Statistical methods in biology are strongly encouraged: they are quantifiable, unbiased, and usually the only ones that are accepted as evidence.

However, it doesn't seem like the whole picture.

Say, someone wants to prove the hypothesis that wings are used for flight. He (or she) conducts an experiment: takes two groups of birds, ties up wings of all birds in one of the groups, then throws them out of a window and gets statistically significant data that birds with tied wings do not fly.

However, strictly speaking, he didn't prove that birds use wings to fly: only that birds with tied up wings didn't fly when thrown out of window. Or, as they say, correlation does not imply causation.

To tie the hypothesis to the results, he might include a paragraph in the discussion that "wings have large area, there are muscles that can flap them etc" - in other words, explain how wings could be used to fly. However, this explanation isn't statistical, or quantifiable: he just wrote a "story", perhaps used some common sense - but last time I checked, it's not evidence.

And that's pretty much what happens in every paper on biology: when they try to prove some mechanism, they include an experiment results and stats for a single aspect that is related to the mechanism (which, obviously, cannot explain the mechanism on its own), but then, when it comes to discussion, they just drop all the numbers and start explaining the phenomenon "logically" and eventually come to the conclusion. To be honest, they could've done no experiment at all and just went to explanation straightaway.

So what I wanted to ask:

  1. Is there such a method of proof in biology as "proof by explanation", or something like that - when the author just explains the causual effects from, say, physics perspective and comes to a conclusion? Perhaps, even something similar to mathematical proof? Does this constitute valid evidence?
  2. Are there any formal requirements for such proof? Perhaps, some way to quantify it or reduce to a "formula" of some sort?
  3. Say, I wanted to publish a paper about some new protein that affects testosterone production. If I did no stats at all, or experiments like trying to remove the gene for the protein from some individuals and seeing what happens in comparison to control group, but instead just extracted and described the structure of the protein, identified the binding sites to some testosterone precursor, and explained how it can affect the testosterone production chemically, would it be a valid paper?
  • 1
    $\begingroup$ Your hypothetical paper on a novel protein that describes the structure and binding sites for a testosterone precursor would be an excellent paper. It just wouldn't be about the effect of the protein. There is more to biology than causal relationships. $\endgroup$ – De Novo Dec 22 '18 at 7:18
  • $\begingroup$ The right policy is to exactly state what is observed and not make any far fetched conclusions. It is a matter of academic practice and therefore your question is not really on-topic here. Academia may be a better place for your question. $\endgroup$ – WYSIWYG Jan 9 '19 at 15:09

You are missing a much simpler and more common method. film birds in flight, then demonstrate mathematically that the wings as used are providing the lift that is enabling flight. That is direct evidence for your hypothesis. You can further test it by clipping the wings or putting the birds in a vacuum, or a hundred other strange experiments. You may want to look into biomechanics, the types of studies you are looking for are far more common. How a limb works is not really a statistical question but more of a dynamics one.

Statistics is a secondary way of testing, used when more direct methods are not possible or terribly impractical. to use your last example yes that would be a perfectly valid way to determine a proteins function, there are limitations reasons we don't usually determine protein functions this way, not just becasue often we discover the function before isolating the protein), but even when it is you also want to test it in other ways that is true of anything in science. This one reason science prefers multiple lines of evidence since one line can compensate for the flaws in a different line.

"Proof by explanation" has specific connotations in biology that may not be what you are really asking about*. Often it refers to unverifiable narrative explanation which is not acceptable in science. Also you never prove anything in science, (even more so with statistics) we shy away from the term proof because it brings to mind absolute proof which cannot exist. In science you can demonstrate something or show it is the most likely known explanation but you do not "prove" it. Research is not one off events but cumulative and constructive, the more lines of evidence the firmer a concepts standing.

  • What I thinks you are really asking is something much more fundamental to how science works and what empiricism and verification is. You may have better luck with this on a philosophy of science discussion, than a specific field of science since you seem more confused about how science operates. science has both experimental and descriptive research each with its own value.
| improve this answer | |
  • $\begingroup$ "Often it refers to unverifiable narrative explanation which is not acceptable in science" Well I understand that - but how does one make "unverifiable" narrative a verifiable narrative? Stats allow making verifiable, formal evidence out of unverifiable counting. But there seem to be no formal procedure to "proof by explanation/description". "you never prove anything in science" Well I understand that, I just don't know how to call it properly.- no-one ever talks about this method. $\endgroup$ – A.V. Arno Dec 22 '18 at 9:01
  • $\begingroup$ A narrative is just a description of a sequence of connected events, Anything from plate tectonics, to a description of how a beaver builds a dam can be a narrative, the important thing is that the narrative be testable. Does it make predictions, for the bird example the hypothesis birds use their wings to fly makes several predictions including wings generate lift, birds without wings will not fly, and the simplest that birds fly. Is model of the universe presented by the narrative empirically distinct from one in which it does not occur, AKA invisible dragon test. $\endgroup$ – John Dec 22 '18 at 15:04

Not the answer you're looking for? Browse other questions tagged or ask your own question.