Clitics

The two-level word

Word-hood looks unitary in English: take a string like cat, write it with spaces around it, and you have a word. But linguists routinely uncover cases where a string acts like one word for stress, vowel-harmony, and pronunciation but multiple words for syntax — and vice versa. Clitics live exactly in that gap.

Take English I'll. Phonologically it is one syllable, one stress unit, written without a space. Syntactically it is two words: a pronoun I and an auxiliary will, each separately accessible to syntax (you can negate I will not but contraction can't span negation: *I'lln't isn't possible). The contracted 'll is a phonological dependent — it has no stress of its own, cannot end an utterance — but a syntactic head, occupying the auxiliary position in the clause.

Proclitic and enclitic

The direction of leaning matters. A proclitic attaches to the front of its host (leans right); an enclitic attaches to the end (leans left):

• Proclitics: French j'aime ("I love") — je leans onto aime. Italian l'amico ("the friend") — l' leans onto amico. Greek ο άνθρωπος ("the man") — definite article ο leans rightward.
• Enclitics: Latin Senātus populusque ("the Senate and the People") — -que leans onto populus. Italian dimmi ("tell me") — pronoun mi leans onto imperative di. English they've — 've leans onto they.
• Mesoclitic (rare): European Portuguese future and conditional inflection lets clitics insert between the verb stem and the inflectional ending: dar-te-ei ("I'll give you") = dar ("give") + te ("you") + ei ("will-1SG").

Whether a clitic is proclitic or enclitic is a fixed property of that clitic in that language; speakers cannot reverse direction freely.

Diagnostic tests for clitic-hood

How do you tell a clitic from a word, and a clitic from an affix? Zwicky & Pullum (1983) proposed a battery of tests:

• Clitics attach promiscuously. English possessive 's attaches at the right edge of an NP regardless of category: the queen of England's hat. Affixes attach only to stems of a specific word class.
• Clitics show no allomorphic idiosyncrasy. Plural -s has irregulars (oxen, mice, geese); possessive 's has none. -ed has irregular pasts (went, sang); the contracted 'd doesn't.
• Clitics don't trigger stem-changes. English plural can change stems (knife → knives, leaf → leaves); the genitive clitic doesn't (knife's, leaf's).
• Clitics often have a wider phonological host range. The clitic 'll follows pronouns (I'll), nouns (the dog'll), and even some prepositions (here'll do).
• Clitics outside affixes. When stacked, clitics sit further from the stem than inflectional affixes — walk-ed-'ll would be illegal; the clitic must come last.

French object pronouns: a worked tour

French has two parallel pronoun series — strong (used after prepositions, in isolation, in clefts) and clitic (used immediately before the finite verb). The two never compete for the same slot:

Person	Strong (free)	Subject clitic	Direct obj clitic	Indirect obj clitic	Reflexive clitic
1SG	moi	je	me	me	me
2SG	toi	tu	te	te	te
3SG.M	lui	il	le	lui	se
3SG.F	elle	elle	la	lui	se
1PL	nous	nous	nous	nous	nous
2PL	vous	vous	vous	vous	vous
3PL	eux / elles	ils / elles	les	leur	se

Clitics cannot stand alone. Try answering "Qui as-tu vu?" ("Whom did you see?") with just me — it's ungrammatical. The legitimate isolated form is moi. The two paradigms are completely separate, used in completely different syntactic environments. English doesn't have this split: me works both leaning ("he hit me") and isolated ("Who? Me?").

French clitic clusters

French verbs can host stacks of clitics in a rigid fixed order:

Subject — ne — me/te/se/nous/vous — le/la/les — lui/leur — y — en — Verb — pas

Je    ne   le   lui   en    donne     pas
1SG   NEG  3DO  3IO   PART  give-1SG  NEG
"I don't give any of it to him."

Tu    me   le   donneras
2SG   1IO  3DO  give-FUT-2SG
"You will give it to me."

Elle  s'   y   est    habituée
3SG.F REFL there is    accustomed-PP
"She has gotten used to it."

Six clitics at once is grammatical. The order is rigid — speakers cannot say *je lui le donne (indirect before direct), even though either order is semantically clear. The fixed sequence is one of the strongest arguments that clitics form a single phonological-syntactic complex, not a series of independent pronouns.

Wackernagel's Law: second-position clitics

In 1892 Jacob Wackernagel observed that across the older Indo-European languages — Vedic Sanskrit, Ancient Greek, Latin, Hittite, Old Irish — sentential clitics consistently land in the second position of the clause, regardless of what occupies the first position. The first position can be a single word or an entire phrase; the clitics line up immediately after it:

Language	First constituent	Clitic(s)	Rest	Translation
Latin	Senatus	populusque	Romanus	"the Senate and the Roman People"
Latin	arma	virumque	cano	"arms and the man I sing"
Greek	ouk	esti moi	...	"there isn't to me..."
Hittite	nu	=mu=ššan	kuwapi	"and to me where..."
Serbo-Croatian	Marija	mu ga je	dala	"Maria gave it to him"
Czech	Petr	se mi	nelíbí	"Peter doesn't appeal to me"
Pashto	plōrēž	=mē	ze rāwṛəm	"yesterday I brought it"

Modern Slavic languages — especially Serbo-Croatian, Czech, and Slovene — preserve a remarkably faithful Wackernagel system. The clitic cluster goes in second position even if it splits a noun phrase: Tom su mu bratom (literally "Tom are him brother" = "they are Tom's brother to him") puts the clitics inside what semantically is a single phrase.

English's quiet clitic system

English has fewer clitics than Romance, but they are systematic and carry productive grammatical weight:

• 's = is, has, or possessive. She's tall, she's gone, Mary's hat. The three are homographs but synchronically distinct.
• 're = are. We're here, they're tired. Only attaches to plural pronouns and some plural NPs in casual speech.
• 'll = will. I'll go, they'll see. Productive across pronouns and many NPs (my friend'll come).
• 'd = had or would. He'd left, I'd prefer. Ambiguous between past perfect and conditional.
• 've = have. They've finished, could've been. Often spelled "could of" by non-experts because the contraction sounds identical.
• n't = not. can't, don't, isn't. More affix-like than the others — it triggers stem-allomorphy (will → won't, not willn't).

The English possessive 's is the most syntactically interesting. Affixes attach to a stem; 's attaches to the right edge of an entire NP. The Queen of England's hat places the genitive on England, not on the head noun Queen. That promiscuous attachment is the hallmark of a clitic, not an affix.

Worked example: from word to clitic to affix

Clitics are typically the middle stage in the grammaticalisation pathway full word → clitic → affix. A textbook trajectory: Latin had a free preposition de ("from") that combined with case-marked nouns. Vulgar Latin lost most case endings, leaving de as a near-affixal marker of genitive and ablative meaning. Modern French de is sometimes analysed as proclitic; it forms a tight phonological unit with following determiners (de + le = du, de + les = des) — a phonological coalescence that wouldn't happen with a fully independent word.

Another well-trodden path: Latin had future-tense forms amabo, amabis, amabit ("I will love" etc). These derive historically from ama + a clitic form of habeo ("I have to love" → "I will love"). The original clitic habeo fully fused into the verb. The same construction with a different ordering and slower fusion gave Romance future tenses (Spanish amaré = amar + he).

Adjacent phenomena

• Special clitics vs simple clitics. Zwicky's distinction: simple clitics are reduced versions of free words placed in the same position as those words ('ll for will); special clitics are placed in dedicated positions distinct from the corresponding free form (Romance object clitics, which precede the verb where the strong pronoun would follow).
• Endoclitics. Clitics inserted inside a word, splitting morphemes. Pashto and some Mayan languages show genuine endoclisis — extremely rare and theoretically fraught.
• Climbing. In Italian and Spanish, an object clitic of a non-finite verb may "climb" onto the finite verb above it: Italian lo voglio vedere = voglio vederlo ("I want to see it"). The clitic moves out of its base clause for prosodic reasons.
• Pro-drop and clitics. In Italian and Spanish, subject clitics aren't even needed; the verbal inflection carries person and number. French requires subject clitics (lost the inflectional richness), Italian doesn't.
• Auxiliary contraction restrictions. English contraction is blocked at clause edges: *What is it the boy's? (intended as "what is it that the boy is?") is bad because the auxiliary can't host the clitic across an empty position. Diagnostic for syntactic structure.

Why clitics matter

• Two-level theories of word-hood. Clitics force the distinction between phonological word and morphosyntactic word.
• Diachronic syntax. Modern Slavic Wackernagel placement is a 4000-year archaism preserving Proto-Indo-European clitic syntax.
• L2 acquisition. Romance clitic systems are notoriously hard for English speakers — order and placement bear no resemblance to English equivalents.
• NLP and parsing. Word tokenization in clitic-rich languages requires splitting at clitic boundaries; off-the-shelf English tokenizers fail on French j'aime or Italian dimmelo.
• Theoretical syntax. Clitic placement diagnostics motivate verb movement, head movement, and the entire framework of feature-driven cliticization.
• Diagnostic tool. Wherever you find a clitic, you find a tightly local syntactic relation — clitics rarely climb past tensed clause boundaries.

Common pitfalls

• Confusing clitics with abbreviations. Written contractions like don't aren't an orthographic shortcut for do not — they reflect a real reduced spoken form with its own phonology.
• Treating Romance object pronouns as plain pronouns. French me and English me look superficially equivalent but behave very differently. Me in English answers a question alone; French me cannot.
• Assuming clitics are always reduced auxiliaries. Many clitics are full pronouns in their language's core grammar, not phonological reductions.
• Mixing up clitic order across languages. French has dative-after-accusative (le lui); Spanish has dative-before-accusative (se lo). Italian distinguishes the two patterns with separate forms (glielo bundles them). The orderings are language-specific even within Romance.
• Calling all reduced forms clitics. Phonetic reductions in fast speech (gonna, wanna, lemme) are sometimes treated as clitics, sometimes as casual phonological forms — diagnostics are not always sharp.
• Ignoring the gradient. Clitics live on a continuum between full word and affix. Many morphemes are intermediate or actively grammaticalising; treating clitic-hood as binary loses the diachronic and gradient picture.