LINGUIST List 6.642

Fri 05 May 1995

Sum: Morphological vs syntactic ergativity

Editor for this issue: <>


  • Bert Peeters, Sum: Morphological vs syntactic ergativity

    Message 1: Sum: Morphological vs syntactic ergativity

    Date: Wed, 3 May 1995 14:52:37 +Sum: Morphological vs syntactic ergativity
    From: Bert Peeters <>
    Subject: Sum: Morphological vs syntactic ergativity

    Original query:

    Reading through the book notes in *Language* 1990 (the things we do...), I stumbled across a BN on Anne Cooreman's *Transitivity and discourse continuity in Chamorro narratives*, by Thomas E. Payne (vol. 66:3, pp. 631-632). Payne reports Cooreman as saying that - I quote - "ergativity in Chamorro (...) is rather marginal, restricted to morphological marking in realis constructions and certain complement clauses. Cooreman claims there is no evidence of syntactic ergativity." Now, here is a possibly dumb question from someone who knows more about ergativity in superficially accusative languages than about the same in ergative languages or anywhere else for that matter. The question is: What's the difference between morphological and syntactic ergativity?

    Thanks to all of the following for replying: Leo Connolly Arantza Martinez Etxarri Randy J. Lapolla Alexis Manaster Ramer Amoena B. Norcross Tom Payne One person who wished not to be mentioned by name

    Most respondents referred back to either one of the following sources:

    Dixon, R.M.W. 1979. "Ergativity". *Language* 55. 59-138. Dixon, R.M.W. 1994. *Ergativity*. Cambridge: CUP.

    On p. 16 of the book, the following is said (S = intransitive subject, A = transitive subject, O = transitive object):

    "In summary, the terms 'ergative' and 'ergativity' - and 'accusative' and 'accusativity' - may be used: 1. To describe the ways in which the syntactic functions of predicate arguments are marked in simple transitive and intransitive clauses, i.e. whether S is marked in the same way as O and differently from A (an ergative arrangement) or whether S is marked in the same way as A and differently from O (an accusative arrangement). This is 'morphological' or 'intra-clausal' ergativity/accusativity ans it relates to the marking of syntactic relations in derived structure (not at any level of underlying structure). 2. To describe syntactic constraints that a language may place on the combining of simple clauses into complex sentences, by coordination, subordination, complementation, etc."

    Here are some of the more substantial answers (slightly edited)

    The distinction goes back to work on relational grammar in the early 70s, e.g., Woodbury on Eskimo, and perhaps independently to the work of Anderson and/or Dixon and/or Silverstein. The idea was that case marking and agreement were somehow more superficial than the other things. In fact, they have turned out not to be as superficial as once thought. However, the distinction remains a useful one, since it does look like these two kinds of ergativity are often quite independent of each other.

    Name withheld on request: Morphologically ergative languages have case-marking such that the subject of most/some intransitives matches the transitive object, not the transitive subject as in 'nominative' languages. The assumption (at least in a P&P framework) is to assume that these verbs are unaccusative, and somehow the case-marking gets to be done as if the argument was an object - e.g. the case gets assigned before movement to subject position where the equivalent of nominative would be assigned. Complications may arise if case marking also varies according to other factors. Cf. work by R.M.W. Dixon, and also by Hale and Bittner. The notion of "syntactically ergative languages" is most controversial. Dyirbal seems to be a good example (cf. R.M.W. Dixon's books; also A. Marantz, On the nature of grammatical relations, MIT 1984 - Marantz later refuted by Chomsky). The idea is that although they look like morphologically ergative languages so far as case-marking is concerned, the reason is different. The reason, again couched in a P&P framework, is that the transitive verbs are coded in the opposite way to what we do: i.e. these languages put the agent-like role internally, and the patient-like role externally. Most intransitives are coded to have just an external role, and this will get case-marked like the transitive external role ie the patient-like one, the 'subject'. The arguments are based on syntactic processes that depend on subjects, e.g. coordinate VP structure reduction, missing subjects, PRO subject, equi, imperative deletion, etc.

    Essentially, morphological ergativity is when there is a morphological case distinction between ergative and absolutive, like the Latin morphological distinction betwen nominative and accusative. Syntactic ergativity is when syntactic processes are sensitive to the ergative- absolutive distinction. For example, some languages allow relativisation on absolutive arguments, but not on ergative arguments. My favorite example is from Yup'ik where coreference in coodinate clauses is based on the absolutive. So if you say: "He ate the bug and got sick" it is the *bug* that got sick, even though pragmatics would seem to make the other reading more probable. Syntactic ergativity is logically distinct from morphological ergativity, though the two often go together. In 1976 Stephen Anderson claimed that ergativity is usually just a morphological phenomenon. Since then, syntactic ergativity has surfaced in various places, but it does still seem to be the case that most syntactic processes that are sensitive to grammatical relations in morphologically ergative languages are oriented to the nominative/ accusative distinction. There are no examples that I know of where a morphologically nominative/accusative language exhibits syntactic ergativity.

    Morphological ergativity refers to case-marking (or the equivalent). In accusative languages such as English and French, the subject of a transitive agentive verb gets the same treatment as the subject of an intransitve (one-place) verb:

    I ate a banana. I died.

    In a morphologically ergative language, it's the patient NP (our DO) that is treated the same as the subject of a one-place predicate. The agent (or whatever) gets some other treatment. In a lousy imitation (note that the verb has the same morphology as with the one-place predicate):

    I died. A banana ate by me.

    Syntactic ergativity is quite rare. Most morphologically ergative languages treat the Agent (or whatever would be the subject in a "typical" accusative language) about the same as the subject of the one-place predicate: it can undergo raising, equideletes with other Agents and with the sole NP of a one-place predicate, while the Patient, despite its subject-like "look", equideletes only with other Patients or not at all. In other words, the syntax of these languages works much as in English or French, without regard to case-marking. Such languages are said to be syntactically accusative. There are, however, some ergative languages (Dyirbal is the most consistent example I know) where the patient of such sentences is not only morphologically the subject, but also acts like one syntactically. This seemingly obvious behavior is quite rare. I know of no languages which are syntactically ergative without also being morphologically ergative. Nor should there be any, in my opinion. I use a case grammar analysis, which in brief is as follows: 1. Deep cases (thematic roles) are ranked. Agents outrank patients (themes), while patients (theme) outrank such oblique categories as Place. 2. Accusative languages such as English normally make the highest- ranking NP the morphological subject. However, even though experiencers and beneficiaries (possessors and "typical" indirect objects) rank right below Agent, some verbs in some languages prefer to make the patient the subject. (Cf. English _own_ with Beneficiary as subject, but _belong to_ with Patient as subject.) 3. In some languages a higher-ranking non-subject shows subject-like properties that the lower-ranking morphological subject does not have. Usually this involves only position, but in Icelandic the darned things can undergo raising and equidelete with those subjects which rank highest in their own clause. This causes many scholars to claim that such NPs are then the "real" or "syntactic" subjects. I say rather, the properties in question in a given language are not subject properties at all, but pertain rather to the highest-ranking NP argument *as such*. The morphological subject -- the traditional subject, and the only one that cannot be called something else -- usually has syntactic properties of its own, such as equideletion and control of verb agreement. The exact distribution varies from language to language. In English, for instance, high-ranking non-subjects seem to have no syntactic salience whatsoever. But they do in German (position) and Spanish (position and often elimination of verb agreement). In Icelandic, they're nearly the whole show. Morphological subjects there are syntactically significant iff they are also the highest ranking NP argument. 4. Ergative languages make the Patient (if present) the morphological subject, even in the present of an Agent, Experiencer, or Beneficiary. But in most such languages the higher-ranking NP argument has most of the syntactic "subject" properties (controlling reflexivization; equideletion; can be raised). That is, the properties in question in most ergative languages pertain to the Highest-Ranking argument and hence apply only to those morphological subjects which are also (with Theme- location verbs or the like, or after passive or "anti-passive" has applied) the highest-rankign arguments. But in Dyirbal, and perhaps a few other languages, the morphological subject is the whole show, as it is in English. 5. There are thus two centers of syntactic salience: the morphological subject, and the highest-ranking NP argument. But where they coincide, as they normally do in accusative languaes, syntactic ergativity is impossible. It can occur only where, and when, the morphological subject (if any) is not the highest-ranking NP argument. (So-called "active" languages, with no uniform "subject" morphology, are best analyzed as lacking subjects altogether. All so-called "subject" properties then must, of course, pertain to the highest-ranking NP argument.)

    If you're interested in a language whose morphology is thoroughly ergative but whose syntax is thoroughly nominative-accusative, there is Tzotzil (a Mayan language I've had the opportunity to do a fair bit of work on). There's a good pedagogical grammar of the language published in Spanish by UNAM Press (Haviland, 1981). I recently translated it into English. xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx Dr Bert Peeters Department of Modern Languages (French) University of Tasmania GPO Box 252C Tel. (002) 202344 +61 02 202344 Hobart TAS 7001 Fax. (002) 207813 +61 02 207813 Australia Email: