Editing Clitic (section)

===Comparison with affixes===
Although the term "clitic" can be used descriptively to refer to any element whose grammatical status is somewhere in between a typical word and a typical affix, linguists have proposed various definitions of "clitic" as a technical term. One common approach is to treat clitics as words that are prosodically deficient: that, like affixes, they cannot appear without a host, and can only form an accentual unit in combination with their host. The term ''postlexical clitic'' is sometimes used for this sense of the term.<ref>Klavans, Judith L. On Clitics and Cliticization: The Interaction of Morphology, Phonology, and Syntax. New York: Garland Pub., 1995. Print.</ref>

Given this basic definition, further criteria are needed to establish a dividing line between clitics and affixes. There is no natural, clear-cut boundary between the two categories (since from a [[Diachrony and synchrony|diachronic point of view]], a given form can move gradually from one to the other by morphologization). However, by identifying clusters of observable properties that are associated with core examples of clitics on the one hand, and core examples of affixes on the other, one can pick out a battery of tests that provide an empirical foundation for a clitic-affix distinction.

An affix syntactically and [[phonology|phonologically]] attaches to a base [[morpheme]] of a limited [[part of speech]], such as a verb, to form a new word.  A clitic syntactically functions above the word level, on the [[phrase]] or [[clause]] level, and attaches only phonetically to the first, last, or only word in the phrase or clause, whichever part of speech the word belongs to.<ref>{{cite book
 | last = Zwicky
 | first = Arnold
 | author-link = Arnold Zwicky
 | year = 1977
 | title = On Clitics
 | publisher = Indiana University Linguistics Club
 | location = Bloomington}}
</ref>
The results of applying these criteria sometimes reveal that elements that have traditionally been called "clitics" actually have the status of affixes (e.g., the Romance pronominal clitics discussed [[#Romance languages|below]]).<ref name=Spencer/>

Zwicky and Pullum postulated five characteristics that distinguish clitics from affixes:<ref name=Spencer/>
* Clitics do not select their hosts.  That is, they are "promiscuous", attaching to whichever word happens to be in the right place.  Affixes do select their host:  They only attach to the word they are connected to semantically, and generally attach to a particular part of speech.
* Clitics do not exhibit arbitrary lexical gaps.  Affixes, on the other hand, are often lexicalized and may simply not occur with certain words.  (English plural -s, for example, does not occur with "child".)
* Clitics do not exhibit morphophonological idiosyncrasies.  That is, they follow the morphophonological rules of the rest of the language.  Affixes may be irregular in this regard.
* Clitics do not exhibit semantic idiosyncrasies.  That is, the meaning of the phrase-plus-clitic is predictable from the meanings of the phrase and the clitic.  Affixes may have irregular meanings.
* Clitics can attach to material already containing clitics (and affixes).  Affixes can attach to other affixes, but not to material containing clitics. That is, an affix may appear between a stem and a clitic, but a clitic may not occur between a stem and an affix to that stem.
An example of differing analyses by different linguists is the discussion of the possessive marker ('s) in English.  Some linguists treat it as an affix, while others treat it as a clitic.<ref>{{Cite book|title=Clitics: An Introduction | first1=Andrew |last1=Spencer| first2=Ana R. |last2=Luis |isbn=9781139560313 |url= https://books.google.com/books?id=wLYgAwAAQBAJ |date= 2012 |publisher= Cambridge University Press |pages=292–293|quote=There are two alternatives that have been explored in recent literature.}}</ref>