From AkanaWiki
Jump to: navigation, search
To Be Continued...
4pq1injbok is still working on this article. The contents are incomplete and likely to undergo changes.

DLNAF (a codename; endonym currently unknown) is a Dumic language spoken in the southern coastal regions of Tatakā, between the Potɑnsʉti and Jouki Stəy domains, circa 0YP.



The tables include Romanisation, in italics.


labial dental alveolar palatal velar
stop p p t t ts c č k k
fricative voiceless f f s s ʃ š
voiced v v z z ʒ ž
sonorant nasal m m n n ɲ ň
oral w w r r j y

/k/ is [x] before /t/. Since /kt/ is the only licit surface-level phonemic cluster of stops, this means no stop clusters occur phonetically.

Nasals assimilate in place to following obstruents. Stops after nasals, though not fully voiced, have a later onset of voicelessness than stops in other positions.

In JS-influenced varieties, nasals in posttonic or complex codas can be realised as vowel nasalisation alone, and coda /ɲ/ can be nasalisation plus [j].

[ɾ] varies freely with [r] as a realisation of /r/.


front back
high i i u u
low æ a ɒ o

The allophonic ranges of the low vowels are generally larger than those of the high ones: cardinal [ɛ ɔ] occur as tokens of /æ ɒ/, but cardinal [e o] aren't found as realisations of anything.


The maximal syllable is CCGVGNC, where G is a glide /w j/ and N is a nasal. A maximally elaborate onset is seen in skwo 'fall' perfective or styim 'languages' abs pl, and a maximally elaborate coda in nownc 'nine' or ksowmp 'during the time'. In two successive syllables, the -NC slots of the former and the CC- slots of the latter may not all be filled, which is to say that the longest possible cluster, glides excluded, is -NCC-.

Complex onsets cannot decrease in sonority, nor complex codas increase, where the sonority hierarchy is j w > r > m n ň > p t c č k f s š v z ž. Also, /z ʒ/ are not licit codas.

/w/ does not occur adjacent to /i/ or /u/. /j/, however, occurs freely in these positions.

Hiatic vowels are licit but rare, as in paá 'shell' or aokwó 'irrigate' perfective. In composition i u become y w when next to a vowel in most circumstances (with i > y having higher priority).

The following table shows the statuses of two-element consonant clusters.

  • Bold clusters are allowed.
  • Cells with an entry in lightweight font indicate how the cluster in question is repaired, if formed in the morphology.
  • Empty cells are pairs of consonants which the morphology resists bringing together, whether by vowel epenthesis or preventing vowel deletion. I call these irreparable clusters.
_p _t _c _k _f _s _v _z _m _n _r _w _y
p_ p ft ps kf pf ps pf ps pr pw py
t_ ft t c č kt tf c č tf c č tr tw ty
c_ sp st sk cf c č cf c č cw cy
č_ šp št šk čf c č čf c č čw č
k_ kf kt ks k kf ks kf ks kr kw ky
f_ p ft ps kf f ps v ps fr fw fy
s_ sp st sk sf s š zv z ž sw sy
š_ šp št šk šf s š žv z ž šw š
v_ p ft ps kf f ps v ps vr vw vy
z_ sp st sk sf s š zv z ž zw zy
ž_ šp št šk šf s š žv z ž žw ž
m_ mp nt nc ňč nk mp nc ňč mp nc ňč m n ň mpr mw my
n_ ntr nw ny
ň_ ňw ň
r_ rp rt rc rk rf rs rv rz rm rn r rw ry
w_ wp wt wc wk wf ws wv wz wm wn wr w wy
y_ yp yt yc yk yf ys yv yz ym yn yr yw y

Dialectally, the r+nasal clusters are also treated as irreparable; those in stems in my lexicon insert a u.

Clusters of more than two consonants are allowed as long as they are syllabifiable and all successive pairs of consonants are allowed. The only subtlety is that nasals are deleted before a fricative-stop cluster.

The citation form I use for obligatorily possessed nouns (see below) may appear to violate phonotactics, but this is only because the citation form is an artificial construct shorn of a prefixed syllable which is always present. I use an initial hyphen to indicate the status of these nouns' roots as bound morphemes. Thus -mpaš 'foot' has its illegal initial cluster made unoffensive in forms like cimpaš 'my foot'; and the apparently floating stress in -´mon 'mother' is always in fact moored to a syllable as in cimon 'my mother'.


Absent clitics, stress falls on one of the last two syllables of the word. The coda of an unstressed final syllable, if not empty, can only contain a single /n/. Subject to these rules, the position of stress is weakly contrastive. My Romanisation marks it with an acute accent if it falls on a final syllable where it might not have, as in paá.

Clitics do not move the stress, e.g. kawpun 'wolf' retains its antepenultimate stress in káwpun=i 'is a wolf'. I will usually Romanise words with clitics solid (káwpuni), and leave the stress marks on if the stress isn't where expected.

Loan adaptation

Jouki Stəy is the greatest contemporaneous source of loanwords in DLNAF, notably for cultural terms. Below are the rules in brief for how its sounds are adapted, excluding resolution of impermissible clusters.

JS source p t ts k f s VsV v Cv đ j Cj m n n# r
borrowed as p t c k f s VzV v Cw1 z ž Cy m n n# r

1 If this would produce the sequences wi wu, they are repaired to uy u.

JS source a [ɑ̃] i u ei au ou əi əy əu
borrowed as a o i u ay2 aw2 ow iy2 uy ow

2 In an important older stratum these often become a.



The most visible morphophonological alternation in DLNAF is jostling. Many suffixes, especially of -C(V) shape, induce jostling on their stem. The general rules for jostling are as follows.

  • In a stem whose stressed vowel is low, a glide w or y preceding this vowel is deleted. Otherwise, nothing happens on or before the stress.
  • In a stem whose stressed vowel is high, the stressed vowel is deleted unless this would bring together an irreparable consonant cluster. If deletion forms a cluster which is unsyllabifiable but not irreparable, copies of the deleted vowel are inserted one position to the left or to the right of its former position, or both, as necessary; the total effect is therefore metathesis. (Insertion to the right is rarer, for historical reasons).
  • A stem with final stress ending in a consonant other than w y gains an interstitial vowel between stem and suffix. This is i after palatals or labiodental fricatives č š ž ň f v, and u elsewhere.
  • A stem with a post-tonic high vowel replaces it: i becomes a, and u becomes ow.


A number of stems jostle not exactly as described above, but following other subregularities:

  • Some stems in -y take an interstitial i.
  • Some stems in a low vowel insert a voiced fricative before it, and some in a glide replace the glide with a voiced fricative.
  • A few stems with a stressed a turn this to i.

Unlike the parallel processes in certain other Dumic languages, jostling is applied cyclically to stems to which multiple jostling suffixes are added. Thus stuy 'language', absolutive singular, forms by successive jostling the absolutive plural styim and from it the genitive plural stiymuň.

Another commonality of several suffixes is an intrusive t: when added to a stem with penultimate stress, these suffixes insert an extra -t- between base and suffix. An example, illustrating how I will cite these, is the relativiser and nominaliser (-t)-á. The antipassive -zota, -tota is subject to a similar alternation except that the -t- replaces the suffix-initial consonant.

There are other morphophonological processes in DLNAF, but none of the same generality. I will discuss them below when they become relevant.

The noun

The noun contains the following morphological slots.

-1 0 +1 +2
possessive root number case


The possessive prefixes are formally similar but not identical to the free pronouns, for which see below. Several show or induce alternations.

sing. dual trial plur.
1st excl. c- cita- ciš- cim-
1st incl. k- kuy- kum-
2nd ma- mata- may- mam-
3rd masc. ko- kota- koš- kom-
3rd fem. tun- tunta- tunči- tumu-
indef. i-

The third person singular prefixes, including the indefinite, cause hardening of their base. Hardening replaces a voiced non-nasal initial with a voiceless one, and inserts a consonant before an initial vowel, usually as follows.

basic initial v z ž r w y zero
hardened initial f t c s p š č before i; elsewhere k

There is only one (significant) class of lexical exceptions, these being vowel-initial words that insert p rather than k or č. Relics of hardening are also visible on the second members of some old compounds, and in some obscure prefixed forms.

The prefixes which end in a consonant, other than tun-, sometimes insert a vowel before the stem, i for c- and the trials, and u for k- and the plurals. E.g. -mpaš 'foot' forms cimpaš 'my foot'. This is usually for phonotactic reasons, to ensure irreparable or unsyllabifiable clusters are not formed: for these purposes the m of the plural is treated as unable to occupy the N slot in the syllable structure, only the final C slot. Moreover k- becomes ku- before a stem in k-, as it would be invisible otherwise. Of less clear motivation, c-, k-, and the plurals perform this insertion before a base-initial unstressed vowel. In the same contexts as the plurals insert a vowel, i- becomes vi-.

Any vowel clusters that result from possessive prefixation are resolved by collapsing two identical vowels to one or oa, ao to o, or else changing i to y, or else changing u to w. As an exception, c- added to a stem in unstressed i gives ca. For example, -icita 'pair of eyes' forms cacita 'my eyes', mataystam 'the eyes of you two', tunčistam 'the eyes of them three (fem.)'

In general, stress never retracts onto a possessive prefix: e.g. to va 'water' is formed mavá 'your water'. However, in forms of -´mon 'mother' the stress is always on the possessive prefix.

Some nouns are obligatorily possessed, body parts and kin terms mostly. These must always appear with a possessive prefix. The indefinite possessor, which renders 'somebody's', is a particularly useful choice with these: for instance, the force of vipicita lit. 'somebody's (two) eyes' is not too different from 'a pair of eyes'.

Possessors indexed by these prefixes are normally animate. When there is an overt possessor noun phrase which is animate, DLNAF shows double marking, genitive case on the possessor plus one of the above prefixes. Inanimate possessors forgo the prefix. Thus anasowžaň kopayňiy chief-gen 'the chief's age', but kfoň wayňiy tree-gen age 'the tree's age'.

An exception is found with metaphorical uses of obligatorily possessed nouns, which take one of the third person markers, masculine or feminine as determined by the metaphoric use in question. The prevailing pattern is that if the prototypical metaphorical possessor is large, one gets the masculine; if small, the feminine. So kfoň kompašim tree-gen 'the tree's roots (lit. feet)'.


The only inflectional number contrast in the noun is that between singular and plural; this is a smaller set of contrasts than found in the pronouns. The singular is unmarked, while the plural is marked by the jostling suffix -m. Exceptionally, it converts a posttonic u to o rather than ow: so 'wolf' has sg kawpun, pl kawpom.

The inflectional plural still appears on nouns modified by a numeral or other sign of plurality, e.g. kawpom fira 'three wolves'. In nullar contexts, however, the singular is demanded, e.g. kawpun čipšič 'no wolves' (lit. 'no wolf').


DLNAF exhibits three cases: absolutive, ergative, and genitive. The absolutive is unmarked, while the suffix of the ergative is jostling -ko and that of the genitive is jostling .

Inanimate nouns do not form an ergative.

The genitive is the case governed by all postpositions. The of the genitive often assimilates in place to the initial of a following postposition. E.g. the genitive ikataň of ikata 'city' appears with assimilation in ikatán tay 'from the city' and ikatam=p 'in the city'.

The genitive case is used for titles: anasowžaň Ňakašú chief-gen Ň. 'Chief Ňakašú'. Lacking possessive prefixes, this cannot be confused with a possessive construction.


Personal pronouns show a greater range of number contrasts than nouns: in addition to the singular and plural they decline also in a dual and trial.

Personal pronouns are used only for animate referents. Among them the first person contrasts clusivity; number in the inclusive is interpreted in the obvious way, the series lacking a singular and starting with the dual kuta 'I and thou'. The third person contrasts masculine and feminine; the masculine dominates in mixed-sex groups.

Case on pronouns exhibits the same contrasts, and generally the same functions, as on nouns: but for instance their genitive is less used bare, since possessive prefixes suffice. The next table gives the absolutive forms.

sing. dual trial plur.
1st excl. ci cita ciš cim
1st incl. kuta kuy kum
2nd ma mata may mam
3rd masc. ko kota koš kom
3rd fem. town townta townč tom
'who' va
'what' ron

The case morphology shows some peculiarities. The first person singular ci is unchanged by jostling when case morphs are added, producing ciko, ciň. The feminine singular takes no excrescent u, forming townko, towň, while the feminine dual and trial townta and townč have jostled stems in main vowel u, e.g. ergatives tuntako and tunčiko. The trials in y (like the others) take excrescent i. The remainder jostle regularly, though forms such as čiko, čiň, these belonging to the first exclusive trial, might not be straightaway recognised.

The interrogative pronoun ron 'what' can be used as an indefinite 'something', for unknown realis or irrealis referents or in the antecedent of a conditional. [example] Va 'who' lacks such uses; 'someone' is rendered cant '(a) person'.

The verb



DLNAF verbs show a robust contrast between perfective and imperfective aspect. Each has a characteristic suffix. The perfective suffix is jostling and has allomorphs , -yó, -wó, -pa; the imperfective suffix is -kay, which becomes -kaži- when jostled. The usage of these suffixes is as follows.

  • The ordinary behaviour, that of most underived verbs, is for the imperfective to be formally unmarked and the perfective to show its suffix.
  • In some verbs both the imperfective and perfective are suffixed. A few underived verbs come here, like impf yinkay ~ pf iynwó 'flee, escape' and impf zonkay ~ pf zompa 'turn out (to be)'. Better represented are inchoatives from adjectival roots, not otherwise characterised except by the aspect suffixes: thus ažaň-č 'old' forms impf ažankay ~ pf ažampa 'grow old'.
  • In some verbs the unsuffixed stem is perfective while the imperfective is suffixed. These include inceptives in -siv (impf -sikfay) and cessatives in -momp (impf -monkfay).
  • Some verbs appear in only one aspect, which is always unmarked: e.g. verbalised adjectives have no perfective.

Of the allomorphs of the perfective, typically appears replacing final unstressed a or after palatals or glides, -yó after f v c, and -wó after other final consonants of the stressed syllable, while -pa is usual after unstressed syllables other than those taking . [examples] There are deviations from this scheme: e.g. the perfective of zafi 'drink' is zafyó.

The forms -ó -yó -wó of the perfective all become -wo- when jostled, discarding the variation in glides. An exception is that perfectives in to roots in posttonic -a jostle to -awo-: [example]. This is notable as a rare instance where jostling doesn't simply apply cyclically but is sensitive to the underlying makeup of its input.


DLNAF exhibits five moods. The indicative is unmarked. The others are all marked by jostling suffixes: of these, the imperative -čin, subjunctive II -so, and tentative -m are invariant, while the subjunctive I is -y on a stem with ultimate stress and -c on a stem with penultimate stress.

The imperative is incompatible with all other inflectional suffixes, so that there are really only semantic rather than morphological grounds for classing it as a mood. The agent, always second person, is left implicit; that is, it is the ergative argument of a transitive but the absolutive of an intransitive that gets omitted.

The imperative is polite, even cordial, in tone; a ruder construction uses the indicative. One might well entreat one's guest mawčin 'sit!' in the imperative, but would reserve for an inferior the order

lit. 'you are now sitting', indicative.

The tentative in main clauses renders epistemic uncertainty, i.e. 'probably'.

There is also a strongly dubitative construction built off the negative of the tentative with a dangling moma ≈ 'but':

lit. 'though we (two) probably didn't kill him, ...'; has the force of 'just maybe we (two) killed him'.


Corresponding to the three cases are three relativisers, which formally result in nouns; see the syntax section for their usage. The ergative relativiser is (-t)-žira, the genitive jostling -pa, and the absolutive jostling (-t)-á. (-t)-á and -pa also have derivational uses (see below).

When the intrusive t of the absolutive relativiser is absent, a glide is inserted following the same rules as the perfective, giving allomorphs -á -yá -wá. The absolutive relative of a perfective in -(y,w,)ó is in -(a)wá. It follows that the aspect contrast is neutralised in absolutive relatives of some verbs, like suk 'fall', perfective skwo, abs rel of either aspect skwa.

There is variation in how the absolutive relative is formed to stems in unstressed final -a, between -atá, which follows the normal rules for intrusive t, and simple -(y,w,)á, imitating the perfective.


There are two participles formed directly to the verb root, differing in aspect but both indeterminate in voice. The imperfective participle is formed in jostling , the perfective in -ká. E.g. suk 'fall' forms imperfective participle skuč ≈ 'falling' and perfective participle suká ≈ 'fallen'.

Clitic verbs

The copula has a full form ži, which inflects normally aside from not changing when jostled, and a clitic form =i, which is =y after vowels, and converts a stem in -w to -u=y (with hiatus). The clitic is imperfective indicative and can take no inflection, and is further restricted in that it can be used for assertion of class membership and location, but not for assertion of identity.

So the clitic appears in

'she is a mother'.

But its tentative mood counterpart town imon žim 'she is probably a mother' cannot use the clitic, and neither can town cimon ži 'she is my mother'.

A location example is:

'he is under the blanket'

The verb soc 'say' also possesses a clitic form, =c. It only appears on hosts which phonotactically allow its addition (if stress is ignored). The form =c takes no other suffixes, and is indicative, but is indifferent for aspect and can be used with either perfective or imperfective force. The host of =c is its object, which must be speech but may be either direct or indirect. See the Speech section in Syntax, below, for examples.

Coincidentally, both of these clitics share forms with allomorphs of the subjunctive I suffix, but there is little opportunity for confusion: even if the syntax should chance to be unclear, jostling disambiguates, and failing that stress position. So from tonta 'put' the subjunctive I is tontac but '(they) say "... put"' is tóntac.

The adjective

The bare stem of the adjective is its basic predicative form: ažan 'is old'. All adjective stems are penultimately stressed, unless monosyllabic. The attributive is formed with the suffix , as ažaňč 'old'. This suffix is not jostling, and in fact all adjective stems are of such a shape that appending is phonotactically valid, once the stress is moved to the ultima. In this sketch I cite adjectives in the attributive with a hyphen, along the lines of ažaň-č.

The predicative bare stem carries the default value of all verbal categories, being for example indicative. To cast predicative adjectives in other categories they are verbalised with the formant -č-. For instance ma 'well-behaved, prudent, &c' forms the imperative mačičin 'behave!' (whose first i is a product of jostling). These verbalised adjectives are defective even so, in that they appear in the imperfective only. Also, verbalising -č- cannot appear without at least one further suffix, so mač can only be the attributive form of ma, not any verbalised form.

Minor categories


Postpositions are generally stressless. Arguably many or all of them are clitical; the case is clear for =p 'in' whose form isn't phonotactically valid if freestanding. Aside from =p, though, I write them as separate words.

In the semantics of the spatial postpositions there is no distinction between static and dynamic senses: the postposition serving for 'in position X' also renders 'to position X'.

The lists of senses of the individual postpositions here are not comprehensive.

  • ama: 'facing, across from', 'concerning, with regard to, about', 'in exchange for, for (a price)'
  • ay: 'away from, far from'. Contrasts with tay roughly in deictic centre: in X-ň ay X is near the deictic centre, in X-n tay X is far from it.
  • čama: 'under', 'as, in the role of, (changing) into', 'in (a language)'
  • čaš: 'colliding with, into', '(turning) over, (knocking) down', 'sending into disarray, awkwardly or disorganisedly in'
  • čin: instrumental 'with, using'
  • čir: 'on, onto (the top of)', 'all over, around (an area)'
  • čira: benefactive 'for', 'for the purpose of'
  • oska: 'made of'
  • =p: 'in, into', 'during (a period of time)'. This only occurs following a word which phonotactically allows it as an extra coda consonant, as all genitive case forms do; the genitive invariably assimilates to -m. In other phonological contexts use vina.
  • ra: dative 'to'
  • ta: 'at, to', 'at (a point in time)', 'alongside', 'on, onto (a vertical surface)'
  • tawn: 'without'. This is a recent borrowing from JS, and in some parts of the speech community is not in use. Natively 'without' is rendered rather with the adjective vimpavyi-č 'empty, free (of)' which can take a genitive noun, as in ňišpaň vimpavyič satowčin 'a blanket without holes'.
  • tay: 'from'. See note at ay. The static sense 'arrived' of this postposition is only found in some fixed expressions.
  • vina: 'in(to) the middle of'. This postposition is also the surrogate for =p when the latter is phonotactically impossible.
  • viy: 'near (but not in)', 'out'


Numbers are uninflecting; they serve as cardinals and ordinals without change in form (though with change in syntax).

The basic numbers are ka '1' — vič '2' — fira '3' — zata '4' — fa '5' — šima '6' — tat '7' — kupu '8' — nownc '9' — ko '10' — cič 'hundred(s)' — kyako 'thousand(s)'. One-digit multiples of powers of ten are formed by catenation, lower factor first: vič ko '20', fira cič '300'. Even the expressions for '100' ka cič and '1000' ka kyako carry a multiplier of one; however, ko stands alone for '10' and *ka ko is not found. Sums of these numbers are again expressed by concatenation, largest term first, with the single variation that 'ten' appears as kow if it precedes a units digit. Thus fira cič vič kow zata '324'.

When counting connectedly, "one, two, three...", there are a few sandhi effects seen among the numbers. '9' becomes nows, and '7, 8' may metathesise to tak, tupu. The form nows for '9' even sometimes escapes from this context and sees general use; this is less common for tak and less yet for tupu.


The derivational affixes listed here are not all productive, but they are at least synchronically visible. They produce irregular formations to greater and lesser degrees, which I have not attempted to catalogue here (see instead the lexicon).

Forming nouns

-ta, jostling, forms duals of noun stems. It is improductive, and fails to combine with some stems where it would seem to semantically belong. So alongside -pwa 'hand' forming -pata 'pair of hands', there is -mpaš 'foot' forming no dual, and 'my (two) feet' can only be cimpašim (vič).

As an independent stem, the dual takes inflectional number normally. Thus contrasted are the plurals -vacum 'single eyes' and -istam 'pairs of eyes'.

Some anomalous deployments of the dual are on the quantifiers psuta 'one of the two', munconta 'both'.

(-t)-á, jostling, makes a deverbal or deadjectival noun referring to the absolutive argument. The same morpheme is a relativiser; see the description above for its regular allomorphy. However, the intrusive t tends only to appear on verb stems, or adjectives with a posttonic coda n; in other adjectives á will supplant a final low vowel and glide a final high one. Some old formations are in -ta without final stress.

(-t)-run, jostling, makes agent nouns, usually from verbs.

-pa, jostling, forms nouns of place to stems of any class. Its productive use is confined to a few subcategories, such as naming of buildings or similarly-functioning spaces, e.g. sowčipa 'shack where fish are dried' from sowč 'fish'.

-čin, jostling, forms instrument nouns from verbs. It is also found in nouns like satowčin 'blanket' and tompačin 'pounder, drumstick' with no evident base.

-ňiy has degree nominalisation of adjectives as its only productive function: ku-č 'healthy' forms kuňiy '(degree of) health'. Of course, these readily transfer to less abstract senses.

-zači mostly forms characteristic nicknames on adjectives and nouns: Mažizači 'White' (after hair colour, say, or a favourite garment), Towzači 'Nose' (after a big one).

Forming verbs

(-t)-siv is an inceptive and -momp (jostling) a cessative. Both are deverbal and fully productive, being the normal ways to express 'begin to V' and 'stop Ving'. -t-siv contracts as usual to -civ.

Zero-marking forms inchoatives from adjectives. These however are characterised by taking both aspect markers explicitly.

The allomorphy of the perfective in these inchoatives is different to usual. Adjective roots do not jostle. Monosyllabic adjectives invariably take -pa: so ku-č 'healthy' makes perf kupa (and impf kukay) 'become healthy, get better'. Longer adjectives that are vowel-final take , replacing a low vowel and gliding a high one; those that end in n take -pa. So oyvi-č 'sweet' has perf oyvyó (impf oyvikay) 'turn sweet', and ažaň-č 'old' has perf ažampa (impf ažankay) 'grow old'.

Forming adjectives

-ki, jostling, has been extracted from adjective borrowings from JS and put to use forming adjectives especially of human qualities. This function is reasonably clear for instance in saynaki-č 'quarrelsome, fractious' which is deadjectival, its base being sayna-č 'other, different' (via constructions where it serves for 'of different opinion').

(-t)-uži, jostling, forms denominal adjectives 'having N'. The initial u merges with a stem-final i or u introduced by jostling to give respectively yu and u.

-vyi, jostling, forms denominal adjectives 'like N'.


Noun phrases

Most modifiers, including adjectives, participles, and ordinal numbers, precede the head noun. Cardinal numbers and other quantifiers such as čipšič 'no', pus 'some', išač 'many', and koy 'all' follow it.

Noun phrases in apposition are juxtaposed. The construction is most common for providing background information on a new referent, and as such is typically found in the absolutive, the case most typical for new referents; then both nouns appear without case morphology. In appositions in other cases, the second noun inflects regularly, but the behaviour of the first is variable. [example]

Postpositions follow the whole noun phrase, which provides one of the main pieces of evidence that they are not cases, even phonologically dependent ones like =p 'in':

'in three cities'

The main nominal conjunction is fi, which on its own will be interpreted 'and'. Both conjuncts inflect for case.

To give a lesser degree of thematic foregrounding to one of the conjuncts, it may be extracted from the noun phrase together with the following fi and removed to the position before the verb proper to adverbial elements. Case marking ensures that the loose conjunct can be restored to the correct argument.

'I planted rice, as (incidentally) did A.'; ≈ 'I planted rice with A.'

Two words, which follow a whole NP conjoined with fi, fill the role of 'or', mownta for free-choice contexts and ka (the number 'one') for others.

'I saw either a wolf or a lion by the fields'. (Not *'... a wolf and one lion'.)

Relative clauses

Relative clauses are internally headed. That is, the relative clause, with the head noun inside unextracted, appears whole in its place in the matrix clause. The relativising suffixes on the verb identify the role of the head noun within the relative clause: there are three, corresponding precisely to the cases. With respect to the matrix clause, the relative clause is a complex nominal, and takes case in the usual fashion. So in

'the wolf I saw was drinking water'

the role of 'wolf' is ergative in the matrix clause but absolutive in the relative. As such the clause is nominalised with absolutive relativiser and then gets ergative case marker -ko. Here is an example parallel to the last one of relativisation on a genitive:

'the wolf whose track I saw was drinking water.'

The head of a relative clause cannot lie within a subclause. (One sign of the reanalysis of =c, below, is that this construction can appear inside a relative clause.)

Relative clauses can be contrasted with participles. Participles never take arguments, nor mood. Beyond that the choice is essentially stylistic, with participles usually yielding more frozen, conventionalised senses.


Direct and indirect speech have the same syntax; they differ rather in mood, subjunctive II for indirect and a mood licit in main clauses for direct. The speech itself is typically an absolutive object coming in its usual clausal position.

'she tells me that we're out of honey'
'she says to me "we're out of honey"'

When no hearer is specified it is common for an argumentless ra to appear between speaker and speech anyway, to demarcate the two for easier parsing, especially if the speech is long; in this use it not dissimilar to a quotative marker. This ra can even appear with syntactically parallel verbs not of speech (like those of thinking).

'then I say: "our (honourable) chief, whose glory will never fade, has arrived"'

An innovative pattern allows the subject of the clitic verb =c 'say' to be dropped when it is a third person pronoun. Thus, the clitic shades into acting almost like a marker of hearsay evidentiality.

'they say the chief has arrived'