Basic concepts: falsifiable claims.

Here's another basic concept for the list: what does it mean for a claim to be falsifiable, and why does falsifiability matter so much to scientists and philosophers of science?

Actually, it's not just falsifiable claims that the science crowd cares about, but also falsifiable theories. Let's start with claims because they're easier.

Claims make assertions about how things are (or were, or will be, or could be under different circumstances). Here's a claim from my post on arguments:

Britney Spears is from Mars.

A falsifiable claim is one for which there is some observation (or set of observations) we could make that would show us that the claim is false. If we did make this observation, essentially we'd have to conclude either that the claim in question was false, or that our observation was a bad one. Trying to hold on to both the truth of the claim and the goodness of the observation would saddle us with a mind-blowing contradiction.

Is there some observation (or set of observation) that could potentially demonstrate that the claim, "Britney Spears is from Mars," is false? Here's one: Keeping Ms. Spears' mother under constant surveillance for the duration of Britney's birth and making careful note of which planet she was on during this time. If the planet of Britney's birth is Earth, we've falsified the claim.*

Of course, at this stage of the game, Britney's birth is years in the past, so our most practical options for observations that might falsify the claim probably have less to do with surveillance and more to do with collecting other kinds of evidence about the planetary location of Britney's birth -- hospital records, witness accounts, etc. The observation we're looking for that might falsify the claim needn't be a direct observation of the opposite ("Dude, I witnessed the birth of Britney Spears, and it wasn't on Mars!"). The observation of other states of affairs that are incompatible with the claim is enough to do the job.**

"Britney Spears is from Mars" is not only a falsifiable claim (because there are observations we could make that would tell us if the claim was false), but, assuming that the rabid Britney fans have already done the legwork to line up information establishing that her home planet is not Mars, it's a falsified claim -- one that has been established to be false. There are loads of falsifiable claims that are not as yet falsified. Sometimes this is because the claim is true, but sometimes it's because the necessary observation just hasn't yet been made. For example:

Al Gore is a mammal.

is a claim that has not been falsified and that (I'm willing to bet) will not be falsified, seeing as how the available evidence (including the fact that he has hair) seems to support the truth of the claim. On the other hand,

On February 2, 2007, Venus will depart from its orbit and collide with California.

is a claim that has not yet been falsified because, seeing as how February 2, 2007 is still in the future (as I type this), we can't make the relevant observation yet. When February 2, 2007 rolls around, skywatchers and those of us who live in California will have the opportunity to make an observation that (I really hope) will falsify the claim.

It is very important not to confuse falsifiable, falsified, and false!

  • A claim that is falsifiable can be true -- in which case, attempts to make an observation that falsifies the claim will come up empty.***
  • A claim that has been falsified is known to be false (because there has been an observation that demonstrates that the claim must be false). And, it must have been a falsifiable claim -- one for which there was observable evidence we could turn to in order to test the claim -- else we could not have falsified it.
  • An unfalsifiable claim might be true and it might be false. However, there's no possible observational evidence we could turn to in order to demonstrate that the claim is false.

So, for example, I might claim:

An omnipotent non-material deity exists.

Presumably, there's a fact of the matter about the existence of such a deity -- an omnipotent non-material deity exists or it doesn't. Yet, at least within my understanding of materiality and omnipotence, there isn't an observational test we could apply to determine whether the claim is false. (One of the options you can exercise if you're omnipotent, I would suggest, is to hide your footprints.) This claim, then, is unfalsifiable -- there's no observable state of affairs that would demonstrate that this claim is false. Not being able to prove it's false doesn't mean that the claim is false, just that no observation we make could settle the question.

Why this concept matters to scientists is that they are trying to build a good account of the world (or of some system of set of phenomena within the world). A constraint on "good account" here is that we can indicate that the scientific story about this bit of the world fits with the empirical data (i.e., what we've observed and measured) from the world. If a piece of empirical data conflicts with one of our claims about the world (and if we've taken all the necessary step to make sure it's not a faulty piece of data), we're supposed to take that claim out of the story. We can't hold onto both the falsfied claim and the piece of data that falsifies it, or the scientific story loses its credibility as an account of the world.

At least, that the simplified version of how falsification in supposed to work in science. As we'll see when we press on to theories, in practice it may not be so straightforward.

(If you're curious, while you wait for the theories post you might have a look at my earlier discussion of how falsification fits into Karl Popper's efforts to draw a clear line between science and pseudo-science.)

*Here, I'm taking "is from Mars" to mean "was born on Mars". If you have some different notion of "is from Mars" in mind, you'd want to make that explicit in order to figure out what sort of observables -- Are her parents or grandparents from Mars? Was she conceived on Mars? Did she go to grade school on Mars? -- are the relevant ones for the purposes of testing the claim.
**Another important point: the observation that you could make to demonstrate that the claim is false doesn't need to be easy to make, only possible. It complicates matters, then, that science keeps shifting the boundaries of what we entertain as possibilities.
***Complication alert: We have to be on guard against wonky measurements. This is one good reason to be wary of tossing out a claim on the basis of a single observation, at least if there's any possibility that the observation in question could have been flawed.

More like this

Perhaps you plan to cover this in a subsequent post, but I think it's valuable to recognize that falsifiability is related to predictive value.

If a claim is not falsifiable, then it has no predictive value. It is compatible with all possible observations. If it is falsifiable, then it must make at least one potentially testable prediction. That is, it predicts that a falsifying observation will not occur.

And of course, science is largely about predicting the world around us accurately. So, whether or not falsifiability solves the demarcation problem, I contend it's an essential element of science.

Falsifiability might seem like a straightforward concept but it appears to be a difficult one for those with a creationist predisposition. Case in point is a 65-page and counting attempt to explain the concept to creationist AFDAVE on the richarddawkins site:

What happens when we leave the absolutist world of true and false for the probabilistic world of statistics?

Does claiming that something is unlikliable count?

no observation we make could settle the question.

This applies for isolated claims IMO. As you say, when we go on to theories it becomes less straightforward. Here we can have different theories where unobservable claims are settled.

For example, imagine comparing variants of Newton's gravity theory, and no other physics is relevant. Say we compare the usual with ones that have several different strength relationship which can't be experimentally dissolved in principle. (For example by requiring that the difference is in a decimal larger than what we can resolve in the observable universe.)

But by parsimony, we choose the usual theory. The claims of difference, or claims of having several fields, are settled without direct falsifiability.

(My inability to come up with a fully realistic variant here, or requiring using the observable universe to resolve possible falsifiability, may point to that this claim isn't about our physics. But perhaps it is enough if it is a possible claim. :-)

By Torbjörn Larsson (not verified) on 01 Feb 2007 #permalink

This subject is not even close to the phrase "beating a dead horse". Instead, it's as if the elusive horse (Falsifiability) is being chased after and at certain times gets hit, but is far from being dead. My point: thank you for recycling a topic that needs to be reviewed in order to see science, and at times the world, for what it is.


By Jesus Sanchez (not verified) on 01 Feb 2007 #permalink

Thanks for this post, and the one on "argument". I have a question similar to the one that Lab Lemming brings up, for both falsifiability and argument.

I would argue that the "best" model for our knowledge of the world and the assertions we make about it is fundamentally a probabilistic one. However as probabilities approach certainty, the proper way to reason about them approaches the true/false binary nature of logic as a "limit". So in some sense, one of the roles of scientific investigation is to push up the probabilities of various assertions about the world into the range where we can legitimately use the rules of logic to explore their consequences. Easier than evaluating Bayes Theorem on everything.

Am not sure if this is coherent at all - but it does seem that the gap between certainty and probably is generally glossed over.

While there is a clear difference between probabilistic and universal generalizations, it's worth noting that reasoning with probabilities is usually a deductive affair. It's not the principles of reasoning/argument that are different so much as the content of the propositions about which we reason/argue.

By bob koepp (not verified) on 05 Feb 2007 #permalink

I would argue that the claim "On February 2, 2007, Venus will depart from its orbit and collide with California" is falsifiable on February 1. We know exactly where Venus will be on February 2, at least with as much confidence as we know that Brittany Spears is not a Martian.