Constructional Approaches to Argument Marking in German
This thesis examines certain complementation patterns in German, showing that they are amenable to an analysis in terms of Construction Grammar and can be considered independently meaningful. The point of departure is the Visibility Condition of the early Principles and Parameters approach, which is found to be essentially a formalization of the functionalist view that case and other argument markers express semantic roles. It is this position that is defended here. However, in view of fundamental criticism regarding the theoretical foundations of the Chomskyan approach, especially its innateness hypothesis, and problems connected with the notion of abstract case, another framework is chosen for the actual analysis. As a sign-based approach, Construction Grammar (CxG) turns out to be an ideal approach for the task in hand: the search for symbols of semantic roles. The choice of framework is further justified by recent experimental research suggesting that CxG has potential for accounting for first language acquisition.
On the basis of Welke’s (e.g. 1994; 2002) critical discussion of the concept, the view of semantic roles prevailing within CxG is shown to be misguided, and Welke’s “significative-semantic”, i.e. representational, notion of semantic roles is adopted. Perspective is found to constitute another semantico-pragmatic category potentially expressed by case. Because of their clearly defined content, Welke’s perspectival roles (e.g. 2002) are preferred to Goldberg’s (1995) notion of profiling.
A usage-based CxG approach is chosen. Lax use of the concept of entrenchment is criticized on grounds that it obscures the role of the rationality of the language user. The need for CxG to account for the nature of language as a social phenomenon is emphasized, resulting e.g. in a redefinition of constructions as form-social value pairs. Parallels between constructions, linguemes (Croft 2000) and memes (e.g. Blackmore 1999) are explored and rejected because of the mechanistic notion of learning adopted by memetics.
The main thrust of the thesis is theoretical. Corpus studies are not conducted; the empirical contribution of the thesis lies in radical theoretical reanalyses of known data. One such radical reanalysis is proposed for German prepositions of prepositional objects (PO-Ps). While most of them can be analyzed as argument markers lexically selected by predicate heads, some of the PO-Ps appear to have turned productive and can be analyzed as argument structure constructions (ASC) in the sense of Goldberg (1995). For instance, the PO-P ‘auf’ occurring e.g. with the verb ‘warten’ displays behavior typical of such constructions: combining it with a verb like ‘sich freuen’ results in a new, coerced verb sense.
The process whereby PO-Ps become productive is regarded as a case of grammaticalization. The process involves three stages: 1) the storage of a local P with a full lexical meaning as part of the head governing it, 2) loss of the concrete local meaning of the P due to storage as part of another sign, 3) resemanticization of the P with an abstract meaning amounting to the emergence of an ASC. Resemanticization being involved, grammaticalization is a controversial label here. Alternative analyses such as exaptation are considered and rejected, and a wide definition of grammaticalization as the emergence of productive means for the expression of grammatical meanings is proposed. In terms of CxG, this amounts to the emergence of a partially or completely schematic construction. Notably, the suggested three-stage grammaticalization scenario can be argued to be involved even in the case of French ‘ne ... pas’, a textbook example of grammaticalization.
The argument structure of the ‘auf’-construction, as well as those of two further cases of productive PO-Ps, is discussed in depth. In this connection, the principles proposed by Goldberg (1995; 2006) for the combining of heads and ASCs are criticized, and a more central role for pragmatics is advocated. The mechanisms that can lead to the productivity of a PO-P are discussed at length. The most plausible scenario is found to be the high token frequency of a verb + PO-P pair combined with at least some occurrences of the PO-P with semantically similar verbs. This hypothesis is based on Goldberg’s (2006) experimental findings regarding the generalization of argument marking patterns in child language acquisition. The modes of prototypical categorization involved in the generalizations leading to the emergence of productive PO-Ps would require further study.
Besides PO-Ps, the thesis investigates the structural cases nominative and accusative in depth, showing that some of their manifestations express both semantic and perspectival roles, while others only symbolize perspectival roles. A survey of some previous approaches that consider the expression of semantic roles by combinations of argument markings to be a viable option is followed by a criticism of the notion of structural case. The “structuralness” of cases is claimed to form a continuum. Rather than being reducible to devoidness of meaning, it is argued to come down to the ability of (combinations of) cases and other argument markers to act as a model for the argument marking of smaller or larger semantically defined verb classes.
It is shown that the case pair nom-acc expresses at least the role constellation Agent/1-Patient/2 in German, the numbers referring to Welke’s perspectival roles 1st and 2nd argument. That is, nom-acc constitutes the form of an ASC with the semantics Agent/1-Patient/2. Due to differences between the properties of semantic and perspectival roles, as well as hyperanalysis tendencies (cf. Croft 2000), it seems plausible that such a construction can split into a nom-acc construction still expressing Agent/1-Patient/2 and another one that only expresses the perspectival roles 1st and 2nd argument. It is argued that such a construction already exists in German and is responsible for the (only) apparent meaninglessness of nom and acc. Its existence forms a cornerstone of a linking approach that is developed.
The proposed approach to linking aims at making do with a minimal amount of lexically determined perspectival information and accounting for the tendencies of semantic roles to carry certain perspectival roles. It is argued that ASCs can adequately capture such tendencies. Individual verbs can in most cases be fused with ASCs solely on the basis of their participant roles. ASCs, which in any case symbolize both semantic and perspectival roles, can then provide the verb participants with perspectival roles. Only verbs with idiosyncratic assignments of perspectival roles to semantic roles need to display lexically determined perspectival information. If they do not lexically specify the argument markers of their participants as well (i.e., specify lexically selected PO-Ps or lexical cases), they can be fused with the solely perspectival nom-acc construction.
The plausibility of the proposed linking solution is corroborated by the explanatory potential it has with respect to reduction tendencies in morphological case systems. It is hypothesized that the possibility of fusing verbs, solely on the basis of perspectival roles, with a nom-acc: 1-2 linking construction can lead to the loss of lexical cases. Such a reduction in morphological case oppositions can in turn contribute to the loss of morphologically distinct nom and acc as well. Stages in the disappearance of morphological case in English in fact fit in well with this scenario. The result is a positional argument marking system like that of present-day English and a preference for a syntactic processing mode (cf. Bornkessel 2002).
