Etymologiae, Book I, chapters 28-29: De analogia to De etymologia
Chapter XXVIII. On analogy Analogy is the Greek name; in Latin it is called the comparison or proportion of like things. Its force is that what is doubtful is referred to something like it which is not doubtful, and uncertain things are proved by certain ones. The comparison of analogy is gathered in eight ways: that is, by quality, comparison, gender, number, figure, case, the endings of like syllables, and likeness of times. If any one of these is lacking, it is no longer analogy, that is, likeness, but anomaly, that is, outside the rule, as lepus and lupus. The whole agrees, but they differ in case; for it makes lupi and leporis. By rule, when you ask whether trames is masculine or feminine, limes is like it in every declension, and it will be masculine. Likewise funis, whom you believe to be of uncertain gender, panis is like it in declension, and it will be masculine. Likewise from the comparison of positives, as if you say doctus, magnus: they are positives and like one another. It is also done through the diminutive; for example, funiculus shows that funis is masculine, as marmusculum indicates that marmor is of the neuter gender. For the gender that is in the principal word is usually in the diminutive. But not always, as pistrinum, pistrilla. But because we ought to know the declension from the position, that is, the primitive, we gather the gender from the diminutive.
Chapter XXIX. On etymology Etymology is the origin of words, when the force of a verb or noun is gathered by interpretation. Aristotle called this symbolon, Cicero adnotatio (annotation), because it makes the nouns and verbs of things known by setting down an example; for instance, flumen (river), because it grew by flowing (fluendo), is named from flowing.
The knowledge of it often has a necessary use in interpretation. For when you see whence a name arose, you understand its force the more quickly. For the inspection of every thing is plainer once its etymology is known. Not all names, however, were imposed by the ancients according to nature, but some also according to pleasure, as we too sometimes give names to slaves and estates as it pleases our will. Hence it is that the etymologies of all names are not found, because some received their words not according to the quality with which they were born but according to the choice of the human will. The etymologies of names are either given from a cause, as reges (kings) from ruling (regendo) and acting rightly (recte agendo); or from origin, as homo (man), because he is from the ground (humo); or from contraries, as lutum (mud) from washing (lavando), though mud is not clean, and lucus (grove), because, dark with shade, it gives little light (parum luceat). Some also have been made from the derivation of names, as prudens from prudentia; some also from sounds, as garrulus from garrulity; some, born of Greek etymology, have been declined into Latin, as silva, domus. Others too have drawn their words from the names of places, cities or rivers. Many also are named from the speech of different nations. Hence their origin can scarcely be discerned. For very many names are barbarous and unknown to Latins and Greeks.