This is a general and brief question. Is anyone familiar with a language which can be largely ambiguous with respect to whether the construction is active or passive, to the degree that in some cases it is only determined pragmatically?
The term passive is a grammatical one relating to syntax and morphology. It does not relate to a verb having a 'passive' meaning. The English metalinguistic term passive is unfortunate and persistently leads to misunderstandings with regard to what a passive is.
It's often possible to reverse the semantic roles of a subject and an object in a sentence merely by replacing the verb (and adjusting some pronouns or prepositions):
- Bob gave a book to Ben.
- Bob received a book from Ben.
In (1) Bob, the subject, is a donor and Ben, contained within a preposition phrase complement, a recipient. In (2) these roles are reversed to the extent that Bob, still the subject, is now the recipient and Ben, still contained in a preposition phrase complement, is now the donor. However, this reversal of thematic roles, and the fact that receive doesn't require any action on Bob's part, does not make sentence (2) a passive version of (1). Indeed it is not a passive. It is still an active sentence, just like (1).
Verbs—or one-word story frames, as I like to think of them—in conjunction with other syntactic or morphological aspects of the clauses they appear in, assign specific semantic roles to the denotees of specific phrases, usually on the basis of their grammatical relations. So in:
- X punched Y
the verb punch along with the active voice construction it occurs in assigns the subject, X, with the role of assailant and the object, Y, with the role of victim. Note that the term active here is also a grammatical and not a semantic one. We can use an active voice construction with a verb that does not denote any kind of action:
- X survived Y (As in Bob survived his nephew)
This sentence indicates that X didn't die before Y did, not that X undertook any sort of action.
What a passive clause does, is assign the subject with one of the semantic roles that's normally assigned to a different phrase (or rather the denotee of a different phrase) in an active voice construction. Compare (5) and (6):
- Bob punched Ben.
- Bob was punched.
In (5) the subject, Bob, is assigned the role of assailant and the object, Ben, the role of victim. In (6) the subject, Bob, is now assigned the role of victim. This role would be assigned to the denotee of the direct object in an active voice clause. Note that (6) has less expressed arguments than (5). The verb punch doesn't have a direct object here, and the semantic role of assailant, normally aligned with the subject, hasn't been reassigned to any other phrase. In most languages a passive clause normally reduces the valency of the verb. In other words the verb usually takes one less complement than it does in an active voice clause. This is not always the case. In Japanese, for example, the semantic role normally assigned to the subject may, in a passive clause, be aligned with a phrase appearing as a complement of the verb. (In English we can, of course, add the missing entity back in using a preposition phrase adjunct: Bob was punched by Ben).
The Original Poster's question
The Original Poster asks whether there is a language where it is largely ambiguous whether a construction is active or passive and this can only be determined pragmatically. If 'active' and 'passive' clauses were homophonous, first of all that language would have to be one where a passive did not normally reduce the valency of the verb. Secondly it would have to be one where one couldn't tell whether a phrase was the subject or, for example, the object of a clause on the basis of either syntax (word order or number of words) or morphology (any kind of case marking, verbal inflection, particles etc). This would seem to put a huge load on the cognitive system of the listener. It would also make it difficult to differentiate some verbs with opposite meanings. If:
- Preceded Ben Bob.
... could mean either Bob preceded Ben or Ben preceded Bob, then it would be difficult to differentiate the meanings of precede and follow. This is not necessarily fatal, however. For example, many language use the same verb to indicate both lending and borrowing.
However, the biggest obstacle to the existence of a language where passive and active constructions are homophonous is the term construction. To say that a language makes use of an active construction is to say that it has clauses which have specific grammatical properties which indicate that the clause is active and that tell the listener what kinds of semantic roles to assign to which phrase. To say that it has a passive construction is to say that it has a clause type whose grammatical characteristics differentiate it from an active clause and tell the listener to assign the semantic roles in a specific way which is different from what is found in a corresponding active voice clause. If a language had polysemous clauses which could be faithfully translated into another language as either an active or a passive clause with the correspondingly different semantics, this would mean that that language specifically lacked either an active or a passive 'construction'.