Evaluative adverbs belong to the class of parenthetical adverbs, together with speech act adverbs (e.g. ‘honestly’), agentive adverbs (so-called ‘subject oriented’, e.g. ‘kindly’), and connectives (e.g. ‘therefore’). Parenthetical expressions are constituents whose semantic con­tribution does not get inserted in the main semantic content. Rather, they provide some sort of comment on either (part of) the content of the clause or the speech act as a whole. In the case of evaluative adverbs, this is shown clearly by examples such as (1) and (2). In (1a), where the adverb occurs inside the antecedent of a conditional, it is intuitively clear that the presence of the adverb does not modify what condition is expressed; compare (1b), where the quasi-synonymous evaluative adjective has a quite different effect. [8, p. 188]

1. ‘If, unfortunately, Paul goes and sees Marie, she will be furious.’
2. ‘If, it is unfortunate that Paul met Marie, it is tragic that he insulted her.’

This data suggests that evaluatives and other parentheticals have a dis­tinct illocutionary status that sets them apart from ordinary constituents.

Most existing descriptions of parenthetical expressions hypothesize a strict correlation between the special illocutionary status of parentheticals and a special prosodic status: parentheticals are supposed to receive a so-called ‘comma intonation’, which may be modelled by saying that they form a distinct prosodic phrase. Such a description might be cor­rect for some kinds of parenthetical expressions. For instance, it seems that appositive relative clauses are both illocutionarily and prosodically distinct from restrictive relative clauses.

(2) a. Did Paul hire the author Mary likes?

1. Did Paul hire the author, who Mary likes?

However this shows that the situation is different in the case of adverbs: for adverbs, illocutionary parenthetical status is a lex­ical property, that is independent of prosody. That is, most adverbs can receive either an integrated or a ‘comma’-type prosody, irrespective of their illocutionary status. This is exemplified in (3) with a sample of parenthet­ical (connective, evaluative, agentive) adverbs and in (4) with a sample of nonparentheticals (modal, frequency, manner).

(3) a. ‘Nevertheless, my brother came.’

1. ‘Fortunately, my brother came.’
2. ‘My brother kindly came.’

(4) a. ‘My brother probably came.’

1. ‘My brother came often.’
2. ‘My brother came quickly.’

This data shows that we need to make a strict distinction between the illocutionary status and the prosodic properties of an adverb.

There is a large consensus that evaluatives provide a commentary on content, rather than being part of the content of the sentence. This is challenged by Bach, who contrasts evaluatives (which he calls ‘assesives’) both with adverbs such as modals and with a number of expressions which he classifies as ‘utterance modifiers’ (usually called ‘speech act modifiers’). The reason is that, although they differ from modals in contributing a different proposition from the main one, they can be found in embedded sentences, as in (1), which prevents them from being considered utterance modifiers. However, Bach himself does not propose an explicit analysis. The first question is: what is the status of the commentary? and the second one: how can it be implemented in a formal grammar? As to their status, evaluatives have been considered either to constitute a speech act, independent of the main one , or to convey conventional implicatures.

The two speech act analysis cannot be maintained. It shows that the utterance of a parenthetical (an in particular, an evaluative) does not have full assertoric force. Evaluatives can express commitments of an agent distinct from the speaker, at least in reportive contexts.