Sentences in the active voice can be converted to the passive voice by - amongst other maneuvers - moving the direct object into the position originally occupied by the grammatical subject.
Does this make it the new grammatical subject of the sentence? If so, what is the original subject now called? If not, how are subject and object(s) classified after the conversion?
The former object becomes the new subject. That is clear -- the new subject has all the properties one could reasonably associate with a subject. Number agreement with the verb and subject raising from complement sentences, for instance.
What happens to the former subject is a more interesting question. Here are 3 theoretical answers:
I. It may not be represented overtly at all, in which case the original subject was an abstract thing called "UNSPEC" (short for unspecified). This is argued by McCawley in The Syntactic Phenomena of English.
II. It may become part of a manner adverb with "by". This is argued by Chomsky in Aspects of the Theory of Syntax. The reasoning here is that we need to explain the "fact" that only verbs that are subcategorized to co-occur with manner adverbs can be passivized. Unfortunately for Chomsky's proposal, this is not so. Cf. "Ohio is bounded on the north by Lake Erie."
III. In Relational Grammar (by Postal and Perlmutter in various articles), the former subject is a chomeur, meaning it no longer bears any grammatical relation in the clause. I like this theory, because of an interesting generalization discovered by McCawley (reported in the same reference as above).
McCawley undertakes a detailed examination of the structure of passive sentences, and finds that the passive by-phrase can turn up in several different places -- in fact, anywhere it will fit into the derived passive structure.
Now, for those theories that associate the internal structure of clauses with the order in which verb-functions apply to their arguments (like Categorial Grammar and HPSG) and further associate this order with grammatical relation (like HPSG and my theory, 2psg), McCawley's fact is predicted.
Since a by-phrase chomeur bears no particular grammatical relation to the verb, the order in which the verb-function applies to this argument is arbitrary, so you should get a number of different derived structures in the passive.
I believe I am the first to have noticed this convergence between McCawley's theory and Relational Grammar.
You have this sentence:
"She was kissed by his lips"
Syntactically the subject is "she". This concerns the order of the words in relation to the verb: she is the subject because it is positioned in the place of the subject. Also you have many other reasons to argue that this is the subject, like the morphological one: you can see that the verb "was" is conjugated in singular to correspond with "she", and not with "his lips", which is in plural.
Semantically -concerning the meaning of the words- you can say that the agent (not the subject, but the agent), which is the thing that is performing the action, is his lips, as they are, in the meaning, the things that are performing the action onto her. However, I insist, this is not what is called "subject".
So, in the sentence:
"His lips kissed her"
The subject (which is evaluated syntactically [concerning the order of the words] and morphologically) is "his lips", and the agent (which is evaluated semantically [concerning the meaning]) is also "his lips", they coincide in this case, but that is not always what happens.
Diagram here of Passive Version Of Ditransitive Active
Note: The diagram represents the passive version of the active ditransitive clause:
My aunt gave Ed a pair of shoes.
Note:
In the passive, the direct and indirect object of the active clause are internal complements of the verb (i.e. are constituents of the VP), but “a pair of shoes” is not internalised because it is an internal complement within the VP in the corresponding active clause: its status remains the same, where “internalised” implies a change.
