In the shower one day I had a flash of inspiration and finally knew how Kijeb tense-aspect-mood marking should be organized. I’ll try to manage to describe it without access to my dictionary.

Many aspects (sic!) of the following wholly or partly obsoletize things said on the Kijeb page on FrathWiki.

The order of the markers (slots) are as follows:

Root-aspect-(inverse)-(mood)-(tense)-(person)-(person)

Items in parentheses may be absent/zero-marked. Intransitive verbs have only one person marker while a transitive verb has two, but one or both of them may be zero-marked or haplologized. When there are two person markers their relative order is determined by the animacy hierarchy as follows

^
|
| First person
| Second person
| Third person animate (singular)
      | proximal
      | medial
      | distal
| (Third person animate plural)
      | proximal
      | medial
      | distal
| Third person inanimate (singular)
      | proximal
      | medial
      | distal
| (Third person inanimate plural)
      | proximal
      | medial
      | distal
|
v

so that markers for persons higher in the list always precede markers for persons lower in the list. Which one is agent and patient is signalled by the presence of the inverse marker if the lower person is agent.

There is no Algonkian-style obviative marking if both arguments are 3rd person animate, but there is a system of free pronouns marked for increasing remoteness from the speaker.

I call them “proximal — medial — distal” and translate them somewhat loosely as ‘this — that — yon’ (in German ‘der hier — der da — jener’. In practice these can be used to serve obviation in that they rank differently, as amended in the list above. They can also be used as a kind of ‘articles’ but what’s more any new referent is automatically assigned a place in the proximal— medial—distal hierarchy’ so that an Algonkian obviate will normally correspond to a Kijeb distal in relation to a medial corresponding to a non- obviate or sometimes a Kijeb medial or distal in relation to a proximal corresponding to a non- obviate.

Moreover there is an enclitic marker -ya which is attached to a non-animate agent NP of a higher- ranking patient (expressed or unexpressed) or to an animate patient NP of a a lower-ranking animate agent (which also may be unexpressed) — i.e. it is a kind of inversion marker attached to NPs. The distributional pattern arises because it is attached to the ‘most inverted participant’ and it can only be assigned once. In practice this means that since inanimates cannot be grammatical subjects but are inflected in the instrumental case -r there arises a virtual ‘ergative ending’ -rya used when the agent is inanimate and ranking lower than the patient, and a virtual ‘accusative ending’ -ya which is used on animate patients when the agent is also animate and ranking lower than the patient. Not surprisingly this is restructured into a split ergative case marking system in the later Sohlob languages, so that they have

• an unmarked absolutive-nominative used for all patients of inanimate agents, for inanimate patients of animate agents and for animate agents.

• an ergative marked by -Vl (where /l/ is the regular reflex of the [ʎ] /rʲ/ //rj// of the Kijeb -r-ya) used for inanimate agents.

• an accusative marked by -Vy used for animate patients of animate agents.

I’ve also finally determined that the word order is verb initial. The order of other constituents is pretty free, since inverse marking on the verb and distance marking on free pronouns generally take care or sorting out agent-patient relations. Quite possibly the slot after the verb is taken up by the topic!

The verb-first rule is probably absolute. The other constituents may follow in any order except that the topic comes in the second place — which incidentally helps alleviating the lack of overt obviation marking, since the topic will normally be the non-obviate!

I have long known that there are three different verb stems, each formed by adding one of the three vowels -a- -i- -u- to the root (in Sohlob these give rise to alternations in the root vowel), and I have known that these express distinctions of aspect rather than of tense for almost as long. Now I also know for sure what the aspects are, namely:

• Punctual or instantaneous (-a-).
• Progressive or durative (-i-).
• Habitual (-u-).

I don’t know if this is an odd bunch of aspects, but I want to be a bit different. Besides the speakers of Kijeb and Sohlob are perhaps not even human.

The Punctual is somehow basic. For example the imperative coincides with the Punctual stem and there is no such thing as a Progressive or Habitual imperative. It us likely that the -i- and -u- aspect morphemes go back to some -CV- form where the C was some palatalized consonant for the Progressive marker and some labial or labiovelar consonant for the Habitual marker. Prehistorically Kijeb verbs are marked by concentration of stress on the root syllable and loss of vowels in other syllables, with subsequent sandhi effects on the resulting consonant clusters, turning strings of tense, mood and person markers into portamanteau morphs and/or morphemes consisting only of consonants and with several combinatory allomorphs. The above-mentioned -CV- aspect markers would first have fused with the stem-final -a- into diphthongs -ai-, -au-, which then even later were monophthongized to -i-, -u-, the forms seen in Kijeb. The net result is that for any Kijeb verb the aspect marker is a fused part of the stem.

Not all verbs can form all three aspects. Verbs which intrinsically express states for example can only form the Habitual. Prehistorically there were several morphemes added as prefixes (in Kijeb mostly fused with the root) and suffixes which assign Aktionsart, iterativity, inchoativity etc. to the verb — semantics which restrict which aspects a verb can form. For example iteratives cannot be Punctual and inchoatives can only be Progressive. Sometimes different aspects encode meanings which are lexicalized as different verbs in European languages, such as Punctual ‘step’ and Progressive ‘walk’.

In pre-Kijeb these were a regular system of prefixes and suffixes for Aktionsart, but the regular loss of unstressed vowels in verb forms and subsequent sandhi effects make the processes seem almost suppletive in Kijeb. Consider what happens to the cusative prefix *p- when it ‘fuses’ with different root-initial consonants by following the p row in the Kijeb sandhi table at FrathWiki

(Initial mb and ñgw aren’t possible, so they become m, ñw!)

I now think that the punctual aspect is actually zero-marked, while the habitual and progressive in pre-Kijeb were marked by -CV- syllables where the C was some voiced palatalized consonant for the durative and some voiced labial for the habitual, becoming *-yV- and *-wV- in the first phase of vowel quashing and -i- and -u- in the second phase. The punctual -a- may have been exapted back from forms where the present marker became -ar-. The imperative probably was special because it was normally a full utterance in itself and thus escaped vowel-quashing. In any case the reduction of the number of possible person markings was as much a matter of restructuring as of sound change.

As said above inverse, mood’ tense and person markers (in that order) are partly fused with each other, so that not all the underlying morphemes are overtly present in all forms. Here there are two important differences to what is currently stated on the FrathWiki page:

• There are two tenses, present and non-present, the present being marked by -r- (with allomorphs -ar-, -ir-, -ur-) and the non-present being unmarked.
• There can be formed both present and non-present irrealis forms. NB that the non-present irrealis would be appropriate to translate both a past subjunctive and a future indicative. The irrealis has, especially with the present, has a strong tone of evidentiality, the irrealis being used for hearsay, assumptions, reported information, reservation etc. Conversely the present realis has a rather strong tone of assertion.

There are three modes: indicative, imperative and irrealis. The imperative is ‘marked’ by the absence of TAM markers and the irrealis by a morpheme. All imperative and some irrealis forms coincide with some other forms in the conjugation. The indicative mood is unmarked.

The person markers are prehistorically derived from the same pronominal roots which are used as possessive suffixes as per the FrathWiki page

but were heavily transformed by automatic phonological processes which affected the long final string of consonants resulting from the draconic loss of unstressed vowels in pre-Kijeb (actually the old version of Kijeb) verb forms.

1. first loss of unstressed vowels
2. {mm, nn, ŋŋ, rr} > {mb, nd, ŋg, dr}
3. second loss of vowels (originally secondary stressed)
4. the whole C(C…) string was devoiced
5. only coronals allowed as wordfinal consonants and clusters: velars became coronals and labials were lost except m > n.
6. tt > tst
7. {n, r, s} > {n̩, r̩, s̩} / C_C (including tst > ts̩t!)
8. haplology
9. tst > st / {V, n̩, r̩, s̩}_ (i.e. when /s/ was not [s̩])
   1. {n̩, r̩, s̩} > {an, Vr, is}

All this leads to surface forms which can be pretty ambiguous, with a good deal of isomorphy, but in spite of this the language tries hard to remain a pro-drop language. If however free pronouns are present the verb still takes the person endings, in spite of their originally being cliticized pronouns.

The funny thing is that I had made up all the phonological rules and their internal order beforehand. The only innovation was the concentration of stress on the verb root with subsequent loss of vowels and voilà a dull agglutinative string of enclitic morphemes was turned into something interesting!

The syllabicization of -n-, -r-, -s-, later resolving into a -VC- sequence, when stranded between two consonants, is somewhat more complicated than the above table suggests: -n- always becomes -an and -s- always becomes -is- and the inanimate person marker -r- always becomes -ur but the present tense marker -r- variably becomes -ar-, -ir-, -ur- depending on the (original) features of surrounding sounds — evidently a syllabic /r̩/ could be palatalized /rʲ̩/ or labialized /rʷ̩/ in addition to plain. The changes that applied were in order:

• The irrealis marker has several allophones -t-, -s-, -tis- of which -s- coincides with the inverse marker, which comes in the slot before the irrealis marker, and the person marker mentioned immediately below has identical allophones, so that a surface form may be ambiguous..
• The first person singular, the first person plural exclusive and the second person plural are marked by identical forms -t-, -s-, -tis- which also coincide with the forms of the irrealis marker.
• The secend person singular and the first person plural inclusive are both marked by zero.

• Third person markers distinguish between singular and plural and between animate and inanimate but not between the different animate sub-genders masculine, feminine, rational and irrational. The markers are as follows:

  • Animate singular: -n.
  • Animate plural: -nt, -ns-, -st-.
  • Inanimate singular: -ur.
  • Inanimate plural: -tur.

• When agent and patient belong to the same gender and both are third person singular a single ending corresponding to the plural of the appropriate gender and person is used.

• When both agent and patient are the same gender and both are third person plural only a singel plural ending of the appropriate gender is used.
• There is a reflexive marker, the same for all numbers and genders, which is also -n, but is distinguished by always coming in the second person marker slot while the animate singular marker always comes in the first person marker slot. When a third person animate singular agent acts on itself only a single -n appears on the surface.

These last three phenomena are what somewhat inaccurately is referred to as *haplologization*. It goes back to the fact that the third person plural forms are historically reduplications of the singular forms, so that e.g. *-na-na > *-nna > *-nda > *-nd > -nt and *-ru-ru > *-rru > *-dru > *-drʷ > *-trʷ > -tur. The list of “third persen enclitic pronouns” — now actually possessive suffixes — makes this clear.

All this leads to surface forms which can be pretty ambiguous, but in spite of this the language tries hard to remain a pro-drop language. If however free pronouns are present the verb still takes the person endings, in spite of their originally being cliticized pronouns.

BTW I’ve decided to change the sandhi rules somewhat: *pp and *kk now give f and x.

Tags: Conlanging, grammar, Kijeb

Related posts

• Rhodrese articles and indefinite pronouns (0)
• The Rhodrese indefinite article (0)
• Mærik: Deriving nouns from verbs and the word amn (1)
• The embarrassing origins of Kijeb (0)