional criteria of logical coherence, predictive power, physical unification, and ontological simplicity.

Scientific anti-realists are impressed by the fact that many (perhaps an infinite number) potential hypotheses can explain any particular set of natural patterns. They try to conclude that the probability of any one of the hypotheses being accurate must be very close to zero.

Pierre Duhem (1861-1916), the French philosopher and scientist, was this type of scientific anti-realist. He argued that some sciences, such as chemistry and physics, can never reasonably claim to have proven that their hypotheses are true. Only direct or instrumental observation can prove the existence of hypothesized entities, under the best of experimental conditions, but chemistry and physics try to describe entities that cannot be observed (what we have here been calling Type III or Type IV entities). Duhem realized that hypotheses about Type III and IV entities can never be proven true, since there are always other possible hypotheses that could explain the same observed natural patterns. We have already explained why the nature of abductive inference prevents proving that its conclusions are true. Since there are always many potential hypothetical explanations for the same observable phenomena of nature, scientific hypotheses are "underdetermined" by the currently available evidence. Duhem's underdetermination argument is simply the philosophical point that no hypothesis about a Type III or Type IV entity (the "unobservables") can be proven true. Duhem went farther than this point about abductive logic, however. Duhem claimed that since the available evidence cannot determine which scientific hypothesis about unobservables is correct, no scientific hypothesis about unobservables should ever be believed. In other words, even a highly confirmed scientific hypothesis cannot be reasonably believed at all. Science cannot ever get us even close to the truth.

The scientific realist replies to this argument by pointing out that empirical adequacy by itself does not produce much real credibility. For any given hypothesis, it is suspiciously too easy to artificially generate a rival hypothesis that can explain exactly what the given hypothesis can already explain. But can the artificial hypothesis cohere with the larger theory of which it must be a part? Can it make any more unexpected predictions? Does it offer any physical unifications, or only disunifications? Does it only increase ontological simplicity? There are far fewer rival hypotheses that can be artificially generated which meet these severe criteria -- and scientists continually try to create them and test them too.

5. Hypotheses about unobservables that continually increase their empirical adequacy and cohere well with larger theories and paradigms, while revealing a simpler reality behind the phenomena, deserve greater credibility. If the scientific anti-realist tries to continue to argue that even the possibility of a few alternative hypotheses which meet this much higher standard should restrain how much credibility any of them receive, this is not a new argument, but only repeats the unoriginal observation that abduction cannot ever prove any hypothesis. But we still can reasonably have some degree of belief in the most successful hypothesis available now.

What is our alternative to provisionally accepting the best hypotheses we have now? Consider what the scientific anti-realist is claiming -- if we really shouldn't think that any hypothesis about unobservables is believable at all, then we are denying that there is any underlying cause for natural patterns. Without underlying causes, it is a sheer miracle of chance that natural patterns persist and a complete mystery of luck that many hypotheses successfully anticipate more natural patterns. The scientific realist can offer a further argument at this stage:

6. If the entities described by highly confirmed hypotheses really exist, that would explain why science's best hypotheses are so empirically adequate. And science's best hypotheses are very empirically adequate. Therefore, the entities described by highly confirmed hypotheses really do exist. This argument for scientific realism has the form of an abductive inference: If P, then Q -- and Q is true -- so P is also true. Of course, this abductive argument cannot prove its conclusion. There may be other possible explanations for why science can produce hypotheses that have high empirical adequacy. The scientific anti-realist does have an alternative explanation for the existence of highly confirmed hypotheses.

Objection Four by the Scientific Anti-realist

The scientific anti-realist can argue as follows: It is reasonable to believe that there are hidden portions of nature where unobservable causes for natural patterns do exist. But we should not believe that any highly confirmed hypothesis approximately describes these unobservable causes. For Type III and Type IV entities, their only properties relevant to the empirical success of hypotheses about them are those properties which require these entities to obey certain natural laws. After all, it is only the natural laws that are used for formulating predictions For example, it doesn't matter what an electron is really like so long as its behavior always obeys certain laws. In fact, the entity that really does exist, if it isn't an electron, only has to obey those same laws, since it really has been the laws themselves that enjoy high confirmation. But it possible to imagine many alternative sorts of entities quite different from the electron that still obey the same laws that electrons are supposed to. Therefore, while unobservables exist, and science is reasonable to postulate unobservable entities, we should not think that the Type III or Type IV entities postulated by highly confirmed hypotheses really do exist.

What is at stake here? If this argument cannot be refuted, its victory leads to the type of rational empiricism called pragmatism: only natural observable patterns, the directly or indirectly observable entities causing those patterns, and hidden unknowable causes really exist. Pragmatic realism is a kind of philosophical naturalism, and it is a candidate for being the most reasonable type of philosophical naturalism.

Reply: The scientific realist should reply that even though it is possible to imagine alternative entities for playing the role of an electron, the hypothesis about the electron's existence only describes electrons in terms of the laws they all must exactly obey. The only properties that science attributes to electrons are the properties of obeying those particular laws. In other words, an electron simply is whatever entity obeys all of the electron laws. Any alternative imagined entity that still obeys the same laws is an electron too! Therefore, there are really no alternative hypotheses about alternatives to electrons -- there can only be rival hypotheses about genuinely different entities that obey somewhat different laws, and those have to be tested just like any hypothesis. It is reasonable to believe in the existence of unobservable entities postulated by highly confirmed theories.

In order to refute objection four by the pragmatic realist, the scientific realist here tries to claim that a hypothesis's conception of a postulated unobservable entity consists only of the laws that entity must always obey. But this claim has a steep price. If correct, this claim implies that all of the real work done by a hypothesis to make predictions is actually done by the natural laws postulated. The unobservable entity itself has no role to play in scientific method. Of course, whatever unobservable entities really exist do supply causes for natural patterns, and that is a crucial role. But only the natural laws -- those numerous equations -- proposed by hypotheses are actually used for formulating predictions about natural patterns. These natural laws are what are really being tested in the experimental method, and according to pragmatic realism, highly confirmed natural laws deserve credibility -- nature really does display habitual regularities as (approximately) described by scientific laws.

But is scientific realism the better choice over pragmatic realism? The pragmatic realist approves the existence of verified Type I and II entities, grants the existence of unobservable entities responsible for natural patterns, and agrees that science should propose and test hypotheses about unobservables. Scientific realism additionally claims that it is reasonable to believe that highly confirmed hypotheses fairly accurately describe unobservable entities. But what descriptions are given by such hypotheses? The only descriptions that matter are descriptions of the natural laws these entities obey, and the pragmatic realist already encourages belief in those highly confirmed natural laws. So what remains of any real difference between pragmatic realism and a reasonable scientific realism? Perhaps none. You can read more about "Pragmatic Naturalism".