Prepared for Oxford
Encyclopedia of Linguistics, 2nd
by Stephen R. Anderson, Yale University
clitics. [This entry is concerned with proclitics and enclitics, bound elements which, in their phonological behavior, resemble affixes, but in their grammatical function resemble independent words. It comprises two articles:
An Overview; Pronominal Clitics
For related topics, see Morphology, article on Morphology and Syntax; see also Affixation; Inflection; and Words.]
‘Clitic’ is a term conventionally applied as a generalization of the traditional categories proclitic and enclitic (for clitics appearing at the beginning and end, respectively, of an associated host word or phrase). The interest of the category derives from the fact that clitics appear to partake both of the properties of independent words and those of affixes. For references, see Anderson 1992, 2000, Everett 1996, Kaisse 1985, Klavans 1985, Zwicky 1977,1985, and Zwicky & Pullum 1983.
This single category conceals two logically distinct linguistic notions, each with its own research history. The first of these is rooted in traditional grammar: in standard descriptions, enclitics and proclitics are identified as accent-less words (or particles) which depend accentually (or ‘lean’: hence the name, from Greek kli:no `lean’) on an adjacent accented word, and form a prosodic unit together with it. Apart from their defining accentual characteristics, these clitics might show other behavior as a class: for example, Wackernagel 1892 showed that, in several ancient Indo-European languages (and probably in proto-Indo-European as well), clitics clustered immediately after the first word of their containing sentence, regardless of their individual function.
A second research tradition originates in the related observation that some ‘little’ words in some languages appear in positions where the normal syntax of the corresponding syntactic category would not be expected to put them. Most notable among these elements are the pronominal clitics found in the Romance languages and elsewhere [See article on Pronominal Clitics below], and since these are also generally unaccented, the term ‘clitic’ came to be used to describe that behavior as well.
Since the unusual properties of clitics (in both of the senses identified above) are bound up with their ambiguous status between affixes and words, we may consider some criteria which distinguish affixes (determined, bound, reduced) from words (undetermined, free, full):
(a) The typical word, but not the typical affix, has an independent accent.
(b) The phonological shape of a word must be listed in the lexicon, but the phonology of an affix is described in general by saying how the shape of some stem is altered (so that affixes can have ‘process’, as well as affix, realizations) [See Affixation].
(c) Separate, language-specific restrictions can govern the possible phonological shapes of words vs. affixes. In particular, affixes, but not words, are often non-syllabic.
(d) Syntactically, words belong to (lexical) categories, i.e., word classes; but the assignment of affixes to such categories is problematic. [See Parts of Speech.]
(e) Syntactic rules introduce word classes as co-constituents with other syntactic categories; but an affix is syntactically dependent, described by rules which locate it by reference to syntactic elements (e.g., on Nouns, on the head of VP, in the first word of S, at the right edge of NP).
(f) For each affix, morphological rules specify the class of words with which it can occur, and the properties of the resultant combination; but the syntactic rules distributing words typically make reference to phrasal categories rather than to word classes. From this, it follows that affixes are typically very selective in the word classes with which they occur, but words are unconstrained with respect to the word classes that happen to occur adjacent to them.
(g) Syntactic rules cannot alter morphological structure. In particular, syntactic rules cannot allow a word to interrupt a stem + affix combination; a word attached to such a combination must have edge position.
(h) Syntactic rules which introduce a lexical category are blind to the morphology and phonology of its co-constituents, but rules which introduce an affix may be contingent on such properties of its stem. From this it follows that there can be arbitrary gaps and morphophonological idiosyncrasies, including suppletion [q.v.], in the set of stem + affix combinations, but not in the set of word + word combinations.
(i) Alternative orders of words within a constituent are common, but the ordering of an affix—with respect to its stem, and to other such affixes—is fixed (although the same affixes may combine in different orders to express different meanings: e.g., the passive of a causative is not the same as the causative of a passive, and the affixers involved may reflect this in their ordering).
Various mixtures of these properties are possible. Syntactically dependent words, or quasi-clitics, are words by all but criterion (e)—e.g., accented Latin adverbs that are located in ‘second position’ in the sentence. Prosodically dependent words, or ‘leaners’, are words by all but criterion (a)—e.g., the English infinitive marker to. Phrasal affixes are affixes by all but criteria (f) and (g), e.g. the possessive markers in English and Finnish.
A number of types of bound words can be distinguished, although the analytic status of the distinctions is not clear. Optionally bound words, like the English auxiliary clitics, are in stylistic alternation with independent words in the language, and so behave like words on criterion (e); other bound words behave like inflections in this respect. Permutable bound words, like Tagalog particle clitics, have some ordering freedom with respect to one another, and so behave like words on criterion (i); other bound words then behave like affixes. Head-bound words, like the object pronoun clitics in most Romance language, attach to words from a single class (verbs), and so behave like inflections on criterion (f); other bound words here behave like words. The Finnish particle clitics belong to none of these subtypes; they are words on criteria (b) and (d)-(g), but affixes on criteria (a), (c), (e) and (i)—and they can be taken as unmarked representatives of the set of bound words.
From these considerations, we can see that it is useful to distinguish two dimensions of clitic-hood: a phonological one, and a morphosyntactic one. The phonological sense of ‘clitic’ is that of an element which, in contrast with normal lexical items, is prosodically subordinate to adjacent material. The morphosyntactic sense is that of an element whose positioning within a larger syntactic construction is determined by principles other than those of the language’s normal syntax, principles which bear a close similarity to those of morphological affixation. Following Zwicky 1977, we can refer to items displaying these two sorts of property as Simple clitics and Special clitics respectively. We can note that most special clitics are also prosodically dependent, and hence simple clitics as well, though this is not always the case: the Latin adverbs referred to above would appear to be special, but not simple, clitics.
Phonologically dependent elements can be regarded as segments, syllables, perhaps even metrical feet, which do not have a lexically assigned organization into a prosodic word [See Metrical Phonology.] As a result of this orphan status, they are necessarily incorporated into an adjacent word by the language’s rule(s) of ‘Stray Adjunction’, and form a prosodic unit with that word. The principles of Stray Adjunction often result in a clitic’s attaching to a word with which it has no (or even a counter-intuitive) syntactic affiliation. In Kwakw’ala, for example, Determiner clitics at the left edge of nominal expressions attach to the word on their left, resulting in a situation where each constituent appears to terminate in a marker for the case and deictic status of the following nominal.
The morphosyntactic dimension of clitic status has been treated in various ways within different theoretical frameworks. Syntacticians have commonly regarded Pronominal Clitics [q.v.] as syntactically homogeneous with other nominals, generated in the corresponding positions and then moved (e.g., to a position immediately preceding the inflected verb of the clause in French) by syntactic mechanisms that are somehow sensitive to the item’s clitic status. Others, however, have preferred to treat clitics of this sort as introduced directly in their surface position within the phonological form of a sentence by mechanisms closer to those of morphology than to those of syntax (reflecting their similarity to affixes as opposed to words).
In this connection, we can observe that there is a fairly narrow range of positions in which special clitics can appear. These positions can be characterized by three basic parameters (following Klavans 1985, Anderson 1992):
(a) The clitic is located within the phonological realization of some specific syntactic constitutent;
(b) The clitic can be located with respect to the leftmost element, the rightmost element, or the structural head of the constituent within which it occurs; and
(c) The clitic may precede or follow the element with respect to which it is located.
These positions are essentially the same as those that can be occupied by affixes within a word, providing some of the motivation for preferring an account outside of normal syntax. The familiar category of ‘second position’ clitics identified by Wackernagel 1892, for instance, is parallel to infixes that come immediately after a word initial segment, vowel, syllable, etc. (e.g., Kamhmu srnal ‘ear lobe adornment’; cf. sal ‘to place in ear’). Each can be identified by fixing the domain (sentence, word), the anchoring position (initial word, initial consonant) and the position with respect to that element (following, in both cases). The fact that the initial element with respect to which second position clitics are placed is at least in some cases (such as Hittite, some forms of Serbo-Croatian, and most of the examples cited by Wackernagel 1892) is the first word, rather than the first syntactic phrase, within the domain further suggests that mechanisms other than those of the syntax might be required for the description of special clitics and their placement (cf. Anderson 2000). Similarly, the fact that clitics typically display fixed order (as long as meaning remains unchanged), even in languages with considerable freedom of syntactic word order makes them seem more affix-like (cf. criterion (i) above).
The semantic interest of clitics arises from the fact that a syntactically dependent item (like ’s in the man with the hat’s problems or The man with the hat’s going) is semantically interpreted with respect to the whole constituent on which it is dependent, not the word to which it happens to be attached.
Stephen R. Anderson and Arnold M. Zwicky
Anderson, Stephen R. 1992. A-morphous morphology. Cambridge: Cambridge University Press.
Anderson, Stephen R. 2000. Towards an optimal account of second position phenomena.'' in J. Dekkers, et al. [eds.], Optimality Theory: Syntax, Phonology and Acquisition Oxford: Oxford University Press, pp. 302-333.
Everett, Daniel.1996. Why There Are No Clitics. Dallas, TX: Summer Institute of Linguistics and the University of Texas at Arlington Series in Linguistics,
Kaisse, Ellen M. 1985. Connected speech: The interaction of syntax and phonology. Orlando, Fla.: Academic Press.
Klavans, Judith L. 1985. The independence of syntax and phonology in cliticization. Language 61.95-120.
Wackernagel, Jakob. 1892. Über ein Gesetz der indogermanischen Wortstellung. Indogermanische Forschungen 1.333-436.
Zwicky, Arnold M. 1977. On Clitics. Bloomington, IN: Indiana University Linguistics Club.
Zwicky, Arnold M. 1985. Clitics and Particles. Language 61.283-305.
Zwicky, Arnold M., & Geoffrey K. Pullum. 1983. Cliticization vs. inflection: English n’t. Language 59.502-513.