TAG and Topology

Classical phrase structure tries to collapse syntactic and ordering information. However, this conception of the syntax of language is erroneous because it supposes that word order is always an immediate reflection of the syntactic hierarchy and that any deviation from this constitutes a problem, denoted by terms with negative undertones like scrambling. Modern linguistic frameworks propose a double structure consisting at least of the basic syntactic structure (valency, functor-argument structure, f-structure, deep structure – we will use the term syntactic dependency) and a linear structure (surface structure, c-structure, phrase structure, precedence rules, ...); some frameworks like LFG put this duality at the basis of their system, some others, like HPSG unify the different structures (e.g. SYNSEM and DTRS) in one sign. While the functional subcategorization give rise to few discussions, the different phrase structures proposed for linearization constitute the bone of syntactic contention. One reason seems to be that surface phrase structure attempts to represent as much dependency information as possible, even though there is a large consensus on the impossibility of phrase structure to represent dependency in a satisfying manner. All these approaches appear to be caught in the transformational thinking which supposes two closely related structures: an already ordered deep structure transforms after some movements into and a surface structure which still carries functional information. For English this attempt can go quite far, for languages with case however, the word order serves mainly to represent other communicative goals than functional recovery. Since Evers (1975) proposed a quite flat phrase structure for German, the follow-ups didn’t look any better from a functional point of view, because they all fail to include full functional information in the phrase structure tree.