A Survey of Nominative Case Assignment by Adpositions
Alan Libert, Department of Linguistics, University of Newcastle, Callaghan, email@example.com
The Government-Binding rule of Case assignment by adpositions is "NP is oblique if governed by P" (Chomsky 1981:170), and this is consistent with the fact that in some of the best known languages, such as Latin and German, objects (or complements) of adpositions are marked for some case other than the nominative. One may note the remark by Beard (1995:240) that "Ps do not occur without Case Marking on the nouns which they accompany in inflectional languages". However, one does sometimes find objects of adpositions which are in the nominative. In this paper I shall present a survey of the circumstances in which this seemingly unusual phenomenon occurs, as a first step to an attempt to account for it.
One must bear in mind the distinction between abstract case and morphological case (or between case assignment and case realization). If one accepts the GB notion of abstract case, then it is certainly possible that objects of adpositions which appear in the nominative case are assigned some abstract oblique case by a P, but that this abstract case does not correspond to an overt oblique case (as happens to objects of verbs in English). If this is so, then it is not a question of nominative case assignment by adpositions, but rather of nominative case realization, but the phenomenon is still of interest. However, one might argue that some examples of it present evidence against the notion of abstract case, or at least that we need some principled way to determine whether an apparently nominative (or caseless) element has been assigned abstract nominative case or an abstract oblique case.
Here I am not looking at situations such as that in English where nouns lack an overt accusative or oblique case, and hence appear to be in the nominative after prepositions, but rather at situations where an oblique case form is available, but is not used. The rest of this paper consists of a presentation of the types of such situations which I have discovered, and a few preliminary remarks on how one might account for them.
Nominative Assignment "by Mistake"
Let us first consider what we might call case marking errors in languages which, unlike English, have morphological oblique cases on nouns. In Classical Greek prepositions take the genitive, dative, or accusative cases. However, in later Greek there are a few examples of prepositions with nominative objects, as in the following example from a papyrus of the Ptolomaic era with the preposition apo 'from', which usually takes the genitive:
(1) apo 'peliote:s [sic] from east.wind-NOM (cited in Mayser 1934:367)
This is very rare and is presumably a consequence of the breakdown of the case system of Greek. In Modern Greek as well occasionally there may be nominative objects of prepositions, as illustrated by Thumb (1912:101), although I am somewhat dubious about whether they are such.
Portuguese is like English in that only pronouns show case, and indeed only some pronouns do; in Brazilian Portuguese prepositions have been observed to occur with the nominative of these pronouns, rather than the expected non-nominative forms (Sebastian Drude, personal communication 1998).
We also find "errors" of this sort in two ancient Semitic languages, Akkadian and Ugaritic, where prepositions "should" take genitive objects but are found with nominative ones. The loss of the genitive itself is not so strange, but one may think it surprising that the nominative, rather than the accusative, so commonly replaces the genitive in its function of marking prepositional objects. A point of interest about these languages, and one which adds to the apparent oddity of the just mentioned fact, is that, unlike many case languages, they have some nominative forms involving discernible suffixes, so we cannot simply say that the prepositional objects have lost a case; rather, they clearly do have an overt nominative case marker. Such data may argue that in at least these instances we should not speak of abstract oblique cases being assigned and being realized as nominatives (or caseless forms). We may note that in at least most of the languages mentioned the case system was in decline, and case errors occur in other structures as well.
In English there are case errors when pronouns occur as conjuncts; some of these errors are when the conjoined NPs are objects of prepositions. Examples are given in (2)-(3).
(3) For both Steve and I, our marriage in 1979 was a second chance (Evening Standard, 30 June 1992:2, cited in Johannessen 1996:674) (3) ... between you and I ...(Sobin 1997:319)
Given that this also happens outside of prepositional phrases, such instances are perhaps to be accounted for by some property of co-ordinators rather than by some process particular to prepositional objects.
Among the errors made by aphasics are those involving the substitution of one case for another. It can happen that the object of an adposition is marked nominative rather than with the required case. An example from German is given below, where the article is marked for nominative rather than accusative.
(4) fuer ein Mund for a-NOM mouth (Stark and Dressler 1990:419)
In further research I intend to compare the type and frequency of these errors with the errors in normal language just described in order to see whether there are any common features.
Nominative Assignment in "Standard" Languages
We now turn to nominative objects of adpositions which might not be considered the result of errors, i.e. deviations from a certain normal state, although this is of course an arbitrary judgement.
I know of no language in which all objects of adpositions are nominative (again, considering only languages which have an overt case system). However, there are languages in which some such objects are marked nominative. We shall first look at languages where some adpositions have nominative objects (and other adpositions have objects in another case), i.e. where the choice of case seems to be a lexical matter, and then languages where it depends on some feature of the object or on some semantic factor.
In many languages adpositions differ in which case they take, for example in Latin some prepositions take the accusative while others take the ablative (and some take either, depending on the meaning). At least a few languages have the same sort of differences among adpositions, with one of the possibilities being the nominative case. Albanian has some prepositions which require accusative objects, and some which require ablative objects, but also two prepositions which take nominative objects, te (or tek) 'at, to' and nga 'from' (I thank Wayles Brown for pointing this out to me). Similarly, in Spanish segun 'according to' takes the nominative, unlike most prepositions (thanks to James L. Fidelholtz and Matthew L. Juge for this information).
In ergative languages the basic function of the nominative, marking subjects, is covered partly by the ergative and partly by the absolutive; prepositional objects bearing these cases could be the equivalent in ergative languages of the nominative objects under discussion, and might be seen as equally unusual, as perhaps shown by the following quote from Haspelmath (1993:22) on the Lezgian postposition patal 'for': "This postposition ... inexplicably takes an Absolutive argument." Several other postpositions in this language also have absolutive objects. My intuition is that ergative objects of adpositions would be more unusual than absolutive objects, but note that Mel'chuk (1988:208) uses nominative to refer to the absolutive, since it is "the most unmarked case".
In some languages there is a split in case marking, depending on what type of constituent is the head of the NP. In at least some Turkic languages several postpositions take the nominative of nouns, but the genitive of some pronouns. For example, in Turkish the postpositions gibi 'like', ile 'with', kadar 'as much as', and ichin 'for' act in this way; other Turkish postpositions take objects marked with one of the following cases: nominative (only), dative or ablative. In (5)-(6) are sentences illustrating the behaviour of ile:
(5) kim-in ile gittiniz? who-GEN with you-went 'with whom did you go?' (Lewis 1967:86) (6) vapur ile gittiniz boat-NOM with you-went 'you went by boat' (ibid.)
A few remarks are in order here. First, what I am calling the nominative in Turkish has a significant difference from the nominative cases of languages such as Latin and Greek: this case is also borne by direct objects if they are not "definite" (v. Lewis 1967:35-6 for details), i.e. it covers some of the territory of the accusative of many other case languages. Lewis calls it the absolute rather than the nominative. One may wonder whether this difference has any connection with the facts just described, and whether there is a language with a more typical nominative where these same facts hold.
Second, two of these postpositions, ile and ichin, also turn up as suffixes, namely -yle and chin (and variations), so it is not entirely clear whether we are in fact talking about postpositions (although they are affixed to a genitive stem rather than a nominative stem when the relevant pronouns are involved, with the exceptions to be mentioned immediately below). Finally, the pronouns which generally appear in the genitive before adpositions sometimes (namely "colloquially" (Lewis 1967:85)) bear nominative case: Lewis (1967:86) says, "This is particularly frequent with kim; instead of kiminle, kimin ichin, and kimin gibi 'with whom?', 'for whom?', 'like whom?', one hears kimle, kim ichin, and kim gibi, the last being a more respectable solecism than the first two". This may be seen as the same kind of "error" as those in Greek and other languages described earlier.
The pronouns which are (generally) genitive here are ben 'I', sen 'you (singular/informal)', o 'he, she, it, that', biz 'we', siz 'you (plural/formal)', bu 'this', shu 'that', and kim 'who?'. A fact complicating any attempt to account for this phenomenon is that when these pronouns take the plural ending -ler/-lar they are nominative, not genitive, e.g. onlar gibi 'like them' (Lewis ibid.).
In some languages there is a split with respect to the case taken by an adposition based not on the syntactic category of the object, but on semantics: the choice between the nominative and some other case depends on the meaning to be expressed; this is similar to the situation in Latin where several prepositions can take accusative or ablative objects. This appears to happen with two postpositions in the Turkic language Bashkir: saqlI means 'like(= the size of)' with the nominative, but 'up to, until, till' with the dative; tiklem has the latter meaning with the dative, and means 'the size of' with the nominative (glosses from Poppe 1964:41).
We find something similar in several Indo-European languages. For example, Russian has two prepositions, za and v, which general ly take oblique cases, but which apparently take the nominative in certain circumstances, although they could perhaps be analyzed as something other than prepositions in those circumstances (I thank Daniel E. Collins and Keith Goeringer for drawing my attention to these facts).
Towards an Account for Nominative Adpositional Objects
One may say first that nominative marking of objects of adpositions is not that unusual a phenomenon; the feeling that it is odd may be a result of the fact that it is quite uncommon in the case languages most familiar to traditional linguistics. I know of no a priori theoretical reason why such objects should not bear nominative case; there is presumably not the potential for confusion with sentence subjects as there would be if verbal objects in a free word order language were marked nominative rather than accusative.
When a language has different Ps taking objects in different cases, one of which is the nominative, the account for this would presumably be the same as the account for why some verbs in German and Latin take objects in different cases than the accusative: it is a lexical property of those verbs that they assign dative, or some other case. Therefore it is marked in the lexical entry of Albanian nga 'from' that it assigns nominative. Chomsky's rule for case assignment for Ps would have to be modified in any event to handle languages where more than one oblique case can be assigned by Ps; one might suggest something like "P assigns accusative unless otherwise specified by its lexical entry". I see no principled reason to assume that nga assigns an abstract oblique case, which is not realized as such, given the fact that other Ps in the language assign oblique cases which do show up as oblique morphological cases.
Splits in case marking based on the nature of the object are somewhat more difficult to account for, but note that again this is not a problem unique to nominative adpositional objects. An account for the Turkish-type split in terms of lexical properties of individual elements is possible, though unattractive. Recall, however, that the split does not occur with all postpositions (and indeed not all postpositions taking the nominative, although there are only two of these which do not exhibit the split, one of which usually has an infinitive as an object, and the other one being "obsolete except in archaizing poetry" (Lewis 1967:85)). Thus, in any case, there will have to be reference to lexical properties of particular postpositions. The problem is with those pronouns which are involved in the split: if some of them are not to be lexically specified as to which case they bear in certain contexts, a more general mechanism will have to be found. At the moment I know of no such solution; I am not convinced that the discussion in Kornfilt (1984:61, 225-7) will account for all the details.
However, one may note that there are other splits between pronouns and nouns, and even between different kinds of pronouns, involving either abstract or morphological case. Examples of these include the difference between English nouns and pronouns, or between who and what, the former having distinct nominative and accusative forms, the latter not. (It is interesting that the Turkish equivalents display a difference with respect to case marking before the postpositions under discussion, kim 'who' being genitive while ne 'what' is nominative.) If such splits can be attributed to other properties of these elements, the latter might be invoked in an explanation of the complicated facts of Turkic languages (this idea is due to Anne Robotham). That is, some properties of pronouns, or of some pronouns, may make them more or less likely to be assigned oblique case, or more or less likely to lose morphological case.
A similar kind of account can perhaps be extended to the case "errors" due to language change or disordered language. One might think that when adpositional objects lose their originalcase marking, due to the decline of the case system of a language, they are marked for some other case, usually not nominative. However, perhaps some kinds of elements are more vulnerable to nominative marking than others, in normal and/or aphasic language; although they too would usually be marked with the accusative or another non-nominative case, perhaps a certain combination of factors makes it possible for them to be nominative. It would therefore be worthwhile to collect and analyze a large number of such "errors" to determine whether they are more likely to occur in some circumstances than others, e.g. are objects of certain prepositions more likely to bear nominative than others, or are certain types of objects more susceptible to such unusual case marking than others?
As has been shown, although it is not a very common phenomenon, objects of adpositions can bear nominative case, perhaps more often than supposed. A detailed analysis of the situations where it happens, and of the constituents to which it tends to occur, may allow us to explain both why it occurs at all, and why it is relatively rare.
I thank George Horn, Christo Moskovsky, Peter Peterson, and Anne Robotham for useful discussion.