I will discuss the syntactic derivation of parenthetical qualified adjunct phrases like the underscored example in (i):
(i) Mitt drank two bottles of gin last night, unfortunately rather quickly.
The adverbial complex consists of two adverbial phrases (AdvPs), one of which evaluates, or qualifies, the other. The adverbial complex forms a distinct intonational phrase, as is typical of parenthetical constituents. Within the complex, nuclear (or focal) stress falls on the head of the qualified AdvP (rather quíckly). The entire complex has the force of an independent secondary assertion, similar to that of the underscored paratactic clauses in (ii):
(ii) Mitt drank two bottles of gin last night;
unfortunately he drank them rather quickly.
unfortunately he did it/this/so rather quickly.
I will defend an ellipsis analysis of the adverbial complexes in (i), modeled on Jason Merchant’s account of sluicing and sentence fragment constructions. This involves a combination of extraction and TP ellipsis. Assuming a source structure resembling one of the paratactic clauses in (ii), the qualified AdvP (rather quickly) is extracted from the TP and moved to a position below the qualifying AdvP (unfortunately). The remnant TP is then elided, with the host clause providing the antecedent TP. Although I refer to this as TP ellipsis, the precise hierarchical level of the elided material is tricky to pin down, mainly because of more complex examples.
In (i) both AdvPs are ‘integrated’ within the complex adjunct, and occur in the same linear and hierarchical order that they would occur in as integrated AdvPs in main clauses. In (iii), the order of the two AdvPs within the complex is inverted:
(iii) The rebels have been defeated—decisively, perhaps.
In (iii), nuclear stress still falls on the qualified AdvP (decisively). I will discuss the derivation of the inverted order, which also occurs in main clauses. The chief candidates are (i) right-adjunction of the qualifying AdvP via initial merge; (ii)
movement (of one or the other AdvPs within the adverbial complex) or (iii) an additional instance of TP ellipsis within the elided TP.
Like simple adverbs (both integrated and parenthetical) and Slifting remnants, adverbial complexes (with or without internal inversion) can be ‘niched’ within the host clause.
(iv) Napoleon, probably deliberately, insulted his host.
The analysis of niching is also problematic. While not resolving this decisively, I will point out that niching turns out to depend on the placement of nuclear (or focal) stress on the preceding constituent in the host clause.
Adverbial complexes also provide evidence bearing on the familiar problem of identity in ellipsis structures. In (v) the ‘qualifying’ frequency adverb often gives rise to a quantificational variability effect; I argue that this implicates a definite source for the elided counterpart of the indefinite object DP in the host clause:
(v) Janet has performed over a hundred autopsies, often incompetently.
More complex adverbial complexes are also possible, with both inverted and un-inverted orders:
(vi) Mitt drank the whole bottle, I think probably again unintentionally.
(vii) Mitt drank the whole bottle, unintentionally, again, probably, I think.
The existence of complex adverbial complexes like (vi) provides further evidence supporting the ellipsis approach, but also complicates the problem of determining the identity of the elided constituent. The inverse ordering effect visible in (vi) vs. (vii) takes us back to the question of how the inverted order is derived. To derive the inverse order via leftward movement of the qualified AdvP, one would need a roll-up derivation of the sort advocated by Cinque in his account of integrated adverb order, and by Koopman and Szabolcsi in their account of Hungarian and Germanic verbal complexes. Our parenthetical adverbial complexes, however, seem to allow for far more ordering options than would be expected under such an approach, suggesting that these adverbial complexes may involve multiple applications of ellipsis.