Aside from diachronic accounts of Indo- Iranian and Indo- Aryan (Bloch 1965; Hock 1986a; Bubenik 1996, 1998 and further references therein), there is not a great deal
of data on the history of ergative languages; any deliberations on the history of erga- tivity will therefore be necessarily speculative.
The main question I would like to address here has to do with the diachronic relationship between PP- ergatives and DP- ergatives. I have already hinted at this diachronic progression in earlier chapters. Historically, adpositions have been known to be reanalyzed as case markers in a number of languages (this devel- opment has received great prominence in the grammaticalization literature, e.g., Hopper and Traugott 2003). A typical grammaticalization cycle posited for prepo- sitions and cases is as follows (see Heine and Kuteva 2002; Hopper and Traugott 2003; Kracht 2003, 2008; and see also Zwarts 2005a,b, for a slightly different model):
(1) … > Adverb/ Noun > Preposition > Case > Zero > Adverb > …
On the synchronic plane, some exponents may be ambiguous between case markers and adpositions. The Japanese marker ni presents a good example. Some research- ers analyze this morpheme as a dative or oblique case marker (Shibatani 1977; Ura 1999), while others consider it ambiguous between a postposition and a dative marker (Sadakane and Koizumi 1995; Miyagawa and Tsujioka 2004).
We thus have two separate diachronic tendencies working in tandem: the devel- opment of ergatives from passive clauses with a by- phrase agent (recall chapter 2) and the grammaticalization of adpositions into case markers. The general trajectory of change is as follows: the PP in spec,vP develops subject properties, and the P head in the PP gets reanalyzed as a case marker:1
(2) a. PP b. PP
PERG
overt DP P
silent DPERG
This change leads to the development of PP ergatives with null P heads, and the set of languages in which this change has taken place is precisely the set that manifests syntactic ergativity. Syntactic ergativity, however, is just one element of the cluster of properties, identified in chapters 4 and 5, that characterize PP subjects. The adpo- sition in the morphological structure is no longer interpreted as such; the outer layer (PP) is still represented, but the P head is silent (2b). Because the adposition is mor- phologically visible, the language learner analyzes it as a case marker. This reanalysis results in the emergence of a silent P head (2b); however, this silent adposition still affects the syntactic properties of the entire ergative expression.
The situation presented in (2b) can be stable as long as the ergative’s syntactic properties match its status as a PP, manifested through a cluster of properties (see chapters 4– 6). But the reanalysis can also proceed further, causing the outer P layer
1. I am omitting any functional layers, such as KP (cf. Bittner and Hale 1996a, b), which may be below PP.
to be lost and the ergative to be reinterpreted as a DP (3c). A powerful trigger for such a reanalysis is the invisibility of the P head.
(3) a. PP b. PP c. DP
PERG
overt DP P
silent
DPERG D NP
Once the ergative is no longer analyzed as a PP and is instead interpreted as a case- marked DP, it may lose various properties associated with its PP status and acquire the properties found in DP- ergative languages, including A- bar movement that leaves behind a gap at the extraction site. Little is known about the co- occurrence of these changes, but there is no principled reason why the accumulation of these properties must proceed simultaneously; some properties may appear in advance of others. In sections 8.2 and 8.3 below I will present two languages that may instanti- ate the transition from the PP type to the DP type and discuss the sequencing of changes in their various properties.
The grammaticalization cycle outlined in (1) above also suggests the possibility that new adpositions can develop from case forms; such cases are attested in more familiar languages, such as Ancient Greek (Hewson and Bubenik 2006: 62– 63). This being the case, we should also anticipate the possibility of an evolutionary path from a DP- ergative language to a PP- ergative language. At present, I am not aware of any languages undergoing such a development. The existence (or absence) of such lan- guages would be a good test not only for the PP- ergative and DP- ergative typology proposed here but also for the evolutionary trajectory described in (1) and (3). I have depicted the change as unidirectional for illustrative purposes only.
There is good evidence in the generative grammaticalization literature for Late Merge (van Gelderen 2004)— that is, the notion that grammatical words such as auxiliaries or adpositions are less relevant to theta- marking than lexical words are, and therefore they merge later in the derivation than their fully lexical counterparts.
The opacity of a case form may, however, motivate the presence of a preposition in the phrase structure, thus leading to its introduction into the language. Such a pro- cess would provide the functional motivation (and possibly a trigger) for the rise of a PP ergative out of a DP ergative.
A note of caution is necessary: the interchange between case markers and prepo- sitions is certainly not the only way in which ergativity can arise. Ergative case can develop from a demonstrative expression (as proposed for Georgian; see Kulikov 2009 and references therein), from the reanalysis of neuter forms (see Garrett 1990 for an analysis of Hittite, and Melchert 2007 for a cautionary note), or from the reanalysis of animacy markers (see Martínez 2009 for Basque and Blake 1977 for some Australian languages). The data on such diachronic transitions is scarce, but the few specific languages just mentioned all share DP- ergative properties. That, in turn, may constitute indirect, suggestive evidence in support of idea that PP- ergative languages form a coherent class. Of course, it is too early to celebrate this result as a validation of the proposals advanced here; the few languages available for scrutiny certainly align as expected, but the sample is too small for far- reaching conclusions.