Science is falsifiable. Or at least, this is what I (like many Americans) learned in many of my high school and college science classes. Clearly, the idea has appeal among scientists and non-scientists alike:
But what exactly does “falsifiable” mean? And why is it valued by some scientists, but dismissed or even considered actively harmful by others?
Imagine you are an infectious disease expert investigating COVID-19. You want to determine whether, absent vaccination, COVID-19 always causes at least some lung damage. To prove this claim is true, you would have to check every case and see if every time a patient has COVID, there is also lung damage. And for every case you check, there are more new cases to check.
However, to prove this claim is false, you merely need to document a single case in which someone who previously had COVID has no lung damage. This is an extension of the logical point that to prove a general claim, you need to confirm every instance, but to disprove a general claim, you only need a single counterexample.
The legendary philosopher of science Karl Popper argued that good science is falsifiable, in that it makes precise claims which can be tested and then discarded (falsified) if they don’t hold up under testing. For example, if you find a case of COVID-19 without lung damage, then you falsify the hypothesis that it always causes lung damage. According to Popper, science progresses by making conjectures, subjecting them to rigorous tests, and then discarding those that fail.
He contrasted this with ostensibly unscientific systems, like astrology. Let’s say your horoscope says “something of consequence will happen in your life tomorrow.” Popper argued that a claim like this is so vague, so devoid of clear content, that it can’t be meaningfully falsified and, therefore, isn’t scientific.
Contemporary scholars who study scientific methodology are often frustrated by the implication that science is logically falsifiable. The problem is that scientists can always make excuses to avoid falsifying a claim. The discovery of Neptune is a famous case. Astronomers had noticed irregularities in the orbit of Uranus. One possibility would be that these irregularities violated the theory currently used to explain planetary motion, called Newtonian mechanics, and that this theory should be rejected. At face value, these observations seemed to falsify Newtonian mechanics. But, no one actually argued for this. Instead, they searched for explanations for the irregularities — including the possibility of another planet. Two astronomers, Urban Leverrier in France and John Couch Adams in England, independently used mathematics to predict the location of this previously unknown planet. Astronomical observations by Johann Gottfried Galle confirmed the existence of a planet and, thus, Neptune was discovered.
Put simply, to test a hypothesis, you have to make a bunch of other assumptions, or auxiliary hypotheses. You have to assume that your instruments are working, that you did the math correctly, that you didn’t miss any relevant causes (like Neptune), etc. When something goes awry, you can then choose whether the real error lies in your main hypothesis or in an auxiliary hypothesis.
For an illustration of this problem, imagine you are baking lasagna. You Google lasagna recipes, find a recipe that looks good, and get cooking. You take your lasagna out of the oven, take a bite, and…it tastes terrible. Does this mean you can falsify the hypothesis that the lasagna recipe is good? Not necessarily. Maybe you didn’t follow the recipe correctly, or the olive oil was rancid, or any number of problems other than the recipe itself.
Similar to the COVID example above, we can imagine a scientist arguing that because of poor resolution in a CT scan, lung damage was not detected when it did in fact occur. In other words, the presumed false hypothesis is not that COVID always causes lung damage. Instead, what is allegedly false is the assumption, or auxiliary hypothesis, that the CT scan was detailed enough to detect the lung damage.
This general argument against falsification is sometimes attributed to the philosopher W. V. O. Quine in a famous 1951 article, but it was actually a widely-expressed concern, including by Karl Popper himself. However, Popper thought that features necessary for the testing of scientific claims would be accepted as background conditions by the scientific community and, therefore, falsification could proceed. For example, after it is accepted that the oven temperature is correct and the ingredients are in good condition and measured properly, then one can test whether the lasagna recipe is any good.
Regardless, when a scientist touts the falsifiability of science, it is rare that they are a strict devotee of Popper. (He held some unorthodox views, e.g., we can never actually gain confidence in a theory, we can only eliminate alternatives.) Usually they mean that, unlike some other systems, science makes deliberately clear predictions and actively attempts to disprove claims.
One of the amazing things about science is not so much its tight logical structure — the scientific process can actually be quite messy — but rather, that science aims to test claims and consider countermanding evidence. The sociologist of science Robert Merton referred to this as “organized skepticism.” (Incidentally, despite his reputation for prioritizing logical falsification, Karl Popper was attentive to this social aspect of science.)
Falsification as a matter of scientific practice, rather than logic, is especially significant because humans like to be right. We are inclined to seek out evidence which supports rather than challenges our existing opinions, a well-known phenomenon that is often referred to as confirmation bias. Science fights against this cognitive tendency by encouraging individual scientists to think critically about their own work and for the broader community to be skeptical of each other.
Falsification does not stand alone as the mark of the scientific, and a lot of scientific research aims to confirm claims or to evaluate claims on metrics other than strict truth or falsity. Nonetheless, the willingness and intent to vigorously confront claims with evidence remains a key aspect of the scientific community. This requires attention to the formulation of claims to ensure they are testable. But, even more important is the careful coordination across the scientific community that allows scientific skepticism to lead to productive research.
Edited by Jennifer Sieben and Joe Vuletich