Particles, Lexical Categories and Syntactic Functions
The notion of ‘particle’ or ‘discourse marker’ is a catch-all designation. The class of particles cannot be defined in UG in terms of purely syntactic properties: here I take the point of (Zwicky 1985). But Zwicky’s analysis does not prove/disapprove the existence of particle/discourse markers as a lexical class. I am assuming that the class of particles or several classes of discourse markers can be defined in a classification of lexical categories (traditionally known as parts-of-speech classification).
Parts of Speech and Particles
Parts of Speech (Lat. pars orationis) are categories of lexical items defined in terms of morphological or syntactic behavior of the lexical item in question (Croft 1991). The first step in every parts-of-speech classification is the distinction of content words and function words: content word classes (e.g. nouns, verbs, adjectives, adverbs) are normally open, function word classes (e.g. conjunctions and prepositions) are always closed.
Linguists use to define particles according to a ‘leftover principle’: they start from funding the content word/ function word distinction, then define conjunctions, prepositions and postpositions and finally label the remaining function words that do not conform to definitions provided for conjunctions and Ps, as ‘particles’. This much maligned practice raises doubts whether a universal definition of particles is possible. Arnold Zwicky claims that particles do not represent any particular syntactic category (Zwicky 1985: 284). It is plausible that particles do not belong to any classification, where the class of clitics could be defined, namely, to a classification of word forms (Jakobson 1971) or to a hierarchical classification of syntactic expressions (maximal projections >…> heads). Zwicky invites a great deal of particles back, this time under the cover term ‘discourse markers’. It is essential to check whether discourse markers/particles constitute some lexical class or not.
Discourse markers/particles may be either phonological words or expressions lacking features of phonological words, that is, be prosodically deficient elements, clitics. Clitics are usually analyzed as syntactic heads, non-projecting or ‘non-branching’ elements in syntax (Bošković 2002). There is no clear-cut syntactic distinction of clitic-like and non-clitic particles or direct correlation of prosodic and syntactic deficiency. Prosodically deficient elements are arguable always syntactically deficient, cf. (Zwicky 1977), (Halpern 1996), (Aikhenvald 2002: 57), but phonological words may be syntactically deficient too: most discourse markers, which are phonological words, are unable to project in syntax, just as clitic-like discourse markers, cf. (Thurmayr 1989). On this reason the idea of reserving the term ‘particle’ only for those discourse markers, which are not clitic-like, and, vice versa, the idea of restricting the term ‘clitic’ with those expressions, which are not discourse-markers, is not attractive.
Particles and Adverbs
The notion of particle could be abandoned beforehand if the parts-of-speech classification were just a pre-theoretic concept. However, virtually all world’s languages distinguish several content and function word classes. A major but not a fatal obstacle is that different world’s languages have different types of lexical categories, and their boarders appear to be language-specific, cf. (Croft 1991). Some languages lack adverbs as a lexical class. For the remaining languages the distinction of adverbs vs particles/discourse markers is apparently straightforward: adverbs are projecting elements that head TAM projections, particles are non-projecting sentential operators that take whole proposition as (logical) arguments. Discourse markers may originate from manner adverbs and other kinds of adverbials, but the reversed development – particles evolving into adverbs – is not attested.
Cf. a fragment from (Zimmerling 2009, in press), with illustrations from two languages with a clear-cut adverbial/particle distinction – Modern Danish and Modern Russian.simpelt_hen
For more details, see a draft version of my paper ‘Clitic particles and the typology of 2p languages’, to be published in Cambridge Occasional Papers in Linguistics, Copil, vol. 5. CliticPartZimmerling09
 A more accurate formulation would be that at least one of the content word classes is always open is every natural language, though particular languages may have a limited number of lexical items from other content word classes, e.g., adjectives or adverbs.