The Dative Subject Construction
With the term Dative Subject Construction, we refer to argument structure constructions instantiated by predicates where the first argument is in the dative case and not in the canonical nominative case (cf. Eythorsson & Barðdal 2005, Barðdal & Eythorsson 2012a). Examples are given in (1a-e) below:
(1) Old Norse-Icelandic a. honum er naudsyn him. DAT is.3SG need 'he has a need'
Latin b. mihi necesse est me. DAT necessary is.3SG 'I have a need'
Ancient Greek c. emoi dei me. DAT needs.3SG 'I need'
Old Russian d. nouzda by jemou need be.3sG. suBj him. DAT 'he would have a need'
Old Lithuanian e. iiemus naudinga buwo them.DAT necessary was.3SG 'they had a need'
As argued by Eythorsson & Barðdal (2005: 827-832) and Barðdal & Eythorsson (2012a) who discuss several subject properties in the Germanic languages, the common denominator of the arguments which pass the subject tests in these languages is that they are the first or leftmost argument of their verbs' argument structure. They follow Croft (1998) in assuming that the internal order of the arguments is determined by the causal conceptual structure of each predicate and the force-dynamic relations holding between the participants of the event denoted by the predicate (cf. Croft 1998, Bar5dal 2001b-c). Clearly, for one-place predicates, the only argument of such an argument structure will automatically be its first argument.
As the syntactic subject properties of the ancient and archaic Indo-European languages have only been investigated to a limited degree (Hock 1990 on Sanskrit, Eythorsson & Bar5dal 2005, 2012a on Early Germanic, Fedriani 2009 on Latin, Grillborzer 2010 on Old Russian), it is impossible at this stage to make substantiated claims to a non-controversial subject status of these dative subject-like arguments across the early Indo-European languages, based on their syntactic behavior relative to any subject tests. For some of these languages, however, a subject analysis is more compatible with the data than an object analysis is. Fedriani (2009, 2014) argues for a subject analysis for Latin, as does Grillborzer (2010) for Old Russian. For Old Germanic, moreover, several instances of control constructions have been documented, as in Old Norse-Icelandic, Old Swedish, Early Middle English, and Gothic, in addition to other ordinary subject behaviors, which in turn rules out object analysis for that language branch. Hock argues for a subject analysis for the subject-like dative of the possessive construction, for instance, but not for the subject-like dative of the experiencer construction in Sanskrit.
For the remaining early Indo-European languages, there is no doubt, however, that these subject-like datives are the first or leftmost arguments of their verbs' argument structure constructions, and as such one may regard them as the S argument of intransitives, in Dixon's (1994) alignment typology, where A is the subject of a transitive predicate, O its object, and S the subject of an intransitive predicate. Hence, our notion of subject in this work refers to the status of subject-like datives as being the S argument of intransitive predicates. We also include in our definition and our discussion the subject-like dative of so-called transimpersonal predicates, in the sense of Donohue (2008), although we prefer to use the label oblique transitives. These are predicates which are formally two-place predicates but where the subject-like argument is not in the nominative case, and the second object-like argument may be a nominative (2a), a genitive (2b), a prepositional object (2c) or simply a subordinate clause (2d), as is shown below with one example from four of the five Indo-European language branches under investigation:
(2) a. omnibus amicis meis idem unum convenit (Dat-Nom) all. DAT friends. DAT my. DAT the-same. NOM one. NOM agreed.3SG Latin All my friends agreed on the one same thing' (Pl.Poen.1340)
b. a molodomu esti dobytb (Dat-Gen) Old Russian but young. DAT honor. GEN creates.3SG 'but the young receive honor' (Sl. O Zadon)
c. honum Kkadi til Sighvats vel (Dat-PP) Old Norse-Icelandic him. DAT liked.3SG to Sighvatur. GEN well 'he liked Sighvatur well' (Fm. IV, 8910)
d. Ar nedera man kq noriu darit (Dat-S) Old Lithuanian does not.suit.3sG me. DAT what want do.iNF
'Is it not appropriate for me what I want to do?' (DP 93 13)
There is thus no doubt that a wide array of Oblique Subject Constructions are found across several of the ancient and archaic Indo-European languages. Since we will be showing later in this chapter that argument structure constructions with a dative subject pivot are not confined to experiencer and benefactive predicates, we do not regard these as having the status of E (for Experiencer), as is done in Nichols (2008), but believe that the notion of S is more accurate, following Andrews (2001) and Onishi (2001). However, we will neither refer to these as S nor E here, but continue to refer to these as dative subjects in the remainder of this chapter.
Observe also that the predicate in (2a) selects for the Dat-Nom case frame, and we believe that several, if not all, such predicates could alternate between two argument structure constructions, Dat-Nom and Nom-Dat. In other words, we believe that the two word orders do not represent a basic word order and a socialization of the basic word order, but rather two different basic word orders. The evidence for this is comparative, as such alternating predicates exist in Modern Icelandic (Bernodusson 1982, Jonsson 1997-98, Barðdal 2001b) and Modern Faroese (Barnes 1986), and have been argued to exist in the history of English (Allen 1995), Icelandic (Barðdal 2001b), German (Eythorsson & Barðdal 2005), and Mainland Scandinavian (Barðdal 1998). The hypothesis that both word orders are basic is also confirmed by the fact that, in Modern Lithuanian, both arguments show some subject properties (Holvoet 2009), and in the old languages, the Dat-Nom word order does not seem to be motivated by any specific information-structural properties (cf. Allen for Old English), although this clearly needs further investigation. Therefore, at this point in our research, we take both word orders to be basic, which means that we include in our investigation predicates which are traditionally listed as Nom-Dat predicates in dictionaries and traditional reference grammars, provided of course that they show properties of basic word order variation. The same is true for Acc-Nom and Gen-Nom predicates, although they fall out of the scope at the present investigation.
Before we commence with the comparison of the semantics of dative-subject predicates across the earliest layers of Germanic, Italic, Greek, Slavic, and Baltic, a few words on the problems associated with reconstructing semantics are in place.
-  See Malchukov (2008) and Mithun (2008) for a different use of the term transimpersonal.