The sixth-century Grammarian Priscian defined "vox" as "aerem tenuissimum ictum vel suum sensibile aurium" (air very subtly struck or its sensible effect on the ears), a definition that, as Marcia Colish (whose translation I use here) observes, "deprives the ear of any active role in sensation, leaving it a purely passive recipient of the data impressed upon it by the sensed object." Keep this passivity in mind for what you read below.
Now, while Priscian then divides vox into four categories--articulata, inarticulata, literrata, illiterata--in practice, so far as I've (inexpertly, initially) noticed, two categories predominate in medieval grammatical theory. In his Ars Grammatica, Marius Victorinus writes:
vocis formae sunt duae, articulata et confusa. Articulata est quae audita intellegitur et scribitur et ideo a plerisque explanata, a nonnullis intellegibilis dicitur….Confusa autem est quae nihil aliud quam simplicem vocis sonum emittit, ut est equi hinnitus, anguis sibilus, plausus, stridor et cetera his similia.
Later, drawing on Isidore Etymologies PL 82: 89B, Thomas of Cantimpré writes:
There are two forms of the voice, distinct and indistinct. The distinct is that which, when heard, is understood and written and therefore explained to many and is said to be understandable to many....The indistinct however is that which is nothing but the single sound of a voice cast out, as is the neighing of a horse, or the hissing of a snake, or clapping, hissing, or other such things.
Omnis autem vox articulata est aut confusa: articulata hominum, confusa animalium. Articulata est, que scribi potest ut a, e; confusa, que scribi non potest ut gemitus infirmorum et voces volucrum aut bestiarum
Last night, I explained this to my Animals, Saints, and Monsters seminar, who were unsure/upset about my assertion that, barring the Skeptics, the Western philosophical tradition by and large asserted that animals lacked language. As I realized--realized, in fact, at the very moment I was doing this nutshell version of medieval grammar--the problem with animal noise for Marius, Isidore, and Thomas is not that, for example, the animal vox can't express abstract concepts. For the definition of vox brackets the question of vox's actual content, just as much as Priscian's definition of vox brackets off its content in favor of what happens when, and if, vox strikes the ear. Thus the definition of vox depends not on what it is but on what can be done with it.
For all voices are either distinct or indistinct: the human voice is distinct, and animal indistinct. A distinct voice is one that can be written, such as a or e; an indistinct voice is one that cannot be written, such as the moaning of the sick or the voices of birds and beasts.
Or, to put this in break-up language: it's not you, it's us.
The implication in Marius, Isidore, and Thomas is that if human written language could be improved, animals could be recognized as having "voces articulata." There is, surprisingly (to me, anyway), an implicitly skeptical element concerning animal language in medieval grammar. Recall Sextus Empiricus:
we certainly see animals -- the subject of our argument -- uttering quite human cries, -- jays, for instance, and others. And, leaving this point also aside, even if we do not understand the utterances of the so-called irrational animals, still it is not improbable that they converse although we fail to understand them; for in fact when we listen to the talk of barbarians we do not understand it, and it seems to us a kind of uniform chatter. Moreover, we hear dogs uttering one sound when they are driving people off, another when they are howling, and one sound when beaten, and a quite different sound when fawning. And so in general, in the case of all other animals as well as the dog, whoever examines the matter carefully will find a great variety of utterance according to the different circumstances, so that, in consequence, the so-called irrational animals may justly be said to participate in external reason.[image from dynamosquito's flickr account via Creative Commons License Attribution-Share Alike 2.0 Generic]