According to widespread opinion, there is a fundamental distinction between "observed facts" on the one hand - which are the unquestionable rock bottom of science and should be collected in the greatest possible number and printed in scientific journals - and "mere theory" on the other hand, which is the product of speculation and more or less suspect.  I think the first point I should emphasize is that such antithesis does not exist.  As a mater of fact when you take supposedly simple data in our field - say determination of Qo₂ basal metabolic rates of temperature coefficients - it would take hours to unravel the enormous amount of theoretical presuppositions which are necessary to form these concepts, to arrange suitable experimental designs, to create machines doing the job - and this all is implied in your supposedly raw data of observation.  If you have obtained a series of such values, the most "empirical" thing you can do is to present them in a table of mean values and standard deviations.  This presupposes the model of a binomial distribution - and with this, the whole theory of probability, a profound and to a large extent unsolved problem of mathematics, philosophy and even metaphysics.  If you are lucky, your data can be plotted in a simple fashion, obtaining the graph of a straight line.  But considering the unconceivable complexity of processes even in a simple cell, it is little short of a miracle that the simplest possible model - namely, a linear equation between two variables - actually applies in quite a number of cases.