Abstract
The New Indo-Aryan (NIA) languages generally allow for a range of oblique subjects (cf. Masica 1991, Mohanan 1994 on Hindi). In this talk we focus on dative/experiencer subjects (Verma and K.P.Mohanan 1990) in the light of recent work on Indo-European (IE) by, e.g., Barðdal and Eythórsson (2009), Barðdal (2012), Barðdal et al. (2012) and Barðdal and Smitherman (2012), who rely on in-depth lexical semantic analysis and the notion of constructions as articulated within Sign-Based Construction Grammar (Michaelis 2010) for an analysis of the history of non-oblique subjects in IE. In particular, they argue for an Oblique Subject/Semantic Alignment Hypothesis in contrast to the Object-to-Subject Hypothesis (as articulated, e.g., in Cole et al. 1980, Haspelmath 2001). The Semantic Alignment Hypothesis reconstructs a construction template which pairs up subjecthood properties with dative case and an appropriate thematic role (e.g., experiencer) for Proto-Indo-European (PIE). Modern dative subjects as in Icelandic, for example, are considered to be continuations of an old, pre-existing pattern, rather than a syntactic innovation.
In this paper, we argue that this hypothesis is directly contradicted by evidence from Indo-Aryan (IA). In particular, we offer data from Urdu/Hindi and Marathi and argue for a version of the Object-to-Subject Hypothesis in which a(rgument)-structures can be variably linked to grammatical relations as is standardly assumed in LFG's Mapping/Linking Theory (Bresnan and Zaenen 1990, Bresnan 2001) and in which case markers contribute lexical semantic information to a clause, cf. the constructive case analysis articulated by Nordlinger (1998) and the lexical semantic approach to case taken by Butt and King (2004) and Butt and Ahmed (2011).
The historical evidence with respect to Urdu/Hindi and Marathi shows the following. While the use of oblique subjects is wide spread in both Urdu/Hindi and Marathi (as in most IA languages), Old Indo Aryan (OIA) shows very little evidence for oblique subjects, with the exception of some slight evidence for genitive/possessive subjects (Hock 1991, Hook 1999). The modern case markers in Urdu/Hindi and Marathi are not descended directly from OIA, but were innovated with the development of New Indo-Aryan (NIA), beginning from about 1200 CE (Beames 1872-79, Kellogg 1893, Chatterji 1926, Butt and Ahmed 2011). The historical record for Urdu/Hindi as well as Marathi shows that the first uses of the dative were spatial or purposive, as shown in (1). Note that both Urdu/Hindi and Marathi display a dative/accusative homonymy, something which is also observed frequently in IA.
| (1) a. | ek | qafla | sodagarõ | ka | dɑmıʃq | ko | ja-ta tha | |
| one | caravan.M.SG | merchants | GEN.M.SG | Damascus | Dat/Acc | go-IMPF.M.SG be.PAST.M.SG | ||
| 'A caravan of merchants used to go to Damascus.' (Dehalvi 1804) (Old Urdu) | ||||||||
| (1) b. | tavã | avadhūta | eku | bhiks. e-si | rīgā-lā | |||
| then | monk.NOM | one | alms-DAT | leave-PERF.3.M.SG | ||||
| 'Then, one monk left for (collecting) alms.' (LC:E:16) (Old Marathi) | ||||||||
| (1) c. | dvārāvatī-ye | guṁpha-si | ge-līyẽ | tava | tethẽ | eku | mahātme | ho-te |
| D.LOC | cave-DAT | go-PERF.1.SG | then | there | one | great.soul.NOM | be.PST-3-PL | |
| '(When) I went to the cave in Dvārāvati, there was a saint (great soul) there.' (LC:E:4) (Old Marathi) | ||||||||
Extensions in usage first saw the marking of direct objects (mainly with respect to verbs like 'find' and 'seek'), then that of beneficiaries/goals and from there to experiencer dative subjects. This historical development is particularly clear in Marathi, for which synchronic variability with respect to the subjecthood of dative experiencer arguments was first observed by Joshi (1993) and has more recently been analyzed within Optimality Theory (Asudeh 2001). In particular, the subjecthood status of the dative argument varies according to which class of verbs is involved. We show that this variability is connected to a historical development by which verb classes have undergone a shift in linking possibilities at different historical stages of Marathi. Intransitives like 'be seen/appear', which originally subcategorized for a nominative argument and an optional spatial oblique in Sanskrit began to license the spatial oblique as an experiencer in Old Marathi. This experiencer is realized as a dative subject in Modern Marathi. Transitives like 'pain' or 'burn', which originally displayed a standard nominative-accusative pattern in Sanskrit and maintained this in Old Marathi, now allow for a dative subject in which the subject is the experiencer of the event. Another class of verbs like 'recall, know/perceive/realize, obtain' also originally showed a nominative-accusative pattern in Sanskrit, but here the nominative subject had the semantics of an experiencer. In Old Marathi these verbs show variable patterning of nominative-accusative and dative-nominative with respect to a subset of the verbs, but only nominative-accusative patterning with respect to another subset, the 'recall' class. In modern Marathi, this 'recall' class now shows variable nominative-accusative and dative-nominative patterning, while a subset of the other class now only exhibits dative-nominative arguments. The historical facts thus point to an on-going change in progress by which originally oblique or object arguments are realized as dative subjects, with the change affecting successive classes of verbs over different stages of time.
We suggest that the rise of oblique subjects in IA was licensed by the restructuring of the verbal system, by which almost all of the original verbal inflectional paradigms were lost and replaced by a periphrastic system based on a combination of old participles and new auxiliaries (Kellogg 1893, Beames 1872-79, Chatterji 1926), see also (Montaut 2006). The participles systematically realized their highest argument as an oblique and this opened the door for oblique subjects in IA. The synchronic and diachronic data from Marathi furthermore clearly shows that subjecthood is indeed "acquired" in stages, as originally suggested by Cole et al. (1980). We find no evidence for an invariable construction template that has been available since OIA and which has been passed down to the grammars of the modern languages, contra Barðdal et al. Instead, we argue that it is best to conceive of the historical development as following from the variable possibilities for linking between a-structure and fstructure in dynamic interaction with the lexical semantics associated with individual items (verbs, case markers). With respect to this latter point, we argue that the lexical semantic approach to case as formulated within LFG (Nordlinger 1998, Butt and King 2004, Butt and Ahmed 2011) allows us to express exactly the right kinds of semantic distinctions at the right level of granularity. Changes in the lexical semantic content of individual case markers have consequences for their overall distribution in the language and, in turn, consequences for which kinds of linking patterns are found (e.g., dative-nominative vs. nominative-dative). In contrast, Barðdal et al. assume an invariable linking pattern (a construction) in conjunction with semantically stable dative markers -- assumptions which are not supported by the historical situation for IA.