About the universality of the notion of subject and the description of ergativity

Question

A very common description of ergativity defines it as a morphosyntactic alignment where the intransitive subject follows a pattern similar to the object and dissimilar from the transitive subject --- in opposition to the accusative alignement (like the answers in Accusative or ergative language?).

But this presupposes that the very concept of subject would be applicable in all languages and as the answer of How do linguists define the idea of a grammatical subject? makes the case, this is questionable.

So in the end it seems to me that the definition of ergativity in terms of intransitive/transitive subject and object is hardly satisfying and looks either circular or ethnocentrist. So why is it still in use? And more generally why isn't the fuzzy and dubious notion of subject made obsolete in linguistics and replaced by more rigorous and appropriate concepts when dealing with languages in general?

Answer 1

in the end it seems to me that the definition of ergativity in terms of intransitive/transitive subject and object is hardly satisfying and looks either circular or ethnocentrist. So why is it still in use? And more generally why isn't the fuzzy and dubious notion of subject made obsolete in linguistics and replaced by more rigorous and appropriate concepts when dealing with languages in general?

This paragraph seems to suppose that in all languages that display the phenomenon of "ergativity", the concepts of "subject" and "object" are artifacts of applying an analysis based on non-ergative languages. This really isn't the case, though: many languages have ergative morphology and also other aspects of grammar that make it possible to demonstrate in a rigorous fashion that absolutive-marked NPs in intransitive sentences and ergative-marked NPs in transitive sentences behave alike in cetain contexts or in certain respects. That is, there are motivations for distinguishing the concept of "absolutive-marked noun phrase" from the concept of something we can call the "subject" that arise from the structure of these languages, not just from the imposition of external concepts of subject- and object-hood taken from languages without ergative morphology. Sure, we could call this concept something other than "subject" if we wanted to, but "subject" seems to fit the semantics and has mnemonic value, so why use a different word?

In many (but not all) languages with an "ergative" system of case-marking, the noun phrase that is morphologically marked as ergative case in a transitive clause acts syntactically similar in a number of ways to the noun phrase marked as absolutive case in an intransitive clause. This phenomenon is often called "split ergativity".

So within many particular languages that makes use of ergative marking, we can have quite clear evidence of a concept of "subject" in those languages from other patterns of marking in other contexts (e.g. the difference in case marking in the past/perfect and present tenses in Hindi and Urdu), or from syntactic evidence e.g. the behavior of "pivots".

---

As for the more general typological criteria for applying the concept of "subject" and "object" to a language that only uses ergative morphology and has a fully ergative syntactic system, there might be some fuzziness. I haven't studied the matter, so there might be more subtle grammatical phenomena other than morphology and pivots that we can use to identify the subject. If not, it just means that "subject" and "object" are defined at this level by mainly semantic grouping, which is not that strange for universally applied typological categories. We can distinguish them by the same vague kind of criteria that we use to distinguish "nouns", "verbs", and "adjectives".

At that level of typology, these kind of "universal" classifications are often disputed if we take them to represent some important fact about the structure of all languages, but they are often felt to at least provide some rough method of inter-linguistic comparison. It's easier to remember a few terms that mean slightly different things in different languages than it is to remember a bunch of terms that are all specially designed to refer to specific features of individual languages.

Answer 2

Greenberg's universal #38 ("Where there is a case system, the only case which ever has only zero allomorphs is the one which includes among its meanings that of the subject of the intransitive verb") presupposes the existence of intransitive verbs, and by convention the primary argument of an intransitive verb, to which the verb may agree, is called the subject. All languages that have intransitive (monovalent) verbs will have subjects, which makes this a pretty uncontroversial linguistic topic (as far as topics in linguistics go).

Morphological ergativity is something that can only be described in languages that have morphological case marking. The form that the marker of a NP may take, to mark for the argument of intransitive verbs, will determine what the 'alignment' is; if the marker is used to mark certain thematic roles of transitive verb arguments (such as agent), the agreement will be accusative, and if the marker is used to mark other roles (such as patient), the agreement will be called ergative. In both cases, linguists typically name the agreement after the secondary marker's name (accusative as the marker for P in transitives where S & A have the same marker, and equivalently for the ergative and A marking).

Syntactical ergativity is a different beast, and does not care for the surface phenomena of morphological ergativity (and is equvalently called 'deep ergativity'), but it still hinges on the way the S argument chooses to align with either the P or A arguments.

I do not see where the fuzziness comes in. Robert Dixon's Ergativity is a brilliant read on the nature of morphosyntactic alignment, even though it is quite a hearty read. Defining ergativity as the phenomenon of the S argument aligning with the P argument does not seem circular, as the S/A/P argument system is divorced from this consideration. I do not see which concepts are more appropriate, if intransitivity is taken as a given (unless one is willing to dispute basically all of syntax, as well as Greenberg's typology).