Our primary goal in this chapter has been to develop a methodology of use when reconstructing constructional semantics for dead languages, based on the theoretical framework of Cognitive Construction Grammar in combination with the Semantic Map Model. As semantic reconstruction has hitherto mostly focused on lexical items and morphemes, reconstructing the meaning of larger and more complex and abstract linguistic units, such as argument structure constructions, represents a greater challenge. This task, however, is made possible by the basic assumptions of Cognitive Construction Grammar where the semantics of semantically general constructions are assumed to be derived from the semantics of the predicates instantiating them. A task like the present one is a theoretical impossibility within linguistic frameworks which do not assume that constructions are form-meaning correspondences. Hence, in non-constructional approaches, there can be no meaning to syntactic structures, let alone to abstract syntactic constructions like argument structure constructions.

In order to lay out the method, we have collected and compared predicates instantiating the Dative Subject Construction in Old Norse-Icelandic, Ancient Greek, Early/Classical Latin, Old Russian, and Old Lithuanian. This data collection has uncovered between 50 and 380 predicate types in each of the five branches considered. As several of these predicates are synonymous or near-synonymous, the data collection resulted in a total of 232 sememes, which in turn may be divided into different higher-level semantic categories, i.e. experience-based, happenstance, modal, possessive, and epistemic predicates. The experience-based and happenstance predicates may be further divided into 14 different subcategories in total. A further analysis of all these into narrowly-circumscribed semantic verb classes yielded 49 such semantic classes.

After a comparison of how different kinds of semantic overlap may represent different types of historical scenarios, ranging from an independent development with no overlap to an early common development with a considerable partial overlap, we have shown in Figure 7, based on the 14 higher-level semantic categories, what the five different Indo-European branches have in common and how they differ. We have also carried out a Principal Component Analysis (PCA), of the 49 narrowly-circumscribed semantic classes, for all five language branches, which has revealed that Latin and Ancient Greek are most similar to each other, while Old Norse-Icelandic, Old Russian, and Old Lithuanian show signs of individual developments making these three more distinct from each other. This is expected on the hypothesis that the category, the Dative Subject Construction, has been a dynamic and a productive category throughout early history. The PCA analysis, however, has not revealed any specific clustering of verb classes based on language branch, which is expected given that these languages are genetically related, in turn corroborating the assumption that the Dative Subject Construction is an early Indo-European inheritance. We have therefore suggested a reconstruction of the semantic structure of the Dative Subject Construction, found in Figure 10, based on the core of the semantic category, i.e. the semantic subcategories found in at least four of five language branches. We do not necessarily claim that this reconstruction is valid for Proto-Indo-European, as it is only based on five out of 11 Indo-European subbranches, but believe that it may reflect a common West-Indo-European language stage, if such a stage existed. Through this enterprise, we have shown how the constructional semantics of earlier language stages and dead languages may be reconstructed with the aid of Cognitive Construction Grammar in combination with the Semantic Map Model.

< Prev   CONTENTS   Next >