Science suffers from what is known as the demarcation problem--the problem of demarcating science from anything else. What we'll find during this exploration is that almost any criteria or groups of criteria we propose will either demarcate too loosely by including things that we presume are not science or it will demarcate too tightly and exclude things we presume are science, sometimes doing both at the same time. Many of you think you already know the answer (falsification, right?), but when we take it seriously, we find problems. Overall, the demarcation problem is somewhat the same problem as pornography, and we rightly find it frustrating when people say they only know it when they see it.
To a degree, the demarcation problem doesn't matter. A scientist needs a clear demarcation as much as baseball pitcher needs physics, both the scientist and the ball player can continue doing their respective activities without a clear understanding of demarcation or physics. And lately, established scientific disciplines have been winning the courtroom battles against established pseudo-scientific disciplines such as creationism. Most legal cases which have established the non-scientific merits of intelligent design have relied on falsification which, as we'll see, is neither a necessary or sufficient criteria. But it is a problem, an interesting one at that, and one which captured the imagination of some of the more influential minds of the 20th century. Demarcation matters even though it likely will remain an intractable problem.
For the record, I am not making an assertion that astrology is a science, only that it is very difficult to say exactly what science is.
One of the more effective attempts to answer the demarcation problem developed from a group called the Vienna Circle who put forward a philosophy of logical positivism. Under logical positivism, a meaningful statement was one which could be verified. If it could not be verified, then it was not meaningful and--needless to say--not scientific. A popular example to mock was Martin Heidegger's dictum, ""Das Nichts nichtet." Loosely translated, this means "The nothing nothings." Even when you figure out what Heidegger was talking about, this claim cannot be verified, and positivists would conclude that it's not a meaningful statement. The challenge to verificationism is that it cannot itself be verified. How does one verify the verification criterion to indicate that it's a meaningful statement? Early positivists felt that it was a matter of time before the idea could develop adequate support but this never really happened. There is also a problem of universality: while we may be able to say, "this swan is white", it's not clear that we can verify that "All swans are white for now and all time." The challenge here is that we want to say things like "Helium is and always will be an inert chemical element" and we are denied the ability to verify that. Another difficulty is that evidence is cheap. If every time I walk under a ladder, I get injured, that doesn't indicate that walking under a ladder leads to injuries, in spite of my ability to verify it. One solution to these problems was presented in Karl Popper's attempt at the demarcation problem: falsifiability.
Under falsifiability, we need only attempt to indicate that a particular statement could be false. So long as it is not shown false, the statement may endure. Once it is contradicted, it may be abandoned. Falsifiability is an alluring proposition: it encourages boldness and even a degree of heroism--a form of "what doesn't kill me makes me stronger" model for scientific propositions. However, when we look at what scientists do, abandoning a falsified theory is seldom the reaction. Consider the scientists of the Ptolemaic school who modeled the heavens by placing earth in the center. If earth is in the center and the sun and planets move in perfect circles around it, there were planets which moved forward, then backward, and then forward again across the night sky in contradiction to the basic theory. This anomaly didn't result in the abandoning of Ptolemy: epicycles were proposed which corrected the model to fit the observable data. And before we suspect that these types of modifications are what differentiates a pseudo-science from science, we repeated this tactic in the 1800s to good effect. Long after we accepted the Copernican model of a sun-centered solar system, when planets were not in the location that Newtonian physics predicted they would be, we did not abandon Newtonian physics--we instead posited the existence of another planet which inevitably was discovered and named Neptune. No science is free from anomalies and it's not very useful to say that anomalies alone suggests that a discipline is a pseudo-science.
And this is the problem with falsifiability: it's quite simply not what scientists do. Falsifying a hypothesis doesn't automatically imply a false hypothesis--it may only suggest an incomplete hypothesis. A good example of this is the Copernican model that the sun was the center of the solar system. When it was proposed, it was a less reliable model to Ptolemy's model because it came before the observations of people such as Kepler who improved the Copernican model until it was better than Ptolemy's. Should we then conclude that Copernicus was not engaged in science? Similarly, Galileo's model was inferior to the contemporary Kepler model; was the church right in preventing the spread of his pseudo-science? Clearly, we reject these ideas, so the falsification criterion is too strong, excluding work that we'd like to think of as science.
But we like the falsification criterion, so maybe we can salvage it. As demonstrated, it's not a necessary criterion for science, but is it a sufficient idea for science? Is the fact that something is falsifiable (whether it's shown false or not) a sufficient justification to label an idea scientific? Not really. Astrology is the whipping boy of the pseudo-scientific world but assuming falsifiability is our criterion, then astrology must certainly be a science. A bad science, perhaps. But a science nonetheless. It makes positive claims constantly--just open today's Times and find your horoscope. My horoscope today says, "Now is the best time for you to stand up for something you truly believe in." Maybe I have evidence that today is the worst day like if I'm on crutches--it's hard to stand up at all (I say tongue firmly in cheek). Nonetheless, the statement is capable of being falsified. The fact that a horoscope may be accurate only to the statistical probability of flipping a coin may be merely coincidental, it may just be an incomplete science and further understanding may improve its predictive power. Similarly, I doubt we want to reject meteorology simply because the weatherman has the same reputation as the tarot card reader. If we want to exclude astrology and include meteorology, falsifiability may not be the way to do it. Much as we may like the idea of falsifiability, it isn't the simple answer to the demarcation problem.
Models and Theories: Kuhn and Quine
If we're having difficulty in developing an abstract criterion, perhaps we can approach this more pragmatically. Perhaps science is merely what scientists do and as long as we go along with that, we're operating in a science. This was the insight of two philosophers: Thomas Kuhn and W.V.O. Quine, both approaching the answer in different ways. Let's look at Kuhn first. Kuhn was foremost a historian of science and he presented his case in The Structure of Scientific Revolutions, which is an engaging book that is easy to recommend to non-philosophers. Kuhn adopted a phrase from linguistics ("paradigms") to describe how a science operates. A paradigm is a model for understanding how a particular field works. Within that paradigm, there is an amount of normal science which can be performed: under the Newtonian paradigm, we can go out and look at how fast apples accelerate when they fall to the ground, we can apply it to planets passing through the heavens, and so on. However, from the application of Newtonian principles, some anomalies may result. The normal reaction is to find solutions to these anomalies within the paradigm (like Ptolemy's epicycles) but after a critical point, a new paradigm may emerge which displaces the old paradigm. In the modern case, anomalies at the subatomic level have given rise to a somewhat incompatible quantum mechanics. Kuhn called these displacements "paradigm shifts". This model of paradigm shifts implies that a science is based on both predictive as well as explanatory qualities. If a paradigm resolves an anomaly in an unsatisfactory manner, and a new paradigm accounts for it better, holding on to that old paradigm is non-scientific. So, creationism once may have been useful as a model, but the evolutionary model better explains the differentiation of the species and can predict future differentiation more accurately.
Several problems emerge from the paradigmatic model. For one thing, it requires competition. To go back to an earlier example, I'm not sure there is a science which can better predict on which days it's better for me to stand up for something I truly believe in. Therefore, astrology may provide the superior paradigm for such predictions. For that matter, I'm not sure what theory could compete with Time Cube, and yet I have a hard time accepting it as a science. Also, the reliance on explanatory power seems insufficient. Long before gravity fields were understood, Newton was able to do a great deal of research on gravity. He couldn't explain at all how the moon was affected by earth's gravity, just that it was. It would be difficult to suppose that "God's hand pushes the moon in its orbit" to be the superior scientific paradigm for Newton's time. So, at the same time, Kuhn's paradigmatic criterion includes astrology and excludes Newton, so it's simultaneously too strong and too weak.
In Two Dogmas of Empiricism--arguably one of the most important papers of the 20th century--W.V.O. Quine proposed that there was no distinction between synthetic statements and analytic statements. All statements ultimately are synthetic. Let's break through this jargon to see what this means and its impact on science. Especially since the 1700s, philosophy has recognized a difference between synthetic and analytic statements. A synthetic statement is one which we can demonstrate to be true: Some bachelors eat ramen. An analytic statement is one which is true by virtue of its meaning: All bachelors are unmarried. We have to do research to find out if the synthetic statement is true, while we can think out the truth of analytic statements. Quine argues that this is a presumption. At one time, a witch was defined as an insane woman. Nowadays, a witch is someone who practices wicca. Meanings change over time such that a definition we once held before is different now. And this is something more than mere semantic drift, it's a problem of science. If we doggedly hold on to the idea that a witch is an insane woman, we can be presented with counter-examples that demonstrate the problem with the definition, such as a male witch or an insane woman who is not a witch. Our meaning has changed, not necessarily because society has moved, but because experience forces us to re-evaluate our definitions. "1 + 2 = 3" seems like an analytic statement, but it also is subject to the trial of empirical observation. If we found in nature that "1 + 2 != 3", we would attempt to build a math based around that fact. What matters for Quine is the web of belief and how all of the propositions which we hold relate to one another within that web of belief. Meaningfulness only has relevance insofar as things are difficult to dislodge within that web. So, I have an idea of red to which many different experiences attach: the color of blood, communism, that it possesses a particular wavelength, etc. Therefore red is a highly meaningful concept. Grue, however, is less meaningful since it attaches only to Zork and a philosophical argument that posits its existence between green and blue depending on time. If someone asserts that "red does not exist" it will be difficult for me to dislodge the notion of red from my web of belief and I'm more apt to reject the "red does not exist" proposition. But for grue, I may happily abandon the idea of grue.
The result of this for science is that science must be regarded as a holistic exercise placed upon the entire web of belief, that all scientific statements are subject to an individual's ability to integrate it into their web of belief. In other words, there is little demarcation between science and pseudo-science if a person's web of belief can stretch far enough to accommodate the propositions of both the proposed pseudo-science and observation. As an example of this, it's possible to still be a flat-earth proponent if one dismisses all of the evidence as elaborate hoaxes. To be fair to Quine, he presents pragmatic strategies to limit what type of qualifications we should apply such as the law of parsimony--more often known as Occam's razor. We probably use this term all the time, but to define it, it's the law that indicates that its unnecessary to include more than the minimum which is needed to establish a fact. So, if "evolution" sufficiently explains the differentiation of the species, then "evolution (with God)" is unnecessarily complicated by the factor of "with God". As a needless aside, most people mis-characterize Occam's Razor as "the simplest explanation is best" but the real Razor cuts better: it only removes unnecessary complexity rather than arbitrarily preferring simplicity. But it's not entirely clear that parsimony must necessarily be a fact of nature nor that it must find a strong presence in any web of belief. Certainly, it's easier to navigate within our web of belief if we employ parsimony but Quine can only leave it to individual experience and consideration.
Needless to say, Quine doesn't present a satisfactory demarcation. Quite the contrary, he makes it possible for all pseudo-science to be a science.
Against Method: Feyerabend
Going back to Karl Popper and falsification, one of his students explicitly rejected any demarcation at all. Paul Feyerabend proposed what he referred to as epistemological anarchism. This is both a sociological position as well as philosophical. Any rule which indicates how science must operate is repressive. To put it in its most argumentative form, unless we do exactly the activities in the Western European model, we disparage the many valuable contributions which do not (perhaps cannot) come from that tradition. We reject rain dances, for example, because they do not bring rain but neglect the importance of the rite to the people from that community--a feature which science cannot explore. As a result, Feyerabend was able to detect traces of elitism and even racism with the scientific approach. Instead of dispassionately explaining the world, it's used to re-create it in the scientist's own image, much as both late 19th century German and French scientists attempted to find out who has bigger brains: Germans found evidence of larger Teutonic brains, while their Gallic counterparts claimed the same. In the Americas, experiments consistently showed the diminished capacity of Blacks, Native Americas, and Southeastern Europeans. We may think ourselves enlightened right now because we have a superior notion of science, but maybe that's only because we are in the majority. However, on a more philosophical level, a method of demarcation limits our ability to accept and understand data. Feyerabend reminds us that, at a powerfully fundamental level, information wants to be free.
The above outlines the major theories and proponents of various demarcation positions presented through the 20th century. However, there are other arguments that occur from time to time which also have problems.
The first is that science only deals with the natural as opposed to supernatural phenomena. This begs the question in assuming there is a useful distinction here, let alone what one or the other means. In a simplified example, we live in dimensions 1, 2 and 3. Perhaps God and souls exist in dimension 4, 5 and 6 and future science can show evidence for the existence of these. Even if it be the case that there is no interaction between these two sets of dimensions, it may be true nonetheless even if we cannot, like Russell's teapot, detect anything there. While one could question the usefulness in positing a being in dimension 4, 5, and 6 with whom we cannot interact, we may still be discussing natural phenomena. Further, if we want to argue that we can only do science on 1, 2, and 3, we begin to run afoul of verificationism. And finally, we frequently propose extra dimensions whenever the math requires it for the more demanding calculations in physics; it seems arbitrary to exclude any other phenomena from also functioning in a similar manner. So, the natural versus supernatural distinction isn't as clear as we presume.
Another idea is that science must show progress, but this is clearly not satisfying. Chemistry, by and large, hasn't really progressed too much in the last 100 years. We've filled in corners and supplied some missing details but it is not a robust and active enterprise at the level of physics. Meanwhile, the discovery of new planets does get incorporated in the latest astrological tables, so there appears to be some evidence of astrological progress.
Perhaps we should require that a clear mechanism be proposed in order to say that things are a science. This, too, fails to account for Newton's work in gravity when he neglected to describe the mechanism under which gravity operates. It also bars Darwin from being able to claim his work as scientific because he was unfamiliar with genetics.
Peer review, then? The esteem of the scientific community? Is that what constitutes science? If anything is a pseudo-science, the theories of Trofim Lysenko should qualify, regardless of its popularity in the Soviet universities of its day, so a sociological basis doesn't seem re-assuring.
The motives of the proponents? Can we dismiss the creationist because of their belief in God? In that case, every scientist who works for money has ulterior motives as well, and so it is dissatisfying to dismiss their efforts on the basis of their motives.
"Science must propose laws," some people may argue. If so, then classifying between domain, kingdom, phylum, class, order, family, genus and species become non-scientific enterprises.
On the whole, this sums up why demarcation is a problem. There are good reasons why we might want to dismiss astrology as pseudo-science but that's a far different thing than putting it into actual words which at the same time preserves our feelings with regards to other things that we want to consider science. In some ways, today's fights against creationism and intelligent design in the courts are using yesterday's arguments to discredit it, even though those criteria have themselves been discredited. We may not feel comfortable with that realization, but I never promised you a comforting answer. All I promised was an explanation of that original "probably".