I'm not sure what you would be right about, but you could be right about something. I think you're wrong if you think it's obvious, even if it is true. In A, you claim that John is the subject, and I have no quarrel with that. Then in B you conclude that "parcel" is the subject and John is the IO. But actually, John is the subject. Notice that we can substitute various pronouns and NPs in that position ("I was sent...", "you were sent...", "the children were sent...") and we notice that that first clause governs verb agreement – because that is the subject.

You are mixing syntactic and semantic analyses. The passive has the function of making the post-verbal nominal become the subject, and it exiles what would have been the subject into a "by" phrase (which you can just omit). In the good old days, semantic relations were determined before the various movement rules messed up what is the subject vs. the object.

You gave contradictory analyses to your examples. You assigned subject role in A based on the syntax (order and agreement), but then assigned the IO role to John in B based on semantics. You have to pick consistent diagnostics. This is why people sometimes talk of the "logical subject", when referring to semantic roles (and thereby avoiding the complication that "subject" isn't a semantic role, but "agent" is, as is "recipient").

"Apple" is a direct object, or in relational grammar terms, a direct object chômeur (former direct object, having been usurped by Dative Movement). Functionally speaking, so-called subject/object/indirect object relations do not change. As I said. Formally, though, they do: something becomes the subject, or direct object. I would certainly insist that if one is asking about "subject" as a formal part of speech, that the form of the pronoun and facts about word order and governance of agreement are all and only the valid factors to consider. And if you are asking about functional relations, then you are not asking about formal relations – so you first have to decide what question you are asking; is it about form (syntax) or function (semantics)?

Is the reluctance to accept that "I" is not the subject simply because "I" is a subject pronoun and comes in the position where one expects the subject to come in a statement. If you say that "I" must be the subject because it is a subject pronoun that is to insist that the function must follow from the form. If you say that "I" or "John" must be the subject because it comes first in the sentence it is because you are following a rule which has been formulated without taking into account exceptions. In short, you are saying that because "I" and "John" behave by position and form as if they are the subject they must be the subject.

It is not only the position and form of "I" and "John" that indicate that they, rather than "parcels", are the (syntactic) subject. There is also the fact that we would say "I was sent parcels" and "John was sent parcels", not "I were sent parcels" and "John were sent parcels".

Furthermore, we can say: "John was sent (and I delivered to him) a parcel". It would be strange to call "parcel" the subject of "was sent" but the object of "delivered".

Many years ago I read in a book that the form “John was given an apple” was ungrammatical but good English. The comment was made tongue in cheek. It is an oxymoron, but it does make the point I am trying to get at quite well. When it comes to language change there is something of a paradox as on the one hand a descriptive linguist will say that native speakers do not make mistakes, but on the other language changes due to native speakers “making mistakes” in the sense that they change what went before. I would say that the form was ungrammatical but is now accepted, but that that does not stop it being an oddity which is not amenable to the standard analysis of passive sentences.

I discuss something along those lines here, with an extensive bibliography:    http://hdl.handle.net/2142/101864

It's interesting to think about whether the double-object passive might fit into that...

...paradox as on the one hand a descriptive linguist will say that native speakers do not make mistakes...

Yes, that point is discussed in the paper linked above. But to be clear, the claim isn't that native speakers can't make mistakes, but that what is generated by the grammar of a native speaker (i.e., Competence, in Chomsky's terms) is, by definition, part of the grammar of the language. That is not to say that mistakes never occur: indeed they do, when something goes wrong in processing (i.e., Performance). In principle, we might say that speakers should recognize when they make a mistake in that sense (a "slip of the tongue", the sort of thing they'd self-correct given the opportunity), but sometimes the lines can be blurred. Regardless, and setting aside nuances of how exactly to phrase it, the claim isn't that native speakers perform perfectly, but that their knowledge of the language is by definition representative of the language. (To be clear, this is why Chomsky insists on individuals being the object of study, rather than speech communities, so that we can only really study the internal language of one individual at a time, although we often generalize for convenience to a language like "English".)

To put it another way, the crucial claim is that one cannot make up independent or "logical" or whatever rules and then tell native speakers they are wrong for not conforming to expectations. If the written grammar rule conflicts with the native speaker, then the native speaker is right. That's really the only way this makes sense.

It is accepted that active sentences with an object can be converted to a passive sentence with an agent and vice versa.

Yes.

John was sent an apple?
All the explanations which insist that “John” is the subject seem to be convoluted

"Subject" is typically defined based on the sentence in final form, not its derivation (hence why "passives" have "subjects" at all!), so you're using "subject" to refer to a different concept than most people would mean by "subject". You seem to be focusing on semantics, while most definitions focus on syntax.

and there is no satisfactory explanation of what role “apple” plays.

OK, now we have reached a relevant question:
1. Assume the passivization rule takes the direct object and promotes it to subject.
2. Assume apple is object.
3. Why is John promoted rather than apple?

Good question.

The problem, though, is regarding the structure of the double object construction, not passivization. Passivization then acts as a sort of test:

ACTIVE:
Henry sent John an apple.

PASSIVE:
John was sent an apple.
*An apple was sent John.

There are two ways to try to solve this problem:
1. Our passivization rule is wrong. It isn't strictly direct objects, but something like a prominent object, or something else. It's a non-subject, that's clear. The rest is murkier.
2. In fact, John is the direct object, and apple is something else.

Actually, (2) might be the right analysis, or at least one in which there are two direct objects in the sentence. Double object constructions like that are weird syntactically, and there have been many debates -- still unresolved -- about how exactly to treat them. For example, they seem to be a case of ternary (non-binary) branching structures (sort of like coordination). There's a lot of literature to read if you're interested (and many syntax textbooks take up that example for discussion!). Note that one way we could explain it then is to say there are literally two objects there, and either one might be selected, but somehow because there are two conflicting, the derivation crashes and settles on only the first. Or historical pressures settled on promoting the recipient because it was already easy to passivize the transferred object (apple) from the other ("apple to John") construction.

But there are also reasons to think (1) is the right explanation. Specifically, passivization is not strictly limited to direct objects. For "important" prepositional phrases closely associated with verbs (maybe phrasal verbs, combining verb plus preposition, but it's unclear how exactly to draw that distinction), we can sometimes get odd passivization too:
"That bed was slept in."

Generally it seems like only the "strongest" object can be passivized:
?*The bed was eaten pizza in.
(<John ate pizza in the bed.)
*A box was sent John the apple in.
(<Someone sent John the apple in a box.)

Regardless:

All the explanations which insist that “John” is the subject...

But that's straightforward: a "subject" in English is simply whatever noun phrase is found before the verb with which the verb agrees. (You might phrase it slightly differently or find exceptions to the ordering, but something along those lines works, and it isn't complicated.)

The problem is that you're mixing up "subject" in the original sentence, and subject of the passivized sentence (they are of course not the same!) as well as the subject of some other, distinct passivized sentence (again, not the same).

there is no satisfactory explanation of what role “apple” plays. It clearly cannot be the object because a passive verb is intransitive and cannot have an object...

Passivization involves promotion of one object to subject, not necessarily all objects, and if the verb is ditransitive, it might end up having more going on. Again, these are open questions depending on how you analyze everything, but you can't assume all passive verbs are intransitive.

In fact, it can get even weirder than that. In German, intransitive verbs themselves can passivize:

"Es wird getanzt."
(lit. "It becomes danced.", where 'become' is equivalent to English 'be' in passives)
Meaning something like "Someone is dancing." or "There's dancing."

So passivization can be thought of as a change to argument structure, where valency is decreased by one. Not necessarily intransitivization.

and it is not the indirect object as nothing is going to it.

That's a semantic argument for a syntactic analysis. In fact, we could call it an indirect object syntactically in the sense that it seems to behave as a secondary/minor object, with "John" being the one that is promoted.

Thematic roles do not always correlate to syntactic structure/position. We get the same configuration with different meanings.

For example, consider the argument structure for Spanish amar vs. gustar:

Yo amo pizza.
I.NOM love.1SG pizza
'I love pizza.'

Me gusta pizza.
me.ACC like.3SG pizza
'I like pizza.'

These so-called "dative subject" constructions make everything weird (what's a "dative subject" anyway?-- note that is primarily a semantic argument!), but the point is that the configuration of arguments there is actually relatively straightforward: subject is what the verb agrees with, and some verbs are just semantically backwards in assigning their thematic roles. (I'm oversimplifying for illustration purposes, but that's certainly one way to look at it.)

A far better analysis is to say that the form is an exception to the rules relating to word order and the form indirect object pronouns must take.

Maybe. That is similar to some analyses of "dative subjects" in Spanish and other languages. By analogy you might come up with some good arguments for passivized double-object constructions in English.

It follows that if in this form the indirect object is “mistaken” for the subject that the verb will agree accordingly so that...

Hm, maybe. Again, similar to some analyses of "dative subjects". But personally I still think the issue rests in the question of the structure of the double-object construction, and once explained would tell us why John (not apple) is promoted out of that.

Whilst when considering a language at any given moment you perhaps ought not to look at its history, you cannot get away from the fact that the reason English (and as far as I know no other language) has this form can be explained as I set out in an earlier post.

Connected to the paper I linked above. Of course in that paper I emphasize the importance of such analyses, but I should also emphasize caution: apply only when necessary, to avoid assuming an overly complicated grammar. But you might be on to something here.

Personally I'd think the explanation is rather straightforward historically: we already had a way to passivize the direct objection ("send the apple to John", the dative construction), so we then adjusted the way to passivize the double object construction ("send John the apple") to promote the recipient instead. This actually might have come about as linear reanalysis like I describe in the paper ("look for the first noun after the verb and move it over to subject position", ignoring the rest of the complicated sentence). It's an interesting example to think about!


Thanks for following through with this discussion. It took a bit to get there, but this is an interesting example in terms of why the indirect object is promoted to subject. Again, by definition it then is the subject, because that's how we define subjects in passivized sentences, but there is a puzzle about the analysis, and the historical perspective may be helpful.


[I edited this reply a few times because by now you've convinced me something somewhat unusual is going on here, although it's still debatable exactly "where" that step is located in the derivation.]