I read Agreement by Corbett earlier this year and of course it contains a chapter on person clitics as compared to person inflection as an agreement strategy (Corbett 99–112). You may have noticed before that Ayeri behaves a little oddly with regards to person marking on verbs, insofar as verbs for the most part agree with agents in person and number, whether they are the topic of the clause or not.[1. “Topic” is not to be understood strictly in terms of topic/comment sentence structure (Li and Thompson 1976) here in the way e.g. Japanese or Chinese uses it, but in terms of the “Austronesian alignment.” For an analysis of how Ayeri treats topics vs. subjects, see the article “Flicking Switches: Ayeri and the Austronesian Alignment” (2012-06-27) on this blog.] Sometimes, this person marking even involves case (as a nominal category governed by the verb!), as we will see below. For a start, however, consider these two examples:[1. Some of the Ayeri examples used here come from a list of samples I provided for a bachelor’s thesis at the University of Kent in March 2016, in private conversation, on request. I don’t know what the author made of them – the questionnaire I filled out initially indicated that the thesis was probably on the syntactic typology of fictional languages regarding typical word-order correlations (VO correlating with head-first order etc.). I hope that my reflections here don’t preempt or invalidate the author’s analyses should they still be in the process of writing or their submitted thesis be in the process of evaluation and grading. I would certainly like to learn about their analysis of my examples.]

I think it is uncontroversial to analyze -ya in (1a) as person agreement: Ajān is a male name in Ayeri while Pila is a feminine one; the verb inflects for a masculine 3rd person, which tells us that it agrees with the one doing the greeting, Ajān. Ajān is also who this is about, which is shown on the verb by marking for an agent topic. In the second case, there is only anaphoric reference to Ajān, so you might say that the agent NP is left out, so very broadly, the verb marking here seems to be like in Spanish, where you can drop the subject pronoun:[1. However, we will see that it is probably more complicated than this.]

Example (1b) probably won’t raise many eyebrows either, except that there is also topic marking for an agent there, the controller of which I have so far assumed to be the person inflection on the verb, in analogy with examples like:

This raises the question whether in Ayeri there is dropping of an agent pronoun involved at all, which is why I glossed the person suffix in (1b) as -ya.Ø (-3SG.M.TOP) rather than just as -ya (-3SG.M).

This leads us to consider another characteristic of Ayeri, namely that the topic morpheme on noun phrases is zero. That is, the absence of overt case marking on a nominal element indicates that it is a topic; the verb in turn marks the case of the topicalized NP with a (case) particle preceding it. Pronouns as well show up in their unmarked form when topicalized, which is why I am hesitant to analyze the pronoun in (4b) as a clitic on the VP rather than an independent morpheme:[1. Also, perhaps a little untypically, topic NPs in Ayeri are not usually pulled to the front of the phrase (at least not in the written language; cf. Lehmann 120–122), so topic-marked pronouns stay in-situ; which NP constitutes the topic of the phrase is marked on the verb right at the head of the clause. How and whether this can be justified in terms of grammatical weight (see, e.g., Wasow 95–98) remains to be seen.]

What is remarkable, then, is that ye ‘3SG.F.TOP‘ is the very same form that appears as an agreement morpheme on the verb, just like -ya in various examples above:

This also holds for all other personal pronouns. Moreover, -yāng as seen in examples (3) and (4b) may as well be used as a free pronoun, as well as other such case-marked personal forms:

As for case-marked person suffixes on verbs, I have so far assumed that they are essentially clitics, especially since the following marking strategy is the grammatical one in absence of an agent NP:

The verb here agrees with the patient – or is it that person agreement suffixes on verbs are generally clitics in Ayeri, even where they don’t involve case marking? There seems to be a gradient here between what looks like regular verb agreement with the agent on the one hand, and agent or patient pronouns just stacked onto the verb stem on the other hand:

Especially the middle, transitional category is interesting in that what looks like verb agreement superficially can still govern topicalization marking, which is indicated in column II by an index “1”. Note that this behavior only occurs in transitive contexts; there is no topic marking on the verb if the verb only has a single NP dependent.

As for personal pronouns fused with the verb stem like in the first column, Corbett points out that

In terms of syntax, pronominal affixes are arguments of the verb; a verb with its pronominal affixes constitutes a full sentence, and additional noun phrases are optional. If pronominal affixes are the primary arguments, then they agree in the way that anaphoric pronouns agree […] In terms of morphology, pronominal affixes are bound to the verb; typically they are obligatory […]. (99–100)

This seems to be exactly what is going on for instance in (3) and (7b), where the verb forms a complete sentence. It needs to be pointed out that Corbett includes an example from Tuscarora, a native American polysynthetic language, in relation to the above quotation. Ayeri should not be considered polysynthetic, however, since its verbs generally do not exhibit relations with multiple NPs, at least as far as person and number agreement is involved.

Taking everything written above so far into account, it looks much as though Ayeri is in the process of grammaticalizing personal pronouns into person agreement (Lehmann 42–45, van Gelderen 493–497). Corbett illustrates an early stage of such a process:

What van Gelderen calls the subject cycle, the “oft-noted cline expressing that pronouns can be reanalyzed as clitics and agreement markers” (van Gelderen 493) applies here, and as well in Ayeri. However, while she continues to say that in “many languages, the agreement affix resembles the emphatic pronoun and derives from it” (494), Ayeri does at least in part the opposite and uses the case-unmarked, unstressed form of personal pronouns for what resembles verb agreement most closely. This, however, should not be too controversial either, considering that e.g. semantic bleaching and phonetic erosion go hand in hand with grammaticalization (Lehmann 136–137, van Gelderen 497).

As pointed out above (see example (7)), Ayeri usually exhibits verbs as agreeing with agents and occasionally patients, not topics as such. This may be a little counterintuitive since the relation between topics and subjects is close, but is possibly due to the fact that the unmarked word order is VAP. This means that agent NPs usually follow the verb, and it strikes me as not too unnatural to have an agreement relation between the verb and the closest NP also when non-conjoined NPs are involved (Corbett 180). This conveniently explains why verbs can agree with patients as well if the agent NP is absent. Taking into account that the grammaticalization process is still ongoing so that there is still some relative freedom in how morphemes may be used if a paradigm has not yet fully settled (Lehmann 148–150) also makes this seem less strange. Verbs simply become agreement targets of the closest semantically plausible nominal constituent. Ayeri seems to be shifting from topics to subjects, and as a consequence the bond between agents and verbs is strengthened due to their usual adjacency; developing verb agreement with agents may be seen as symptomatic of this change.

Up to here signs point towards Ayeri’s person agreement in fact being more likely enclitic pronominal affixes, even what I had been thinking of as person agreement before (i.e. suffixes on the verb that only encode person and number, but not case), but can we somehow corrobate this? Corbett offers a typology here:

A pronominal affix, then, is syntactically an argument of the verb but has the morphology of an inflectional form. If we compare this to the gradient given in table 1 above, it becomes evident that I definitely fulfills these criteria, and II does so as well, in fact, in that there is no agent NP that could serve as a controller if the verb inflection in II were ‘merely’ a agreement target. The inflection in III, on the other hand, appears to have all hallmarks of agreement in that there is a controller NP that triggers it, with the verb serving as an agreement target. Moreover, the person marking on the verb is not a syntactic argument of the verb. As example (7a) shows, however, marking of type III permits the verb to mark more than one case role, which makes it slightly atypical, although verbs can only carry a single instance of person marking (Corbett 103). Regarding referentiality, the person suffixes on the verb in table 1, columns I and II are independent means of referring to discourse participants mentioned earlier, whereas the person suffix in III needs support from an NP in the same clause as a source of morphological features to share:

Since person marking of the type I and II is referential, as shown in example (9a) and (b), it can be counted as a cliticized pronoun (103). Pronouns in Ayeri can also refer to non-people – there are both a ‘neuter’ gender for non-people considered living (or being closely associated with living things), and an ‘inanimate’ gender for the whole rest of things; however, since mere agreement as in type III needs support from an NP within the verb’s scope, it does not have descriptive/lexical content of its own, i.e. it only serves a grammatical function (104). As for Corbett’s balance of information criterion, table 1 also highlights differences in what information is provided by the person marking. Nouns in Ayeri inherently bear information on person, number, and gender, and all three types of person inflection on verbs share these features. However, there are no extra grammatical features indicated by the first two inflection types that are not expressed by noun phrases, although under a very close understanding of Corbett, the following example (10) may still qualify as person-marking on the verb realizing a grammatical feature shared with an NP that is not openly expressed by the NP. He writes that in the world’s languages, this frequently is number (105). This, however, does not apply to Ayeri because the only time that verbs display number not expressed overtly by inflection on a noun is in agreement like in type (3a):

As shown above, verb marking of the types I and II is independent as a reference, so there is unirepresentation of the marked NP. In contrast, verb marking of type III requires a controlling NP in the same clause to share grammatical features with, so that there is multirepresentation typical of canonical agreement (106). Note that unirepresentation as outlined here is probably different from pro-drop, as in this case I would expect sentences like (9c) to be grammatical (107). A further property that hinges on types I and II being independent pronouns tacked onto verbs as clitics is that they are not coreferential with another NP of the same grammatical relation, but in complementary distribution, as commonly assumed with pronominals (108). Hence, you can’t say something like either of these two:

However, verb agreement with a pronoun is also not possible even though it would be expectable according to Corbett (109):

In conclusion, we may assert that Ayeri appears to be in the process of grammaticalizing pronouns as verb infletions, however, how far this grammaticalization process has progressed is dependent on syntactic context. Ayeri displays a full gamut from personal pronouns (usually agents) tacked on verbs as clitics to agreement with coreferential NPs that is transparently derived from these personal pronouns. With the latter, there is the complication that coreferential pronoun NPs are not allowed as one might expect, but only properly nominal ones. Slight oddities with regards to Austronesian alignment – Ayeri’s actors bear more similarities to subjects than expected, but still without fully conflating the two notions – can possibly be explained by a strengthening of the verb-agent relationship pointed out as a grammaticalization process in this article as well. What was not discussed here, and consequently saved up for later, are more detailed questions of verb agreement such as resolution and mismatches.