Moving along the grammaticalisation path: locative and allative marking of non-finite clauses and secondary predications in australian languages

5.5 Diffusion of LS and AS outside the core area

Where LS/AS is found outside the immediate Pama-Nyungan area of overlap of LS and AN, as in Wardaman, Jingulu, and Garrwa-Wanyi it is likely that the pattern has diffused from the core Pama-Nyungan area to Non-Pama-Nyungan neighbours, since all these languages are adjacent to Pama-Nyungan languages with LS and AS. Wambaya, which also abuts on to Pama-Nyungan languages, has only LS in an obvious form, with a different object control form which may have an etymological link to an ALLativeALL form. It seems probable that LS in Gooniyandi is also a case of structural diffusion from Pama-Nyungan neighbours: here the neighbours, like Walmajarri, do not have AS so that would explain its absence in Gooniyandi. It may be that Sstructural diffusion may have been involved in Ngumpin-Yapa because only the eastern languages have the pattern fully; Kaytetye as discussed above, similarly falling together with northern neighbours; and possibly in Warumungu where LS is not clearly present (although it may have been historically as discussed).

6. Conclusions

We have described three patterns of grammaticalised local case-marking involving Locative and Allative in languages between northern and central Australia. The LS pattern of locative marking of non-finite subordinate clauses is widely distributed on a world scale, but the AN pattern whereby a locative secondary predication can take ALL marking just in case it is controlled by an object in the clause is highly unusual, perhaps unique, and its explanation remains somewhat elusive. These two phenomena have overlapped and created a hybrid, the AS construction which combines these two patterns to provide a means of distinguishing between subject and object control in non-finite subordination.

Rather than regarding all these phenomena as a result of the diffusion of some high-level general conceptual scheme of ‘switch-reference’ it is better to try to chart the effects of particular types of scheme for distinguishing control in their linguo-genetic descent and diffusion within areas, and how they affected each other.
A preliminary conclusion is that the AN construction was an areal feature in a number of Pama-Nyungan languages which are in part discontinuous today but may have been contiguous in the past. This construction intersected with, and affected the LS subordination pattern to yield the AS construction.
This then leaves the issue of why and how the AN construction itself arose. This means of marking of secondary predications on object NPs is typologically rare as a type of grammaticalisation of local case-marking. In some of the languages in which this occurs it has the special function of singling out the object NP alone as being in the location, rather than the whole event and all the participants²⁰. Thus in a certain sense the action ‘moves’ from the agent to a patient literally as they are in different spatial positions and it may be a form of this Path metaphor which is encoded in the AN and eventually AS constructions. This however is unsatisfactory as an explanation of how this construction actually arose and more research is needed on this question.
Another possible approach to AN could take into account that an apparently similar phenomenon is found in ‘directional locatives’ in Finnic languages . The semantics of these directional locatives seems quite different from what occurs with AN in the Australian languages we have been discussing however, e.g. ‘see’ ‘hear’, ‘spear’ etc. There is no implication in the Finnish example (6) cited earlier that the hunter (subject) is outside the forest (where the object ‘bear’ is) for instance, which would be the case in the most clear-cut cases of semantic differentiation between AN and Locative marking e.g. in Yolngu. However, the common element between AN and the directional locative/fictive motion constructions may be that there is physical separation of the subject and the location of the object at some phase of the event.
There is some evidence in Wakaya, a Warluwarric language, cited above, that a second-phase resultative location of the object separate from the subject is a key element in the motivation of the AN construction. It is possible that the original form of AN in the area formed by Warluwarric and Yolngu at least did embody the use of an ALL form for secondary predication indicating locations of grammatical objects which are perceived as being in a separate location from the subject of the clause. If this results from a biphasic event structure, then this event structure is different from what it is generally thought to be for verbs like ‘see’. A process which may have resulted in AN of this type is analogical extension of the pattern of verbs which do involve a kind of biphasic event structure - motion of the object as with ‘put’ - to other transitive verbs.
The following diagrams compare the kind of event structures involved and how they are handled in the Finnic languages and in those Australian languages which have AN which is not purely driven by grammatical control phenomena but which has a semantic element related to the relative locations of subject and object also. In these diagrams A indicates the transitive subject and O the transitive object. The circles represent the place which is of relevance to the discourse; if A is outside, that means that the subject entity is not perceived as being in the same location as whatever is in the circle.
While clearly the Australian and Finnic phenomena differ in some parameters, they both reflect ways in which a locational case system can be adapted to a particular metaphorical or ‘fictive’ extension of notions of space and co-lllocation in relation to syntactic arguments. It seems likely that when the AN phenomenon in Australia arose in certain Pama-Nyungan languages it did express this kind of differentiation of collocation co-location and separation of subject and object. But as the grammaticalisation of this phenomenon expanded to expressing types of control in subordination, it also began to lose some of its original semantic character at least in some languages. The subordination strategies, once grammaticalised, were then able to spread by structural diffusion to neighbouring Non-Pama-Nyungan languages.


Comparison of locative and allative secondary predication in Finnic and some Australian languages