Agreement with the patient, but not the agent

[dɮ the phoneme]

Most languages have some form of person marking on the verb. Typically, this is either subject agreement or some type of polypersonally agreement. According to WALS, however, there are a scattered few languages (well, about 7% of their sample, appearenty) which mark person agreement only with the patient. Honestly, before looking it up, this was a feature I assumed didn't exist, since I've essentially never heard it discussed. I've certainly never seen it in a conlang.

Mainly for that reason, I'm interested in learning more. However, I have no familiarity with any of these languages, and at least according to Wikipedia, the map is innacurate for a few of them. So, if anyone here knows of any sources which cover this in more detail, I'd love to know what they are. And if anyone is actually familiar with a language that has this feature, I'd be extremely curious to hear how exactly it manifests. Appearently it's somewhat common in Chadic?

[bradrn]

WALS lists sources for each language; from their Khoekhoe source, I managed to find the following rigorous description:

At the end of any verb which is not intransitive, there may occur a patient suffix. Semantically, all patient suffixes indicate who or what is the "immediate patient" of the action expressed by the active verb. By "immediate patient" is meant the "direct object" of a transitive verb, [or] the "indirect object" of a ditransitive verb [...] The object suffixes are the least complicated from [a] syntactic point of view. They are simply "pronomial" suffixes which indicate the immediate patient of the verb by person, gender and number. [..] most of them are identical to the pgn [person-gender-number] suffixes [...] The existence of this of this complete system of object suffixes is interesting because it appears to be entirely unnecessary, since for any of these suffixes on the verb may be substituted a pronoun [...] with no change of meaning. The first example sentence is a good illustration: //'ĩiku ke //'ĩipà kè mùũ. and //'ĩiku ke kè mùũpi. both mean "They (masc., pl.) saw him." In the first sentence "him" is expressed by a pronoun NPà //'ĩipà, and in the second by an object suffix -pi. The construction with a pronoun [...] is the more frequent in texts. The only difference in meaning [...] appears to be one of emphasis, the construction with the object suffix representing a very low degree of emphasis on the "patient" of the verb, in this case "him."

And for Yapese:

[...] we see that the direct object is expressed by a suffix to the verb. This is the way a pronoun direct object of a transitive verb is expressed in Yapese. The suffixes used are illustrated by the following sentences:

Kea guyeeg.
'He saw me.'

Kea guyeem.
'He saw you (singular).

Kea guy.
'He saw him.'

[...]

Note that these object pronoun suffixes are incompatible with the subject number suffixes. In other words, to say something like 'They two saw me' you must say:

Ka ra guyeew gaeg.
'They two saw me.'

with an independent pronoun functioning as the direct object 'me', and not the incorrect sentence:

Ka ra guyeeweeg.

(In general, for cases like this, the Grammar Pile (a large collection of reference grammars) is a great help. It usually contains the grammars referenced in WALS.)

[akam chinjir]

If I'm understanding right, in both of those cases you don't get the object marker on the verb if there's a separate pronominal object. That makes them look more like weak pronouns or clitics than like agreement affixes, imo. (My impression is that it's very common for pronominal clitics to get misidentified as affixes, and it makes me very wary of trusting a source like WALS for this sort of thing.)

[bradrn]

If I'm understanding right, in both of those cases you don't get the object marker on the verb if there's a separate pronominal object. That makes them look more like weak pronouns or clitics than like agreement affixes, imo. (My impression is that it's very common for pronominal clitics to get misidentified as affixes, and it makes me very wary of trusting a source like WALS for this sort of thing.)

I didn't think about that when I read those descriptions, but that does seem to be the case for those two.

One interesting thing I noticed is that two (probably more) languages listed as agreeing with 'Only the P argument' in WALS are ergative. This makes sense as an explanation for this behaviour: if agreement is with the absolutive only, the verb would agree with the transitive patient or the intransitive experience.

[Moose-tache]

The chapter text on WALS blatantly says that they count both affixes and clitics, with no comment on what happens when the clitics are mutually exclusive with normal pronouns. This is probably why they get away with listing Spanish under "polypersonal," presumably because of special pronouns like lo. I wouldn't put much stock in this WALS category at all if you're looking for actual agreement.

[Pabappa]

Some of my early conlangs do this, so I can attest that it causes no major problems with writing basic sentences. However my conlangs marked the noun class of the patient on the verb, not just person, so it was a much wider category than in those natlang examples .... e.g. "to carry a baby", "to carry fruit", "to carry weapons", etc .. and since they underspecify the patient, they *are* used when the patient is named explicitly. Also, the human classifiers are divided by person and gender, so the person marking gets taken care of too.

[Frislander]

Looking at some of the other cases, there's even more weirdness going on with this category. Firstly there's some notably erroneous data points - Warao in particular is especially bad, since the source they cite doesn't actually show it's patient only agreement for several reasons. Firstly, it looks as if Warao operates on the basis of clitic pronouns rather than true affixes, since the person markers are reduced pronouns, but secondly Warao has subject markers as well as object ones! Now the source does say that these are rare in normal speech, but even so this is at least grounds for putting in the 2A or P but not both category". Similarly they mark Anejom as having "P only" when if anything there seems to be only agent marking - object marking only seems to extend to animacy. Additionally, as has been pointed out, other cases include apparently ergative systems, such as Canela-Kraho (from Popjes & Popjes 1986):

i-mã a-kĩn
1-TEMPRY 2-like
I like you

vs.

a-te po kam a-catõc
2-PAST deer at 2-shoot
You shot at the deer

So yeah, just one example of why you shouldn't take WALS at face-value.

[Opalescent Yams]

Am I being dense or is this just marked-absolutative ergative agreement?

[akam chinjir]

Am I being dense or is this just marked-absolutative ergative agreement?

It's not clear that it's agreement at all (see above). But if it's agreement, it's agreement with the P argument, namely the object of a transitive verb, which I guess could be called accusative(-only) agreement, but I'm not sure what would be gained by that. It's not ergative because subjects of intransitive verbs are treated just the same as subjects of transitive verbs (they don't control agreement).

[Opalescent Yams]

It's not ergative because subjects of intransitive verbs are treated just the same as subjects of transitive verbs (they don't control agreement).

Oh duh ok I was indeed being dense.