It is sometimes desirable in theories of
syntax to postulate elements in sentences that are not pronounced. They have
grammatical effects but no phonetic output. These are called
empty categories, and a number of distinct kinds can be discerned: two of them are the
pronouns
pro and
PRO, also called little pro and big PRO.
Roughly the difference is that pro is used for case-marked positions, the subject and object of a main verb, while PRO is used in positions that can't take case, the chief one being the subject of an infinitive or gerund in a subordinate clause.
pro
English does not allow
pro. All such positions have to be filled explicitly, even if only by an
expletive such as 'it' or 'there':
It seems Mary is ill. If a language allows the empty
pro it is called
pro-drop, and this is believed to be a
parameter of
Universal Grammar. Italian, Spanish, Chinese, and Japanese are all pro-drop. The former two have a rich
agreement system, so you don't need a subject pronoun because the subject is adequately marked on the verb (
mangio 'I eat',
mangiamo 'we eat'); however Chinese and Japanese have no agreement at all yet freely omit subjects and objects.
Bantu languages such as Swahili have verbal prefixes for both subject and object, so don't need either of these separately:
alikiandika 'they wrote it'. Such pronoun-less sentences can be regarded as having underlying forms like
pro mangio and
pro alikiandika pro, with the
pro satsifying the need of the verb to mark something as subject or object.
PRO
By contrast
PRO is postulated where it can't be marked, but is still logically a subject. In English it occurs in sentences such as
John tried to leave and
Eating people is wrong. In the first one, John is the subject of 'try', since it's John who's trying, but what John is trying to achieve is the state where John leaves: there is an internal clause with an unexpressed subject. You can separate the two subjects and say
John tried to get Mary to leave, but you can't say this more directly as *
John tried Mary to leave..
This is analysed as John tried PRO to leave, and to indicate that both expressions refer to the same person, that is the PRO is an anaphor referring back to its antecedent, we notate this by using the same index: Johni tried PROi to leave.
The verb try doesn't mark a direct object. One that does is persuade. In John persuaded Mary to leave, Mary is the object of persuade but she is also the subject of to leave. You can separate these two roles: John persuaded Mary that Bill should leave, but you can't do so while still using the infinitive: *John persuaded Mary Bill to leave. With the infinitive the inner subject co-refers with the outer object, and we co-index PRO accordingly: John persuaded Maryi PROi to leave.
The co-reference of PRO with one of its antecedents is called control. Try has subject control, and persuade has object control. A third possibility is arbitrary control, where there is no antecedent. This receives a general reading, something like the pronoun 'one' or 'anyone': It is difficult PRO to leave quietly.
A gerund is a clause that is acting as a noun phrase: Eating in public disgusts John or John dislikes eating in public. The embedded clause has a subject of its own, and this can be taken as PRO with either bound or arbitrary readings: Johni dislikes PROi eating in public with same indices meaning John dislikes John doing it, or Johni dislikes PROj eating in public with different indices meaning he dislikes people generally doing it.
PRO and binding
In
binding theory (q.v. for details) a clear distinction is made between
pronouns like
him and
anaphors like
himself. Anaphors have to be governed by a local
antecedent, roughly one in the same clause, while pronouns can't be. Their distributions are mutually exclusive.
PRO however behaves partly like a pronoun and partly like an anaphor, which is why it can be ambiguous between antecedent or arbitary control. This would be a contradiction, unless we concentrate on the fine print: an anaphor must be bound in its
binding domain, a pronoun must not be bound in its binding domain. For
PRO to be both anaphor and pronoun, it must fail to have a binding domain at all. This result is called the
PRO Theorem.
It accords with the positions PRO is found in, as the subject (and only the subject) of an infinitive or gerund. These are both non-finite verbs. A finite verb agrees with its subject in number, tense, gender, and so on. (The description I've been giving so far is very closely tied to English examples, but this whole discussion is applicable to any language, and details of what agreements occur vary between languages.) The inflection of a finite verb causes its subject to have nominative case. Therefore a non-finite verb cannot assign case to its subject. This is why PRO in this position is not governed, i.e. does not have a binding domain.
There are two other ways of assigning case: verbs and prepositions assign case to their complements (see me and from me), whereas nouns and adjectives cannot directly (*teacher me or *proud me -- they need to interpose the preposition of). In situations where either a verb or a preposition governs a subject position, PRO does not occur. Or to be more honest, is considered not to occur, because there are nasty gaps in the theory at this point that I can't explain.
When a verb in an outer clause governs the subject of an inner clause it's caused Exceptional Case Marking: an example is Mary believes him to be clever. This is equivalent to using a finite clause Mary believes he is clever where the subject he is nominative. In the first example the inner subject him is governed by the verb believe, and you can't use PRO: you can't say *Mary believes to be clever. However, sadly, this does not account for pairs such as Mary expects him to win and Mary expects to win, so I don't know what current wisdom is about those. It looks awfully like a governed PRO to me.
A non-finite clause can be preceded by the complementizer for. This is a preposition so it assigns accusative case (or some non-nominative case at any rate) to the subject: For him to leave is surprising and Mary arranges for him to leave. As the subject position is now governed, it is ungrammatical in standard English (though not in some dialects) to use PRO instead of an overt pronoun: *For to leave is surprising and *Mary arranges for to leave.
I was made somewhat less confused by repeated reading of the relevant sections in:
Ouhalla, J. (1999) Introducing Transformational Grammar, Arnold.
Culicover, P. (1997) Principles and Parameters, Oxford.