From Latin referre "carry back" via French. (Oxford English Dictionary)
In the sense of a person mentioning or talking about something, the verb refer is attested in English since the mid-16th century. In the sense of a word or symbol designating something, it is attested since the mid-17th century. These early attestations are all from ordinary, non-technical usage, not from theoretical studies of language. I imagine it would be very difficult to identify a clear dividing line between such uses and the introduction of refer as a piece of technical terminology.
The OED traces to Bradley (1883) the use of the noun reference for "The action or fact of applying words, names, ideas, etc., to an entity; the relation between a word or expression and that which it denotes." Bradley writes: "Judgment is not the synthesis of ideas, but the reference of ideal content to reality." There is no reason to suppose that Bradley thought of himself as introducing a new piece of terminology; reference would be the natural noun to use, given his characterization earlier in the book (p. 10) of a judgment as "the act which refers an ideal content... to a reality beyond the act." It is interesting that refer here represents a 3-place relation, rather than a 2-place relation; in addition to the "ideal content" and the "reality" it represents, there is the "judgment" which refers the content to the reality.
Referent has a much clearer moment of introduction — in Ogden and Richards (1923):
The word 'thing' is unsuitable for the analysis here undertaken, because in popular usage it is restricted to material substances — a fact which has led philosophers to favour the terms 'entity,' 'ens' or 'object' as the general term for whatever is. It has seemed desirable, therefore, to introduce a technical term to stand for whatever we may be thinking of or referring to. 'Object,' though this is its original use, has had an unfortunate history. The word 'referent,' therefore, has been adopted, though its etymological form is open to question when considered in relation to other participial derivatives, such as agent or reagent. But even in Latin the present participle occasionally (e.g. vehens in equo) admitted of variation in use; and in English an analogy with substantives, such as 'reagent,' 'extent,' and 'incident' may be urged. Thus the fact that 'referent' in what follows stands for a thing and not an active person, should cause no confusion.
The point about the "etymological form" being open to question is that the -ent suffix in referent (which derives from the Latin present active participle ending) would lead one to expect that a "referent" is one who refers, not one who is referred to. After all, a "president" is one who presides; a "dependent" is one who depends, etc. Ogden and Richards defend their terminology by pointing out that even in Latin, the participial ending was not perfectly consistent in interpretation: vehens, the present active participle of vehere "to carry," was used interchangeably with the perfect passive participle vectus, so that vehens in equo can mean "carried on a horse" (that is, riding a horse), rather than "carrying on a horse." This is a pretty weak defense; vehens was one of a very small number of active participles to be interpreted this way, and I don't think referens (the active present participle of referre) was one of them. But Ogden and Richards are correct that English -ent is not consistently agentive in meaning, as their examples of reagent, extent, and incident show.
References
- Bradley, F. H. (1883) The Principles of Logic. K. Paul, Trench & Co.
- Ogden, C. K. and I. A. Richards (1923). The Meaning of Meaning: A Study of the Influence of Language Upon Thought and of the Science of Symbolism, K. Paul, Trench, Trubner & Co.