Voice phenomena: how many properties behind this label?

Speakers’ intuitions, philosophical categories and traditional grammars have long considered some types of sentences, namely passives and middles, as the reversed or semantically synonymous mirror image of another construction type, namely active sentences. As soon as by Aristotle, pathos was proposed as a category that represented a way of approaching the relationship between participants in a situation or event. In the Romance tradition, grammarians tried to found equivalent forms for the Latin passive and middle paradigms, and considered the periphrastic be-Passive translations as functional and categorical equivalents. But the mismatches between meaning and formal properties, as well as the inadequacy of reproducing the categories of Latin grammar into Romance descriptions, were very soon apparent to grammarians. To illustrate the point, let us remember that Nebrija refused, on morphological grounds, to accept the existence of a passive voice in Spanish.1 Two centuries later, Bello also warns about the usual temptation of translating Latin categories, morphologically supported, to Spanish, where the morphological basis does not exist2. In current frameworks where semantics reads off the output structures of the syntactic component within a strict homomorphism, mismatches such as two different –but related– structures being the input for the same semantic representation have always been a challenge. Therefore, in all stages of the theory the focus has been swinging between the two poles: the attempt