PhilSci-L01b: Popper and Lakatos

In short

What makes a theory scientific?

Popper: a scientific theory sticks its neck out. It says certain things will not happen, so the world has a chance to prove it wrong. A theory that fits anything that could possibly happen is not science, however much evidence it seems to have.

Lakatos: nice idea, but real scientists almost never drop a theory just because an observation contradicts it, and they are often right not to. So stop judging single theories against single observations. Judge a whole line of research over time, and ask whether it keeps predicting new things that then turn out to be true.

On the mock exam this is two questions worth 10 marks each: Q2 on Lakatos and Q3 on Popper and Einstein.

The question both of them are answering

Astrology has a stupendous mass of observations. Psychoanalysis has a clinic full of case studies. Physics has experiments. All three can point at evidence, so “it has evidence” cannot be what makes something science. Popper calls astrology’s method pseudo-empirical: it appeals to observation and experiment, but does not meet scientific standards.

Finding what does make the difference is called the problem of demarcation. A demarcation criterion is a necessary condition a field has to meet to count as science: fail it and you are out.

The criterion is not about truth. Science is often wrong, and a pseudoscience can get lucky. Popper wanted a mark of scientific status that holds whether or not the theory turns out true.

And it is not an armchair question. In Lakatos’s examples, the Catholic Church banned Copernicus’s theory as pseudoscience in 1616, and the Soviet Communist Party declared Mendelian genetics pseudoscientific and persecuted its defenders, such as the geneticist Vavilov. Each decision rested on some demarcation criterion, so getting the criterion right has ethical and political stakes.

Part 1. Popper: a scientific theory must be able to fail

The theories that explain everything

Vienna, 1919. Popper is surrounded by four fashionable theories: Einstein’s relativity, Marx’s theory of history (history is driven by class struggle and heads for a workers’ revolution), Freud’s psychoanalysis, and Alfred Adler’s individual psychology (a breakaway from Freud that explains behaviour through feelings of inferiority).

Their fans loved how much they explained: once you “got” Marxism, every newspaper confirmed it; once you “got” Freud, every patient did. Popper’s favourite illustration:

The two men and the child

One man pushes a child into a river to drown it. Another man drowns trying to save a child.

Freud explains the first by repression (a desire pushed out of awareness) and the second by sublimation (the same desire turned into something admirable). Adler explains both by inferiority feelings: the first man needed to prove he dared commit a crime, the second needed to prove he dared to rescue.

Two opposite behaviours, and both theories explain both of them with total confidence.

Popper had worked in Adler’s social guidance clinics in Vienna, and once reported a case to him that did not look very Adlerian. Adler explained it in terms of inferiority feelings without difficulty, without having seen the child. Asked how he could be so sure, he said “because of my thousandfold experience”. Popper’s reply: “And with this new case, I suppose, your experience has become thousand-and-one-fold.”

That is the whole argument in one joke. If every possible case can be interpreted in the light of the theory, then fitting one more case confirms nothing. So the constant fit that impressed Freud’s and Adler’s fans was, for Popper, a weakness: if no possible observation could count against a theory, the world never gets to test it, and its fit with the evidence is worth nothing.

The contrast: Einstein and the eclipse

General relativity says gravity bends light. So a star whose light passes close to the sun should appear slightly shifted away from the sun, by an angle the theory calculates in advance. Normally the sun’s glare hides such stars, so you photograph them during a total eclipse, photograph the same stars at night months later, and compare.

Eddington’s expeditions did exactly this in 1919 and found the shift Einstein predicted.

What impressed Popper was that the prediction could have failed. If the shift had not been there, the theory would simply have been refuted. The theory forbids results that would have looked perfectly normal before Einstein, such as no shift at all. That is a risky prediction, and it is the opposite of Adler explaining every possible child.

The criterion: falsifiability

Popper's demarcation criterion

“The criterion of the scientific status of a theory is its falsifiability, or refutability, or testability.”

To be ranked as scientific, a theory must be capable of conflicting with possible observations.

Popper’s conclusions, boiled down:

• Confirmations are cheap. You can find them for nearly any theory if you look. They only count when they come from a risky prediction: one where, without the theory, you would have expected the opposite.
• A good theory is a prohibition. It forbids certain things from happening, and the more it forbids, the better.
• Irrefutability, which people take for a virtue, is a vice.
• A real test is an attempt to falsify. If the theory survives a serious attempt, it is corroborated.

“The more it forbids, the better” sounds backwards until you try it on the weather. “Tomorrow it will either rain or not rain” forbids nothing. It can never be wrong, and it tells you nothing. “Tomorrow 12mm of rain falls in Amsterdam between two and three in the afternoon” forbids almost everything. It is easy to refute, and if it survives, you have learned a lot. Saying more and risking more are the same thing.

Why falsify instead of verify

Take “all swans are white”. A million white swans do not prove it, because the next swan might be black. No finite number of observations logically gets you to a claim about all swans: this is Hume’s problem of induction.

One black swan, on the other hand, does refute it. The logic is modus tollens: the theory $T$ implies an observation $O$; we observe that $O$ is false; so $T$ is false.

$\frac{T\to O\neg O}{\therefore \neg T}$

So logic lets you refute a universal law and never lets you prove one. Popper builds science on the half that works: make bold guesses (conjectures), try hard to knock them down (refutations), and keep whatever survives, for now.

Corroborated does not mean "probably true"

Corroboration is a report on the past: the theory has survived our attempts to kill it so far. Popper argued that the probability of any theory, scientific or pseudoscientific, given any amount of evidence, is zero: a universal law covers infinitely many cases, and any body of evidence is a vanishing fraction of them. So “well corroborated, therefore probably true” is exactly the inductive step he refuses to make.

Popper’s scorecard

Einstein’s relativity: scientific. It made risky predictions that could have failed.

Astrology: not scientific. Astrologers were so impressed by confirmations that they ignored unfavourable evidence, and made their prophecies vague enough to fit anything. “In order to escape falsification they destroyed the testability of their theory.”

Marx’s theory of history: not scientific, but it used to be. Its early predictions about the coming revolution were testable, and they failed. Followers then re-interpreted both the theory and the evidence until the two agreed again, which made it irrefutable. Popper’s name for this kind of rescue, the conventionalist twist, comes in the next section.

Freud and Adler: not scientific, and never were. No conceivable human behaviour could contradict them. Marxism took a risk and lost; psychoanalysis never took one.

Rescuing a theory, and when it is allowed

The Marxists’ move works for any theory. You can save a refuted theory by adding an extra assumption whose only job is to explain away the bad result, or by re-interpreting the theory until it no longer clashes. Popper calls this rescue the conventionalist twist or conventionalist stratagem (the theory ends up held true by convention, whatever the evidence says). It always works, and that is the problem: it rescues the theory “only at the price of destroying, or at least lowering, its scientific status”.

The key word is ad hoc, Latin for “for this purpose only”. A change is ad hoc when its only job is to absorb the evidence that would have refuted the theory. It has no independent support and makes no new testable predictions of its own.

But Popper does not say every failed prediction kills the theory on the spot, and the modus tollens schema above needs one correction. A prediction never comes from the theory alone. In the eclipse test it also relied on auxiliary assumptions: that the telescopes worked, that the photographic plates were not warped, that the data were reduced properly, that the stars were identified correctly. So the real logic is:

$\frac{(T\wedge A)\to O\neg O}{\therefore \neg (T\wedge A)}$

Modus tollens only tells you that something among the premises is false: the theory or one of the auxiliaries $A$. Logic does not say which. This is the Duhem-Quine thesis, covered properly in L03. Popper grants this logical point (he did not accept Quine’s stronger version, the web of belief in L03). He also admits that accepting an observation report is, in the end, a decision reached by agreement among scientists.

So Popper allows you to blame the measurement instead of the theory, on one condition: you need independent reasons to reject the measurement: say, a later check finds a telescope mirror warped by the heat, something you can establish without assuming the theory. “Otherwise my theory would be refuted” is not an independent reason. With independent reasons, it is ordinary scientific correction. Without them, it is an ad hoc rescue.

"Then I am very sorry for the dear Lord" (mock Q3)

In 1953 Einstein was asked what he would have done if Eddington had not found the shift. He said he would have been sorry for the dear Lord, because the theory is correct.

Would Popper accept that? No, if the measurement was sound. Relativity earned its scientific status by forbidding certain results, such as no shift at all. If one of them had come in, the theory would be falsified and should be given up. Insisting it is correct anyway is dogmatic, the same move as the Marxists.

Could it ever be acceptable? Yes, if there were independent reasons to think the measurement was wrong, because by the Duhem-Quine thesis the blame can always fall on an auxiliary assumption. But Einstein gives no such reason, only his confidence in the theory. So on Popper’s terms, no.

Not science does not mean nonsense

Popper often gets lumped in with the logical positivists of L01 (also called logical empiricists: Carnap and the Vienna Circle). He was arguing against them. Mock Q1 asks about the verifiability criterion, and one sentence of this contrast is worth adding to that answer.

Carnap and the Vienna Circle: verifiability. Apart from the truths of logic and mathematics, a statement is meaningful only if observation could in principle settle whether it is true: “the meaning of a statement is its method of verification”. The line runs between meaningful and meaningless. Metaphysics lands on the meaningless side: pseudo-statements that only look as if they say something.

Popper: falsifiability. A statement is scientific if it could conflict with possible observations. The line runs between science and non-science, and says nothing about meaning. Non-scientific ideas can be meaningful, important, and the seed of later science: most scientific theories started life as myths (the Greek philosopher Empedocles proposed evolution by trial and error). Popper even thinks Freud and Adler may contain real psychological insight, just not in a testable form.

The logic also runs in opposite directions. Verification needs observations to establish universal laws, which the swans showed is impossible, so strictly applied it makes laws of nature meaningless. Falsification only needs observations to be able to knock them down.

Which is better? Falsifiability survives the problem that sinks verifiability, but has weak spots of its own: Duhem-Quine, and Lakatos’s point (next) that scientists rightly keep refuted theories.

Part 2. Lakatos: judge the research programme over time

Clearing the ground

Lakatos keeps Popper’s goal, an objective line between science and pseudoscience, and attacks the way Popper draws it. First he disposes of three older answers.

• Not: science is what people strongly believe. Knowledge is not strong belief. People have believed absurd things with total commitment. The hallmark of science is scepticism even towards your favourite theory: “blind commitment to a theory is not an intellectual virtue: it is an intellectual crime”. A theory’s scientific value is objective, independent of anyone’s mind. A claim can be pseudoscientific though everybody believes it, and valuable though nobody does.
• Not: science is what the facts prove. No law follows validly from finitely many facts (the swans again).
• Not: science is what the evidence makes highly probable. If Popper is right that every theory has probability zero, theories are equally improbable as well as equally unprovable.

That leaves Popper’s falsificationism.

The objection: scientists have thick skins

Lakatos puts Popper’s criterion like this: a theory is scientific if you are willing to say in advance which observation would make you give it up (a potential falsifier), and pseudoscientific if you refuse.

Notice what that does: the same theory would be science for someone willing to name a falsifier and pseudoscience for someone who refuses. So Popper’s test really sorts scientific method from non-scientific method, rather than theories. That sets up Lakatos’s own move: judge the record of a line of research over time, whatever its supporters’ attitude.

Then the real objection, the one that makes this falsificationism naive (Lakatos’s label for Popper’s view as he reads it: a theory must be given up once it clashes with an observation scientists have accepted). It “ignores the remarkable tenacity of scientific theories. Scientists have thick skins.” When an observation contradicts their theory, they do not abandon it. They do one of two things:

1. call it an anomaly and invent an auxiliary hypothesis to explain it, or
2. if they cannot explain it, ignore it and work on something else.

Lakatos’s sharpest line: ask Newtonian physicists which experiment would make them give up Newton, and some would be as stuck for an answer as some Marxists. So Popper’s test, applied to real science, cannot tell Newton from Marx. And the famous stories of a crucial experiment (one test that settles the matter for good) killing a theory were, Lakatos says, “fabricated long after the theory had been abandoned”.

All theories are born refuted

Every major theory has faced contradicting results from day one. Lakatos uses “refuted” loosely here: a well-founded result that seems to clash with the theory.

• When Newton published, everyone knew his theory could not properly explain the motion of the moon.
• The physicist Kaufmann “refuted” Einstein’s special relativity (the 1905 theory, before the general relativity of the eclipse) in the very year it was published.

Both “refutations” later turned out to be spurious: the moon’s motion is consistent with Newton after all, and Kaufmann’s results came from experimental error. A scientist who dropped a theory at the first clash, as naive falsificationism demands, would have thrown away two theories that went on to succeed, because of results that were wrong. That is Lakatos’s point: holding on is sometimes the right call. (Popper’s verdict on Einstein’s “dear Lord” line stays as above: without independent reasons to doubt the measurement, holding on is dogmatic. Keep the two verdicts apart in Q3.)

So “still unrefuted” cannot be the mark of science. Every programme, at every stage, lives with unsolved anomalies.

The fix: the research programme

Lakatos reads Thomas Kuhn (next lecture, L02) as drawing a radical conclusion from the same facts: if refutation does not drive theory change, theory change is irrational, more like a religious conversion. (Kuhn himself denied this.) Lakatos refuses to go there. If that were right, there would be no line between science and pseudoscience, no objective way to tell progress from decay, and no objective standard of intellectual honesty. He wants a middle way: keep Popper’s objective standards while accepting the thick-skinned behaviour just described.

His answer is to change what gets judged. Not a single hypothesis, but a research programme: a series of theories that develop over time around a shared core. Newton’s physics, for instance, is a long series of models of planets, moons and comets, all built on the same laws. A “heuristic” in what follows just means a rule for where to look and what to do.

Parts of a research programme (Newton as the example)

1. Hard core. The basic assumptions that define the programme. For Newton: the three laws of motion plus the law of gravitation. Scientists in the programme agree not to touch it. That makes it irrefutable by methodological decision: evidence could logically count against it, but the scientists agree never to let it. Give up the core and you have left the programme.
2. Protective belt. Everything around the core that can be adjusted: auxiliary hypotheses, initial conditions, theories of the instruments, models of the particular system. The belt is what actually meets the evidence, and it is what gets changed when something goes wrong.
3. Negative heuristic. The rule saying where not to aim: never direct modus tollens at the hard core. Blame something in the belt instead.
4. Positive heuristic. The programme’s problem-solving machinery, a partly worked-out plan for what to do next: which models to build, which anomalies to tackle in which order, which mathematical tools to develop. It “digests anomalies and even turns them into positive evidence”.

The belt is Lakatos’s answer to Duhem-Quine. Since a failed prediction never says where the fault is, the programme decides in advance that it goes in the belt.

A planet misbehaves: both heuristics at work

A planet does not move quite as Newton’s theory says it should. The negative heuristic says: gravitation is not on trial. The positive heuristic says what to check instead: assumptions about how the atmosphere bends light, about light in magnetic storms, and hundreds of others. The astronomer may even invent an unseen planet and calculate its position, mass and velocity.

This really happened, twice, though Lakatos does not name the cases. Uranus strayed from its predicted orbit, Le Verrier and Adams predicted a new planet, and Neptune was found in 1846. Mercury’s orbit did not fit either, the same move predicted a planet called Vulcan, and it never turned up; general relativity explained Mercury instead.

Neptune is an anomaly turned into positive evidence, which is exactly what the positive heuristic is for.

Progressive or degenerating

Having a hard core, a belt and a heuristic does not make something science. Lakatos counts Newton’s physics, relativity and quantum mechanics as research programmes, but also Marxism and Freudianism, each with a stubbornly defended core. And every one of them has anomalies. So he asks which direction a programme is moving in.

Lakatos's criterion

• A programme is progressive when it predicts novel facts: facts nobody expected, or that rival programmes said were impossible, which are then found. In the key’s words, it leads to new discoveries. The theory runs ahead of the facts.
• A programme is degenerating when its theories are made up only to accommodate facts already known. The theory lags behind the facts and runs to catch up.

Progressive examples:

• Halley’s comet. Using Newton’s theory, Halley watched a short stretch of a comet’s path and predicted it would return in seventy-two years (Lakatos’s figure, fine to use in the exam; the real period is about 76), at a specific spot in the sky. It came back as predicted, with Newton and Halley long dead.
• Newton’s unseen planets. Newtonians predicted the existence and exact motion of small planets nobody had seen yet.
• Einstein’s eclipse. The distance between two stars would be different when measured during an eclipse than when measured at night. Nobody had thought to check that before. It is the same test Popper admired, now counted as a novel fact.

The degenerating example is Marxism. “Has Marxism ever predicted a stunning novel fact successfully? Never!” Its bold predictions all failed, and each failure was explained after the event:

• Workers would get ever poorer. They got richer, and a theory of imperialism was produced to explain why.
• The first socialist revolution would come in the most industrialised country. It came in backward Russia.
• Socialist societies would have no revolutions. Berlin (1953), Budapest (1956) and Prague (1968) followed.
• Socialist countries would not clash. Russia and China split.

Lakatos’s summary: trivial confirmations are worthless (every falling stone “confirms” Newton), and refutations are not fatal, because every programme lives in “a permanent ocean of anomalies”. What counts are dramatic, unexpected predictions that come true, and a few are enough.

This makes scientific status a historical question about a track record, and the answer can change: a degenerating programme can recover, a progressive one can run out of steam. “Progressive” and “degenerating” describe a direction of travel, so they apply to a programme over time, never to a single hypothesis.

Patching a theory: Popper and Lakatos compared

• Popper judges a patch at the moment it is made. Is there an independent reason for it, or a new test it could fail? If its only job is to absorb the refuting evidence, it is ad hoc, and the theory loses scientific status.
• Lakatos says patching the protective belt is normal, constant and necessary; it is what the positive heuristic is for. Belt changes are not suspect by default. He asks two things. Does the patch predict anything new? If not, that step is degenerating. If it does, he waits to see whether those predictions come true. The programme is judged by its record of such steps.

On the textbook cases they agree. Le Verrier’s invented planet made a new testable prediction (point the telescope here), so it is not ad hoc for Popper either, and it paid off when Neptune was found. The Marxist theory of imperialism predicted nothing new, so both reject it.

They split on what happens next. Take Vulcan: a legitimate, testable patch whose prediction failed. For Popper, Vulcan’s absence refutes the Vulcan hypothesis, so Mercury’s orbit counts against Newton’s theory again; unless a new, independently testable explanation turns up, the theory stands refuted. For Lakatos one failed prediction does not sink a programme if others come true; he looks at the whole record. They also split on a programme that sits on its anomalies for decades: Popper calls that dogmatic, Lakatos waits for the track record.

How revolutions happen

When there are two rival programmes and one is progressing while the other is degenerating, scientists tend to move to the progressive one. That is what a scientific revolution is, and it is rational. Lakatos adds three qualifications and one jab at Kuhn:

• Loyalty is allowed. It is not dishonest to stick with a degenerating programme and try to turn it around, as long as you keep its record public.
• No instant rationality. A new programme deserves patience; it can take decades to get going. There is no rule that tells you the exact moment to give up.
• No refutation without a better theory. There is no Popperian quick kill by counterexample. You abandon a programme when there is somewhere better to go.
• Against Kuhn: revolutions are not sudden, irrational changes of vision. “On close inspection both Popperian crucial experiments and Kuhnian revolutions turn out to be myths.”

The price of all this: verdicts only come in hindsight. You can say a programme has degenerated for thirty years; you cannot tell a scientist today that she must abandon it.

Popper, Kuhn and Lakatos side by side

Kuhn’s terms (paradigm, puzzle-solving, crisis) are explained in L02.

	Popper	Kuhn	Lakatos
Judges	One theory	Paradigm and community	Programme over time
Mark of science	Falsifiability	Puzzle-solving tradition	Novel predictions
Contrary evidence	Refutes, unless independent reasons	Anomaly, tolerated until crisis	Absorbed by the belt
Tenacity	Suspect if ad hoc	Normal and useful	Rational
Theory change	Reject what fails a serious test	Revolution, no algorithm	Leave a degenerating programme
Crucial experiments	Decisive tests	Rare, reconstructed later	A myth

Lakatos charges Kuhn with having no real demarcation criterion; Kuhn’s own answer is the puzzle-solving tradition.

What to be able to write

• Popper’s criterion, and what made Eddington’s test a risky prediction.
• Marxism (falsified, then rescued) versus psychoanalysis (never testable).
• Blaming the measurement is fine only with independent reasons, because by Duhem-Quine the fault could lie anywhere.
• Verifiability sorts meaning; falsifiability sorts science.
• Lakatos’s four components by name: hard core, protective belt, positive heuristic, negative heuristic (any order is fine).
• Progressive predicts novel facts; degenerating only accommodates known facts. Halley’s comet; Marxism’s theory of imperialism.
• Lakatos against Popper: thick skins, all theories born refuted. Against Kuhn: change is rational.
• Is Lakatos’s criterion objective? For: whether a predicted fact was found is a public matter, independent of anyone’s belief. Against: there is no rule for how long to wait before calling a programme degenerating, so it cannot tell anyone what to do today, and a verdict can be reversed later.

Full model answers for mock Q2 and Q3 are in the first two exam questions below.

Exam questions

Mock exam Q2 (10 points, half a page): Imre Lakatos analyses the development of science in terms of "scientific research programmes". Describe the components of research programmes, and explain what, according to Lakatos, scientific progress consists in.

Key points: Hard core; protective belt of auxiliary hypotheses; positive heuristic; negative heuristic; progressive leads to new discoveries, degenerating only accommodates known facts.

Follow the official key: “(1) Hard core; (2) protective belt with auxiliary hypotheses; (3) positive heuristic; (4) negative heuristic. Progressive programme: leads to new discoveries. Degenerating programme: only accommodates known facts.”

1. Unit of appraisal: a research programme, a series of theories over time sharing a common structure, rather than an isolated hypothesis.
2. Hard core: the fundamental assumptions that give the programme its identity (Newton: three laws of mechanics plus the law of gravitation), held irrefutable by methodological decision.
3. Protective belt: auxiliary hypotheses, initial conditions, theories of the instruments. It meets the evidence and is what gets modified when an anomaly appears.
4. Positive heuristic: the problem-solving machinery, a partially articulated plan for developing the belt (which models to build, which anomalies to digest). Example: the Newtonian checks atmospheric refraction, light in magnetic storms, may invent an unknown planet and calculate its position, mass and velocity.
5. Negative heuristic: never direct modus tollens at the hard core; absorb anomalies in the belt.
6. Progress: cannot be “unrefuted”, since all programmes grow in an ocean of anomalies. Progressive = predicts novel facts (Halley’s comet returning after seventy-two years; Einstein’s eclipse star separations). Degenerating = theories fabricated only to accommodate known facts, lagging behind them (Marxism, theory of imperialism cooked up after the event).
7. Revolutions: scientists desert a degenerating programme for a progressive rival.

Losing marks: calling the hard core unfalsifiable in principle (it is protected by decision), or treating every belt modification as ad hoc.

Mock exam Q3 (10 points, half a page): Popper praised Albert Einstein's theory of general relativity as a truly scientific theory, citing the example of Eddington's expedition. This expedition set out to measure the shift of a distant star during a solar eclipse. Einstein's theory states that large masses, like the sun, curve space-time. Thus, incoming light rays will follow a curved path around the sun and will appear in a different spot as seen from the earth. The theory makes a very precise prediction about this shift. In 1953, however, Einstein was asked how he would have responded to a measurement outcome that did not show this shift: "Then I am very sorry for the dear Lord [Eddington], for the theory is correct." Would, and-or could, this be acceptable for Popper? Explain your answer.

Key points: No, if the measurement was correct: theory falsified, Einstein dogmatic; perhaps the measurement was wrong; Duhem-Quine: blame can always go elsewhere; acceptable only with independent reasons for rejecting the measurement (no ad hoc hypothesis).

Follow the official key:

1. No, assuming the measurement was correct: the theory is falsified and has to be rejected. Einstein shows a dogmatic attitude. Eddington’s test is exactly the risky prediction that makes relativity scientific: it is incompatible with results which everybody before Einstein would have expected.
2. Einstein’s move is the Marxists’ move: re-interpreting to escape refutation, which rescues the theory only at the price of destroying or lowering its scientific status.
3. However, perhaps the measurement was not correct. The prediction comes from the theory plus auxiliary assumptions (telescope optics, photographic plates, data reduction, star identification). Given the Duhem-Quine thesis, scientists can always put the blame elsewhere and retain the theory.
4. For Popper this is acceptable only if there are independent reasons for rejecting the measurement result, reasons other than that the theory would otherwise be refuted. Without them it is an ad hoc hypothesis (a conventionalist stratagem).
5. Einstein as quoted gives no independent reason, only confidence in the theory, so the reply is not acceptable.

Losing marks: stopping at “no”; letting Duhem-Quine excuse Einstein without the independent-reasons condition; saying Popper thinks a falsified theory is proved false and gone forever.

State Popper's criterion of demarcation and use it to classify Einstein's relativity, astrology, Marx's theory of history and psychoanalysis. Why do the last two fail in different ways?

Key points: Falsifiability as the criterion; Einstein scientific (risky prediction); astrology fails (vague prophecies); Marxism testable, falsified, then rescued by a conventionalist twist; psychoanalysis never testable.

• Criterion: “the criterion of the scientific status of a theory is its falsifiability, or refutability, or testability.” A theory must be capable of conflicting with possible, conceivable observations.
• Einstein: scientific. The 1919 eclipse prediction was risky: had the shift been absent the theory would simply have been refuted.
• Astrology: fails. Astrologers were impressed by confirmations and unimpressed by unfavourable evidence, and made prophecies vague enough to explain anything: “in order to escape falsification they destroyed the testability of their theory”.
• Marxism: fails, but not originally. Early versions (the coming social revolution) were testable and in fact falsified; followers re-interpreted theory and evidence to agree, a conventionalist twist that made it irrefutable.
• Psychoanalysis (Freud, Adler): simply non-testable from the start. No conceivable human behaviour could contradict it (the two men and the child).
• So Marxism was once science and was destroyed by its defenders; psychoanalysis never made a risky claim.

What is Lakatos's main objection to Popper's falsificationism, and how does the methodology of scientific research programmes answer it without following Kuhn?

Key points: Scientists have thick skins (tenacity); all theories born refuted; Kuhn’s irrationalism rejected; unit becomes the research programme; novel prediction replaces refutation.

• Objection: Popper’s criterion “ignores the remarkable tenacity of scientific theories. Scientists have thick skins.” Faced with contrary facts they invent a rescue hypothesis for a “mere anomaly”, or ignore it and work on other problems.
• Crucial-experiment stories are “fabricated long after the theory had been abandoned”. A Newtonian asked what would make him abandon Newton “would have been exactly as nonplussed as are some Marxists”, so Popper’s test fails to separate Newton from Marx.
• All theories are born refuted (lunar motion and Newton; Kaufmann and relativity), so “unrefuted” cannot mark science.
• Not Kuhn: Lakatos reads Kuhn as concluding theory change is irrational (Kuhn denied this); then there is no demarcation, no distinction between progress and decay, no objective standard of honesty.
• Answer: shift the unit from the theory to the research programme (hard core, belt, heuristics) and the test from refutation to novel prediction: progressive programmes predict novel facts, degenerating ones accommodate known facts.
• Revolutions are rational: scientists join the progressive programme, with no instant rationality.

Popper and Lakatos both discuss modifying auxiliary hypotheses to save a theory from contrary evidence. Compare how each judges such a modification, using an example.

Key points: Popper judges a patch when it is made: independent support or new testable content, else ad hoc and status lowered; independent reasons required to blame the measurement; Lakatos: belt modification normal and necessary, the step is degenerating if it predicts nothing new, otherwise judged by whether its predictions come true; they agree on the textbook cases (invented planet fine, theory of imperialism not) and split on failed testable patches and long-ignored anomalies.

• Popper: the modification is judged when it is made, by its independent support and testable content. If introduced only to absorb the refutation (ad hoc: for this purpose only, adding no new testable consequences) it is a conventionalist stratagem that lowers scientific status. Blaming the measurement is allowed only with independent reasons.
• Lakatos: belt modification is normal, constant and necessary, since that is what the positive heuristic does. A patch that predicts nothing new is a degenerating step for the programme; one that does is judged by whether its predictions are then found, as part of the programme’s record.
• Example: an invented planet to explain an orbital anomaly makes a new testable prediction, so it is acceptable to both; the Marxists’ theory of imperialism predicted nothing new, so both reject it. They differ on what happens next. A testable patch whose prediction fails (Vulcan) is itself refuted for Popper, and the theory faces the anomaly again unless a new testable explanation turns up; for Lakatos one failure does not sink a programme with other successes. On a programme that sits on anomalies for decades, Popper calls it dogmatic and Lakatos waits for the track record. (Background example, which Lakatos does not name: the planet postulated to explain Uranus’s orbit was found as Neptune, while the same move for Mercury gave Vulcan, never found.)

Losing marks: treating every protective-belt modification as automatically ad hoc.

Contrast Carnap's verifiability criterion with Popper's falsifiability criterion: what each demarcates, its verdict on metaphysics, and the direction of its logic.

Key points: Verifiability demarcates meaningful from meaningless; falsifiability demarcates science from non-science; metaphysics nonsense for Carnap, possibly valuable for Popper; universal laws cannot be verified but can be falsified by modus tollens.

• Verifiability (Carnap, Vienna Circle): a statement is meaningful if its truth can in principle be determined by observation or experiment; “the meaning of a statement is its method of verification”. Demarcates meaningful from meaningless. Metaphysics is nonsense: pseudo-statements.
• Falsifiability (Popper): a statement is scientific if capable of conflicting with possible observations. Demarcates science from non-science. Non-scientific statements are meaningful, sometimes important, sometimes the ancestors of science (Empedocles).
• Logic: strict (conclusive) verification wants observations to establish universal laws, which is impossible; falsification wants observations able to destroy them, which modus tollens permits.

Compare Popper, Kuhn and Lakatos on the unit of appraisal, the demarcation criterion, and how theory change happens.

Key points: Unit: theory, paradigm, programme; criterion: falsifiability, puzzle-solving tradition, novel prediction; contrary evidence: refutes unless independent reasons, anomaly, absorbed by belt; theory change: reject what fails a serious test, revolution, desertion of a degenerating programme.

• Unit: Popper, the individual theory; Kuhn, the paradigm and its community; Lakatos, the research programme (a series of theories over time).
• Demarcation: Popper, falsifiability (a potential falsifier specified in advance); Kuhn, a puzzle-solving tradition (Lakatos charges him with having no real criterion); Lakatos, progressive versus degenerating (novel predictions).
• Contrary evidence: Popper, refutes the theory unless there are independent reasons to doubt the evidence; Kuhn, an anomaly tolerated until anomalies pile up into crisis; Lakatos, hits the belt and is digested by the positive heuristic.
• Tenacity: Popper, suspect when it takes an ad hoc rescue; Kuhn, functional in normal science; Lakatos, rational and often correct.
• Theory change: Popper, reject what fails a serious test; Kuhn, a revolution, no algorithm decides it (Lakatos charges him with irrationalism); Lakatos, desertion of a degenerating programme for a progressive rival, rational but no instant rationality.
• Crucial experiments: Popper, decisive tests; Kuhn, rare and reconstructed later; Lakatos, a myth.

Flashcards

What is a demarcation criterion?

A necessary condition a discipline must satisfy to count as genuine science; failing it makes the discipline non-scientific.

Why did Popper see the explanatory power of Marxism, Freud and Adler as a weakness?

They seemed to explain practically everything in their fields, so confirmations were everywhere and therefore worthless.

What is the point of Popper's example of one man drowning a child and another dying to save one?

Freud (repression, sublimation) and Adler (inferiority feelings) each explain both opposite behaviours, so neither theory takes any risk.

Why did Popper find Eddington's eclipse test of Einstein impressive?

The risk: had the predicted shift been absent, the theory would simply have been refuted.

When should a confirmation count, according to Popper?

Only when it results from a risky prediction: without the theory, we would have expected an event incompatible with it.

Why is it true for Popper that "the more a theory forbids, the better it is"?

Forbidding more gives it more ways to be wrong, so it is riskier and more testable.

What is Popper's criterion of demarcation?

The criterion of the scientific status of a theory is its falsifiability, or refutability, or testability.

What does Popper call evidence from a serious but unsuccessful attempt to falsify a theory?

Corroborating evidence.

Why can no number of observed white swans verify "all swans are white"?

No finite conjunction of instances deductively entails a universal statement (Hume’s problem of induction).

How can one observation falsify a universal law?

By modus tollens: from $T\to O$ and $\neg O$, infer $\neg T$.

How did astrologers escape falsification, according to Popper?

By making prophecies vague enough to explain anything, which destroyed the testability of their theory.

What is the difference between how Marxism and psychoanalysis fail Popper's criterion?

Marxism was testable and in fact falsified, then rescued; psychoanalysis was never testable at all.

How did Marxists rescue their theory of history from refutation, according to Popper?

They re-interpreted both theory and evidence to agree, a conventionalist twist that made it irrefutable.

What does "ad hoc" mean for a modification that saves a theory from refutation?

For this purpose only: its only motive is to absorb the refuting evidence, adding no new testable consequences.

What does an ad hoc rescue cost a theory, according to Popper?

It escapes refutation only by destroying, or at least lowering, its scientific status. The move itself is always available.

When may a Popperian blame the measurement instead of the theory?

Only with independent reasons for rejecting the measurement, beyond the fact that the theory would otherwise be refuted.

Does Popper hold that any negative result must kill the theory?

No: a prediction never comes from the theory alone, so blame may in principle fall on an auxiliary assumption or the measurement.

What is the difference between what Carnap's verifiability and Popper's falsifiability demarcate?

Verifiability separates meaningful from meaningless; falsifiability separates science from non-science.

Does Popper think non-scientific theories are meaningless?

No: they can be meaningful and important, and myths can develop into testable science.

How does Lakatos formulate Popper's criterion of demarcation?

Scientific if one will specify in advance a potential falsifier; pseudoscientific if one refuses.

What is Lakatos's main objection to Popper's falsificationism?

It ignores the remarkable tenacity of theories: scientists have thick skins and do not drop a theory because facts contradict it.

What do scientists do when facts contradict their theory, according to Lakatos?

Invent a rescue hypothesis for a “mere anomaly”, or, failing that, ignore it and work on other problems.

Why does Popper's test fail to separate Newton from Marx, according to Lakatos?

Some Newtonians, asked to name a falsifier, would be as nonplussed as some Marxists.

Why does Lakatos refuse what he reads as Kuhn's conclusion, that theory change is irrational?

Then there is no demarcation, no distinction between progress and decay, and no objective standard of honesty.

What is Lakatos's unit of appraisal?

The research programme: a series of theories over time sharing a common structure.

What are the four components of a Lakatosian research programme?

Hard core, protective belt of auxiliary hypotheses, positive heuristic, negative heuristic.

What is the hard core of the Newtonian research programme?

The three laws of mechanics plus the law of gravitation.

Why is the hard core of a research programme irrefutable, for Lakatos?

By methodological decision of its practitioners, not by logic or meaning.

What does Lakatos call the part of a research programme that absorbs anomalies?

The protective belt of auxiliary hypotheses.

What does Lakatos call the rule never to direct modus tollens at the hard core?

The negative heuristic.

What is Lakatos's positive heuristic?

The problem-solving machinery: a partial plan for which models to build next and which anomalies to digest.

How does a Newtonian handle a planet that does not move as predicted, according to Lakatos?

Changes the protective belt: checks atmospheric refraction or light in magnetic storms, or invents an unknown planet. Gravitation is never on trial.

Why can "still unrefuted" not mark science, for Lakatos?

All theories are born refuted: every programme has unsolved anomalies at every stage.

Which "refutations" of Newton and Einstein does Lakatos cite?

Lunar motion against Newton, Kaufmann’s results against relativity. Both were later shown spurious.

When is a research programme progressive, for Lakatos?

When it predicts novel facts, undreamt of or contradicted by previous or rival programmes: theory runs ahead of the facts.

When is a research programme degenerating, for Lakatos?

When its theories are fabricated only to accommodate known facts: theory lags behind the facts.

Which comet prediction does Lakatos give as a novel fact of the Newtonian programme?

Halley’s comet: from a short stretch of its path, Halley predicted its return in seventy-two years, and it returned.

How did Marxists explain rising working-class living standards, according to Lakatos?

With a theory of imperialism, cooked up after the event.

What is the difference between how Popper and Lakatos judge a modification that saves a theory?

Popper judges it when it is made: independent support or a new test? Lakatos judges the record: did novel predictions follow?

Is a protective-belt modification automatically ad hoc, for Lakatos?

No: it is normal, constant and necessary; it counts against the programme when it predicts nothing new, or when its new predictions keep failing.

Why does Lakatos judge an invented planet and the Marxist theory of imperialism differently?

The planet predicted something new that was then found; the theory of imperialism only explained a known fact after the event.

How do scientific revolutions happen, according to Lakatos?

Scientists desert a degenerating programme for a progressive rival.

What does Lakatos mean by saying his methodology offers no instant rationality?

Budding programmes deserve leniency, may take decades to progress, and no rule says when to give up.

What does Lakatos say Popperian crucial experiments and Kuhnian revolutions turn out to be?

Myths: what normally happens is that progressive programmes replace degenerating ones.

How does Lakatos contrast Kuhn's account of scientific revolutions with his own?

Kuhn, as Lakatos reads him: a sudden, irrational change in vision. Lakatos: rational desertion of a degenerating programme for a progressive one.

What is the difference between Popper's and Lakatos's attitude to scientists' tenacity?

Popper: suspect, since an ad hoc rescue lowers scientific status. Lakatos: rational and often correct, since all theories are born refuted.

What is the cost of Lakatos's position?

Verdicts are only retrospective: you cannot tell a scientist today he must abandon a programme.

Links

• Course: Course overview
• Same week: PhilSci-L01 - Introduction and Logical Empiricism for Carnap, verificationism and the Vienna Circle that Popper is arguing against
• Next: PhilSci-L02 - Kuhn on Scientific Practice for paradigms, normal science and the irrationalism charge Lakatos is trying to avoid
• Related: PhilSci-L03 - Under-determination for the Duhem-Quine machinery behind the ad hoc rescue and behind the protective belt
• Source: Curd & Cover, Philosophy of Science: The Central Issues, chapter 1. Popper, “Science: Conjectures and Refutations”, pp. 3-10 (week 1 reading). Lakatos, “Science and Pseudoscience”, pp. 20-26 (listed under week 2 in the syllabus, taught with the Popper material). Chapter introduction, pp. 1-2.
• Primary texts: Popper, Conjectures and Refutations (Routledge and Kegan Paul, 1963), pp. 33-39, originally a 1953 Peterhouse lecture published as “Philosophy of Science: A Personal Report”. Popper, The Logic of Scientific Discovery (1934, English 1959), Appendix vii for the zero-probability argument. Lakatos, Philosophical Papers, vol. 1 (Cambridge, 1977), pp. 1-7, originally an Open University radio lecture broadcast 30 June 1973.