There are different kinds of human knowledge, and it's useful to sort them out in order to understand what's going on in science. We'll consider the following terms: Fact, Deductive Inference, Inductive Inference, Hypothesis, Multiple Working Hypotheses, Theory, Evidence, Ockham's Razor, Natural Law, and Paradigm. The first few may be a little boring, but hang in there - things get more interesting further down.

fact - a truth known by actual experience or observation. The hardness of iron, the number of ribs in a squirrel's body, the existence of fossil trilobites, and the like are all facts.
    Is it a fact that electrons orbit around atomic nuclei? Is it a fact that Brutus stabbed Julius Caesar? Is it a fact that the sun will rise tomorrow? None of us has observed any of these things - the first is an inference from a variety of different observations, the second is reported by Plutarch and other historians who lived close enough in time and space to the event that we trust their report, and the third is an inductive inference after repeated observations (see below).

deductive inference - a conclusion based on reasoning from accepted premises. Consider a somewhat loaded example: "The earth is a spherical body, a sphere by definition has equal radius in all directions, and therefore the radius of the earth is equal in all directions." We've taken two reasonable premises and reached a conclusion from them with. In this example, the conclusion is slightly flawed because the first premise is only an approximation: the earth is really a prolate spheriod (it bulges toward the equator because of its rotation). Deductive inference can be a powerful tool when the premises are correct, but the example illustrates what happens when one of the premises is flawed.
     (There's no logical fallacy in the example; the Web provides pages and more pages on fallacies of logic.

inductive inference - a conclusion based on repeated observation of fact. Drop a particular kind of ball on a particular floor from a particular height numerous (n) times, and you can, by induction from those examples, make an inference and a prediction about what will happen the next time you drop the ball. However, your prediction is not a fact, in that you won't know by actual observation the result of the n+1th drop until it has happened.

hypothesis - a testable proposition explaining the occurrence of a phenomenon or phenomena, often asserted as a conjecture to guide further investigation. After dropping the ball from one height several times, you may think that dropping it from a greater height will lead to a different response, and you may predict that different response. Your prediction is a hypothesis, and you can test it by changing the height of the drop and observing the result. At that point, you'll have done an experiment to test your hypothesis.
     One important word in the definition of "hypothesis" is "testable". If a proposition contains some component that defies testing or detection, the proposition is not a scientific hypothesis. To continue with the example above, the proposition that "a ball of a particular type dropped from a particular height onto a particular surface will bounce a particular distance" would be a scientific hypothesis, because we can drop that type of ball from that height onto that kind of surface. However, the proposition that "a ball when dropped will bounce a distance determined by the undetectable influence of an undetectable entity" is not a scientific hypothesis because, if we can't detect the entity or its influence, we can't test whether that entity is responsible, or if that entity even exists.

multiple working hypotheses - a method of research where one considers not just a single hypothesis but instead multiple hypotheses that might explain the phenomenon under study. Many of these hypotheses will be contradictory, so that some, if not all, will prove to be false. However, the development of multiple hypotheses prior to the research lets one avoid the trap of narrow-mindedly focusing on just one hypothesis.
     The Web has more on Multiple Working Hypotheses

theory - a coherent set of propositions that explain a class of phenomena, that are supported by extensive factual evidence, and that may be used for prediction of future observations. For our rather trivial example, a theory would emerge only after a huge number of tests of different kinds of balls at different heights. The theory would try to explain why different kinds of balls bounce differently, and it ought to be useful in predicting how new materials would behave if dropped as balls in the same way.

Scientists have produced lots of familiar theories:
- - - Copernicus's theory of the heliocentric solar system,
- - - Newton's theory of gravity,
- - - Einstein's theory of relativity, and
- - - Darwin's theory of natural selection are a few.
Each of these theories draws on huge numbers of facts:
- - - observations of the passage of the sun and planets for the heliocentric theory;
- - - the behavior of the planets, of projectiles, and rather famously of apples for the theory of gravity, and
- - - the existence and location of fossils, as well as the modern distribution and reproduction of organisms, for the theory of natural selection.

    Some people dismiss a given scientific idea with "That's just a theory". They're right - all science can provide is theories. However, those theories have proven quite useful to all of us. Most of us won't step off the top of a building because of the results predicted by Newton's theory of gravitation - and yet it's just a theory. NASA and other space agencies launch space craft to distant planets on the basis on Newton's theory of gravitation and Copernicus's theory of the heliocentric solar system - and yet they're just theories. It's instructive to remember that Copernicus was required by the authorities of his time to preface his work as just a series of "hypotheses", and not even as a "just a theory".

evidence - the physical observations and measurements made to understand a phenomen. Perhaps equally important is what's not evidence: theories aren't evidence, and the opinions of even the most learned scientists aren't evidence.

     Note that evidence is one of the critical underpinnings of a theory (see above). A good scientist or observer of science periodically asks, "What do we think we know, and why do we think we know it?" The answer to the second part should be some sort or sorts of evidence, as defined in the previous paragraph.

Ockham's Razor (a.k.a. Occam's Razor) - a philosophical statement developed by William of Ockham, an English monk who died in 1349. His orginal statement was " non sunt multiplicanda entia praeter necessitatem", or "assumptions are not to be multiplied beyond necessity". In thinking about our hypotheses and theories discussed above, perhaps the best modern statement of Ockham's Razor is

     If we have multiple hypotheses that can explain a thing, we ought to reject the hypotheses that involve agents or processes for which we have no evidence (bearing in mind how we've defined evidence above). Let's say we've observed a large rock in an otherwise featureless area. One of our hypotheses for the presence of the rock might be that an ancient giant threw it there, and another hypothesis might be that glacial ice transported it there. Ockham's Razor tells us to reject the first and retain the second for further consideration, because we have no evidence for ancient giants - they are an unsupported assumption. We do have modern evidence that flowing ice can transport large stones.

     It's not true to say, and William of Ockham wouldn't have said, that "the simplest explanation is the best explanation". The explanation that all matter consists of earth, air, fire, and water was simpler than the explanation involving the modern periodic table of elements, but it was wrong. An even better example is Devil's Tower in Wyoming. Native American legend tells that this landform originated when a huge bear tried to climb a steep mountain to attack an Indian maiden, and the bear's claws scraped the sides of the mountain away. That's a simple explanation, but it assumes the existence of huge bear capable of clawing the sides of a mountain to leave something like Devil's Tower. The bear is an unsupported assumption that would cause most of us to reject the Native American story as anything other than folklore or myth.

natural law - a term rarely used today, at least by scientists thinking about what they're saying. Nineteenth-century science presumed that it could arrive at immutable, absolutely true, universal statements about nature, and these were to be "natural laws". Newton's ideas about gravitation, for example, were considered the "laws of gravity". To continue that example, in the twentieth century Einstein's theory of relativity showed that Newton's ideas needed correction in some cases. Thus it became apparent that it would be wisest to treat even our most trusted ideas, of which Newton's had been one, as theories rather than absolute laws.

paradigm - a way of thinking, commonly so ingrained in people's behavior or thought that they aren't even aware of it. If a theory presents a broad understanding of a phenomenon or problem, a paradigm may be the mindset that causes us to think that the theory matters one way or the other. In a non-scientific example, the Domino Theory was an explicit statement of what many Americans thought would happen if a single country in a given region (e.g. southeast Asia) had a communist government. The implicit paradigm was that the US ought to be, and had to be, involved in a global struggle with another superpower over what kind of political system would dominate the world's governments.

     In science, a major example of a change in paradigms was the change from Scholasticism to Modern Science, roughly around AD 1600. Scholasticism, which assumed that answers to questions about nature could be deduced from ancient texts and philosophical principals, gave way to the modern view of science where induction from accumulated evidence is (or should be) the underpinning of theories. (We talked about this more in the previous section .) When Galileo was threatened by church authorities with torture for his claim that the earth orbits the sun, Galileo and his accusers were not only at odds about an astronomical theory. They were also arguing, if unwittingly, because they were using two very different paradigms: the churchmen were using scholasticism, and Galileo modern science.
     Incidentally, the fact that we only call today's way of thinking "modern science", rather than a distinct name, is a sign of how the users of a paradigm generally don't recognize what they're using. Another change of paradigms came when scientists, or at least some scientists, realized the futility of the search for natural laws, as discussed above.

     This distinction between paradigm and theory can be seen in the earth sciences. For example, the earth sciences have seen major theories of earth movment and mountain building come and go. Into the early 1900s, a static earth was the largely unquestioned model. Continental Drift, the theory of continents plowing through passive oceanic crust, was a controversial theory accepted in the early to middle parts of this century by many if not most geologists in the Southern Hemisphere, and by many in the Northern Hemisphere. It has been supplanted today by the widely accepted Plate Tectonic theory (in which the oceanic crust has a dynamic rather than passive role).
     Implicit behind all this changing theory has been the paradigm that the major goal of the earth sciences should be a theory to account for crustal movement, mountain building, and processes deep in the earth. We may now be going through a paradigm shift: we increasingly expect that the earth sciences should be mostly concerned about cycling of elements and changing conditions at the earth's surface. The paradigm isn't changing our theories, but it's changing our focus from one theory (or group of theories) about one problem to another theory (or group of theories) about another problem.