The relation of speech reproduction or enunciative relation is a relation between two propositional entities p and q as follows: p is of one speaker's responsibility, and it attributes q to another speaker's responsibility. A cover term for q in this configuration is ‘reported speech’; its original speaker is its author. Types of speech reproduction are distinguished by the following criteria:

  1. the degree to which the configuration of p and q is a construction of the linguistic system, including the degree of grammatical integration of q into p,
  2. the exact propositional status of p and q.

Ad #a: Dispensing with any dedicated device of the linguistic system, the speaker of p may just switch roles, as in .

.Now here comes old McWhinney: “Who left the light on again?”

If were a spoken utterance, its speaker might use paralinguistic methods to signal the enunciative relation and/or the author change between p and q. Since it is a written example, punctuation is used to this effect. At any rate, exhibits no lexical or grammatical marking of the relation between the two clauses or their meanings.

Incipient use of recourses of the linguistic system to indicate the enunciative relation consists in explicitly introducing q in p by a speech act verb or a speech-act designating word of another category. In such cases, q may bear an intrinsic relation to that speech act verb (commonly, an effected undergoer). For instance in , said is the speech-act verb and the string in quotation marks is the reported speech, functioning as the direct object of this verb.

.Linda said to Irvin: “I have done my share; now do your duty!”

The reported speech may also contain markers of not being speaker's own speech, for instance a quotative particle, as in ,

.Yístáchiteish-á"Káyísm-i̱."ni̱.
Cabecar1.SGgrandfatherERG3say-PFV[ NEG1.SGgo-IPFV ]QUOT
My grandfather said: “I’m not going.”

Concerning the degree of integration of q into p, the semantic (in-)dependence of q may be distinguished from its syntactic (in-)dependence. The former is the distinction between direct and indirect speech:

In direct speech (or quoted speech, traditionally termed oratio directa), the speaker pretends to render the author's original wording, possibly including paralinguistic features. He conserves the author's deixis and illocution, but does not appropriate them, so they have no effect in the speaker's speech situation. Thus in , the word I in the quoted speech does not refer to the speaker of and instead to Linda; and analogously in . Likewise, the speaker of does not pose a question; he quotes it.

Indirect speech, traditionally called oratio obliqua, is the rendition of an author's speech in the current speaker's terms. The speaker only renders the content of what the author said; he does not vouch for the original wording. The deixis inside indirect speech is the current speaker's.

.Linda told Irvin that she had done her share; now he should do his duty.

This is why, in , she has the same referent as I in , and his has the same referent as your in .

Concerning now the syntactic integration of q into p, the indirect speech of is introduced in a that-clause which is the direct object of the matrix verb. The direct object relation is the same as in ; so a syntactic relation between the expressions coding p and q is not a sufficient condition for indirect speech. Neither is indirect speech necessarily an argument of a communication predicate; it may constitute a sentence or even a series of sentences of their own.

.Gesandte der Helvetier erschienen vor Cäsar, um sich zu beschweren. Römische Soldaten hätten sie grundlos angegriffen ...
GermanHelvetian envoys appeared in front of Caesar to complain: Roman soldiers had attacked them for no reason ...

The final clause of is syntactically independent; it does not depend on the speech act verb of the initial sentence. However, in contrast with the English translation, the final clause signals its status as indirect speech by the subjunctive.

Ad #b: The expression propositional entity in the introductory definition of ‘speech reproduction’ leaves the exact linguistic status of p and q vague. P is, at any rate, part of one speaker's discourse. However, the enunciative relation to q may be contracted by an entire utterance, as in ; it may be comprised by the meaning of a verb of communication in p, as in ; or the clause coding q may be governed by such a superordinate speech act verb, as in .

Likewise, the extension of q is not restricted. It consists of more than one sentence in and ; and in these examples, they even differ in their sentence type. Thus, the quoted items may even be entire utterances, except that they develop no illocutionary force in the quoting speech situation. All of this shows that the enunciative relation is not an intrinsic interpropositional relation as long as it is not integrated into a construction of the system; and this is why speech reproduction has been singled out here.

In , the expression coding the initial part of q is a finite complement clause. However, the enunciative configuration may be grammaticalized even further. In , q is a non-finite complement clause, more specifically, a participial.

.Legatihaecseadsuosrelaturosdixerunt ...
Latinenvoy(M):NOM.PL[ D1.N.ACC.PLRFL.ACCtoPOSS.3:M.ACC.PLreport:PCPL.FUT:M.ACC.PL ]say:PRF:IND.3.PL
The envoys said they would report this to their people ...(Caes. B.G. IV, 9)

By the gradual subjection to rules of the linguistic system described here, the enunciative relation becomes an intrinsic interpropositional relation, viz. the relation of a complement clause to its governing verb. As such, it is a topic of the grammar of the indirect declarative clause.