That's a strong statement. I'd say it 'ceases to be trivially interpretable'. And I think this is part of where the 'the published paper said ___ so it must be true!' disparity between scientists and public becomes more apparent.

In biology when you design a good experiment, run it, come up with a model that fits your experiment, no scientist would ever pretend that their articulated model was how things actually worked. It's close, it's reasonable, but many more publications would be needed to confirm it. A single publication just is not enough to decomplexify the entanglements in biology. Many other scientists will use that model as the basis for their newer experiment, and when their results do not comport, they alter, adjust, or if need be, discard the original model.

The trouble with biology is it might take an entire 'Cell' paper (12 pages of dense publication - ie, many many grad-student-years of work) just to come up with the most simplistic model for what the biological experiment observed. That doesn't mean it wasn't meaningful - it was just much more complicated and less easily well-describable than physics and chemistry experiments. But very much like physics and chemistry, there is a definite trend towards a well-described and predictable system. Models actually get increasingly accurate over time. And this feature might be part of the border in defining a 'robust' science.

> just to come up with the most simplistic model

Yeap. Bioinformatician here. Not only are most published models overly simplistic -- which everyone acknowledges -- but I think our bigger problem is we don't even have a reasonable "meta-model" of biology.

What I mean by that is, imagine experiments were instantaneous and free. How would we then incorporate the results into a mathematical/computational/predictive framework that describes biological reality in a way analogous to physics? AFAIK, we haven't the foggiest clue.

Biology is pretty theory-light and empiricism-heavy, probably because we have good results with things like "I dunno what this mold is secreting, but it seems to be killing the bacteria!"

So our experimental results are pretty precise (although arguably not "robust" between strains/conditions/etc), when people bother to use n>3, but the models and theory are anything but.

> knocking them so far out of balance that it ceases to be a meaningful experiment

Things in physics can sometimes be knocked so far out of balance during an experiment it boggles the mind, while still rendering perfectly sensible observations.

E.g., I once heard the LHC described as "trying to determine how a grand piano works by throwing lots of them one after one down a very long set of well-defined stairs equipped with lots of very sensitive microphones and force sensors".

In other cases, nothing could be farther from the truth. Say you are measuring heat transfer coefficients in high-Re laminar flow through a tube. A single 100-micron step imperfection in the tube wall will render your results pretty much useless.

The distinguishing feature of physics is perhaps that one can know to a very high degree of precision before constructing the experiment which parameters/uncertainties it is very sensitive to.