Sometimes someone comes along and pulls the rug from under your most deeply-held ideas.

This time it is a thesis, entitled Unaccusativity at the Interfaces, by Patricia Irwin. You can access it here, although you might want to finish reading this before you decide whether you have the moral courage to do so.

As far as verb contructions go, English is an accusative language. That is to say, the direct objects of its transitive verbs assume the accusative case. Browsing Huddleston and Pullum shows no trace of any serious alternative, however confuddling English can become when its nouns do not inflect, and its pronouns are in worse disarray.

So we say I hit him; him is the direct object, and is put into the accusative case (and being a pronoun it stands a better chance of actually showing that too, as here). Thus we can analyse that simple transitive sentence as consisting of a subject (I, in the nominative case), a verb (hit), and a direct object (him, in the accusative case).

Other languages may follow different patterns. RMW Dixon (Basic Linguistic Theory vol 2) and others have given full details, but for my purposes I will now distinguish the ergative pattern, in which the subject is placed in a special case known as the ergative, and the direct object is put in an absolutive case, which is usually the same as the nominative case in being unmarked. So if English worked like that and the ergative case of I was mi, for example, an ergative version of that sentence might read Mi hit he.

Relatively few languages are fully ergative. It is more common for them to be split ergative, in which some situations require accusative marking as English does, and others require ergative marking. This may seem even more confusing, but those who speak these languages manage to do so without any confusion. Georgian is an example of a language with split ergativity, as it normally marks accusatively, but when using the perfective aspect (aorist screeve) it marks ergatively.

If you feel that a diagram and more examples might help, the relevant Wikipedia article is here.

Not content with accusative and ergative, Perlmutter unleashed unaccusative and unergative at us, although Dixon disparages the terms.

You will see here that Wikipedia defines an unaccusative verb as an intransitive verb (with no object) whose syntactic argument does not actively initiate, or is not actively responsible for, the action of the verb, or an intransitive verb which treats its argument like the accusative (direct object) of a transitive verb. Claimed examples in English are die and fall, whilst run and resign are claimed to be unergative instead.

It is one of Dixon’s contentions that unaccusativity and unergativity are ill-defined and should therefore be avoided, which is exactly what Huddleston and Pullum do.

Patricia Irwin, in her thesis, defines unaccusativity as a syntactic configuration in which a sentence has no external argument and a single VP-internal (VP = verb phrase) argument requiring structural case. English examples claimed include verbs of change of state, such as break, freeze, and those that denote motion and existence, such as arrive, drive up, (although these are rather different and best left to the thesis to argue).

These are ambitransitive verbs, in that they can be used as transitives (with an object) or intransitives (without an object). There are thus two ways that you can account for their behaviour. Huddleston and Pullum and pretty well everyone else has just accepted that they work as transitives when used transitively, and intransitives when used intransitively. Thus:
I froze him – transitive, so subject (nominative), verb, object (accusative)
He froze – intransitive, so subject (nominative), verb. (An alternative way of viewing this version is to call it an inchoative, but you didn’t really want to know that, I suspect.)

Patricia Irwin views this pair differently, as the direct object of the transitive version is the subject of the intransitive version. She argues that the movement and change in case in the theme (object) of the transitive version in generating the intransitive version is “a deep unaccusativity diagnostic”. She also proposes two other diagnostics, of resultative constructions and there insertion, which I will leave to her thesis too.

If you have stuck with me to here, you will probably be wondering why all this need for unaccusatives when all you need is to consider them ambitransitive? Well it is all to do with bringing the right sort of syntactic analysis out, so that it fits in with syntactic theory.

I think that I will stick with ambitransitives, thank you, and keep the rug under me.