I never cease to be shocked – shocked! – how many scientists don’t know how science works and, worse, don’t seem to care about it. Most of those I have to deal with still think Popper was right when he claimed falsifiability is both necessary and sufficient to make a theory scientific, even though this position has logical consequences they’d strongly object to.

Trouble is, if falsifiability was all it took, then arbitrary statements about the future would be scientific. I should, for example, be able to publish a paper predicting that tomorrow the sky will be pink and next Wednesday my cat will speak French. That’s totally falsifiable, yet I hope we all agree that if we’d let such nonsense pass as scientific, science would be entirely useless. I don’t even have a cat.

As the contemporary philosopher Larry Laudan politely put it, Popper’s idea of telling science from non-science by falsifiability “has the untoward consequence of countenancing as `scientific’ every crank claim which makes ascertainably false assertions.” Which is why the world’s cranks love Popper.

Almost all of today’s philosophers of science agree that falsification is not a sufficient criterion of demarcation.

But you are not a crank, oh no, not you. And so you surely know that almost all of today’s philosophers of science agree that falsification is not a sufficient criterion of demarcation (though they disagree on whether it is necessary). Luckily, you don’t need to know anything about these philosophers to understand today’s post because I will not attempt to solve the demarcation problem (which, for the record, I don’t think is a philosophical question). I merely want to clarify just when it is scientifically justified to amend a theory whose predictions ran into tension with new data. And the only thing you need to know to understand this is that science cannot work without Occam’s razor.

Occam’s razor tells you that among two theories that describe nature equally well you should take the simpler one. Roughly speaking it means you must discard superfluous assumptions. Occam’s razor is important because without it we were allowed to add all kinds of unnecessary clutter to a theory just because we like it. We would be permitted, for example, to add the assumption “all particles were made by god” to the standard model of particle physics. You see right away how this isn’t going well for science.

Now, the phrase that two theories “describe nature equally well” and you should “take the simpler one” are somewhat vague. To make this prescription operationally useful you’d have to quantify just what it means by suitable statistical measures. We can then quibble about just which statistical measure is the best, but that’s somewhat beside the point here, so let me instead come back to the relevance of Occam’s razor.

It’s unscientific to make assumptions which are unnecessary to explain observation and don’t make a theory any simpler. But physicists get this wrong all the time.