Are clitics intermediate between lexemes and affixes in grammaticalization?

It is a widespread stereotype that cliticization is an intermediate stage between free lexeme and affix status, and many overview texts and textbooks present a scale such as the following:

content item > grammatical word > clitic > inflectional affix (> Ø)

For example, Spencer & Luís (2012: §2.1) say that “inflections are generally thought to arise in historical language change through a stage of cliticization”. (This kind of scale is often called a “grammaticalization cline”; see Konvička’s very informative 2019 blogpost about the history of this idea).

But is this stereotype true? At the last International Conference of Historical Linguistics in Oxford  (August 2022), I gave a talk which is summarized in this blogpost. My conclusion was that in order to understand what is happening in “coalescence” in grammaticalization, we need to be clearer about our basic concepts, especially what we mean by clitic. On a reasonable definition of the term clitic, most types of clitics do not become affixes, and most types of affixes did not pass through a clitic stage.

Here is the definition of clitic that I propose in a separate paper:

Clitic: A clitic is a bound morph that is neither an affix nor a root.

This definition relies on earlier definitions of the terms morph, bound, affix, and root. All of these are widely thought to be technical terms of linguistics, but they are rarely defined in such a way that they have clear boundaries.

Few attempts at defining “affix” and “clitic” have been made (Zwicky & Pullum 1983 provide a set of “diagnostic symptoms”, which are not really criteria; cf. Haspelmath 2015). The problems with defining word are well-known, too (Haspelmath 2011a; Tallman 2020). Many authors appeal to “tightness of bonding” or “coalescence” when characterizing affixes, but what exactly is implied by this is not always clear, and multiple dimensions need to be taken into account (Börjars & Harries 2008) – and these dimensions need not correlate with each other (van Gijn & Zúñiga 2014).

But we need such clarity of terms if we want to make claims that can be tested in an objective way. In earlier work, I have defined a morph as “a minimal form”, a bound form as “a form that does not occur in isolation”, and an affix as “a bound morph that always occurs on roots of the same class (=that is class-selective)”. A root is a contentful morph that denotes an object, an action or a property.

I have not seen an alternative definition of “clitic” that can be applied readily to all languages, so in the following, I will assume the above definition. But I should probably mention here that many linguists associate the term “clitic” with “prosodic deficiency”, something that plays no role here. The reason why the above definition makes no reference to prosody is that the various criteria for “deficiency” are quite heterogeneous (and thus not suitable for a straightforward definition); moreover, most affixes are prosodically deficient (in some way), too. So it seems best to say that clitics are nonroot bound morphs which are not affixes (i.e. which are nonselective or “promiscuous” in that they occur on anchors of different classes).

(It is also sometimes thought that clitics differ from affixes in that they are “phonologically attached” to a lesser degree. But what exactly does “attachment” mean? In my 2021 paper, I coined the new term welding to refer to segmental interaction, as when an affix’s shape depends on the root’s shape or vice versa. But in many cases, there is no welding between affix and root, so this cannot be a necessary criterion for affixhood.)

So let us get back to the original question: Is there a continuous scale (or “cline”) from lexicality to affixhood? (or even “discourse > syntax > morphology > zero”, as in Givón 1979?) This may make sense at an intuitive level, but in order to test it scientifically, we need ways of measuring an item’s position on the scale. Rigorous quantitative testing requires clear qualitative comparative concepts. In the Oxford talk, I did not do this, but I suggested (impressionistically) that

  • most types of clitics do not become affixes, and
  • most types of affixes did not pass through an earlier clitic stage

Let us look at the first claim. If clitics are defined as bound morphs that are neither affixes nor roots (as above), then there are quite a few kinds of clitics that evidently do not tend to become affixes, e.g. discourse particles (like English however), 2nd position clitics, coordinator clitics, and question particles:

However, it is true that person indexes do seem to become affixes in some well-known cases, e.g. in the Romance languages. Old Portuguese and Old Spanish seem to attest a clitic stage of the weak person forms (Pescarini 2021):

Modern Romance languages do not allow this sort of “interpolation”: Object clitics must immediately precede the verb, or immediately follow the verb. In other words, they have affixal status (this is largely uncontroversial among Romance scholars).

A particularly good case of gradual grammaticalization is the development from 2nd position clitics to verbal person markers in Polish, described by Andersen (1987):

Andersen actually provides frequency figures showing how the 2nd position usage has decreased in Polish over time:

So person-number clitics do sometimes become affixes, but this is far less common than seems to be implied by the literature.

Now let us look at the second claim, that most types of affixes did not pass through an earlier clitic stage. The most typical affixes are the following:

verbs:  tense-aspect and person affixes

nouns: case, number and gender affixes

I think that most of these do not derive from elements that are usually called clitics, though I have not (so far) tried to document this systematically.

The grammaticalization origin of number and gender affixes is often unclear, though articles seem to play an important role. Articles may be clitics, but they may also be affixes (= occur always on the noun). There do not seem to be clear cases of affixal articles deriving from clitic articles. The Nordic suffixed article does not derive from a clitic (Börjars & Harries 2008): The change *barn hit ‘that child’ > barn-it > barn-et ‘the child’ is a change from free word to affix, but it does not imply any intermediate clitic stage.

Tense-aspect affixes from auxiliary verbs (e.g. Bybee & Dahl 1989), but auxiliaries are very often class-selective. For example, the Spanish perfect auxiliary haber always occurs immediately in front of the verb, and in many languages, this is characteristic of auxiliaries throughout their history.

Case affixes are generally thought to derive from auxiliary (relational) nouns (e.g. König 2011), but such “auxiliary nouns” are very often selective, too. For example, Lezgian k’an-ik ‘under’ (from k’an- ‘bottom’) is originally a noun that always occurs immediately next to its nominal complement:

But as I said above, person affixes sometimes go back to nonselective person forms, as in Romance and Polish. How often this is the case is unclear, as the literature tends to repeat the same kinds of well-known examples from European languages. “Givón’s Cycle” (Givón 1976) has become a kind of stereotype, but in actual fact, person(-number) affixes can arise in a wide variety of ways, as recently highlighted by Seržant (2021). So the clitic > affix change may be much less common in the world’s languages than it appears from the literature.

So the following was the conclusion of my Oxford ICHL talk:

  • Many linguists have the intuition that there are degrees of bondedness – more “tightly bonded” grammatical forms and more “loosely bonded” forms.
  • Often linguists seem to assume that the tightness of bonding (or “coalescence”) is of a phonological nature (they often talk about “phonological boundness”) – but there is no clear conceptual basis for this.
  • Comparative grammar research needs clear comparative concepts, and clitic can be defined clearly only with respect to their indiscriminate behaviour (contrasting with class-selective affixes).
  • It is of course plausible that grammaticalization involves not only shortening (= fewer segments) and boundness (= non-occurrence in isolation), but also some kind of “phonologically close association”, but this has not been formulated clearly yet.
  • At present, the well-known scale “lexeme > clitic > affix” is not more than a stereotype, and we need to replace it by a more rigorous formulation.


Andersen, Henning. 1987. From auxiliary to desinence. In Harris, Martin & Ramat, Paolo (eds.), Historical development of auxiliaries, 21–52. Berlin: Mouton de Gruyter.

Börjars, Kersti & Harries, Pauline. 2008. The clitic-affix distinction, historical change, and Scandinavian bound definiteness marking. Journal of Germanic Linguistics 20(4). 289–350. (doi:10.1017/S1470542708000068)

Bybee, Joan L. & Dahl, Östen. 1989. The creation of tense and aspect systems in the languages of the world. Studies in Language 13(1). 51–103.

Givón, Talmy. 1976. Topic, pronoun and grammatical agreement. In Li, Charles N. (ed.), Subject and topic. New York: Academic Press.

Givón, Talmy. 1979. On understanding grammar. New York: Academic Press.

Haspelmath, Martin. 1993. A grammar of Lezgian (Mouton Grammar Library 9). Berlin: Mouton de Gruyter.

Haspelmath, Martin. 2011a. The indeterminacy of word segmentation and the nature of morphology and syntax. Folia Linguistica 45(1). 31–80. (doi:10.1515/flin-2017-1005)

Haspelmath, Martin. 2011b. The gradual coalescence into “words” in grammaticalization. In Narrog, Heiko & Heine, Bernd (eds.), The Oxford handbook of grammaticalization, 342–355. Oxford: Oxford University Press.

Haspelmath, Martin. 2015. Defining vs. diagnosing linguistic categories: A case study of clitic phenomena. In Błaszczak, Joanna & Klimek-Jankowska, Dorota & Migdalski, Krzysztof (eds.), How categorical are categories? New approaches to the old questions of noun, verb, and adjective, 273–304. Berlin: De Gruyter Mouton.

Haspelmath, Martin. 2020. The morph as a minimal linguistic form. Morphology 30(2). 117–134. (doi:10.1007/s11525-020-09355-5)

Haspelmath, Martin. 2021. Bound forms, welded forms, and affixes:  Basic concepts for morphological comparison. Voprosy Jazykoznanija 2021(1). 7–28. (doi:10.31857/0373-658X.2021.1.7-28)

Lowe, John J. 2016. Clitics: Separating syntax and prosody. Journal of Linguistics. Cambridge University Press 52(2). 375–419. (doi:10.1017/S002222671500002X)

Pescarini, Diego. 2021. Romance object clitics: Microvariation and linguistic change. Oxford: Oxford University Press.

Seržant, Ilja A. 2021. Cyclic changes in verbal person-number indexes are unlikely. Folia Linguistica 55(s42–s1). 49–86. (doi:10.1515/flin-2021-2014)

Spencer, Andrew & Luís, Ana R. 2012. Clitics. Cambridge: Cambridge University Press.

Tallman, Adam J. R. 2020. Beyond grammatical and phonological words. Language and Linguistics Compass 14(2). e12364. (doi:10.1111/lnc3.12364)
van Gijn, Rik & Zúñiga, Fernando. 2014. Word and the Americanist perspective. Morphology 24(3). 135–160. (doi:10.1007/s11525-014-9242-z)
Zwicky, Arnold M. & Pullum, Geoffrey K. 1983. Cliticization vs. inflection: English n’t. Language 59(3). 502–513.

Leave a Reply

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.