Reciprocation strategies and the syntax of reciprocal constructions

Reciprocal constructions in the world's languages are typically associated with unusual syntactic behaviour and other anomalous properties. This thesis examines reciprocal constructions in Icelandic (Germanic), Malagasy (Austronesian) and Swahili (Bantu) in order to: firstly, determine the variability of their non-standard syntax; and secondly, to analyse these constructions within the theoretical framework of Lexical Functional Grammar (LFG) with a view to explaining the relationship between them and the reasons for their non-standard behaviour. These analyses reveal that the syntactic and, to a lesser extent, semantic behaviour of reciprocal constructions arise not from structural features (such as whether the reciprocal construction is formed from a clitic, affix etc.), but from more fundamental reciprocation strategies by which asymmetric predicates are made to describe symmetric situations. I demonstrate that the highly irregular case marking and word order associated with the Icelandic (nominal) reciprocal construction is intimately related to two of these reciprocation strategies. As a result of two extensive surveys undertaken to document its behaviour and history (a corpus based survey of the reciprocal construction in Old Icelandic (c. 1200 CE) and an online survey of its usage by 500 Icelanders), I show that the variable word order and case marking of this construction can be understood as a consequence of a change in its underlying reciprocation mechanism from the compositional to the pronominal – a change that is for the most part indiscernible in its morphosyntactic expression. In contrast to the phrasally formed Icelandic reciprocal construction, the Malagasy and Swahili reciprocal constructions are both verbally marked. As existing analyses of apparently similar constructions in Romance, Hungarian (Uralic) and Chicheŵa (Bantu) are unable to explain their syntax, I propose two alternative analyses. The first is a novel analysis of the reciprocal construction in Malagasy as being derived from a covert application of the pronominal reciprocation strategy. This introduces a mismatch between functional and constituent structure and I show that this mismatch is required to explain the presence of reciprocated control and possession constructions, as well as its surprising similarities to the Icelandic reciprocal construction. The reciprocal morpheme in Swahili is associated with two, distinct, reciprocal constructions: a monadic construction that groups the participants in the relation into the subject NP while losing an object NP; and a dyadic construction that places one participant in the subject NP and another in a comitative phrase. I show that the syntax of these constructions (and the relation between them) results from a reciprocation strategy that derives symmetric lexemes (much like the naturally symmetric verbs in English, e.g., quarrel, argue). Most of these reciprocation strategies require that the participants in the symmetric event be in a single NP. This means the normative syntactic principles that link a predicate to its grammatical functions cannot straightforwardly accommodate reciprocated predicates – requiring that these principles be abused or at least stretched. I show this to be the root cause of the unusual syntactic behaviour associated with many of these constructions. Finally, as the morphosyntactic expression of a reciprocal construction is largely divorced from its underlying reciprocation strategy, I advance the view that typologies of reciprocal constructions should be orientated around these mechanisms rather than structural properties.