Construct marking: Markers on modified nouns to signal the presence of a modifier (some comments on Creissels 2018)

Many languages have a genitive flag in their adpossessive nominals, i.e. a case-marker or adposition on the possessor nominal (e.g. English [Kim’s] money, the roof [of the house]). But alternatively, they may also have a marker on the possessed noun – an antigenitive marker. For example, Ge’ez (an ancient Semitic language of Africa) had an antigenitive suffix -a, as in wald ‘son’, wald-a nəguś [son-ANTG king] ‘the son of the king’.

More generally, languages may have (non-cross-referencing) markers that occur on the noun when it is modified, to signal the presence of a modifier. Such markers are called construct markers by Denis Creissels (2009; 2018), following the tradition of Semitic linguistics (cf. the construct suffix -t in Moroccan Arabic, e.g. xala ‘aunt’, xal-t l-bənt [aunt-CS DEF-girl] ‘the aunt of the girl’). There do not seem to be many languages with a general construct marker, such as -e in Persian (called Ezâfe marker), which is used both with possessor modifier and adjective modifiers, e.g.

ketâb-e târix [book-CS history] ‘history book’

ketâb-e sabz [book-CS green] ‘green book’

Nevertheless, Creissels (2018: §3.4) proposes the following universals (slightly reformulated here):

Universal 1: If a language has a single construct marker, it almost always marks the noun as having an adpossessor (possibly among other modifier types).

Universal 2: If a language has two construct markers, one of them is always an antigenitive marker.

There is not enough evidence to be sure about these universals, because construct markers have not been studied for very many languages, but the proposals are no doubt interesting. And Creissels mentions construct-form constructions from languages of Africa, North and South America, Eurasia and the Pacific. Maybe one of the reasons these markers have not been studied from a comparative perspective is that there has not been any widely known general term (cf. Haspelmath 2021: §3 on the role of transparent terminology).

So Creissels’s terminological discussion is very important, too, and he devotes 3 full pages to it (see also Rießler 2016: §4.5 for related discussion). For some reason, he does not mention the terms antigenitive (e.g. Andersen 1988) and adnumerative (for Russian forms like rjad-á ‘row’ which signal the presence of a cardinal modifier; e.g. Igartua & Madariaga 2016). But he discusses Dixon’s (2010) term pertensive (a cover term for antigenitive markers and adpossessive indexes, i.e. all kinds of adpossessive “head markers”), and he points out that the term “annexed state” in Berber linguistics (also sometimes called “construct state” on the analogy of Semitic linguistics) is confusing because this is really a kind of case (cf. Arkadiev 2015).

Creissels’s definition is slightly different from the one I gave above, so let us look at this. Creissels says that “the two essential characteristics of construct marking are that:

(i) it is obligatory if the noun fulfills the role of head in a given type of noun – modifier construction; (ii) it does not cross-reference features of the modifier that conditions its use.”

It seems to me that one does not need the term “head” here, and one could simply talk about nouns and their modifiers (cf. Haspelmath 2019, where I discuss flagging and indexing in contrast to head and dependent marking). But more importantly, I do not see the need to require obligatoriness. I cannot think of a clear case of an optional construct marker, but if a marker like -t in Arabic (recall xal-t l-bənt [aunt-CS DEF-girl] ‘the aunt of the girl’) were optional, it would still be a construct marker, because it is conditioned by the presence of the adpossessor (and thus signals it).

Creissels also wants to exclude cases where there is a marker that occurs only in (compound-like) binominal constructions, e.g. Amharic bet-ä mädhanit [house-marker medicine] ‘pharmacy’. But the definition he gives does not strictly speaking exclude them, because it does not mention that the head and the modifier have to be “syntactic”. It is quite normal to say that in a compound like flower-pot, there is a head (pot) and a modifier (flower-). It is true that such binominals are often semantically (more or less) opaque, but when they are transparent, there does not seem to be any reason to reject the term construct marker for such elements. Moreover, it has often been noted that the division between morphology and syntax is not watertight (and perhaps primarily based on spelling habits).

However, there is another problem, again related to the problem of identifying “morphological” properties reliably. In order to identify a marker as a head marker or dependent marker, we need to identify its location (its “locus”, as Johanna Nichols says). But this is not always clear – in particular, where is the Persian Ezâfe marker located? Above, I said that it is a suffix on the noun (ketâb-e sabz ‘green book’), but alternatively, one could perhaps say that it is a attributivizing prefix on the adnominal modifier:

ketâb   e-sabz [book ATTR-green] ‘green book’

It is not clear that this is not a better way of thinking of this marker, and indeed Creissels (2018) adopts it, because this marker occurs not only between a noun and a modifying adjective, but also following an adjective if another modifier follows:

ketâb e-sabz e-bi arzeš [book ATTR-green ATTR-without value] ‘green book without value’

So Creissels calls this an adnominalizer (my preferred term is attributivizer), even though he uses the conventional spelling (with -e attached to the preceding word). Identifying marker location is often easy, so linguists generally take it for granted, but such cases remind us that we cannot even take this for granted.

Finally, Creissels’s way of talking is different from the way I have talked so far is by using the abstract noun “construct marking”, where I used the concrete term “construct marker” (which is defined as a kind of morph, Haspelmath 2020). This is because a marker is something that can be objectively observed, whereas “marking” refers to a kind of process and is thus more abstract and potentially contentious. For example, Mende (a Mande language) has an anlaut alternation, as in toko/loko ‘arm’:

tokó [arm] ‘arm’

ndopo-í loko-í [child-DEF arm-DEF] ‘the child’s arm’

What is happening here? Is there a special construct form loko-, somehow derived from the basic form tokó? Or is there a special “absolute” form tokó, somehow derived from the basic form loko-? Creissels notes that the latter view may be more appropriate, not only diachronically, but also synchronically. So do we want to say that construct marking can be subtractive? I would not want to take a stand on this, and from a comparative perspective, it seems best to stay neutral (after all, we want to test universals like those listed above, and this must be possible in an objective way). Other cases are clearer, especially when there is neutralization (e.g. tonal neutralization in Dogon languages like Tommo So).

And of course, we do want to talk about “marking” elsewhere, e.g. “tense marking” and “number marking”, not only about tense-marking morphs and number-marking morphs, because both tense and number can be marked by stem changes (e.g. English sing/sang, mouse/mice). But crucially, which tense or number is marked becomes clear only when there is a marker. If all English present/past pairs were of the type sing/sang, find/found, swim/swam, there would be no basis for saying that “the past tense is marked by vowel change”, because the past-tense form could be the basic form (and the present-tense form derived). And if a marker is obligatory, then one could always say that the absence of the marker constitutes “zero-marking”. All these are interesting considerations, but for typology, it is generally safer to concentrate on concrete forms, i.e. markers.

Finally, there is a (smaller) problem with “construct marking”: It is not fully parallel to “tense marking” or “case marking” in that while tense marking serves to mark (or signal) tense, construct marking does not serve to “mark (or signal) construct”. It marks adnominal modification. So unlike “tense marker”, the term “construct marker” is not transparent. A construct marker is a marker of modification that obeys certain further conditions, and “construct” does not exost outside of “construct marker”. This is a bit confusing, and it seems to have led Creissels to use the odd combination “construct form marker” in addition to construct marker (see also Van de Velde 2017). This does not lead to any misunderstandings, but if we want a clear and consistent terminology, we should also pay attention to such smaller issues.


Andersen, Torben. 1988. Downstep in Päri: The tone system of a western Nilotic language. Studies in African Linguistics 19(3). 261–316.

Arkadiev, Peter. 2015. The Berber “state” distinction: Dependent marking after all? A commentary on Mettouchi & Frajzyngier (2013). Linguistic Typology. De Gruyter Mouton 19(1). 87–110. (doi:10.1515/lingty-2015-0003)

Creissels, Denis. 2009. Construct forms of nouns in African languages. In Peter K. Austin, David Nathan & Peter Sells, Oliver Bond, Monik Charette (ed.), Proceedings of Conference on Language Documentation and Linguistic Theory 2, 73–82. London: SOAS. (

Creissels, Denis. 2018. Binominals and construct marking. In Masini, Francesca & Pepper, Steve (eds.), (A volume on binominal constructions, to appear). (

Dixon, R. M. W. 2010. Basic linguistic theory. Vol. 2. Oxford: Oxford University Press.

Haspelmath, Martin. 2019. Indexing and flagging, and head and dependent marking. Te Reo 62(1). 93–115. (doi:10.17617/2.3168042)

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. The structural uniqueness of languages and the value of comparison for description. Asian Languages and Linguistics (to appear).

Igartua, Iván & Madariaga, Nerea. 2016. Russian lower numeral phrases in typological and diachronic perspective. Rhema. Рема. Москва (

Rießler, Michael. 2016. Adjective attribution (Studies in Diversity Linguistics 2). Berlin: Language Science Press. (

Van de Velde, Mark. 2017. The augment as a construct form marker in Eton relative clause constructions. In Atindogbé, Gratien Gualbert & Grollemund, Rebecca (eds.), Relative clauses in Cameroonian languages: Structure, function and semantics. Berlin: De Gruyter Mouton.


OpenEdition suggests that you cite this post as follows:
Martin Haspelmath (August 26, 2020). Construct marking: Markers on modified nouns to signal the presence of a modifier (some comments on Creissels 2018). Diversity Linguistics Comment. Retrieved July 19, 2024 from

3 thoughts on “Construct marking: Markers on modified nouns to signal the presence of a modifier (some comments on Creissels 2018)

  1. The term *antigenitive* perhaps has the disadvantage of suggesting (to some) that it differs from a *genitive* in a way that is somehow similar to the way an *antipassive* differs from a *passive*. I agree with Dixon that terms like **antiaccusative** or **antiergative** seem cloudy at best.

    So — for adnominal “possession”, anyway — I’ve been advocating for Dixon’s *pertensive* in correspondence with contributors to the second edition of the Uralic Languages volume I’m editing with Riitta Valijärvi.

    • True, “antigenitive” is not fully transparent (like “construct marker”, as I pointed out). But “anticausative” is not fully transparent either, and neither is “antipassive”. The main advantage of “antigentive” is that it has already been used (by a few people) and is partially transparent. As for “pertensive”, Dixon uses it in the sense of “antigentive or adpossessive index”, and such a term may well be useful. It is not a competitor of “antigenitive”.

      • and maybe an advantage of the *anti-* of *antigenitive* is that it implies some kind of opposite of *genitive*, which it is in the sense the locus is reversed — in Dixon’s terms, on D rather than on R.
        Other variables, though, are difficult to build into any term. I’m thinking of genitives that index more than the presence of a possessum (i.e., features of D, like number or gender), or pertensives/antigenitives that index features of R (other than “third person”).

Leave a Reply

Your email address will not be published. Required fields are marked *

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