Jeremy W Bowman

Flowers come in many shapes and colors and smells, and many of them seem to "hide" their nectar away at the end of a long tube where it's hard for insects to reach. Why? Why not simply make it as easy as possible for insects to reach the nectar, and thus to come into contact with the pollen, and hence to carry it off to other flowers which can then be fertilized?

Is it not to their advantage to have as many insects visit them as possible?

The answer is No. Flowers need to be visited by insects that will move on to other flowers of the same species (thereby carrying pollen from one to the other). If all flowers had their nectar at the same depth, and all insects had proboscises of the same length, then each insect would carry a large mix of pollen, but few grains of the right type. So the flowers would get less in return for the nectar they supply.

By specializing in color, shape, smell, depth at which nectar is stored, etc., flowers diversified along with their insect (and other) "clients". This is a good example of co-evolution. If a rare flower has to make contact with another rare flower on the other side of the forest, it pays to keep the nectar somewhere only a long-distance flyer can reach. A hummingbird or a hummingbird moth may be the best "client" for that purpose.

In general, male flowers of a given species cannot look much different from female flowers of the same species, because they must attract the same "clients".

That example of "a priori" reasoning actually helps to illustrate something important about the scientific method. Its technical name is the "hypothetico-deductive method", and goes like this:

First, a scientist comes up with an explanatory hypothesis H (usually about things that can't be observed directly).

Second, an experimenter deduces some observational consequence O from H (in concert with other assumptions and auxiliary hypotheses).

Third, a test is performed to see if O is true.

Fourth, if O is indeed found to be true, then H is "confirmed" (or "corroborated" if you prefer a weaker-sounding word). But if O is found to be false, then H (or one of the assumptions and auxiliary hypotheses) is rejected as having been "falsified".

In the flower example, we might start off with the hypothesis that in general, the appearance of flowers is determined by what sorts of insects they primarily need to attract. This hypothesis (together with some other uncontroversial assumptions) leads to the "a priori" conclusion, above, that the male and female flowers of any species must look alike. It's important to notice that only deduction is involved in making this inference. No observation of the outside world leads us to the "a priori" conclusion.

Once we have that conclusion, we are in a position to test it. To test it, the best we can do is look around us and see if we can see any male flowers that do look different from female flowers. If we can't find any, the original hypothesis (that the appearance of flowers is determined by their "clients") is corroborated in the sense that we have a better reason to believe it's true.

Our reason for believing H (if O is true) is that it has cleared a "hurdle" that was set for it. The more varied and/or difficult the hurdles that H can clear, the more of an "unlikely coincidence" it would be for H to be able to clear them, yet for H still to be false. So we can guess that H is probably true. (It's never anything better than an educated guess, by the way.)

The above is called the "hypothetico-deductive" method because it crucially depends on an observational consequence O being deduced from the hypothesis H.

Now, it is a remarkable fact that many psychologists and statisticians do not see the scientific method as being like that at all. (I will be addressing their misconceived ideas of science quite soon, in another page on this website.) They think science is really all about extrapolating from some pre-existing "data". Well, they are wrong. We should not believe statistics when all that is done with it is extrapolation. We should believe statistics when it is used for testing. Statistical hypotheses can be tested, and frequently are tested: an observational consequence is derived from the original hypothesis using statistical methods — and it is subsequently tested, and the result is found to fall, or else not to fall, within some predicted range.

As far as I know, no one has any complaint with statistics being used in that way. What I'm complaining about is the use of statistics to derive a hypothesis from pre-existing data without any testing. Our reason for believing a hypothesis arrived at by that method is much, much weaker than our reason for believing a hypothesis because it has passed a test. That's why so many people insist on "falsifiability" as a criterion of scientific rigor.

To return to flowers being "shaped" by insect preferences: if flowers are attractive to us because they have been given an attractive form by insects, then human and insect ideas of attractiveness probably do not differ very much. Our high ideals of beauty are probably not all that far above the relatively low brows of the insect brain.