Mapping the pragmatic field: in how many ways?

1 What are we looking at?

Let us begin with a cross-linguistic sample that illustrates the type of issues that any formal approach to the syntactization of the discourse has to consider. Example (2) is from Romanian and (3) from Sasson Arabic (Akkuş and Hill 2021).

Ross (1970) provides examples as in (1) as evidence that a non-lexical category of the nominal type must operate in the structure to bind the anaphoric myself in the root clause. This nominal category indicates a super-ordinate non-lexical I, i.e., speaker, higher than CP, which constrains the form of the anaphoric pronoun (i.e., first vs. third person). Accordingly, we infer that formal analyses must count with the representation of the speaker, which would provide a null pronominal pro binding the anaphor. Assuming that the representation of the speaker is systematic, the example in (2) must also count with a silent speaker in its structure. In addition, this example further displays a speaker oriented particle (i.e., vai) that describes the speaker’s mindset (i.e., emotional venting to no one in particular). The vocative phrase in (2) provides the referential information regarding the addressee, and an injunction (i.e., the particle hai) by which the speaker pushes the addressee into action. So a formal analysis must also count with an addressee feature within the same space as the speaker (outside the CP), since neither vocatives nor injunctive particles belong to the argument structure of the clause. Notably, the utterance in (2) may be either a direct address or You-centered self-talk. The outlook may become even more complex, as in (3), where both the speaker and the addressee have lexical implementations, and the referential content of these lexical items can be switched between speaker and addressee on the condition that the common ground knowledge involves an acceptable age and an endearment degree. Hence, some feature reflecting the common ground is also active in the derivation, being obligatory for the grammaticality of address formulae as in (3).

2 Representations

How do neo-performative and interaction-based approaches deal with these constructions? The formal representations for (1) and (2) are shown in (4) (formalization according to Miyagawa (2022: 90) and Hill and Miyagawa (2024: 8)) and (5).

(4)

In (4), the features [speaker] and [addressee] may probe for two goals: a nominal category that provides referential information for the speaker or the addressee (e.g., the vocative phrase), and/or a lexical item (phrasal or non-phrasal) that describes the speaker’s mindset (e.g., vai), or injunctions the speaker may impose on the addressee (e.g., hai), giving some indications of the relevant common ground.^[1] The Commitment Phrase, adopted from Krifka (2019, 2023), links the discourse participants to the proposition by connecting the speaker’s knowledge (i.e., [+/− evid]) and his/her commitment to the truth of the proposition (i.e., [+/− commit]) with the clause typing feature of C. This combination is responsible for switches between declarative and interrogative clauses or the generation of rhetorical questions.

Casting the same data in MW’s framework, we obtain the representation in (5) (formalization according to MW).

(5)

According to the discussion of ma and oh in MW’s examples (57) and (58), speaker oriented particles (equivalent to vai in (2)) belong to the RespP in (5), as they are considered reactive responses, (turn-taking is not applicable when there is no addressee involved). Vocatives and injunctions belong to the addressee grounding, whereas Ground_spk constrains the referential indexing (e.g., myself vs. himself in (1)) and verb selection in self-talk construction (e.g., +/− verbs of cognition). Considering MW's examples in (14), null pro is also presumed in the Spec positions of the Ground phrases.

From a comparative perspective, Table 1 states that the hierarchies are different insofar as the addressee phrase is higher than the speaker phrase for MW, which is the reverse of what appears in the so-called neo-performative analyses. However, looking at (4) and (5), this contrast loses strength: it appears that SpkP in (4) subsumes the feature of RespP_self of (5), whereas the feature of RespP_other in (5) is reinterpreted as a side effect of feature combination under CommitP in (4). That is, SpkP is higher than AddrP in (4), but it has the speaker’s mindset feature that is associated with RespP_self in (5).

In light of these observations, the contrast that warrants further research and discussion is not whether the addressee is higher than the speaker, or the preference in labels, but what exactly the speaker oriented particles like vai or ma stand for, and whether RespP_self and RespP_other accounting for the turn-taking, which is also relevant to clause typing, deserve a separate projection mapping a dedicated feature, as in (5), or whether the same effects can be derived from a conjunction of existing features within the hierarchy, as in (4).^[2] For a further comparison, Giorgi (2023) put forth another analysis that takes into consideration turn-taking as a formal feature, but she locates it at the bottom of the hierarchy (see her tests on It. ma ‘but’). As for the speaker oriented particles, they are generally analyzed as hierarchically higher than the addressee in all other independent approaches to the pragmatic field (Giorgi 2023 a.o.; Haddad 2020; Portner et al. 2019).

Incidentally, the Ground_addr > Ground_spk order proposed in (31) does not receive a clear justification. The reader understands that I-centered self-talk has a deficient pragmatic field featuring only Ground_spk, and that You-centered self-talk triggers the addition of Ground_addr, but it is not clear why they have the order in (31) since no tests on actual data are provided (e.g., linearization constraints) to support this hierarchy. However, the reader could infer that, in a bottom-up approach to syntactic derivations, hierarchical complexity is increased by adding to the existing structure (e.g., rather than by head splitting or having a top-down derivation). Be it as it may, this inference gets challenged when we come to the examples (57) and (58), which contain speaker oriented particles (i.e., ma equivalent to Rom. vai). These particles are justified in (55) through the possibility of a RespP_self > Ground_spk structure. If that is the case, when this structure is turned into a You-centered self-talk construction, don’t we expect, for consistency, to have the hierarchy by addition, as Ground_addr > RespP_self > Ground_spk? One solution would be to consider these speaker oriented particles merged directly in Ground_spk in I-centered self-talk, as descriptive of speaker’s mindset (along the lines in (4) above) rather than reactions to the discourse. The problem for the hierarchy is that, in You-centered self-talk data, vocative phrases may appear, as in (57b), and they are lower than ma and oh, which is contrary to the prediction in (55). Are these particles moving to RespP_other? Should we rather count with a RespP_other > Ground_addr > RespP_self > Ground_spk hierarchy? At least an explanation is expected for why it is possible to insert Ground_addrP in-between RespP and Ground_spkP (i.e., turning (55) into (5) above), but not in-between Ground_spkP and CP in (31).

For the conceptual aspect, the following statement is made in MW (p. 189f.): “Wiltschko’s interactional structure differs from the other two approaches in that it does not contain a dedicated (Speech) Act Phrase, the conceptual argument being that speech acts are constructed and hence cannot define grammatical categories.”

In this respect, as a general observation, there is no speech act feature in the neo-performative analyses. The proposal in SAP analyses is that some interjections, such as hai in (2) have verb-like properties and introduce pragmatic roles (speaker, addressee and theme for the CP). However, SAP entails a constructed speech act reading. This is stated in Haegeman (2014) (i.e., “the speech act layer [is] a syntactically encoded interface between the utterance and the discourse”) and in Miyagawa and Hill (2025: 591): “SpkP, AddrP and CommitP make up the syntactic representation of the speech act (SAP)”. So the reading arises compositionally, which is the case for all the SAP analyses (since Haegeman and Hill 2013).

Part of the confusion for the reader when it comes to the author’s comments on Table 1 stem from the fact that the term of comparison is limited to Speas and Tenny’s (2003) proposal on SAP structure. However, there were subsequent adaptations and improvements of that analysis, which are not considered in the discussion. In particular, Speas and Tenny’s attempt to account for clause typing by manipulating the SAP field has been abandoned immediately in the literature, due to challenging data (Gärtner and Steinbach 2006).

As a result, MW’s statement that neo-performative analyses assume the SAP structure as the locus of illocutionary force does not hold: the discussion around the analysis in (26) does not reflect the state of the art in neo-performative analyses, which are essentially cartographic and associate illocutionary force with the clause typing in ForceP, as explicitly stated in e.g., Coniglio and Zegrean (2012) following Rizzi (2004). The omission from this discussion of the recent versions of the SAP model also qualifies the following statement as inaccurate: “unlike the interaction-based framework, there is no layer of structure dedicated to regulating turn-taking [in neo-performative models]” (MW: 212). As explained for the structure in (4), the SAP analysis was amended by introducing a CommitP which regulates the switch between assertions, questions and commands. In fact, Hill and Miyagawa (2024) provide a detailed discussion of how turn-taking can be switched off and on in interrogatives.

To conclude this section, what MW considers a conceptual contrast between neo-performative and interaction-based approaches may in most part be reduced to terminological rather than substantive differences when the comparative discussion takes into consideration the recent developments of SAP approaches and of other new studies on the pragmatics-syntax interface. More collaborative (rather than competitive) research is needed in this area, irrespective of the formal model adopted, as consensus should be reached with respect to the interpretation of discourse particles and the set of formal features that trigger the merge of these particles in the derivation of clauses.

3 Extensions

One issue that was tangentially discussed in MW’s paper concerns the importance of vocative phrases for the understanding of the syntactic organization of the interactional field. The relevant aspect for this paper was the location of vocative phrases in the clause hierarchy. As argued for the example (32), the fact that vocatives are disallowed in I-centered self-talk brings evidence that an I-centered self-talk construction “does not license any linguistic phenomena that require the grammatical representation of an addressee (vocatives, imperatives, and addressee oriented discourse markers)” (MW: 205).

This observation can prove valuable for the future, when attempts are made to transfer the hierarchy in (5) as such to the nominal domain, to represent the internal structure of vocative phrases. In particular, MW’s observation that vocatives are incompatible with I-centered self-talk also entails that the left periphery of regular vocative phrases lacks a feature catering to the speaker, so the vocative does not qualify as a goal for the clausal probes available in an I-centered self-talk structure.

For more information, it has been shown (Bernstein 2023; Hill 2014, 2022; Stavrou 2014 a.o.) that vocative phrases are DPs with an extra-layer that maps features of conversational pragmatics, among which the feature relevant to the addressee is crucial for the conversion of a regular DP into an address (i.e., the 3rd person value of D becomes a 2nd person in vocatives). This is in line with the observation in MW regarding the affinity of vocatives with imperatives and other constructions that involve a Ground_addr/AddrP: the [addressee]/[2nd pers] feature of the vocative phrase qualifies as a goal for the clausal [Ground_addr]/Addr in (5) or (4), so vocative phrases merge directly in the relevant Spec position.^[3]

Considering this background, MW’s observation that constructions with I-centered self-talk ban vocatives indicates that there is no equivalent to the Ground_spkP in the internal structure of vocatives (no goal for probes available on the clausal spine). Hence, the eventual attempt to extend the hierarchy in (5) to the nominal domain should proceed with caution, as the type of formal features seems to be different for the two pragmatic fields (i.e., CP and DP).