Performance Grammar

Gerard Kempen and Karin Harbusch


Performance Grammar (PG) is a psycholinguistically motivated grammar formalism. It aims to describe and explain intuitive judgments and other data concerning the well-formedness of sentences of a language, but at the same time it contributes to accounts of syntactic processing phenomena observable during language comprehension and language production.

Garrett (1975) identifies two stages of syntactic processing: an early `functional' and a later `positional' stage. This distinction has since been adopted by most students of language production (e.g., see (Bock and Levelt 1994)). Accordingly, we assume that syntactic tree formation in PG is a two-stage process. First, an unordered hierarchical structure (`mobile') is assembled out of lexical building blocks. The key operation at work here is feature unification, which also delimits the positional options of the syntactic constituents. During the second step, the branches of the mobile are temporally arranged by a `read-out' module that realizes one positional option of every constituent.

Download the most recent version COMPASSIII (cf. guided tour through COMPASSIII). The user can select words from the lexicon and drag them into the workspace. Here treelets can be combined by moving a tree to an appropriate footnode. It desired treed can be teared appart again. The word order component is realized by the grey quared boxes where the user sorts all filled grammatical function elements into (even accross several levels as for promotion). System reacts with feedback on a background in a traffic-light-inspired color to any action by the student.


Literature on the formalism, its features and its complexity

Literature on applications based on the formalism


For freely available JAVA-implementations of a parser and a generator, respectively, contact one of the authors.

Gerard Kempen
Karin Harbusch

free tracking