Relevance theory and hierarchical structure

church, the progression of personal salvation. However that may be, the sentence is a paragraph with different conceptual steps. In general, at the boundary between steps we would expect to find sentence boundaries as well, but here we do not. Under what conditions does this happen? This is an area for further study.

2.6.6 Relevance theory and hierarchical structure

“Relevance-based approaches are concerned with processes of discourse understanding rather than the structure of discourse” Blakemore 2001:113, that is, about “tracing the hearer’s route in the interpretation of the speaker’s intention” p. 12. What the theory has had to say about discourse has been mainly concerned with such on-line processing issues as “relevance relations” and “discourse markers” Blass 1990, Matsui 1993, Unger 1996, Blakemore 2001. In regard to coherence, the general conclusion is of the form “process not structure”: “intuitions about the coherence of texts result from the ways texts are processed in the mind, rather than reflecting actual cognitive entities” Unger 1996:420. The conclusion in regard to discourse organization is of the same form: “discourse organisation is best explained pragmatically as a by-product of processing texts following the relevance-theoretic comprehension procedure.” Unger 2001:135f.. Since relevance theory RT focuses on process, its treatment of discourse structure has ranged from frankly dismissive to something-we-don’t-talk-about. Blakemore 1988, 2001 mentions discourse structure only in relation to surface form, associating it with the code model of communication. She fails to discuss the possibility of structure in conceptual organization, upon which discourse topicality and thematicity are based in this and certain other treatments. It is common to find an explicit or implicit claim that, since relevance relations need to be recognized anyway, there is no need to recognize hierarchical discourse organization in addition Unger 1996:428, 2001:136; Blakemore 1988; Blass 1990:260. Relevance and analysis of discourse structure are presented as an eitheror choice: “discourse is best approached in terms of process rather than structure” Wilson 1998:70. The position of Unger is categorical: “there is no evidence for hierarchical organization in discourse. …the notion of hierarchical discourse structure was argued to be unwarranted” 1996:403, 426; “there is no evidence that discourse is hierarchically structured…discourse is not hierarchically structured” 2001:135, 181. 69 Nevertheless, there is in relevance theory an implicit recognition of hierarchical discourse structure. It is common to find a recognition of whole texts as linguistic units; the recognition of structural parts of texts is rarer, but also present. Sperber and Wilson 1995:216 mention that “the classic discourse topics are titles and picture captions, whose role is precisely to give access to encyclopedic information crucial to the comprehension of the accompanying texts and pictures.” 70 Gutt goes farther: in his analysis of Matthew 2, he states that “the structure of this chapter is rather straightforward. It consists of four narrative sections…” 1991:73. In his analysis of Matthew 2 he states that implicit contextual information is sometimes needed to rightly understand not only individual utterances, but also paragraphs and larger discourse units pp. 92f.. Unger 2001:9f. also recognizes that whole texts can have characteristic properties: “the interpretation of utterances within a text or discourse may depend not only on properties of the particular utterance, but also on properties of the type of text or discourse it occurs in …the understanding of texts depends in important ways on genre knowledge.” Hierarchical structure is shown in the possibility that “discourse types may be embedded” Unger 2001:129. Unger himself embeds a discourse type in quoting the parable of the persistent widow Luke 18:1–8; p. 139. In an extended discussion of Isaiah 5.1–7 he describes “the connectivity of this text,” treating it as a discourse unit. He goes on to speak of such a discourse unit as a “complex ostensive stimulus,” commenting that a 69 Unger’s arguments against hierarchical discourse structure Unger 1996, summarized in 2001:179–182 deal either with occasional indeterminacies and transition phenomena or with theory-internal assumptions. These deserve a detailed reply, but that would go beyond the scope of the present treatment. It is worth noting that Unger freely makes reference to the hierarchical discourse structure of his own text; see on Example text 44 in §3.6.5. 70 Similarly, Blakemore 2001:112 states that “consistency with the Principle of Relevance explains disambiguation in both isolated utterances and extended texts.” whole text can be “seen as a special case of a complex ostensive stimulus” p. 168. In these ways, Unger implictly utilizes the notion of hierarchical structure in discourse. The most striking thing about his claim that discourse is not hierarchically structured is that he makes that claim within two genres—the academic dissertation and the academic article—which have an very clear form of hierarchical structure. He himself refers to this, citing, for example, “the previous section” and “the new section” of his own composition Unger 1996:427f.. Be that as it may, as relevance theory deals more and more with textual issues, there appears to be a growing recognition that both whole texts and structural parts of texts have global properties. This admission is made in cautious ways, however, perhaps because since relevance theory focuses on sequential processes, it is not well set up to explain how text can have global properties or to deal with sequential processes in terms of discourse units, unless hierarchical structure is frankly acknowledged. I know of no reason why any approach that focuses on process should be dismissive of structure, or why the two strategies could not complement each other. Relevance theorists recognize grammatical structure on the utterance level, but in discourse find it hard to see anything but process. A focus on process can be just as reductionistic as a focus on structure: recognizing one does not diminish the need for recognizing the other. Until relevance theory begins to more clearly recognize the place of discourse structure, what it has to say about discourse organization must remain quite limited. In particular, this means a great loss for the study of discourse thematicity and topicality, since they are commonly realized by the construal of relevance—the “management of expectations of relevance” Unger 2001:2—over stretches involving different individual acts of communication, hence often over discourse units. As Unger 2001:178f., 182 explains, “the organisation of discourse is intricately linked to intuitions of ‘aboutness’, i.e. to intuitions of what the discourse segments and the discourse as a whole are about. …this ‘aboutness’ of discourse is a natural consequence of the cognitive and communicative principles of relevance.… Indeed, Sperber and Wilson’s 1995 account of topic and focus…can be seen as an explication of the pre-theoretic notion of ‘aboutness’ in terms of relevance. In essence, the topic provides access to contextual information which allows the addressee to compute the intended cognitive effects of the focus, i.e. realise what the focus is ‘about’. …intuitions of ‘aboutness’ underlie the notions of topic and focus, and also of discourse organisation.” Discourse topicality and thematicity should therefore be a prime area for the application of relevance on this point, see also Sperber and Wilson 1995:216. Relevance theory has asked why, given productive sequential processes of discourse interpretion, there is any need to additionally recognize hierarchical discourse structure. But the question could also be asked why discourse units should not be expected to occur, given not only the pervasive evidence that we find for it, but also the broad universal human tendency to conventionalize functions and realize them as formal units. In relevance theory terms, this kind of “structuralizing” makes the scope of themes, topics, speaker’s communicative intentions, and other phenomena more readily accessible to addressees. That is, structural hierarchy in discourse is predictable not only in terms of how language works on other levels, but also in terms of relevance theory itself. Suppose we were to think of discourse units as a means of directing addressees “to a particular set of assumptions” Blakemore 1988:248. • There is, for example, Hinds’ 1979:136 notion of paragraphs as “a unit of speech or writing that maintains a uniform orientation,” so that in a narrative paragraph, for instance, there would be continuity if not actual uniformity along the orientation dimensions of time, place, participant roles, and related events §§2.3.1, 2.6.3. • There is also the role of a discourse schema for a particular discourse unit, integrating its separate parts in a unified view of what is going on in the text or what the speaker is getting at §2.2.4. In relevance theory terms, the addressees’ emerging perception of the speaker’s current discourse schema is part of their “assumptions,” and the particular way that an utterance relates to its schema is an important part of the “relevance relation” between the utterance and its context. • There is moreover the “processing shift” effect that one finds at the boundaries of discourse units: “once a processing shift has occurred, information represented in the previous substructure becomes less available” Gernsbacher 1985:344. In relevance theory terms, at paragraph boundaries the text’s set of assumptions is reset: formal signals “save the addressee wasted effort by alerting him to the fact that a switch in contexts is about to take place” Wilson 1998:71. 71 For example, such boundaries tend to be relatively opaque to referring expressions §2.6.3; Unger 1996:425 cites Hofmann 1989 in this regard. • Finally, there is the notion of a discourse unit as a kind of “homogeneous context” by virtue of its being the textual span of a discourse topic. This has to do with the fact that the set of assumptions is not reset within a paragraph as it is at boundaries, but is only updated in a nondiscontinuous way. Within such a context Sperber and Wilson 1995:216 describe a special type of relevance that integrates concepts with the topic, linking them to “the conceptual address associated with that encyclopaedic entry” see discussion in §2.2.3. 72 Given properties such as these, discourse units could easily have an interpretation as bundles of assumptions that are used in text processing. The linguistic and conceptual signals of such units both reflect the speaker’s processes of production and cue the addressees’ processes of interpretation, helping them distinguish between “same bundle of assumptions vs. different bundle of assumptions.” In relevance theory terms, “if speakers wish to constrain the interpretation that hearers recover, then they must constrain their choice of context by making the necessary assumptions immediately accessible to them thus ensuring their selection under the principle of relevance” Blakemore 1988:248. The bundling of assumptions by means of discourse units is one way that this apparently happens. For this reason the occurrence of discourse units at various levels—that is, the existence of hierarchical discourse structure— could be predicted by relevance theory: “in many cases these structures are themselves determined by considerations of relevance” Blass 1990:77; “the reason for discourse structuring is the principle of relevance itself” Unger 1996:426. 73 Such remarks nourish a hope that relevance theory, in combination with a some conceptual view of discourse structure, would yield insightful analyses. For this to happen, from the point of view of the present treatment, two adjustments would be required in the way that relevance theory is currently practiced: • first, a recognition of qualitative as well as quantitative distinctions in relevance, since the relevance of a theme or topic to component parts of a discourse space involves intrinsic interest on the part of the speaker, hence attention management as well as knowledge management; 74 • second, a recognition of the conceptual structure of discourse to deal with what the present treatment treats in terms of schemas and with what Sperber and Wilson 1995:216 speak of as “homogeneous context.” Such adjustments would be enrichments to the theory, not revisions of any of its fundamental tenets. 71 In Unger’s 1996:431 framework, “paragraph breaks are not treated as breaks between units, but as transitions between utterances where some kind of radical change of context than usual occurs.” He cites several kinds of formal phenomena that commonly occur in such a place. One wonders what a discourse unit would have to look like to be recognized as such. 72 Blass 1990:78, echoed by Unger 1996:426, explicates the “homogeneous context” of a paragraph in terms of sequential conntectivity: “information made easily accessible by the interpretation of the first utterance is used in establishing the relevance of the second; information made easily accessible by the interpretation of the second utterance is used in establishing the relevance of the third, and so on indefinitely.” However, this explication would just as well fit a text with no paragraphs at all; it makes no reference to the fact that at paragraph boundaries the set of assumptions is reset and that within a paragraph it has at most only continuous updating. 73 Although relevance is involved in this kind of structuring, it should probably not try to take of all the credit for it. Chafe 1994:138, for example, suggests that one motivation for paragraph breaks relates to how many semiactive concepts can be held in memory at a single time; that is, assuming Chafe’s “semiactive concepts” include processing assumptions, memory limitations may need to be calculated into relevance processes. 74 Categories of attention management which are dealt with in the present treatment—activation, attention, interest—do not appear in the index of Sperber and Wilson 1995, nor is there bibliographic reference to recent work in such areas.

2.7 Easy comprehension practices, literary techniques, and practices of other kinds