The psychological reality of syntactic dependency relations

Resume I review some evidence that dependency structure is psychologically real (and by implication that phrase structure is not), I start by comparing WG dependencies with MTT surface syntax and deep morphology. I then consider the following kinds of evidence: Dependency distance: dependencies tend to be short, and longer dependencies tend to be harder to process. Dependency direction: consistent direction (head-initial or -final) is cognitively simpler but harder to process, so mixed directions are functionally motivated. Dependency classification: grammatical functions are sub-types of dependency, and require the same kinds of cognitive machinery as non-linguistic relations such as kinship relations. Dependency prototypes: dependency types (e.g. ’subject’) are prototypes which combine observable and unobservable features in bundles which allow exceptions. Dependency parsing: dependencies are added one at a time to the head word, rather than by the addition of extra ’mother’ nodes. Dependency lexicalisation: many dependencies are stored as relations between lexical items. Dependency learning: every dependency can be learned by induction from adjacent word pairs, and its properties can be elaborated through experience.