Prosodic and Syntactic clitics
Linguistics is plagued by the fact that many of its basic terms are regularly applied in two or more different meanings: it is really disturbing that such easy-looking descriptions as prosodic clitics, syntactic clitics, prosodic/syntactic approaches to clitics may be associated with different content and refer to different traditions in clitic studies. During the last 5-7 years I have been defending the following relevant distinctions and arguing for the following terminological conventions:
- Prosodic clitics are elements that cannot form a phonological word (= a tact group) when used in isolation.
- Syntactic clitics are elements that take syntactic positions that cannot be filled by any prosodically non-deficient category. [I have been told that this definition of mine is too ‘liberal’, but I am ready to defend it].
- Prosodic theories of clitics are those theories that take into account the possibility that clitics sharing one and the same set of syntactic features may have different phonetic and prosodic features (e.g. be stressed/unstressed, + High tonal / – High tonal, do not exceed a maximum of n syllables/ n+ m syllables etc.). All other theories of clitics, including those calling themselves ‘prosodic’, are, in my mind, syntactic, or at least, not genuine prosodic.
I admit that the proposed distinctions are induced by my own research interest – the integration of clitic typology into word order typology. I highly appreciate other perspectives differing from mine. I like Aikhenvald’s idea that with a scalar approach (clitics occupying a mid-way position from independent word items to fully grammaticalized affixes) the class of clitics in a given language might be heterogeneous – some of the them resembling affixes, some of them similar to root morphemes. I think that this view is compatible with my own analysis, since I am arguing against a unified treatment of all syntactic and prosodic clitics in a language too. I am less sure about Cardinaletti-Starke’s theory of syntactic deficiency which posits an intermediate class of syntactically deficient, but prosodically more or less normal forms – ‘weak pronouns’ & ‘weak adverbs’ – between standard ‘strong form’ and standard clitics: I suspect that such an analysis largely depends on how we define relevant properties of, e.g. Romance pronominal clitics or whether we decide to exclude Romance subject pronouns or Italian loro from this class. But I highly appreciate this theory too and am benefitting from its applications in works by other linguists.
Homophony and Segmental Corruption
A ternary distinction of strong forms, their weak counterparts and clitics is often substantiated by an argument that real clitics should not be homophonous to any other items and their segmental structure should be different from that of the ‘strong’ forms. I do not consider this argument quite convincing. Of course, there might be languages that withhold a contrast of two or more series of weak-stressed elements: it has been maintained that it is the case for some German dialects, where one can find triplets as DENN (‘strong form’) ~ denn (‘weak particle’) ~ dn (‘clitic’), while some positions are associated only with the shortest forms like dn and arguably cannot be filled by forms retaining a vowel. But I do not agree that one may not identify a ‘weak form’ as a clitic in a language with only one relevant grade of prosodic reduction, if it is found in a position, where it can be realized both as /denn/ and as /dn/, and if its behaviour conforms to certain predictions made in clitic typology (for instance, if it is clusterized in a rigid order). The criteria central for a procedure I would suggest are the element’s inability to form a phonological word and to be highlighted with a marked phrasal accent: syncope and apocope show stress conditions and/or phonotactic restrictions rather than syntactic and communicative factors in play. I realize that segmental basis of an element (vowel reduction, restrictions on the onset, outset and number-of-syllable) might be one of the tools of defining the class of clitics in some particular language. I am just protesting against fetichizing such tools and projecting them into UG, if they cannot be validated cross-linguistically. And I would not deny such forms as, e.g. Lat. autem, ergo, item the status of clitics in Plautus’ s texts, even if the scribes did not write them down as /=atm, /=rg, =itm/ in a clause-second position.