Can a whole population be viewed as a sample?

Let's imagine a species S that is almost extinct on earth. It displays a variation in color: red and blue. A researcher manages to weigh all the specimens left and concludes that red S:s weigh more than blue S:s. Since this is the whole population, and not a sample, would a statistical test (such as a t-test) still be applicable to see if this difference is statistically significant or just random? Or is it not sound to perform tests in observational studies when the whole population is known?

If a test is applicable would it require a view that the extant population is a "sample" of the extinct+extant population, or even the "natural kind" of the species (which could be viewed as an infinite pool from which live specimens are draws, much like dice rolls are drawn from an imaginary pool of dice rolls with a well defined probability distribution)?

I don't know if the question makes sense, I'm testing this idea against the background that the tests in all my text books are exemplified using samples.
The population of Species S, as defined by those existing now, is infinite. It includes all that have lived in the past and all that are alive now and all that will or could live in the future-an infinite number. Much of science and statistics depends on this view.
All quantities of anything are samples; 2 or 102 or 2 zillion.


Less is more. Stay pure. Stay poor.
If you have t he full population than just compare the means if they are different, than they are different. If the species is going "extinct" there may be selection bias, so you may have a sample of the target population.