Kerstin Fischer: A Semantic/Pragmatic Model of Particles - What Does it Take? A model of the semantics/pragmatics of particles first of all has to account for the different interpretations of the items of the class. This means that it has to describe precisely and exhaustively the different readings a particle may get in its different structural contexts. Secondly, it also has to account for the relationship between these different readings. Just listing the different interpretations treats the items under consideration as homonymous, leaving unexplained how the interpretations observable are related, learnable, interpretable, etc. The problem that research has focussed either on finding an abstract semantic meaning of a given particle, failing to account for its functional variability, or on describing the whole list of their pragmatic functions without accounting for the relation between these different readings, Hentschel & Weydt 1989 have called the particle paradox. While finding a common abstract meaning may not be the only way to indicate the relatedness of the interpretations of a given item (see for instance the approaches taken in Östman 1981, Schiffrin 1987, Mosegaard Hansen 1998, Fischer 2000), showing the connections between the different interpretations of an item seems to be an essential requirement for a semantic/pragmatic model of particles. Thirdly, it would be desirable if also the relationship between the different items regarded as particles would be clear, if we could explain why a given particle lexeme may function in different word classes and what the criteria are to regard particular particle uses as instances of particular word classes. The question is thus also what makes a particle a discourse particle, a modal particle, what an adverb or a conjunction. A fourth requirement may be that the model to be developed should be unified such that the different requirements are a natural consequence of the interaction of the components of the model. For instance, the criteria for the word class memberships should not be different from those which guide their distribution and their pragmatic interpretation. A model of the meanings and functions of particles will be presented that fulfills all of these requirements. It is based on the combination of three components: an invariant meaning which is common to all occurrences of a given particle, a set of communicative domains to which the invariant meaning aspects refer and that are word class specific, and a set of linguistic constructions (Goldberg 1995, Fillmore & Kay 1995), form-meaning pairs, which are language dependent. Using the example of German doch which may function as discourse, answer, and modal particle, as well as a conjunction, it will be shown how the invariant meaning, which is described in NSM terms (Wierzbicka 1991, 1992), can motivate the different interpretations possible. What interpretations doch may get is determined by the constructions it may enter. These constructions, which describe the structural contexts in which a particle may occur and which identify the communicative domains to which the particle meaning refer, are word class specific. To sum up, the model proposed accounts for all four requirements: It accounts for the whole functional spectrum a particle may fulfill and for the relation between the different readings by motivating the individual interpretations by the invariant meaning. It also provides the relationship between the items of a word class by specifying the different reference elements which distinguish items from different word classes. Finally, it fulfills the requirement that the definition of the word classes be a natural consequence of the semantic/pragmatic model. Universität Hamburg, Fachbereich Informatik, Arbeitsbereich Natürlichsprachliche Systeme