The Problem of Underdetermination

Gig Φ Philosophy: Published April, 2026

What is it?

The problem of underdetermination concerns how we actually justify scientific reasoning, especially when selecting between competing theories for the same finite set of observation, evidence, or data.

We often think of the scientific method as a flawless machine: you test a theory, and if the experiment fails, you throw the theory in the trash. But when we look closely at scientific practice, a massive problem emerges. Scientific theories are "underdetermined", meaning that a finite set of observation, evidence, or data does not provide an outcome unique to only one theory.

🕸️ Duhem’s Holism &
The Auxiliary Web

The French physicist, mathematician, and philosopher Pierre Duhem pushed back heavily against strict falsificationism. He introduced what we call Holist Underdetermination.

Duhem’s core claim is that we never test specific, individual hypotheses in isolation. When people enter scientific endeavors, they bring along a massive web of background assumptions about the world. Before testing a theory (let's call it T), scientists make allowances for all of the other hypotheses they hold as true, which are known as auxiliary hypotheses (A1, A2, A3... An).

Think of Newton's theory of universal gravitation. On its own, Newton's theory does not logically imply a testable prediction about where a planet will be at a specific time. To make a prediction, you also need Newton's laws of motion, initial conditions like the mass of the sun, and theories about how measuring instruments like telescopes and clocks work. Because observation is inherently "theory-laden," physicists must translate practical facts into theoretical facts using theories about their apparatus.

So, scientists aren't just testing: If T, then O (Observation). They are actually testing: If (T & A1, A2, A3... An), then O.

FREE PRIMARY READINGS

Newton-Smith "Underdetermination of Theory by Data"

"Commentary - The Duhem-Quine Thesis and Underdetermination"

Thought of the day: Next time you see a scientific theory overturned, ask yourself: did the core theory fail, or did the web of assumptions holding it up simply become too messy to defend?

🤷 The Ambiguity of Falsification

Here is where things get messy. What happens if your experiment yields a false result?

Because the theory alone doesn't entail the observation, logic alone does not tell us where to pin the blame. All we know is that at least one element in our massive collection of theories and assumptions is false.

The Medical Analogy: Imagine you go to the doctor with a headache. Rather than taking you apart piece by piece, the doctor has to hypothesize the cause and go through a series of tests, using the process of elimination to arrive at a diagnosis. Science works exactly like this!.
The Christmas Lights Illusion: We like to imagine science is like fixing a string of broken Christmas lights: you test each bulb individually to determine which needs replacing, and you know you found the right one because the string lights up. Science does NOT work like this, even though we often act as if it does!.

Because the experiment does not designate which specific auxiliary hypothesis is incorrect, we can simply modify the auxiliary hypotheses to make the larger theory work. If (T & A1, A2) fails to yield the observation, we just modify our assumptions to (T & Ax, Ay) to get our desired result.

⚰️ The Death of the "Crucial Experiment"

This leads to a radical conclusion: there is no such thing as a "crucial experiment". Traditionally, a crucial experiment is defined as a single test whose outcome is predicted differently by two different theories, definitively confirming one and refuting the other.

Because empirical statements are interconnected, they cannot be singly disconfirmed. No experiment can conclusively falsify a physical theory. If we wish to hold a particular theory as true, we can always adjust the accompanying auxiliary statements, making modifications dependent on our background assumptions.

(Bridging into broader philosophy: The philosopher W.V.O. Quine famously expanded this idea into the "Duhem-Quine Thesis". Quine argued that this underdetermination applies not just to physics, but to the entirety of human knowledge—creating a giant "web of belief" where any statement can be held true come what may, provided we make drastic enough adjustments elsewhere in the web).

🧠 "Good Sense" in Science - if strict logic can't definitively kill a bad theory, how does science ever progress?

Duhem argued that it ultimately comes down to "good sense". While you can theoretically keep saving a dying theory by adding complex, arbitrary modifications (like "tangled-up stays" holding up "worm-eaten columns"), eventually it becomes childish and unreasonable to maintain. Good sense tells scientists when it is time to raze the tottering building and adopt a new, more elegant system.