Home » Academics » Nerds' Corner

Abduction and the Limits of Science, Part 3: The Falsification of Competing Theories

Welcome to My Third Article in This Series … Please Fasten Your Chinstraps

In part 1 of this series, I argued that scientists rely inextricably on abductive reasoning (i.e., "inference to the best explanation") to formulate and defend their explanations of the world; then I argued that abductive reasoning is formally identical to the logical fallacy Affirming the Consequent. In part 2, I argued that, if we use either simplicity or explanatory power (or a combination of the two) as the criterion by which we pick the best among competing theories, we risk embracing a theory that isn't true.

In this article, I'm going to address the claim that we can determine the best among competing theories by testing them — i.e., by subjecting them to experimentation and rejecting any theory found to be false.

Donald Sutherland Again

Part 2 of this series made repeated reference to a story in which you and a friend had an afternoon of Frisbee thwarted by a flooded field and were subsequently debating three competing theories for why the field was so wet:

• The rain theory, according to which the field had been flooded by heavy rain.
• The sprinkler theory, according to which the sprinklers had been on all night.

• The Donald Sutherland theory, according to which Donald Sutherland and Gary Wasserfeld flooded the field with a garden hose.

In the story, these theories were weighed in relation to the following observations:

• Everything for miles around the field is a little wet.
• Everyone is indoors.
• There are dark rain clouds nearby.
• The parking lot above the field has considerably less water on it than the field does.
• The drainage ditch below the field also has considerably less water in it than the field.

For the sake of referencing accessible examples, we'll use this scenario again in this article.

Falsification and Competing Theories

So, here's the idea: We can't formulate our theories without relying on a logically fallacious form of reasoning, and we can't guarantee the truth of our theories by retaining only the simplest or most powerful of them. But perhaps we can subject them to tests. Perhaps we can figure out what our theories entail and run experiments to see if any of them are false — and if we find one that is false, then we can discard it and invest more confidence in the remaining theories. Or, if we find that all of them are false, we can formulate new theories.

For example, regarding the flooded field story above, if the sprinklers ran all night, the city water district would have a record of it. We can test the sprinkler theory, then, by checking with the city water district. If it doesn't have a record of the sprinklers running all night, then we know that the sprinkler theory is false. We can also subject the rain theory and the Donald Sutherland theory to similar tests.

To borrow a metaphor from Imre Lakatos, visualize a theory in one corner squaring off against an observation in the other corner. If the observation contradicts the theory, then the theory has been falsified, simple as that. The theory must then be rejected.¹

But this image is misleadingly over-simplistic, as the history of science and a little reflection bear out.²

Auxiliary Hypothesis

For starters, as a matter of historical fact, theories that find themselves up against observations that appear to falsify them are often successfully protected by what Lakatos calls auxiliary hypotheses — hypotheses that appeal to acknowledged facts about the world in order to resolve the tension between the theory and the observation that appears to falsify it.³

In our story about the field, the fact that there is very little water in the parking lot appears to falsify the rain theory. But the rain theory is well protected by the auxiliary hypothesis that the parking lot isn't very wet because it's above the field — the water ran out of the parking lot, down the hill and onto the field, the hypothesis goes. This auxiliary hypothesis resolves the tension between the theory that heavy rain flooded the field and the observation that the nearby parking lot isn't very wet.

Theories Upon Theories Upon More Theories

An even larger problem for the possibility of falsifying competing theories arises when we reflect on the practical consequences of the fact that, in order to determine whether or not a given observation could falsify a given theory, we must rely on other theories — or, at least, other beliefs we can easily call into question.

For example, in order to determine that the city water district's records could falsify the sprinkler theory, we must assume that the city water district keeps records; we must assume that it keeps records of the sprinklers on this exact field; we must assume that these records are accurate; we must assume that employees at the city water district would tell us the truth if we asked them; we must assume that we would know how to interpret the data if we look at them ourselves; and so on, and so forth.

Click to expand image

If an employee at the water district reported not having any record of the sprinklers running all night, instead of concluding that the sprinklers didn't run all night (i.e., that the sprinkler theory is false), we could conclude that the records weren't accurate, or that the employee lied to us, and so on. What's worse, any attempt to test these conclusions would involve the exact same problems as our attempt to test the sprinkler theory in the first place. If we ran up against an observation that appeared to falsify our conclusion that the employee lied to us (for example), instead of rejecting it, we could reject one of the assumptions its apparent falsification rests upon.

Click to expand image

Or, for a better example — one that falls within the natural sciences — consider J.B.S. Haldane's famous remark that evolutionary theory would be falsified by fossilized rabbits in the Precambrian, the geological time period preceding the Cambrian Explosion.⁴

A rabbit in the Precambrian would be a serious problem for evolutionary theory because of the astronomical improbability of organisms as complex as rabbits evolving from single-celled organisms fast enough to be in the Precambrian. But now suppose there really is a fossilized rabbit in the Precambrian, and suppose that a bunch of paleontologists stumble upon it. What would happen?

First, how would they be certain they were looking at a fossil? Second, how would they be certain they were looking at a fossil of a rabbit (or another organism of equal complexity, since that's really the issue)? Third, how would they be certain that it was really in the Precambrian?

None of these questions could be answered conclusively just by looking. Our paleontologists would have to rely on sophisticated instrumentation and abstruse biological, physical, chemical and geological theories to form any conclusion at all about a biologically complex fossil supposedly in the Precambrian or something in the Precambrian that's supposedly a biologically complex fossil.

If their instruments aren't working correctly, our paleontologists might form the wrong conclusion; if they use these instruments incorrectly, they might form the wrong conclusion; if they misunderstand the theories by which they're interpreting their data, they might form the wrong conclusion; if they do understand these theories but misapply them, they might form the wrong conclusion; then there's the possibility that these theories aren't correct in the first place.

If an actual paleontological study produced the conclusion that a fossilized rabbit was found in the Precambrian, really, what do you think would happen?

Would our paleontologists reject the entire edifice of evolutionary theory, or would they conclude something less significant had gone wrong somewhere else? Perhaps their instruments weren't working correctly, or perhaps they were using them incorrectly. Or, perhaps they misapplied one of the theories somewhere. By picking one of these options (or any number of other options) they could conclude that what appeared to be a fossil rabbit in the Precambrian was really something else — not really a fossil, not really a fossil of a rabbit (or any other complex organism), not really in the Precambrian, and so on. And by doing so, they could resolve the logical tension between evolutionary theory and their observation.

The lesson is this: In general, any time observation O seems to falsify theory T, if we choose, we can point to one of the assumptions that allows us to view O as falsifying T and reject it instead of T. In fact, if we're desperate enough, we can even reject the assumption that we aren't hallucinating — we can conclude that we're seeing things and thereby reject O instead of T.

The Moral of the Story

Noting that theories are arrived at via abductive reasoning, which commits the logical fallacy Affirming the Consequent, Karl Popper concluded that any acceptance of a theory must involve a free decision.⁵ But Popper thought theories could be objectively falsified and therefore objectively rejected, leaving behind only theories with a high likelihood of being true.

What Lakatos points out is that rejecting a theory also involves a decision. Theories are only contradicted by observations against a backdrop of assumptions about the bearing of those observations on the theory. Since we can always choose to reject one or more of these assumptions instead of the theory itself, we are never forced to view on observation as contradictory to the theory.⁶

This isn't to say that science isn't capable of producing true theories or that choosing between scientific theories is like choosing between flavors of ice cream. Theories are arrived at via abductive reasoning and, as we saw in "A Fancy Riddle and Three Kinds of Reasoning, Part 2" in teammate 3's successful attempt to figure out the color of his own X, abductive reasoning can be profoundly helpful in leading us to the truth. But it's not infallible. Dogmatism, bigotry, political agendas, false worldviews, conspiracy theories, selfish personal aspirations, intellectual laziness and a million other things can all get in the way.

C O F F E E  S H O P

---

If we must rely on other theories to falsify theories, can you ever really know what's correct?

Join the discussion!

The moral of the story, then, is this: Contrary to the view that the scientist is mechanical and detached — that the method does the work rather than the person — the success of the scientist involves an irreducible ethical component. Scientific theories are formulated, accepted and rejected via abductive reasoning; in the hands of scientists or scientific communities with significant character flaws, however, abductive reasoning is unlikely to be reliable. The reliability of science itself, then, rests at least in part upon the moral condition of its practitioners.