More or less all semantic reconstruction within historical linguistics is focused on reconstructing the semantic content of lexical items or morphemes (cf. Dyen & Aberle 1974, Blust 1987, Fox 1995, Fortson 2003, Zorc 2004, Urban 2011). Within grammaticalization studies the focus has mostly been placed on internal semantic reconstruction (cf. Traugott 1986, Smith 2007, inter alia). In contrast, little or no effort has been put into the historical-comparative reconstruction of the semantics of larger and more complex constructions such as abstract argument structure constructions. The reason is presumably that, while scholars do not agree on how to reconstruct semantic content and are still working on uncovering the mechanisms and directionality of semantic change, one will of course not embark on the even more complicated undertaking of reconstructing the semantics of units larger than words. Another reason is, of course, the fact that most current linguistic frameworks do not assume that constructions are form-meaning correspondences, let alone that abstract syntactic constructions have a meaning of their own. This, however, is unproblematic on a cognitive construction grammar approach where the semantics of argument structure constructions is regarded as being derived from the semantics of the predicates instantiating it.
In this chapter, therefore, we attempt to approach the problem of reconstructing the semantics of larger units than words, in this case argument structure constructions, with the aid of Cognitive Construction Grammar and the Semantic Map Model (Croft 2001, Haspelmath 2003, Cysouw, Haspelmath & Malchukov 2010). On a constructional approach, all linguistic units from the level of morphemes are regarded as form-function correspondences, and this includes larger and more complex units such as argument structure constructions. Like all other constructions, argument structure constructions can be divided into two semantic types, i.e. SEMANTICALLY SPECIFIC and SEMANTICALLY GENERAL constructions (Tomasello 1998, Croft & Cruse 2004: 253-254, Barðdal 2001b, 2004, 2007, 2008, Barðdal , Kristoffersen & Sveen 2011, Barðdal & Eythorsson 2012b), also sometimes referred to in the literature as semantically non-compositional and semantically compositional (cf. Goldberg 1995: 13-16, Croft 2001: 180-184, Wulff 2008). The difference between the two is that the semantics of the first type is not a function of the semantics of the parts, while the semantics of the latter one is. Or, in other words, the first type is semantically irregular, where the meaning part has to be idiosyncratically attributed to the form part on a construction-specific basis, while the second is semantically regular with the meaning of the whole being derivable from the meaning of the parts.
It is the second type of construction that we will be dealing with here, where the semantics is taken to be derived from the lexical verbs instantiating a construction (Goldberg 1995, Barðdal 2008, 2011). In this respect, the notion of verb classes becomes important, as different subclasses of verbs instantiating an argument structure construction may be regarded as representing different subconstructions of a construction (Croft 2003, Barðdal 2004, 2006, 2007, 2008, 2011, Barðdal , Kristoffersen & Sveen 2011). This is why a comparison of the lexical semantic verb classes which instantiate an argument structure construction across related languages is useful for semantic reconstruction. The lexical semantic verb classes function as the unit of comparanda, and on the basis of presence or absence of such verb classes, mapped onto semantic space, a reconstruction can be carried out. It is thus semantic fields, mapped across different regions in semantic space, that are reconstructed for earlier proto-stages. This is how Cognitive Construction Grammar, in combination with the Semantic Map Model, aids in the reconstruction of larger, more complex, and abstract units like argument structure constructions for proto-stages, more specifically in the reconstruction of their semantics. Such a reconstruction has already been successfully carried out for the Ditransitive Construction in Germanic (Barðdal 2007).
We now proceed to a comparison of the lexical semantic verb classes which instantiate the Dative Subject Construction in the five Indo-European language branches under investigation.