| A heuristic notion of "structural affinity" is developed, as an equivalence relation holding between sentences of natural language which (i) consist of the same stock of lexical morphemes and (ii) are structured by approximately the same net of "grammatical relations". Structural affinity manifests itself in sentence variants differing with respect to (a) the order of constituents, (b) the prosodic form of a constant surface order, (c) the possible "doubling" of constituents (e.g., dislocations, "subject doubling"), (d) the ellipsis of constituents (e.g., constituent answers), (e) a "global" change of construction (e.g., clefts, pseudo-clefts, passive), (f) the possible combinations of these. Data from French, Italian and Spanish are considered. Based on this notion, the Affinity Thesis can be formulated: For each sentence of natural language of the "unmarked" "normal" form (hitherto concentrated upon in formal grammars) there is an immense number of variants related to each other by "structural affinity" (an "affinity set"). This constitutes a considerable challenge for formal approaches adhering to the Compositionality Principle such as Montague Grammar. To meet this challenge, the possibility is investigated of associating the variants of an "affinity set" with different topic-focus structures. A number of attempts at defining notions such as theme and rheme or topic and focus are reviewed critically and rejected. The question test is accepted as the only reliable basis for determining a heuristic notion of topic and focus. Starting from this notion, a model-theoretic semantics is developed for topic-focus structures and for direct questions on the basis of Thomason's intentional logic. It is shown, as a by-product, that functional application is unsuited for either of these two domains, and the need of a new semantic entity, the "informatum" (alongside "contents" and "referents") is demonstrated. A surface-compositional treatment--including a model-theoretic semantics--of various phenomena of "structural affinity" is then given, as an extension of surface grammars for basic fragments of the aforementioned Romance languages. In the process, a number of new descriptive devices are introduced for Montague Grammar such as rules of derivation operating on whole analysis trees. |
