Split-S or lexicalized detransitivization
The vitality of the detransitivization process, as illustrated in the previous section, suggests an alternative analysis in which all SA verbs are still transitive verbs. If this analysis is accepted, the ‘SO’ verbs would be the true intransitive verbs in the language, and the person-marking pattern could be described as ergative (since the true intransitives treat their subjects as if they were transitive objects). In this view, the detransitivizer prefix is not derivational (i.e. the reflexivized verb is not a derived verbs stem), but inflectional.
This analysis, though appealing at first sight, has problems. The main points against it are:
The existence of (apparently very old) exceptional non-detransitivized SA verbs. There are 5 to 7verbs (depending on the language) which have the morphological properties of an SA verb without any hint of detransitivizing morphology, not even diachronically; they are listed in Table 1. The existence of these verbs suggests the existence of a lexical category SA verb independently of the morphologically detransitivized verbs.
Table 1. Non-detransitivized SA verbs in Cariban languages.
The existence of semantically anomalous pairs. There seem to be, in each language, several language-specific cases of detransitivized verbs with meanings that diverge strikingly from the meanings of their transitive sources, as seen in numerous examples above, highlighted in Table2. Such semantic anomalies are expected in a derivational process which creates new words, but are problematic for an analysis which considers the detransitivizer to be an inflectional process. The Panare example as-ahka ‘to shave, comb’ is particularly interesting: not only is it semantically divergent (the source transitive ahka means ‘to cut O’), but there is a second reflexive form, with a different allomorph of the detransitivizer prefix, for the regular reflexive meaning ‘to cut oneself’.
Table 2. Semantically anomalous cases of detransitivization in Cariban languages
The unpredictability of the surviving argument. As we have already seen (cf. Table 1), even when the meaning of the detransitivized verb is very close to that of the transitive source, there are certain idiosyncratic details, such as which argument (A, O, or both) survives. Like semantic unpredictability, this is more consistent with derivation than inflection.
Loss of source transitive in certain cases. There are also several cases of detransitivized verbs which have lost their original transitive sources (1–2); these cases have become synchronically basic (though diachronically they must certainly have had a transitive source at some earlier stage of the language). The detransitivizing morpheme could hardly be an inflection on a stem which is otherwise unattested in the modern language.
(1)
(2)
‘Basic’ meanings in the SA category. A look at any of the lexical tables (Tables 6–10) reveals a number of detransitivized verbs that represent ‘basic’ concepts (e.g. laughing, dancing, singing, vomiting, etc.). The basic word for these concepts appears to be, in each case, the detransitivized verb, which creates a mismatch between morphological and semantic complexity. For instance, in Panare, atachiima [atatwiima] ‘dance’ is a detransitivized stem derived from achiima [atwiima] ‘dance with O’ (Mattei-Muller 1994:30); the transitive source, though morphologically simpler (in that it does not have the detransitivizer prefix), is semantically more complex (cf. Payne 1990 for the original observation and more examples).
The different factors listed above basically suggest that it is not a good idea to analyze all SA verbs as simple cases of transitives with reflexive inflectional morphology. Even though etymologically reflexivized, the synchronic meaning is not always reflexive; a number of stems have lexicalized, to varying degrees, and must be considered synchronically to be independent lexical items.
One might argue that this lexicalization, though important, is not so massive as to be a major factor in the analysis of the verb system. Since the number of lexicalized items varies from language to language, it is possible that, for some of them, a ‘list of irregularities’ analysis might prove more useful. However, at least for some of them, a considerable number of verbs have lexicalized, to the point that the ‘list of irregularities’ analysis would seem out of place.
For instance, in Katxuyana, a number of SA (reflexivized) verbs with rather ‘basic’ meanings (‘bleed’, ‘yawn’, ‘get ashamed’, ‘grow’, ‘work’, etc.) were found, for which other Cariban languages usually have SO verbs. Upon closer inspection, these Katxuyana SA verbs are formed in two ways:
Detransitivizer + [SO root + Causativizer/Transitivizer]. SA verbs formed in this way are based on old (sometimes not synchronically occurring) SO roots that were first transitivized and then detransitivized (e.g. from an SO ‘to be ashamed’, an SA ‘to make oneself be ashamed’ is derived). At some point, they acquire the same meaning that the original SO verbs had; the SO verbs are then replaced.
(3)
(ii) Detransitivizer + [N root +-ka ‘remove-verbalizer’].
The-ka verbalizer derives a transitive verb from a noun; the final meaning is that of removing the noun from its possessor, which becomes the O (e.g. from kami ‘blood’, kamii-ka ‘draw blood from O, bleed O’). When such a stem is detransitivized, its meaning becomes ‘remove N from oneself’ (e.g. ot-kamii-ka ‘draw blood from oneself’), but many such complex stems have developed a new middle-like meaning, more like ‘produce (spontaneously) N’ (e.g. ot-kamii-ka ‘bleed’, for instance from a wound). These new SA “middle verbs” apparently have replaced older SO verbs with identical meaning, formed by adding the SO verbalizer-ta ‘produce, be characterized by N’ (e.g. Katx. kamii-ta ‘menstruate’, etymologically ‘bleed’ as well). While the opposing SA and SO forms have drifted apart semantically, at least both still exist in combination with kamii ‘blood’; in many cases the etymologically prior SO form has been lost, leaving behind only the new SA middle (4). Similar patterns can be found in other Cariban languages as well (5–6).
(4)
(5)
(6)
The Katxuyana data suggests that a number of SA roots have replaced SO equivalents. It is not known to what extent this process has affected the Katxuy ana lexicon, but in the limited lexical database available, roughly halfof“basic” intransitive verbs are SA, half SO. In Panare, this replacement seems to have gone much farther: out of a sample of 120 intransitive verbs, 80 (67%) were SA (cf. Examples in 7). Notice that, in spite of this large number of ‘lexicalized’ SA, the detransitivizer prefix still exists and is used as the main reflexive marker in prototypical reflexive situations.
(7)
Apparently, as far as the amount of lexicalization is concerned, Panare has gone further than Katxuyana (which is already a step ahead of languages like Tiriyó, Wayana, etc.), as in the diagram in Figure 1.
Figure 1. Lexicalization scale
