A term in Semitic grammar for a reduced form of a noun when it indicates a thing possessed. In European languages generally, it is the possessor who is marked by a special genitive case, and the thing possessed is untouched: the man's horse, Latin equus viri. In Semitic the word 'horse' is put in the construct state.

The construction used is schematically 'horse the-man-of'. The possession precedes the possessor (in typology this is symbolized NG, as opposed to GN), and it has no definite article on 'horse'. The horse is necessarily definite already: it is the one possessed by the noun.

When possessive relationships are nested, all but the last element are construct and all but the first are genitive: 'head horse-of the-man-of'.

In Semitic, adjectives follow their noun (typologically NA), and agree in definiteness as well as gender, number, and case. So 'the big man' is structured as 'the-man the-big'. In possessive constructions the adjective can't go between the two nouns: so in 'horse the-man the-big' it can be either the horse or the man that's big. This could be ambiguous unless there are other agreement cues: for example, in 'horse the-woman the-big', the adjective will be either masculine or feminine depending on which noun it agrees with.

The possessive personal pronouns are expressed by suffixes, so 'my horse' is 'horse-my'. The base they attach to is a reduced form essentially the same as the construct state, though there might be some variations. The pronoun suffixes are complicated and I won't discuss them in detail here.

The state that isn't the construct state is called the absolute state. Now for specific languages.


The following is a description of Biblical Hebrew. The same principles apply to Modern Hebrew but I don't know the detail. The phonetics of my transcription will have to wait till I node more on Biblical Hebrew, but it is worth noting that although a and â are different vowels, in some respects they function like short and long versions of the same vowel. The same is true of e and ê.

Hebrew illustrates the shortening effect well. Even in the Biblical period the case endings had been lost, so there is no genitive, i.e. nothing corresponding to the 'of' in 'horse the-man-of'.

The word sus 'horse' can't shorten, so it just loses its article (hassus 'the horse') in the construct state: sus hammelek 'the king's horse'. The plural of 'horse' is susim, construct state susê.

The word dâbâr 'word' illustrates a common two-syllable pattern. Hebrew had and has a strong final (usually) stress, and earlier vowels often get reduced. So dbar hammelek 'the king's word'. We also see the reduction in the absolute plural: dbârim. In speech the initial db- might have been separated by a light vowel called schwa, but this is unnecessary to the description. However, in the construct plural even further reduction would give *dbrê, but this being quite unpronounceable, the actual form is generated by adding back a helping vowel: dibrê hammelek 'the words of the king'.

Another common pattern of masculine noun is called segholate, from the name of the vowel e (called seghol) in the second syllable. These have the unusual property of non-final stress. Examples are yeled 'boy' and sêper 'book', as well as melek 'king'. In the singular their construct is the same as their absolute. Their absolute plurals are ylâdim and spârim. The construct plurals reflect the difference in first-syllable vowels: yaldê and siprê.

These represent the basic pattern of masculine nouns. Feminines almost always end in , plural -ot, as for example malkâ 'queen', plural mlâkot, or susâ 'mare'. The construct singular is malkat. The appearance of a t here is an oddity of Semitic generally; in Arabic it's known as the ta marbuta. The construct plural is malkot.

Some irregularities are worth mentioning. Most of these are in the nature of shortening the vowel or diphthong, but the words for 'father' and 'brother' are remarkable in all the Semitic languages for actually increasing in the construct state. I'll only give the singular, except for the extraordinary form for 'mouths'.

bên 'son', constr. ben hammelek 'son of the king'
shêm 'name', const. shem
bayit 'house', constr. bêt leh!em 'house of bread, Bethlehem'
?ishshâ 'wife', constr. ?êshet
?âb 'father', constr. ?bi
?âh! 'brother', constr. ?h!i
pe 'mouth', constr. pi
piyyot 'mouths', constr. pipiyyot
As there are no case endings, it is only the possible variation of gender that can disambiguate adjectives:
sus hammelek hatt!ob 'the good horse of the king, the horse of the good king'
susat hammelek hatt!ob 'the mare of the good king'
susat hammelek hatt!obâ 'the good mare of the king'
Semitic languages have dual number: susayim '(two) horses', susâtayim '(two) mares', of which the construct states are respectively susê (same as the plural) and sustê.

A possessed noun is often equivalent in effect to an adjective: har qodesh '(mountain of holiness) holy mountain', ?ish ?lohim '(man of god) godly man'. This has importance in exegesis when combined with the fact that ben 'son of' can mean 'member of': son of Israel = Israelite; son of man = human being; son of god = divine being.


Most Arabic nouns don't change phonetically, but just lose the article in the construct state: baytun 'a house', albaytu 'the house', baytu rrajuli 'the man's house'.

In the dual, nouns end in -aani (nominative) or -ayni (accusative and genitive). In the construct state these are reduced to -aa and -ay. The "sound" masculine plural ending -uuna becomes -uu. "Broken" (irregular) plurals and feminine plurals behave like singulars.

This is complicated by the effect of liaison, with long vowels shortened before the vowel of the article, which drops out. It'll be clearer if I leave out examples.

Some of the same words as in Hebrew are irregular: ?abun 'father' becomes ?abuu, and ?akhun 'brother' becomes ?akhuu. Arab men are often known as their son's father, such as Abu Bakr. famun 'mouth' can be regular or have irregular construct fuu. These lengthened vowels are case endings, so nominative ?abuu has accusative ?abaa and genitive ?abii.

Two words occur only in the construct state: kiltaa, fem. kilataa 'both of'; and dhuu, fem. dhaatu, a word very vaguely meaning 'possessor', but often used with nouns to form adjectives: dhuu &ilmin '(possessor of learning) learned', dhuu maalin '(possessor of money) rich'. It has irregular dual dhawaa, fem. dhawaataa, and plural dhawuu, fem. ?uuluu (sic).

The construct state can also govern an entire clause, as in shahaadatu ?an laa ?ilaahu ?illa llaahi 'the confession that there is no god but God'.


There was no article in Akkadian but there were case endings. All short case endings are omitted in the construct state: beelum 'lord', beel biitim 'lord of the house'. When phonetic reduction would leave an unpronounceable consonant cluster, a short vowel is reinserted: shiprum 'work', shipir beelim 'work of the lord'. The words abum 'father' and akhum 'brother' are again irregular: abi beelim 'father of the lord'.

As with the other languages, adjectives follow both nouns. Case endings may distinguish which one they qualify: shar maatim dannim 'king of the great land', shar maatim dannum 'great king of the land' (with -um nominative agreeing with shar). An alternative construction is to use sha instead of a construct phrase: sharrum dannum sha maatim 'great king of the land'.

The construct state is used before a clause: awaat iqbuu 'the word he said'.

Other Afro-Asiatic languages

The Semitic family is one branch of the Afro-Asiatic family. I've had a look through my books to see how much something like the construct state is used in the other branches, and the answer seems to be a little bit, in fixed expressions, but not thoroughly the way it is in Semitic.

Ancient Egyptian, with no articles, normally used the preposition n 'of', as in per n nb 'house of the lord', but a simple apposition was used in a few closely-knit expressions like nb 3bdw 'lord of Abydos'.

Hausa also has the word order GN, and uses a particle between the nouns agreeing with the thing possessed: masculine in gida na sarki 'the chief's house', feminine in saniya ta sarki 'the chief's cow'. These often contract into a form that superficially resembles a construct state: gidan sarki, saniyar sarki.

In Somali there's a small class of expressions like af-Soomaaliga 'Somali (the Somali language)'. If this had normal grammar you would expect the article 'the', -ga or -ka, to be attached to the word af 'language'.

Outside Afro-Asiatic

Just having order GN and putting the two nouns in apposition isn't enough to count as construct state. Welsh does that, as in y cornell y tref 'the corner of the town', but both words appear in their normal form complete with articles. A construct state is a special state of the first noun.

The Nilo-Saharan language DhoLuo or Luo is the only one I know of that does have a construct state, outside Afro-Asiatic, but there's no reason why it shouldn't occur elsewhere. The DhoLuo example is quoted in several of my books because of the phonetically complicated nature of the changes. In some words voiceless sounds undergo lenition, in others voiced sounds become voiceless, and in others nasals gain an extra plosive. There are also some less natural changes, and some have changes in the advanced tongue root feature of the vowel.

lep 'tongue', constr. lew mon 'woman's tongue'
ot 'house', od
kItabu 'book', kItap
udo 'ostrich', ut
chogo 'bone', chok
opuñ 'heel', opuñj
taya 'lamp', tach
bul 'drum, bund
chOng 'knee', chongg
lUth 'stick', ludh

Log in or register to write something here or to contact authors.