The hypotheses stand or fall together

The Duhem–Quine thesis from Wikipedia.

The Duhem–Quine thesis argues that no scientific hypothesis is by itself capable of making predictions.[2] Instead, deriving predictions from the hypothesis typically requires background assumptions that several other hypotheses are correct; for example, that an experiment works as predicted or that previous scientific theory is sufficiently accurate. For instance, as evidence against the idea that the Earth is in motion, some people objected that birds did not get thrown off into the sky whenever they let go of a tree branch. Later theories of physics and astronomy, such as classical and relativistic mechanics could account for such observations without positing a fixed Earth, and in due course they replaced the static-Earth auxiliary hypotheses.

Although a bundle of hypotheses (i.e. a hypothesis and its background assumptions) as a whole can be tested against the empirical world and be falsified if it fails the test, the Duhem–Quine thesis says it is impossible to isolate a single hypothesis in the bundle.

I have not come across this before but it seems intuitively correct, at least in a vernacular form.

It is analogous to controlling variables. When we want to compare two phenomena which are created via different processes, it is important to control for confounding variables. For example, say there are two populations each earning the same amount of money per capita and each having the same savings rate. After several generations it is observed that Population A has a much higher wealth accumulation than Population B. Given that they have the same income and savings rates, it would be easy to jump to the conclusion that Population A must be manipulating the system in some nefarious way to their benefit. Outrage!

But there is a hidden assumption that the only two relevant variables are income and savings rate. But the above described outcome could be the result of differing marriage and family practices. If Population A always marry and always have a single child, over time there will be an increasing concentration wealth (one child inherits the wealth accumulated from two parents). If Population B only rarely marry and those that do marry have many children then over time there will be an increasing diffusion of wealth (many children divide the inheritance of a few parents). If you don't control for the confounding variable of marriage and family traditions, then your initial comparison is simply misleading.

With Duheme-Quine you have to test all the hypotheses collectively to derive a view of validity. Similarly, with statistical comparisons you have to control all the variables, not just the ones you think are relevant.