I was wondering to what extent the syntactic distribution of so-called “datives of agent” and that of “ablatives of agent” is different. For example, besides appearing in verbal contexts (e.g., Proelium ab equitibus commissum est), ablatives of agent can also be found in non-verbal constructions like the following ones (the first example is an Ablative Absolute construction and the second one is a dominant participle construction):
a. Commisso ab equitibus proelio, ... (Caes. Civ. 1.41).
b. post civitatem a Lucio Bruto liberatam (Cic. Phil. 5,17)
My question is whether “datives of agent” can also be found in these non-verbal contexts. My intuition is that they can’t: i.e., in an example like His rebus deliberatis mihi,… the dative mihi cannot be understood as an "agent" but only as a beneficiary. In contrast, in a verbal context like Mihi hoc est deliberatum, the dative pronoun can indeed be understood as the one who has carried out the deliberation. Hence its traditional label of "dative of agent".
Something similar, I think, happens when dealing with non-verbal passive contexts that include a gerundive (NB: the following examples are not Ablative Absolute constructions but exemplify the same restriction above). Again my intuition, which could of course be wrong, is that only by-phrases with Ablative case are possible in these (non-verbal) contexts:
a. De mercenariis testibus a suis civitatibus notandis… (Cic. Ad Fam. 3, 11, 3)
b. De provinciis ab iis qui obtinerent retinendis… (Cic. Ad Fam. 12, 22).
Interestingly, contrary to what one could infer from reading Latin grammars where it is asserted that the dative is the agent in periphrastic passives with gerundives, a dative is not to be found in those non-verbal passive contexts containing a gerundive (e.g., cf. the well-formedness of the (b) example above with the ill-formedness of *De provinciis iis-dat qui obtinerent retinendis). Notice that these two examples are interesting since they involve a gerundive in a clearly passive context and the alleged "dative of agent" is impossible!
As is well-known, the "agent" in gerundive passives is often said to be expressed by a dative (e.g., Carthago delenda est nobis). Some exceptions to this rule are typically “explained (away)” in Latin grammars by saying that the ablative of agent is possible to differentiate complements (e.g., Quibus est a vobis consulendum (Cic. Man. 6)) or to maintain a syntactic parallelism (e.g., Nec, si a populo praeteritus est quem non oportuit, a iudicibus condemnandus est qui praeteritus non est (Cic. Plan. 3, 8)). However, as evidenced by the two previous examples from Cicero's letters, it seems that this traditional description is not accurate/complete enough. One could perhaps object that an alleged example like *De provinciis iis qui obtinerent retinendis is to be avoided for ambiguity/processing reasons: the dative iis could be interpreted as associated to the ablative provinciis, causing a "perceptual muddle". However, I think this simplistic functionalist explanation, which is, by the way, the one provided by Philip Baldi (1983: 21-22) in his work "Speech perception and grammatical rules in Latin" (in H. Pinkster (ed.) (1983). Latin linguistics and linguistic theory. Amsterdam: John Benjamins), is wrong and does not account for the structural prohibition I'm dealing with here: my intuition is that an example like De provincia iis qui obtinerent retinenda, where there is no such ambiguity, would also be ill-formed due to structural/syntactic reasons: i.e., so-called "datives of agent" cannot appear in non-verbal contexts.
So the relevant conclusion seems to be that, as far as their syntactic distribution is concerned, unlike ablatives of agent, so-called “datives of agent” always (?) require the syntactic presence of verbal forms: (i) typically, the verb esse in constructions related to so-called "resultative passive", possession and passive periphrastic: e.g., Mihi hoc deliberatum est (cf. Habeo hoc deliberatum) and Mihi currendum est (cf. less classical Habeo currendum) and (ii) less typically, other verbs: e.g., neque cernitur ulli (Verg. Aen 1, 440).
Is this conclusion correct?
EDIT (Jan. 8th, 2022)
One could think that the following examples do show that a "dative of agent" can appear in a non-verbal context. However, please note that these contexts can be claimed to involve an elliptical verbal form: cf. [esse] below. So I do not consider them as good counterexamples to the claim put forward above.
Consilii nostri, ne si eos quidem qui id secuti sunt non paeniteret, nobis paenitendum [esse] putarem (Cic. Fam. 9.5.2).
Quod tanti Tib. Gracchum fecisset, ut, quidquid ille vellet, sibi faciendum [esse] putaret. (Cic. Amic. 37).
See Pinkster (2015: 305, OLS, vol. 1) for similar examples: Caesar maturandum [esse] sibi existimavit (Caes. BG. 1.37.4) // Cum omnes censerent primo quoque tempore consulibus eundum [esse] ad bellum (Liv. 27.38.6).