The overall diachronic picture
After examining detransitivization and its interaction with the split-S case marking pattern in Cariban languages, one is struck by the impression that what really is going on, rather than a semantically based split in the intransitive verb class, is the semantic evolution of a reflexive marker into new semantic areas (cf. Kemmer 1993). Apparently, the morphological behavior of the SA verb class is an accidental consequence of the fact that detransitivization, as far back as we can reconstruct, entails all the morphology as typical of SA verbs. The alignment of person-marking prefixes appears not to be driven by any semantic forces in the language; it is as though they were being dragged by the evolution of the reflexive marker. Apparently, most of the typologically common areas have been covered: reflexives as ‘middle’, ‘medio-passive’, ‘antipassive’, ‘passive’, etc. (cf. Kemmer 1993).
In at least some languages, the process of lexicalization of detransitivized stems has led to the creation of a new verb class. To the extent that this affects the overall grammar of the language (which would otherwise only have SO verbs, i.e. it would be ergative), the emergence of this new verb class can be seen as an instance of grammaticalization. As often happens in cases of grammaticalization, it is sometimes very difficult to agree on ‘how much’ lexicalization is necessary to posit a new class, rather than just a list of exceptional cases; the analysis of individual languages will probably vary according to the preferences of the analyst. Since so many Cariban languages remain almost totally unknown, it is not possible to say what range of variation this phenomenon might show. The possibility cannot be excluded that, in some as yet undescribed Cariban language, the SA class might have grown beyond the size it has in Panare, thus becoming overwhelmingly predominant, while the former SO class has become a small set of exceptions, or perhaps even disappeared completely. More descriptive work is urgently necessary.
Another important point involves the relationship between this process of ‘class emergence’ and the marking of grammatical relations. Since the two verb classes show very little semantic coherence, they tend to look more like morpho logical classes (similar to, e.g., the different Romance conjugation classes in are,-ere,-ire, or the different Germanic strong verb classes) than like the semantically based classes found in typical split-S languages. The morphological differences in these cases, and apparently also in Cariban languages, serve no purpose other than being ‘class markers’, without any obvious semantic or functional load.
The person-marking alignment, so strikingly split-S in its form, appears thus to be epiphenomenal to the evolution of detransitivization. In fact, the irrelevance of the person-marking system in this evolution can be seen in the fact that a Cariban language like Makushi, in which the original set of person markers was lost due to other factors (cf. Gildea 1998:18–21, 28, 161–182 for a description of this change), can nevertheless continue to lexicalize detransitivized verbs without any obstacles (cf. examples in the previous section). The fact that, for some yet unknown historical reason, detransitivization and A-marking prefixes became connected seems to be the only reason for the case marking to have shifted from a hypothetical earlier ergative pattern (with all truly intransitive verbs being from the SO category) to a split-S pattern, with the possibility of an eventual nominative pattern arising, if the trend continues. We are thus apparently confronted with a language family in which the case-marking system has changed, so to speak, by accident, producing a pattern that, albeit certainly present, carries no meaning by itself. One feels like saying that, just as the existence, in a given language, of a word that is pronounced /hæt/ does not imply that it means ‘hat’, the discovery of a certain formal pattern in the grammar of a language does not necessarily entail that it has the same meaning or functional motivation that similar patterns have in other languages. The relationship between meaning and form is more complicated than that.
