Prefixing Synthesis Sketch

Since I mentioned my background conlanging project on Twitter recently, I thought I’d sketch out what I was aiming for here. The general idea at the beginning was to build something that looked a bit like a cross between Mayan and Athabaskan: shared features like synthetic, head-marking, verb-centric, with a broadly Mayan~Mesoamerican phonology and verb and something like the complex, templatic but heavily prefixing verb of Athabaskan. But at the same time, I wanted to avoid too many sequences of verbs ending up starting with the same consonants and syllables, which gets a bit repetitive but is an easy place to end up if you have a mostly agglutinative morphology.

1. The Template

1.1. Tense, Aspect, Mood, Negation and Subordination

The attempted solution to increase variety combined two basic ideas:

  1. Sequencing of TAM markers, e.g. narrative and backgrounding “tenses”
  2. A complex mapping of inflectional categories to templatic slots to minimise the chance of the same morpheme always starting the verbal word

I’ll present the template here, and then dive into the why. The core inflectional prefix template is as follows:













Figure 1: Core Verbal Prefix Template

Here each column represents a mutually exclusive set of prefixes. Most labels represent a single prefix, but Aɢʀ1 and Aɢʀ2 each represent a set of possible agreement prefixes. Aɢʀ2 is obligatory, while Aɢʀ1 is optional and only occurs in certain transitive configurations involving either reflexivity, or action on a first or second person object.

This prefix chain fits within a larger verbal word that looks something like this, although I won’t talk about the other components right now:

ɪɴғʟᴇᴄᴛɪᴏɴᴀʟ.ᴘʀᴇғɪxᴇs – (ɴᴏᴜɴ.ɪɴᴄᴏʀᴘ) – (ᴘʀᴇᴠᴇʀʙ) – (ᴠᴏɪᴄᴇ) – (ᴅɪʀᴇᴄᴛɪᴏɴᴀʟ) – ᴠᴇʀʙ+ɴᴜᴍʙᴇʀ (=Aᴅᴠᴇʀʙɪᴀʟ.ᴄʟɪᴛɪᴄs)

Returning to the inflectional prefixes, since many of the TAM categories are mutually incompatible, the final matrix of possible inflections looks something like this:

Past PerfectiveAɢʀ1-Aɢʀ2-i-Aɢʀ1-nu-Aɢʀ2-Aɢʀ1-Aɢʀ2-óh-Aɢʀ1-nu-Aɢʀ2-óh-
Future (Perfective)Aɢʀ1-Aɢʀ2-eb-Aɢʀ1-nu-Aɢʀ2-eb-Aɢʀ1-si-Aɢʀ2-Aɢʀ1-nu-si-Aɢʀ2-
Prefixal Inflection Patterns

These can be approximately divided into forms which typically express foregrounded vs background information, and forms which are used for events on the main timeline in narrative vs those which are not. Narrative sequencing is primarily expressed by the perfective and the consecutive (for subsequent clauses in the chain). The imperfect covers senses such as the asserted progressive present, immediate future and habitual. Purpose and adversative clauses are also focal.

-Nu- is primarily a background marker. Clauses marked by -nu- do not advance the main timeline but provide framing for other clauses, with senses like when, after, while, since, …, as well as marking a variety of complement clauses. Similarly, the conditional describes the background within which the focal clause(s) hold.

I imagine that in many cases these subordinate clauses will be reinforced by adverbial clitics or particles, but I haven’t really got that far yet. I definitely need to organise my thoughts on the use of these forms a bit so maybe I’ll do a follow-on post as the thing evolves.

A few further notes:

  1. The distinction between imperfective and future is neutralised in the negative
  2. There is no negative consecutive, negative past perfective or imperfective forms are used
  3. Consecutive, conditional, purpose and adversative clauses are inherently linked to another (main) clause, and do not take the subordinator nu-
  4. Whether there is a need for a distinction negative purpose and adversative is unclear / under consideration… I feel like there is a semantic difference between “so that not”/”lest” and “so that”/”lest not”, but it is fairly marginal.

1.2. Person Agreement Markers

The verb agrees with the subject and object, but only in person, including a separate inclusive-we category. Number agreement happens separately via suffixes which I won’t describe here. This pattern is not too uncommon in the wild, but is most frequently seen in North and Mesoamerica. It occurs in, for example, Sierra Popoluca, Winnebago, Wichita, Caddo, Pawnee and Mixtec (see Cysouw). Outside of the Americas, this split between four persons with separately marked number agreement is less common.

The person categories tracked for subject and object are 1 (I, we excluding you), 2 (you, you all), 1+2 (I, you, and possibly others), and 3. Third person is effectively a default in the absence of any other marker, which means that in transitive clauses there are 3 x 3 = 9 basic possibilities, since each of the first person (I) and second person (you) can be part of the subject, part of the object, or neither.

In intransitives, and transitives with third person objects, the agreement pattern is that Aɢʀ1 is empty and Aɢʀ2- agrees with the subject via the following prefixes:

1 = First exclusive = I, we but not youš- (initial, non-final), -î (as final prefix), ch-/tz- (after o’)
2 = Second = you, you allm- (initial, non-final), -n (as final prefix), b- (after o’)
1+2 = First inclusive = I and you (and others)k-
3 = Neither I nor you, he, she, it, theyr- (initial, not past perfective), zero (elsewhere)
Person markers

Aɢʀ1 is used in transitive clauses and is filled when either:

  • a third person acts on a speech act participant (marked by er-), in which case Aɢʀ2 agrees with the object
  • a speech act participant acts (SAP) acts on another SAP, in which case Aɢʀ1 is cha- and Aɢʀ2 agrees with the subject
  • the clause is reflexive or middle voice, in which case Aɢʀ1 is o- ~ o’- ~ of-, sometimes with a hardening effect on a following fricative

Honestly, this part I’m not sure about because Aɢʀ1 is a mixture of object marking (reflexive, SAP>SAP) and subject marking (3>SAP). Having started out somewhere slightly different (I’ll get to that later), I basically ended up reinventing a more or less direct-inverse voice system, with er- acting as an inverse marker or obligatory passive and flipping the controller of Aɢʀ2 from subject to object. Anyway, the transitive agreement matrix looks like this:

1cha-š--š-o’-š-, och-
2cha-m--m-o’-m-, ob-
Transitive Agreement Patterns

In this table I represent affixes by one allomorph, but see the previous table for other realisations. Aɢʀ1 and Aɢʀ2 are written separately because they can be separated by other morphemes.

1.3. Some Morphophonology Notes

A few major processes which affect the concatenation of prefixes:

  • Deletion of first vowel in hiatus in most cases (a few morphemes have allomorphs to resolve hiatus), e.g. -nu-a- to -na-
  • A general sibilant harmony process applies to fricative and affricates (similar to in Navajo and some Mayan languages)
  • Coda stops are not allowed word internally, and are generally resolved via progressive copy vowel insertion

2. Natlang Notes

2.1 The Road Not Taken: Agreement in Hixkaryana

I had originally thought to have a system more like Hixkaryana, which manages to have agreement in subject and object person via a small set of monosyllabic prefixes (with number marked separately by suffixes in the same manner I described above). The Hixkaryana system looks like this, according to Derbyshire:

IIuro mɨ-amna mɨ-mɨ-mɨ-, o-, ow-
I+IIIamna o-amna nɨ-amna nɨ-
IIIro-o-kɨ-amna y-(-O) nɨ-,
(+O) y-
Hixkaryana Verb Agreement

In the Hixkaryana system, of the 13 possible transitive combinations, 8 are marked by a single consonant or syllable (the vowel ɨ is typically lost before another vowel). Of the remaining cases, most relate to exclusive we, which can only be expressed by an obligatory independent pronoun with a verb showing third person agreement. The remaining case is 1>2, which requires an independent first person singular object pronoun.

In this system, the main ambiguity (apart from number for 2nd and 3rd persons, which is optionally marked by separate suffixes) is the prefix kɨ-, which marks bot I>II (“I … you”) and III>I+II (“He/she/they … us”). It originally seemed like a pretty good starting point for the structure of my Aɢʀ1 and Aɢʀ2 affixes.

So why didn’t I go this route? Basically because the most important Aɢʀ2 markers can end up buried a few syllables into the word, and I thought a smaller number of contrasts would be better. In my final paradigm there are basically three overt markers in Aɢʀ2, since the third person is mostly unmarked, and they are fairly distinctive in terms of place and manner. But cutting down on the contrasts in Aɢʀ2 means that Aɢʀ1 needs to be a bit more generally load-bearing in more marked combinations like 3>SAP. I guess I just like to overengineer things.

2.2. Coptic: Variable Affix Ordering

One key property of the template in figure 1 is that the main agreement slot, Aɢʀ2, precedes some tense/aspect/mood markers and follows others. A good example of a natlang with this property is Coptic, which developed a complex series of auxiliaries as it evolved from Ancient Egyptian. Auxiliaries either occur clause initially, with a subject NP potentially occurring between the auxiliary and the verb, or directly preverbally. Crucially, when these auxiliaries show agreement, the agreement marker occurs in the same location as a subject NP would: as suffixes or infixes to clause-initial auxiliaries, but as prefixes to preverbal auxiliaries.

Prefixal agreement primarily applies in the present and future tenses: ti-na-sôtm 1sɢ-ғᴜᴛ-hear “I shall hear”, while most other tenses take suffixal agreement: a-i-setm ou-hroou ᴘʀᴇᴛ-1sɢ-hear a-voice “I heard a voice”.

2.3. Subordinate and Consecutive Forms

The original core idea here was to have a consecutive inflectional form meaning “… and then” so that sequences of perfective clauses don’t look inflectionally identical. This kind of clause chaining in a VO language is straightforwardly attested in languages like Swahili, although it is definitely more common in verb final languages. The following example of narrative -ka- clauses is from Wikipedia:

The consecutive tense is mainly used with the past tense -li- in narrating a sequence of events whereby -li- is used for the first verb and -ka- for subsequent verbs. It roughly carries the meaning "and then" and makes the use of na "and" or halafu / kisha "then" essentially redundant. Where context is clearly past, a narrative may also be begun with -ka-.

Alipomwona nyoka, alivua shati akalitupa juu ya nyoka akamkanyaga.

a-li-po-mw-on(a) nyoka, a-li-vua shati a-ka-li-tup(a) juu y-a nyoka a-ka-m-kanyag(a)

cl1-pst-cl.16.rel-cl.1-see snake(cl9), shirt(cl5) cl1-cnsc-cl5-throw top(cl9) cl9-gen snake(cl9) cl1-cnsc-cl1-trample

when he saw it snake, he took off shirt he then threw it (the shirt) on top of snake he then trampled it (the snake)

"Upon seeing the snake, he took off his shirt, threw it over the snake and then trampled on it."

The decision to expand to other subordinate clause types was not really based on a VO language. It was originally based more on West Greenlandic, which is mostly verb final. West Greenlandic has the following inflectional “moods”, some of which mark speech act types and others adverbial subordinate clauses and clause chains:

Indicative – declarative main clauses

Interrogative – questions

Imperative, Optative – orders, requests, suggestions, wishes

Causative – subordinate clauses of reason, cause, and consecutive action (“when”, “because”, “since”). Also used as a narrative past tense.

Conditional – used to mark the protasis of conditionals

Contemporative – subordinate clause of simultaneous action (“while”), clauses of manner

Participle – complement clauses

Since West Greenlandic is a polysynthetic language, it also has a number of non-final suffixes which can be added to these “moods” to add additional nuances or clarify the intended meaning of the subordinate clause where necessary.

In terms of the current conlanging project, perfective -nu- would be broadly equivalent to the West Greenlandic causative, and imperfective -nu-a- = -na- would be equivalent to the contemporative and/or participle forms.

An example of a more verb-medial language which inflectionally distinguishes subordinate verb forms is Ojibwe. Ojibwe has three inflectional orders: the independent order for main clauses, the conjunct order for various kinds of subordinate clause as well as content questions, and an imperative order for commands. When the conjunct order is used in adverbial clauses, a bound preverb and/or an independent conjunction is generally used.

Another VO language example would be Alto Perené, which marks purpose, adversatives, result, locality, temporal overlap, conditionals, relative clauses and some complement clauses by verbal affixes or clitics, in addition to a number of alternative constructions involving independent conjunctions.

3. Putting It Together

So in the end, hopefully this produces a certain diversity of prefix shapes without resorting to a lot of ad-hoc fusion. The core distinction between perfective and imperfective verb forms results in differences in shape, such as šaûa’ /ʃawaˀ/ 1-(ᴘғᴠ)-ʜɪᴛʜᴇʀ-be “I went” compared to aîaûa’ /ajawaˀ/ (ɴᴘꜰᴠ)-1-ʜɪᴛʜᴇʀ-be “I’m going”. In narrative, there is at least a distinction between the first verb and following verb(s):

šicha' tze'îsaq    /ʃit͡ʃaˀ t͡seˀjsaq͡χ/

š-i-ch-a' cha-e'-î-saq

1-ᴘꜰᴠ-ʜɪᴛʜᴇʀ-be SAP-ɴᴀʀʀ-1-see

"I came and saw you"

I’m at the point now where I need to actually create some vocabulary and test it before deciding what needs tweaking.