One of the most famous phenomena in argument flagging is differential object marking (DOM), or split P flagging: a situation where some P-arguments bear a special accusative flag and others don’t, typically depending on referential prominence (definiteness, animacy, topicality, etc.). The following examples are taken from my 2021 paper on split argument marking:
This phenomenon has been widely studied in Spanish (and many other Romance varieties), in Hindi-Urdu (and other Indic and Dravidian languages), in Slavic languages, in Turkic, in Hebrew, and so on. In much of this work, the focus is on the precise conditions for the presence of the accusative flag, but there is also the question of universality: At least since Bossong (1991) and Aissen (2003), it has been widely known that there is a general tendency for objects to be marked differentially if they are referentially prominent. The purpose of the present blogpost is to point out that the first clear statement of the differential object marking universal can be found in Moravcsik (1978), a stunning paper that is not as well known as it should be.
One might say that this is only of historical significance now, and I admit to some subjectiveness in my motivation for writing this post: Edith Moravcsik was my professor in my first year of university in 1982-83 when she introduced me to Greenbergian typology, and we have been good friends ever since. But maybe this paper has been unfairly overlooked, also because it is written in such a timeless manner (it was certainly not fashionable at the time, when most syntacticians focused on transformations). Moravcsik (1978) states the generalizations very clearly, without speculating about possible explanations for the tendencies that she finds. By contrast, Bossong’s (1985: 3-8; 1991) general considerations do not include a clear statement of what is universal, and neither does Aissen (2003), who focuses on OT mechanisms that can simultaneously describe p-languages and explain general tendencies (as is typical of generative grammar). Bossong’s and Aissen’s explanations are ultimately on the right track, I believe, and they are broadly compatible with earlier observations by Comrie (1977), Dixon (1979: §2.3.2), and other earlier statements. But what exactly is universal? In Haspelmath (2021: 131), I formulated the universal as follows:
The split P flagging universal
If a language has an asymmetric split in P flagging depending on some prominence scale, then the special flag is used on the prominent P-argument.
This is thus an implicational universal, as noted by Bossong (1985: 3), and also implicitly by Moravcsik (1978), who starts by saying that “in some languages”, P-arguments “admit of alternative case markings”. Moravcsik used the precise expression “segmental markers of participant functions adjacent to the noun” (for the more recent flag), and she talked about “case marking alternatives”, where we now use the terms split marking or (more generally) differential marking (I try to distinguish between split marking and divided marking for grammatically vs. lexically conditioned differential marking, but these two types are not often distinguished). It seems that the “split” terminology goes back to Dixon (1979), while the term “prominence scale” was adopted from Aissen (1999; 2003). By asymmetric split flagging, I mean a split in which one of the marking patterns has more segments (see Witzlack-Makarevich & Seržant 2018 on the difference between symmetric and asymmetric DOM; the universal above makes no claims about symmetric DOM, which is quite rare).
Moravcsik’s statement of the universal is quite similar, but broader (in that it includes adverbial/locative and partitive alternations, as well as affectedness as a conditioning factor):
Moravcsik’s (1978: 283) DOM universal
“If there is any semantic difference between an accusative marking and either an adverbial, or a partitive, or a nominative marking (to the exclusion of passivization), this semantic difference will be related either to definiteness, or to animacy, or humanness or to the degree of affectedness of the Object, with the accusative, by definition, marking the more definite (rather than the less definite), the animate or human (rather than the inanimate or non-human), and the totally affected (rather than the unaffected) noun phrase.”
This statement is stronger than my formulation above in that it claims that relevant semantic differences are restricted to referential prominence or affectedness, but I prefer to leave affectedness aside, because it concerns the event meaning. I also leave partitive marking (a kind of symmetric DOM) aside, because it seems to be far less widespread than asymmetric DOM, and it is not yet clear what is universal about it. If this is so, we can also leave out “by definition” – then there is no danger of the accusative to be taken as a genitive or locative (Moravcsik did not have a notion of P-argument and thus had to define “accusative” in such a way as to make reference to referential prominence and affectedness, as seen in her note 1).
If we modify Moravcsik’s statement in this way, we get the following:
“If there is any semantic difference between an accusative marking and … a nominative marking (to the exclusion of passivization), this semantic difference will be related … to definiteness, or to animacy, or humanness …, with the accusative … marking the more definite (rather than the less definite), [and] the animate or human (rather than the inanimate or non-human), … noun phrase.”
This is very similar to my formulation above, though definiteness and animacy/humanness are only two aspects of referential prominence. Moravcsik does not note that it is often topical objects that get special accusative marking (this was highlighted by Iemmolo 2011 and Dalrymple & Nikolaeva 2011), and she does not include nominality (the preference for accusative marking in personal pronouns was perhaps too obvious).
In one passage, Moravcsik (1978: 275) mentions “the conditions that correlate with the differential marking [emphasis mine] of direct objects”, but her focus is not on introducing novel terminology (and I do not know if Bossong coined the term independently, which is quite possible). Her paper instead contains a very rich discussion of the precise conditions under which languages use split accusative marking, and her examples come from a wide range of languages (mostly from Eurasia, but also some Eskimo, Warlpiri, and Amharic).
Thus, Moravcsik’s paper is very clearly the first classic of the literature on differential object marking, and it would be good if it were treated as such. Bossong (1985: 3) says that he introduced the term differential object marking in Bossong (1982), and he does cite Moravcsik (1978), but only in very general terms (in his n. 19), without giving her the credit she deserves (I feel), and without mentioning that she had already talked about “differential marking of objects”. (Bossong also gives credit to Comrie 1979; 1981 in general terms, and in the 2nd edition of his book, Comrie 1989 did use the term “differential marking”; however, it was not present in Comrie 1977; 1979; so Moravcsik was first.)
But in addition to being a classic, Moravcsik’s old paper can perhaps invite us to think again about the way in which we try to explain the general patterns, because there seems little doubt that the DOM generalizations are indeed closely related to the locative alternation patterns and other facts relating to affectedness (e.g. English alternations of the type She swam across the river / She swam the river). Many linguists cite Hopper & Thompson (1980), and there is a more recent book by Næss (2007) on “prototypical transitivity”, but there does not seem to be any widely known (let alone generally accepted) explanation for the trends that we see. Generative approaches tend to focus on particular languages (such as Spanish, or Palauan), and they sometimes try to explain differential object indexing simultaneously with differential object flagging (e.g. for Neo-Aramaic), which does not seem right (because they do not fall under the same generalizations). Thus, the patterns of accusative vs. non-accusative flagging or absence of flagging are still a challenging topic for comparative linguists.
Aissen, Judith. 1999. Markedness and subject choice in Optimality Theory. Natural Language & Linguistic Theory 17. 673–711.
Aissen, Judith. 2003. Differential object marking: Iconicity vs. economy. Natural Language & Linguistic Theory 21(3). 435–483.
Bossong, Georg. 1982. Historische Sprachwissenschaft und empirische Universalienforschung. Romanistisches Jahrbuch 33. 17–51. (doi:10.1515/9783110244908.17)
Bossong, Georg. 1985. Differentielle Objektmarkierung in den neuiranischen Sprachen. Tübingen: Narr. (https://zenodo.org/record/4697660)
Bossong, Georg. 1991. Differential object marking in Romance and beyond. In Kibbee, Douglas & Wanner, Dieter (eds.), New analyses in Romance linguistics, 143–170. Amsterdam: Benjamins. (https://www.rose.uzh.ch/dam/jcr:ffffffff-c23e-37d9-0000-00006e1a9200/Bossong_80.pdf)
Comrie, Bernard. 1977. Subjects and direct objects in Uralic languages: A functional explanation of case marking systems. Études Finno-Ougriennes 12. 5–17.
Comrie, Bernard. 1979. Definite and animate direct objects: A natural class. Linguistica Silesiana 3. 3–31.
Comrie, Bernard. 1981/1989. Language universals and linguistic typology: Syntax and morphology. Oxford: Blackwell.
Dalrymple, Mary & Nikolaeva, Irina. 2011. Objects and information structure. Cambridge: Cambridge University Press.
Dixon, R.M.W. 1979. Ergativity. Language 55. 59–138.
Haspelmath, Martin. 2021. Role-reference associations and the explanation of argument coding splits. Linguistics 59(1). 123–174. (doi:10.1515/ling-2020-0252)
Hopper, Paul J. & Thompson, Sandra A. 1980. Transitivity in grammar and discourse. Language. Linguistic Society of America 56(2). 251–299. (doi:10.2307/413757)
Iemmolo, Giorgio. 2011. Towards a typological study of differential object marking and differential object indexation. University of Pavia. (PhD dissertation.) (https://zenodo.org/record/4388380)
Moravcsik, Edith A. 1978. On the case marking of objects. In Greenberg, Joseph H (ed.), Universals of human language, vol. 4: Syntax, 249–289. Stanford: Stanford University Press. (https://zenodo.org/record/4688136)
Næss, Åshild. 2007. Prototypical transitivity. Amsterdam: Benjamins.
I wonder if the obviative of Algonquian languages might not be a counterexample: in Algonquian, the most prominent argument is the proximate one (which must be unique, and is suffixless for instance in Ojibwe or Cree) while less prominent arguments take the obviative suffix. In examples like (1), animosh ‘dog’, the proximate form, does not have a suffix, and serves as intransitive subject of niiskikaa- and transitive subject of migin- “bark at”.
(1) Mii igo ezhi-niiskikaa-d ‘aw animosh migin-aad ini makwa-n.
PRED EMPH THUS:IC-raise.hackles(VAI)-CONJ:3SG:PROX DEM:PROX dog:PROX bark.at(VTA)-CONJ:3SG–>3′ DEM:OBV bear-OBV
That dog’s hair stood up as he barked at the bear.
If the bear would be considered more prominent, it could be put in proximate form (makwa), and “dog” would receive obviative case (animosh-an) (verbal indexation would also be affected). (1) would become (2) (I did not recheck this example with native speakers, so there may be problems in it, but it is sufficient to make my point).
(2) Mii igo ezhi-niiskikaa-nid iniw animosh-an migin-igod a’aw makwa.
PRED EMPH THUS:IC-raise.hackles(VAI)-CONJ:3′ DEM:OBV dog-OBV bark.at(VTA)-CONJ:3’–>3SG DEM:PROX bear:PROX
Obviation can be viewed as a type of differential argument marking, except that it is not restricted to a particular syntactic function (subject vs object), and here clearly the most prominent argument is the one that is NOT marked.
This is a very interesting point! Obviative marking has never been mentioned in this context, but maybe it should be! I certainly need to think about it. Do you know anyone who suggested that an obviative marker is a kind of accusative marker?
I recall a discussion with Lynn Drapeau, where we both agreed that obviative is a case marker, but it is not usually the way it is presented in Algonquian linguistics.