Varieties of Naturalism

                              Scientific Method

Scientific Method is used to construct scientific knowledge about nature. Knowledge is increased through careful observation and logical inference. Observation, inference, and knowledge are almost always fused together to some degree. For example, much of our observation of the world is recognition of the familiar, which is observation informed by knowledge. Also, most of our observations of objects in the world are informed by inference, because the information from our senses in quite superficial by itself. Using vision, we only really see surfaces, shadows, colors, patterns: but we observe objects having depth and volume and texture. Using hearing, the various noise we hear become the sounds of things near and far, like approaching cars or hidden animals. Our ordinary everyday experience of the world is a type of knowledge, which we can characterize as practical reliable knowledge. Although this knowledge of ordinary experience is often mistaken, it works well enough for our daily activities. Examples are gathering vegetables and cooking them for a nutritious meal, or weaving cloth and sewing it to make clothing. This webpage explains the scientific method. You can also proceed to "Naturalism and Science" and read a webpage about "Scientific Progress".

Empiricism and Rationalism

Philosophers who believe that experience is the source and ultimate justification for all knowledge are called empiricists. Some empiricists have looked to experience to provide a higher type of knowledge than practical reliable knowledge, a type of knowledge that is infallible and certain, which will never turn out to be false. But not all empiricists search for certain, perfect knowledge -- we might call those who do undertake this search "extreme empiricists". There have been few extreme empiricists, since there are serious problems with trying to find reasonable cases of experiences that give perfect knowledge about the world. These problems are so severe that other rival philosophers have concluded that experience by itself cannot be a source of perfect knowledge at all. Indeed, most empiricists do accept that experience needs help from reason to establish knowledge about the world. However, experience (even with help from reason), can never establish perfect knowledge about the world (as will be explained below). A philosopher searching for perfect knowledge will conclude that experience cannot play any role in perfect knowledge. What other source of perfect knowledge is possible? The alternative to experience is reason, and philosophers who emphasize the large role that reason must have for knowledge are called rationalists. Some rationalists, searching for perfect knowledge, will use only reason to find knowledge, and we might call them "extreme rationalists". As it turns out (also to be explained below), reason by itself cannot establish any perfect knowledge about the world. That is why there have been few extreme rationalists in the history of philosophy. Most empiricists have decided that experience needs a little help from reason to establish knowledge, and most rationalists have concluded that reason needs a little help from experience to establish knowledge about the world. Debates between these empiricists and rationalists are surveyed by this article about "Rationalism vs. Empiricism". If both sides assume that perfect knowledge must be the quest, then both sides must fail. Experience and reason can indeed be artificially separated from each other, in the philosophical imagination (again, far from our ordinary experience in which observation, inference, and knowledge are partially fused together). By artificially separating experience from reason, extreme empiricists and extreme rationalists actually destroy the possibility of knowledge about the world. That is why most philosophers conclude that both experience and reason are needed for knowledge about the world, and the difference between empiricists and rationalists comes down to different estimates about how much experience and reason contributes to knowledge.

In the extreme empiricists' philosophical imagination, experience is "purified" of anything that might admit the possibility of error and illusion, and the empiricists announce the discovery of a realm of "sensations" or "sense data" that can never prove false. Example: "There is a bright point of light." In this example, a person making this judgment is claiming to observe something and describe it so narrowly that she can never be shown to be wrong. If instead she claimed, "There is a star in the sky", this judgment could conceivably turn out to be wrong, because we can imagine how further investigation could show that what this person really experienced was not a star (but instead a planet, or an airplane, etc.). The problem with pure sensations, even when described in infallible ways, is that they cannot help establish knowledge about the world. Knowledge consists at least of judgments about the world expressed in propositions of some public language. If pure sensations are expressed in judgments, they either (1) fail to be about the world, and instead are about some realm of pure experience (just lights and colors and noises and tastes, etc.); or (2) they try to be about the world but begin to suffer from the possibility of error and illusion (e.g. is that really a circle of light, or maybe an ellipse -- and is it red, or reddish-orange? etc.). Furthermore, anything like scientific knowledge about the world would at minimum consist of judgments about the regular behavior of objects and events in the world. Yet pure experience cannot establish these sorts of judgments because of the "Problem of Induction": even though a series of experiences may have common features, and appear to present a pattern, it is impossible to have perfect knowledge that this pattern would continue into the future. Empiricism's quest for perfect knowledge through experience alone can therefore only lead away from knowledge about the world and can never produce anything like scientific knowledge. In the 20th century, scientific anti-realists have generally preferred types of Empiricism (like positivism's view that science can only describe patterns of phenomena).

On the other side, in the extreme rationalists' philosophical imagination, reason must have a method of inference for establishing perfect knowledge. The only method of inference that promises to prevent all possibility of error is deduction. Deduction is a careful relation between premises and a conclusion, designed so that if you know that the premises are all true, you can also know that the conclusion is true. So long as the premises remain true, the conclusion can never turn out to be false, and your knowledge of the conclusion is perfect knowledge. You can read an advanced article about "Classical Logic" here. The difficulty with deduction is that a person's perfect knowledge of conclusions depends on perfect knowledge of the premises. How can a person perfectly know the premises? Well, perhaps other deductive arguments show that each of the premises are knowably true. Ok, but those additional arguments must have their own additional premises, which all need their own deductive arguments to justify why they can be known to be true, and so forth, and so on -- are an infinite number of arguments needed for any knowledge? That seems strange, since no person could hold an infinite number of arguments in their mind, and thus can never be assured that perfect knowledge is achieved. There are two other alternatives: (1) perhaps some premises can be known to be true without any argument (see "Foundationalist Theories of Epistemic Justification"), or (2) perhaps some special conclusions can serve as premises for other arguments, which in turn prove conclusions that serve as premises justifying those special  conclusions, so that only a finite number of arguments are actually needed (see "Coherentist Theories of Epistemic Justification"). Rationalists have usefully developed the foundationalist or coherentist alternatives, and these developments are very important for scientific method and realism, so they will be discussed further in sections below. However, extreme rationalism is a dead-end because pure deductive inference (nor inductive or abductive inference either -- more about these below) cannot establish any perfect knowledge about the world. Reason by itself can form perfectly coherent systems of thought, but there is no way to determine which system must be true, and most are quite compatible with the natural world. In other words, pure reason's truths are either (1) not about the natural world at all, or (2) somehow they are true about all possible worlds. Most rationalists therefore admit that reason needs some information from experience in order to produce knowledge about the actual natural world (thus agreeing with most empiricists that experience needs some assistance from reason).

The endless debates between extreme empiricism and extreme rationalism are inconclusive, because each side can show why the other side must be inadequate. Experience by itself cannot be a path to perfect knowledge about the world, but reason alone cannot establish any knowledge about the world either. Most philosophers turn away from this fruitless debate and take a compromise position that could be called "Rational Empiricism": knowledge about the world is created by experience and reason working closely together. However, rational empiricism is a philosophical position that admits that perfect knowledge about the world is never possible. There may be types of perfect knowledge, but none of them can be about the actual natural world. Since scientific knowledge is about the natural world, then scientific knowledge cannot ever reach perfect knowledge, and therefore any scientific knowledge is less than certain -- instead, scientific knowledge, even at its best, is always fallible (could be exposed as false in the future) and revisable (could be improved or entirely replaced with better scientific knowledge). The scientific method itself is not a case of perfect knowledge either -- rather, scientific method is a tool that can be (and has been) modified and improved through regular use and testing. Furthermore, although there is general scientific method that is explained here, each of the sciences uses its own specific version of the scientific method that works best for that science.

We can now ask this question: what is the relationship between the knowledge of ordinary everyday experience and scientific knowledge? Is scientific knowledge a quite different sort of knowledge from ordinary reliable knowledge? This introduction to scientific method takes the position that scientific knowledge is also a kind of reliable practical knowledge, but vastly improved: the reliability and practicality of scientific knowledge is far greater than that of ordinary everyday knowledge. Also, the scientific method depends on ordinary experience, but often must improve that experience for its own purposes to become scientific observation.

Scientific Observation

A person makes a scientific observation by properly using an approved instrument (one that has the confidence of the scientific community) for focus and/or measurement to carefully experience a thing or event that is public (could be observed by others too), and the person makes a record of the observation using a description that is precise (the thing or event is described in a more formal way than ordinary language, using special concepts and categories to increase discrimination and accuracy). The best kinds of scientific observations are designed to be both precise and public: these observations are described using concepts and categories specially designed and used by a scientific community of people, all trained for making good observations. Example: Measuring the movement of a planet across the sky from night to night across two years' time. The astronomy community designed a system of concepts and categories (right ascension and declination) for describing the exact position of any object in the sky. Using this system, any trained and careful observer will be able to accurately record the position of a sky object. When a community of scientists all use the same system for observation, and are well-trained to perform observations using this system, the community has established the possibility of scientific objectivity: scientific knowledge about natural objects and events within experience. This scientific objectivity, which provides reliable and practical information about objects and events, is the starting-point of scientific method and makes science possible. The scientific method uses experience to produce knowledge, but not just any sort of experience: only scientific observation counts. Of course, scientific observation is still fallible and revisable, since scientists make errors and misjudgments even when sincerely trying to do their best. The best kind of scientific observation is highly objective by being repeatable and durable: lots of scientists have been able to make the same observation (or almost the same, within a reasonable amount of error) over long periods of time.

Observation, inference, and knowledge are always fused together to some degree. This is true for ordinary experience, and it is true for scientific observation. For example, when astronomer Tycho Brahe observed and recorded the positions of the planet Mars during the late 1500s, he used quadrants and sextants. Brahe's observations enjoyed a high level of scientific objectivity because of their precision and replicability, which was possible by using his excellent instruments. Only an instrument already approved by a scientific community, which agrees on how that instrument is correctly used, can be used to make scientific observations. The scientific community endorsed the use of Brahe's instruments because it understood how they worked and agreed upon how they should be used. Brahe was able to make scientific observations of Mars because his experience was enhanced by inferences from what he saw using his sophisticated instruments to make conclusions about the precise position of Mars in the sky, and these inferences depended on his knowledge about his instruments worked. The 1600s witnessed the introduction of the high-powered telescope, which further transformed astronomy. Astronomers quickly accepted the telescope because they acquired reliable and practical knowledge about telescopes, and they also possessed a well-established theory about how a telescope worked, from the science of light and optics.

A scientist's own senses can qualify as scientific instruments. For example, a scientist's own eyes can be adequate instruments for making scientific observations. Unless Brahe's eyes were adequate instruments for using his instruments properly, his observations would not have been accepted as scientific by the community of scientists. The trained eyes of a botanist are used to make scientific observations about the structures of flowers. The trained ears of a ornithologist are used to make scientific observations of bird calls, as another example.

Scientific observations are observations about natural objects that really exist. How do scientists know that their observations are truly of things that really exist? Rational Empiricism would say that a valid observation is an experience aided by inference and knowledge. A scientist must have an experience of an object that includes its "identifying qualities". The scientist already knows what qualities a certain thing must have, which identify it. In the observation of a thing, the scientist looks for its identifying qualities, and when those identifying qualities are observed, the scientist infers that it is indeed that particular thing which is observed. This inference could be mistaken (maybe other things also have that quality), so a scientific observation remains fallible, like any knowledge. Scientists reduce the possibility of mistaken identifications by having rigorous tests for several qualities that uniquely identify particular things.

Science makes a useful distinction between a thing's qualities that depend on their being observed ("perspectival" qualities), and a thing's properties that exist regardless of whether they are being observed ("independent" qualities). Perspectival qualities only exist when organisms are perceiving them: colors, sounds, tastes, textures, etc. Independent qualities exist even when they are not being perceived, although we naturally detect them using perception: shapes, mass, size, etc.

Also, science makes a useful distinction between those perspectival qualities which can be observed by the unaided senses ("directly observed" qualities), and those which can only be observed through mechanical instruments ("instrumentally observed" qualities).

Finally, some independent qualities are detected only by inference from other instrumentally observed qualities.

Five kinds of qualities are therefore discriminated by science, according to the method of their identification:

(1) the directly observed perspectival quality (DOPQ).  Examples of using a DOPQ: a chemist identifying a mineral by its color, a ornithologist identifying a bird by its song, and a geologist identifying a rock by its texture.

(2) the instrumentally observed perspectival quality (IOPQ).  Examples of using a IOPQ: an astronomer identifying a red giant star by its flickering color through a telescope, and submarine sonar operator identifying a surface vessel by its amplified propeller noise.

(3) the directly observed independent quality (DOIQ).  Examples of using a DOIQ: a paleontologist identifying a fossil bone by its shape, and an oceanographer identifying a tide by the water height.

(4) the instrumentally observed independent quality (IOIQ).  Examples of using a IOIQ: an official of the bureau of weights and measures using a standard gallon container to identify a full gallon of gasoline from a station pump, and an engineer using calipers to measure the size of a machine part.

(5) the instrumentally detected independent quality (IDIQ).  Examples of using a IDIQ: a physicist identifying a metal by calculating its density from its measured volume and weight, and a geologist identifying an iron ore by measuring its magnetic attraction.

In actual scientific practice, identifications often use a combination of two or more of these means. There is one kind of natural entity which must receive additional scrutiny: that entity which, due to the scientific conception of that entity, can only be identified by one or more IDIQs, and thus cannot be identified by any direct or instrumental observation. Examples of such non-observable entities are black holes, the force of gravity, and the curvature of space-time. Evidence for such entities must always consist of the detection of their effects on scientific instruments.

Logical Inference

There are three types of logical inference: deduction, induction, and abduction.

Deduction: If the two premises are both true, then the conclusion must necessarily be true. Conversely, if the conclusion is false, then one or both of the premises must also be false. Deduction is the only method of inference that is capable of proving that a proposition is true. You can consult this article on "Deductive Reasoning" and another on "Logical Consequence".

Induction: The two premises describe the qualities of a sample from a larger group, which suggests a pattern. The conclusion states that the pattern will continue. If the two premises are both true, then the conclusion has some (perhaps small) probability of being true (somewhere between 0% and 100% probability). The degree of probability depends on the size of the sample, the size of the larger group, and the method used to select the sample. You can consult this article on "Statistics" and another on "Induction". Induction can never prove that a proposition is true. That is because it is always conceivable that a pattern will change or stop at some point in the future: this is the "Problem of Induction". Although induction cannot lead to truth, it remains very useful so long as it is done carefully to avoid "Faulty Generalization".

Abduction: The two premises state what is known now about a situation. The conclusion is a hypothesis about how that observed situation came to be that way -- a hypothesis that tries to explain the current situation in terms of some other hidden situation that hasn't been observed. An abductive inference has this form:

Understanding abduction is essential to understanding scientific method. Abduction is science's only way of suggesting novel explanations for the observable events and things in nature. Deductions do not give explanations: their conclusions only restate, in a rearanged way, what is already stated by the premises. Inductions do not give explanations: their conclusions only make predictions about the future. Abductions do give explanations: their conclusions are statements about things or events that have not yet been observed, or will never be observed, which are responsible for the facts stated in the premises. In the first example about the beans, the two premises state what is known now: all the beans from this bag are white, and these beans are white. What might explain where these white beans came from? Well, they might have come from that bag, where all the beans are white too. If these beans really did come from that bag (an event that was not observed by the person making the abductive inference), that would explain where the beans came from. In the second example about the sun and gravity, the two premises state what is known now: a force exerted by the sun will keep its planets in orbit, and planets are in orbit around the sun. What might explain why these planets go in orbit around the sun? Well, the first premise states that if there was a force exerted by the sun then the sun will keep its planets in orbit. What force could be exerted by the sun? Well, a force of gravity, if it really existed (but it has not been observed), would explain how the sun could exert a force on its planets. If there really was a force of gravity exerted by the sun, that would explain how the sun keeps its planet in orbit.

A scientific explanation in general has the following abductive form: These facts are known to be the case now; if some presently hidden thing really exists or did exist, then the known facts would have to be true; therefore, some presently hidden thing really exists or did exist.

There are four basic types of hidden things or events (hereafter collectively called "entities") that play roles in explanations by science. Let us call them Type I entities, Type II entities, Type III entities, and Type IV entities.

Type I. An entity which could be observed directly, and identified by its DOPQ or DOIQ. Example: Did these white beans come from that bag? -- Well, the explanation is "hidden" in the past. Perhaps someone observed where those beans came from. When a Type I entity is hypothesized by an abduction, that hypothesis can still be proven to be true by actually directly observing it.

Type II. An entity which could be observed instrumentally, and identified by its IOPQ or IOIQ. Example: Did a fault line cause that earthquake? -- Well, the explanation is "hidden" under the ground. Perhaps someone can instrumentally observe the fault line using seismology equipment. When a Type II entity is hypothesized by an abduction, that hypothesis can still be proven to be true by actually instrumentally observing it.

Type III. An entity which could be observed by some new instrument not yet invented, and identified by its IOIQ. Example: Did the very early universe have a certain structure? -- Well, we now have no instrument that can make any good observations of the very early universe. Perhaps someone will invent a far more powerful telescope. After the invention of the needed instrument, the Type III entity changes to a Type II entity. When a Type III entity is hypothesized by an abduction, that hypothesis can never be proven to be true, until the needed new instrument is invented.

Type IV. An entity which could not be observed by any instrument because it can only be identified by one or more IDIQs, and thus cannot be identified by any direct or instrumental observation. Examples of such non-observable entities are black holes, the force of gravity, and the curvature of space-time. Observed evidence for such entities must always consist of the detection of their effects on scientific instruments. Sometimes science advances through both a theoretical and instrumental advancement, so a Type IV entity can be converted into a Type III or Type II entity. For example, until the 20th century, science had to classify atoms as Type IV entities, but now large atoms can be instrumentally observed. When a Type IV entity is hypothesized by an abduction, that hypothesis can never be proven to be true, and it is never reasonable to believe with 100% certainty that this entity really exists.

 

The Six Steps of Scientific Method

The Scientific Method has three stages and six steps. In the first stage, the "observation stage", there are two steps which describe how science begins with scientific observation and then uses induction to formulate a law of nature. In the second stage, the "hypothesis stage", there are two steps which describe how science uses abduction to postulate one or more hypothetical entities (from among the four Types I-IV) to explain what has been observed in stage one. In the third stage, the "testing stage", there are two steps which describe how science uses deduction to test the hypothesis from stage two against more scientific observations and against rival hypotheses.

Stage One: Observation

Step One: Phenomena. Using established scientific knowledge, new scientific observations of a pattern of events are recorded.

Step Two: Natural Law. Using induction, this pattern of events is believed to continue into the future, and this pattern can usually be expressed as a regularity or habit of nature (sometimes as a "law of nature" like an equation, as well).

Stage Two: Hypothesis

Step Three: Explanation. Using abduction, a hidden entity of Type I, II, III, or IV is postulated as the explanation for the regularity of nature found in step two.

Step Four: Prediction. For a Type I or Type II entity: its predicted existence can be tested by direct or instrumental observation, so long as its characteristic IOPQ or IOIQ identifiers have been agreed upon.
                                    For a Type III or Type IV entity: using deduction the actual existence of this hidden entity implies that it must be responsible for other unexpected patterns of events also, besides those observed in step two and other patterns already recognized by science. These other unexpected patterns are the hypothesis's predictions. To be optimally useful, a prediction should be a "risky" prediction: (1) very unexpected (ideally, forbidden by a rival hypothesis); (2) very specific (vague predictions are suspicious because they are too easily confirmed); and (3) not very difficult to test by experiment in the next stage.

Stage Three: Testing

Step Five: Experiment. For a Type I or Type II entity: its predicted existence can be tested by scientific observation, so the needed observations are attempted.
                                    For a Type III or Type IV entity: using established scientific knowledge and deduction, experiments are designed and conducted to find out whether any of the predicted patterns of events from step four can be scientifically observed.

Step Six: Verification, Confirmation, or Falsification. For a Type I or Type II entity: if its existence is looked for and successfully verified by scientific observation, then the hypothesis is verified as true (although there may be additional entities that are also contributing causes to the patterns of events). If its existence cannot be established, then science can return to step three to try again.
                                    For a Type III or Type IV entity: if a predicted pattern of events is scientifically observed in an experiment, then this positive result is a "confirmation" for the hypothesis. A confirmation makes it reasonable for belief in the postulated hidden entity to marginally increase. If a series of predicted patterns are all confirmed, and none are disconfirmed, belief in the postulated hidden entity can become substantial, but should never reach 100% certainty. If a predicted pattern of events is looked for and found to not exist, then this negative result is a "disconfirmation" for the hypothesis. Unless a disconfirmation can be explained by human error (in the prediction, or in the experiment design, or in the observation), this disconfirmation makes it reasonable for belief in the postulated hidden entity to marginally decrease.

Under certain circumstances (where a prediction is carefully deduced, the experiment is well designed, and no scientific knowledge involved in step five can be reasonably faulted instead of the hypothesis) a disconfirmation makes it reasonable for scientists to conclude that the hypothesis is proven false and the entity does not exist. The inference to such a negative conclusion has a valid deductive form, superficially similar to that of abduction, which is called "modus tollens":

If P, then Q.

But Q is false.

Therefore, P is false.

Let P be the hypothesis "this hidden entity exists" and Q be a prediction deduced from this entity's existence. If this prediction is discovered to be false by an experiment, then Q is false and therefore P must also be false: that hidden entity does not exist. This "falsification" will force science to return to step three to either modify the hypothesis or to entirely abandon the hypothesis for some other alternative hypothesis. You can read more about scientific experiment in "Experiment in Physics".

This account of scientific method, by emphasizing the crucial role played by abduction, explains why scientists take a realistic approach to explaining nature's processes. The entire point of abduction is to postulate the real existence of some sort of entity. Likewise, the entire point of the entire scientific method is to show how to reasonably increase or decrease belief in the real existence of postulated entities. Some philosophers, whose arguments will be explored below and in "Scientific Realism and Truth", have concluded that "anti-realism" is correct: no version of scientific method legitimately shows how to reasonably modify belief in the real existence of postulated entities. Anti-realism is a philosophical position about the legitimacy of scientific methods: do the accomplish what they propose to accomplish? Because anti-realists answer "No", they frequently make the additional argument that scientists should not take a realistic approach to explaining processes, and they therefore would dispute the account of scientific method given here. But such dispute is entirely premature and irrelevant. This account of scientific method explains what scientists are actually doing when they attempt to explain nature's processes, and accurately portrays the scientific "natural ontological attitude" (a phrase from Arthur Fine) of tentative realism towards postulated entities. Scientists really do postulate the real existence of entities, and modify their degree of belief in those entities, using this scientific method (or minor variations upon this method). Even if anti-realism could be persuasively established by philosophical argument, that does not affect the usage of scientific method itself. Anti-realism is a protest against realism, and this account of scientific method explains where that realistic attitude originates.

This account of scientific method is designed to equally apply to both the physical and social sciences. Some naturalists have questioned the social sciences' very claim to scientific status, claiming that they have a similar yet distinct scientific methodology. Do the social sciences hypothesize about hidden entities like the physical sciences? Can the social sciences make and test predictions under experimental conditions like the physical sciences? If the answers to these questions are negative, then the social sciences use some other methodology besides the one presented here. However, this naturalist believes that the answers to these questions can be affirmative (regardless of the current status of the social sciences), for reasons impossible to explore here.

Genuine Scientific Hypotheses and Theories, and the Demarcation Problem

We have already defined the genuine scientific observation. A genuine scientific hypothesis is a hypothesis that is designed to explain a natural pattern already scientifically observed, and is testable by the scientific method, outlined above. The statement of the natural pattern discovered in stage one is not a hypothesis, although clumsy use of words sometimes results in labeling a scientific law as a hypothesis. The discovery of a natural pattern is not an explanation -- it is what needs an explanation. It is possible to "explain" a single event by holding a natural pattern responsible (the leaf fell off the tree because trees lose their leaves in the fall). However, a pattern of events cannot really explain why one event in that pattern did happen. While much of science is focused on discovering and describing natural patterns, a field of study does not become a scientific field until it proposes and tests hypothetical explanations for natural patterns.

A hypothesis is scientific when it is treated by investigators as scientific: when it is developed and tested using the scientific method. For example, the Greek philosopher Democritus promoted the idea that all natural things were made of tiny invisible atoms that cannot themselves be divided. But the arguments which Democritus and fellow atomists gave for this idea only had deductive forms and appealed to allegedly "necessarily true" premises, and they did not experimentally test their idea. The atomic theory did not become scientific until the late 1800s.

There are many ways that people can prevent their hypothesis from being genuinely scientific. For example:

(1) Offering an explanation for phenomena that have not been scientifically observed;
(2) Postulating an entity that has no clearly defined identifying qualities;
(3) Postulating an entity that would not be responsible for any unexpected natural patterns;
(4) Postulating an entity so contradictory against established scientific knowledge that experimental testing is impossible;
(5) Refusing to deduce predictions from the supposed existence of the postulated entity;
(6) Ensuring that any predictions are either vague, difficult to experimentally test, or unsurprising;
(7) Ignoring any prediction that turns out to be false;
(8) Modifying the hypothesis just enough to be able to afterwards "predict" a bad experimental result.

A "theory" consists of several hypotheses that are interrelated and support each other in order to provide a fuller explanation of a range of phenomena in some field (chemistry or astronomy or psychology or archaeology, etc.). For example, the theory of natural selection in biology consists of a large number of hypotheses about organisms and how they interact with their environment. A theory is scientific so long as all of its hypotheses are scientific. It should be noted that items of knowledge from logic or mathematics are used in a theory, but they are not hypotheses, since they are not postulates about hidden entities. However, logical or mathematical principles may be modified or replaced within a theory, if the theory's development requires these changes. For example, a theory may be able to explain some new natural patterns only if it uses a different mathematical or logical system. In this sense, mathematical and logical principles could be considered as "testable" against scientific evidence, because a theory's ability to explain the evidence can occasionally require modifying these principles. But there is no way to directly test any logical or mathematical principle against evidence -- by themselves, apart from all hypotheses, these principles make no claims about nature and they are compatible with any natural events.

We can also ask whether a scientific theory can ever lose its status as scientific. Some philosophers, including Karl Popper, have argued that a scientific theory must continually be used to generate new predictions and be tested. But this is not a reasonable standard, since most of the established body of scientific knowledge no longer receives serious experimental testing. One serious event can cause a once scientific theory to lose its status: A pattern of nature is discovered which the theory ought to be able to explain, but the scientific community ignores this need and does no inquiry into whether the theory really can explain it. A theory which ignores new patterns of nature will likely be replaced eventually, because a rival scientific theory will emerge which does succeed in explaining the new patterns and gain credibility quickly with these successes.

A "paradigm" consists of several theories that are interrelated and support each other in order to provide the fullest explanation of the widest range of phenomena in some field. For example, biology's current paradigm is evolution, which incorporates theories about natural selection, reproduction by genetic inheritance, DNA mutation by random errors, and other theories about living organisms. A paradigm is scientific so long as all of its theories are scientific. Each scientific field is typically dominated by one paradigm for a time, when a large majority of scientists in that field accept only this paradigm. Occasionally, a field may have multiple scientific paradigms competing for dominant status, and at other times a field might have no scientific paradigm that is accepted by even a significant minority of scientists.

The only requirement that a theory must meet to be scientific is the requirement that the theory's hypotheses are all designed to explain scientifically observed natural patterns and they are testable by the scientific method. However, there are some additional criteria which enhance the scientific value of a theory or paradigm. These additional criteria are often labeled as the "pragmatic criteria." This label is misleading, because the entire scientific method seeks pragmatic confirmations of the empirical consequences of hypotheses. The most important additional criteria are:

1. Logical Coherence. There should be a very high degree of logical coherence among a theory's established hypotheses, and among a paradigm's established theories. If there are logical contradictions between established principles of scientific knowledge, then those contradictions should be eliminated. All other things being equal, a scientific theory with fewer or no internal logical contradictions is a better scientific theory. Some scientific revolutions occur because scientists notice such contradictions and resolve them by dramatically changing previously established principles of knowledge. For example, Einstein developed portions of his theory of relativity by noticing that the constant speed of light (required by electrodynamics) is incompatible with the principle of additive velocities (required by classical mechanics). Einstein resolved this incoherence by replacing or revising much of classical mechanics. You can visit a website about science and "Thought Experiments".

2. Predictive Power. There should be a very large number of predictions to be made by a theory or paradigm. There are two benefits to this "predictive power". First, more predictive power means a better chance of becoming highly confirmed (or proven false). Second, all of other things being equal, a theory that successfully explains a much wider range of natural phenomena can be more reasonably persuasive than a theory that can explain only a small range of phenomena. You can visit a website about science and "Predictive Power".

3. Physical Unification. There should be very wide range of phenomena unified by a theory or paradigm. For a while, a science may treat one natural pattern very differently than another pattern, but then a new theory arrives which shows how these two patterns are really the same pattern. For example, Newton's theories of motion unified the motions of heavenly objects with the motions of earthly objects, treating their patterns as all obeying the same basic laws of motion. Another example is how James Clerk Maxwell's theory of electrodynamics showed how visible light is the same sort of photon radiation as all other forms of radiation. Most of modern physics heavily depends on this persuasive power of physical unification. You can visit a website about science and "Unification".

4. Ontological Simplicity. There should be a very small number of entities postulated by a theory or paradigm that are required to explain a wide range of phenomena. If two theories can both explain the same phenomena, yet one theory postulates far fewer entities, that theory appears to be more believable than the other. The value of simplicity is probably rooted in ordinary practical common sense: the simpler explanation is more believable, will probably suffer from fewer internal contradictions, and will be easier to prove false. The more complex theory appears too ad-hoc and too well-designed  -- arousing the suspicion that the theory was really designed to fit all evidence now available and to prevent its falsification. Also, rationality itself seeks unity behind diversity (e.g. "everything is made of atoms" or "there is ultimately only one natural force causing everything"). You can visit a website about science and "Simplicity".

The "demarcation problem" is the philosophical problem of justifying a reasonable standard to judge whether an explanation (a hypothesis, or a theory or a paradigm too) is a scientific explanation, or not scientific at all (such as pseudoscience or religion or mythology, etc.). There is an easy way to seemingly solve the demarcation problem: justify an account of scientific method (such as the six-step method described above) and then declare that only hypotheses that are testable by this scientific method qualify as scientific. Sounds easy -- but the most difficult part is precisely justifying an account of scientific method. Philosophers and scientists have been trying since Aristotle to accomplish this. A philosophical account of scientific method must explain (1) how hypotheses that survive trial by this method are more likely to be true, and also (2) how hypotheses that do not survive trial by this method are more likely to be false. The first task is the philosophical problem of explaining why highly confirmed scientific hypotheses have a better chance of accurately describing real entities -- you can proceed to the issue of "Scientific Realism and Truth". The second task is the philosophical problem of explaining why disconfirmed scientific hypotheses probably fail to describe real entities. Unless the scientific method can at least help scientists to judge which hypotheses are false, science cannot be any help deciding what reality is like.

Pierre Duhem (1861-1916), the French philosopher and scientist, argued that hypotheses about Type III or Type IV entities -- the "unobservables" -- can never be proven false. His argument started from the fact that no hypothesis can really be tested by itself, apart from the larger theory of which it is a part. Duhen, and other philosophers since, are concerned with the worry that a hypothesis cannot really be tested and hence never proven false (or never shown to be probably true either). It is true that a hypothesis cannot be properly tested without also using some other items of established scientific knowledge, as mentioned in step five, "experiment". For example, an experiment to test for whether water exists on Mars depends on already established knowledge about how to detect water using scientific instruments: the signs that water will give from a distance, the effects of those signs on an instrument, how the instrument works, etc. When an experiment is designed, the logical form of the experimental inference is deductive. In the following table, the deduction on the left illustrates the reasoning when a prediction is experimentally confirmed, while the deduction on the right illustrates the reasoning when a prediction is experimentally disconfirmed.

In the second example of a disconfirmation, the premises 1-4 cannot all be true. At least one of them must be false, assuming no experimental error. But which one? Remember, all scientific knowledge is fallible. Just because the purpose of the experiment is to try to test the new hypothesis, this does not mean that only the hypothesis can be shown to be wrong. Any premise, any knowledge, used in the design and execution of the experiment can be held responsible for being false. Reasoning only says that at least one, and perhaps more than one, of the four premises in this inference must be false. Reason and logic cannot identify which is false. Of course, if scientists decide to trust the other premises rather than the new hypothesis, a disconfirmation makes it reasonable for scientists to conclude that the hypothesis is proven false and the entity does not exist. But this reasonable conclusion depends on the scientist's decision to trust prior knowledge. It is also possible for scientists to protect the new hypothesis by deciding that one of the other premises must be false instead. Karl Popper's philosophy of falsificationism demanded that scientists must always discard the new hypothesis, but neither logic nor actual scientific practice requires this drastic approach. Scientific method must permit scientists to make judgments about which parts of theories should be changed. Because any single hypothesis needs assistance from other parts of a larger theory in order to be tested [what is now called the "Duhem-Quine thesis"], only entire theories really confront experimental evidence. Theories will gradually change over time as scientists selectively judge which parts require modification or replacement in order to continue to make successful predictions. Likewise, paradigms will gradually over time as its component theories are modified. Sometimes, a paradigm is suddenly replaced by a new paradigm -- in what is called a scientific revolution -- as philosopher and historian of science Thomas Kuhn has described.

Because scientists can protect some hypotheses by discarding others, some philosophers have claimed that experiments do not really determine the validity of a new hypothesis, and hence they doubt whether either scientific evidence or reason really decides which hypotheses or theories should be believed. If the evidence does not control which hypotheses should be believed, what really makes scientific theories any different from other kinds of theories that some people want to believe in, such as religions or superstitions or pseudosciences such as astrology? Science begins to look like any other cultural belief system, sustained over generations by mere persuasion or coercion, and not reason or truth. If this view of science is correct, then there is no rational demarcation between science and any other belief system, and hence scientific "knowledge" should not be permitted to have any greater authority than any other belief system like magic or myth. This view, which is a kind of "Relativism," was championed by the 20th century philosopher Paul Feyerabend. Feyerabend was not a skeptic, since he believed that plenty of fallible practical knowledge is available to people -- so much knowledge, in fact, from so many cultural sources, that it is impossible to find some objective supreme methodological standard to judge which sort of cultural knowledge is superior to any other.

Relativism about scientific knowledge is closely related to scientific anti-realism, so we can proceed to "Scientific Realism and Truth".

All material on this website is copyright 2007 by John R. Shook