Zwicky and Pullum 1983
(Language 59.3:502-513)
Cliticization vs. Inflection: English N'T

This paper is notorious for proposing that the contraction N'T (isn't, hasn't, can't, won't) is actually an inflectional affix, and not a cliticised 'not'.  It is most cited not because of that, but because of the six diagnostics they propose for distinguishing clitics from affixes.  They proceed from the assumption that they are distinct, and take as clear cases of clitics the contractions of auxes (attaching to subjects): he's, we've, I'll); while clear cases of (inflectional) affixes are noun plurals, verbal pasts, and the superlative of adjectives.

A theoretical preliminary is that they assume that word-clitic combinability is largely governed by syntactic considerations, while stem-affix combinability is governed by morphological and/or lexical considerations.

Here are the diagnostics:

A.  Clitics can exhibit a low degree of selection with respect to their hosts, while affixes exhibit a high degree of selection with respect to their stems.

B.  Arbitrary gaps in the set of combinations are more characteristic of affixes than clitics.

C.  Morphological idiosyncrasies are more characteristic of affixes than clitics.

D.  Semantic idiosyncrasies are more characteristic of affixes than clitics.

Two more diagnostics follow from a hypothesis of Zwicky's: that cliticization is post-syntactic.  Thus

E.  Syntactic processes can affect affixed words, but not clitic groups.

F.  Clitics can attach to material already containing clitics, but affixes cannot.

Z+P then illustrate A-D with differences between cliticized auxes (to subjects) vs. noun plurals, verb pasts, and adjective superlatives.

Then: -N'T, by all these criteria, is (inflectional) affixation.