SIL Electronic Book Reviews 2004-013

Argument and overview of the book

This book is an extended argument for the position that there is a properly linguistic level in discourse: the Discourse Mode. Smith intends to counter the tendency in discourse studies to treat discourse phenomena as pragmatic rather than linguistic. Smith claims that this pragmatic turn is an over-reaction to earlier unsuccessful linguistic approaches to discourse. If one takes the text passage rather than the whole discourse, paragraph, or episode as unit of analysis, then it is possible to observe linguistic regularities in adjacent sentences. Text passages, i.e. sequences of at least two sentences, belong to different discourse modes which are determined by the types of discourse entities—events, states, generalizations, abstractions—which are introduced in the passage, and the type of progression from one discourse entity to the next. This is called “semantic progression,” and can be temporal or atemporal in nature. Smith recognizes five discourse modes: Narrative, Description, Report, Information, and Argument. The modes are claimed to be linguistic units because they relate regular patterns of linguistic forms in texts to regular interpretations thereof. While there are aspects of discourse organization that are constituted by the structure of linguistic actions rather than linguistic structure—genre, episodes, etc.—there is an aspect of discourse organization which is properly linguistic: the organization of sentences into passages of different discourse modes. The analysis of the discourse modes is formalized in Discourse Representation Theory (DRT).

Besides making an argument for the existence of linguistic discourse modes and explicating their distinctive features, Smith also investigates the question as to how linguistic properties of passages can shed light on features which are shared by all discourse modes. These features are the expression of subjectivity (in the sense of “point of view” of participants), and the effects of “surface presentation” in the sense of topic-comment and focus-background modulations in discourse. In this way, the book extends beyond the justification and characterization of linguistic discourse modes to a general exploration of the question how far linguistic analysis of discourses can be taken within DRT. However, the author concludes that “surface presentation” cannot be formally expressed with the resources of DRT.

The eleven chapters of this book are organized into four parts. Part one, “Discourse structure,” introduces the approach pursued in this book, the main concepts to be introduced, and the theoretical framework of DRT. Part two, “Linguistic analysis of the discourse modes,” discusses in more detail the aspectual and spatio-temporal properties of the five discourse modes: Narrative, Description, Report, Information, and Argument. Part three, “Surface presentational factors,” discusses the expression of participant’s viewpoint and topic-comment and focus-background modulations in discourse passages of all modes with the techniques of Discourse Representation Theory developed in the previous parts of the book. Part four, “Discourse modes and their context,” summarizes the claims made in the book by presenting complex analyses of passages of natural texts belonging to different modes with the tools developed in earlier chapters, and by relating the concept of discourse mode to coherence-based theories of discourse. Rather than summarizing the content of each chapter—for which there is no space here—I summarize what is explained about the five discourse modes in various chapters of the book.

Narratives introduce events and states into the universe of discourse. The events in narratives are temporally related according to narrative time. Bounded events in narratives advance the narrative time, as well as temporal adverbials. Bounded events are typically expressed by event-clauses with perfective viewpoint (aspect). Construction rules update the Discourse Representation Structure (DRS) of non-initial clauses in a narrative passage; they can be informally described as:

If two adjacent clauses express bounded events and no deictics are associated with the event in the second clause, the reference time of the second is later than that of the preceding one.

If two adjacent clauses express states or unbounded events and no deictics are associated with the event in the second clause, the reference time stays the same.

Descriptions usually introduce states, ongoing events, or atelic events. Text progression is spatial rather than temporal: from one part of the scene to the next. Time in descriptions is static, it doesn’t move dynamically. The DRS construction rules for this mode can be summarized as follows:

The first sentence of a passage that expresses an unbounded or atelic event and has an adverb of location introduces an implicit adverbial in the semantic representation (DRS) which locates the event temporally.

The following sentence(s) of this passage have the same reference time as the preceding one(s) if they express unbounded or atelic entities.

Reports introduce events, states and occasionally general statives (a class consisting of generic sentences such as The lion has a bushy tail and generalizing sentences such as Mary speaks French; p. 73). The time of the situations expressed is evaluated against speech time, in contrast to narratives where the situations are temporally related relative to each other. Linguistic clues for temporal progression in reports are tensed verbs, modals and temporal adverbs.

The Information mode is atemporal in the sense that it introduces facts, propositions and generalizing statives into the universe of discourse; these entities are not located in time or space. The text progresses through a metaphorical space determined by the semantic domain of the text. The DRS rules for progression first identify a so-called “Primary Referent” to be able to track its metaphorical location. The “Primary Referent” is that referent which is the most “salient and significant” referent in a situation; typically, this is “the entity that moves or changes” in events, or “the entity whose location is maintained or asserted” in states (p. 124).

The Argument mode is another atemporal mode. The discourse entities introduced are mostly abstract entities such as states of affairs, facts and propositions, but also general statives. Argument passages progress metaphorically in “location”, i.e., points in the argument. Since arguments essentially make claims and claims can be made in all sorts of linguistic structures, the DRS construction rules for arguments are left relatively vague.

Questions of theory

This book covers an unusually wide spectrum of discourse phenomena within the framework of DRT. Smith makes a serious and successful effort to explain the issues informally in plain language before developing formal analyses. Also, the theoretical framework of DRT is admirably well explained in a way focused clearly on the issues addressed in this book and leaving aside unnecessary details. This makes it easy to follow.

However, the book raises some fundamental questions. There is no space here to go into them all, so I concentrate on the more important issues. On page 1, Smith writes:

I characterize the modes by their linguistic features, that is, grammatical forms with consistent interpretations. The linguistic features of the modes are covert categories in the sense of Whorf (1956). They are not overtly marked but they have characteristic patterns of distribution, and of interpretation. These properties are subtle, but they are demonstrably part of a person’s knowledge of language. The emphasis throughout this book is on grammatical rather than lexical features of discourse.

In other words, linguistic properties (features) are claimed to be grammatical patterns that receive consistent interpretations. But this cannot be right, as there are grammatical patterns that do not receive a consistent interpretation as in the case of the genitive construction in (1):

The genitive construction is without doubt a linguistic property. However, the form does not receive consistent interpretations when uttered: there are seemingly endless possible variations on the particular relation between the head noun and its complement.

On the other hand, there are sentences that when used in utterance receive consistent interpretations even in the absence of linguistic features that account for them, as in the case of so-called “conventional implicatures” such as the scalar implicatures in (2):

The pragmatic implication that Nigel has not more than five children is on no one’s account part of the linguistic pattern, along with other cases of scalar implicatures, yet these interpretations arise quite consistently, as emphasized for example by Levinson (2000).

These examples make it clear that linguistic properties cannot be isolated by pairing grammatical patterns with consistent interpretations. The reason is that linguistic form radically underdetermines the meaning conveyed in any utterance: pragmatic inference is needed to enrich and complement the logical form of utterances to understand the propositions conveyed (Carston 2002). Interpretations, therefore, always contain a pragmatic element. Linguistic properties are those that are computed in the linguistic module to output incomplete logical forms. To argue that a particular property of an utterance or text is a linguistic one, one needs to show that it can be automatically computed with the module’s inherent database—or in other words, that it falls within the domain of the linguistic module. In effect this means that it needs to be shown that the effects of the property in question are fit for input to pragmatic processes rather than explainable as effects of such processes.

Smith does not present any such arguments. Rather, throughout the book the methodology is to model the interpretations of sentences in text passages in terms of DRSs. This means that she in effect tries to capture the truth-conditional semantic content of pragmatically enriched interpretations (“the framework is flexible enough to allow for semantically and pragmatically based information,” p. 56), and with reference to Levinson (2000) she adds that DRSs—the level that gets a model-theoretic interpretation in DRT—typically include pragmatically enriched semantic information. In line with the above quote from page 1, the argument is that to the extent that those interpretations can be constructed by DRS construction rules the aspects of meaning concerned are of a linguistic nature. This is supported by the following statements:

The formal statements give content to my claims about the linguistic basis of intuitions about discourse mode and subjectivity. Interpretations that can be stated in construction rules differ significantly from those that cannot be stated within the theory, as we shall see. (p. 57)

Sentence topic and focus…are problematic for construction rules…. The principles for identifying the topic of a sentence are almost entirely pragmatic and are difficult to formalize. (p. 58)

Inferred perception is difficult to establish formally since it depends on inference and world knowledge. It is beyond the scope of compositional rules. Other kinds of perception can be recognized at the level of grammatical form. (p. 172)

But this line of argument depends upon an inadequate notion of linguistic properties. In consequence, Smith’s argument that discourse modes are linguistic units is unconvincing.

Be that as it may, and let discourse modes be either linguistic units or pragmatically inferred mental structures. In any case, these units must play a causal role at some stage in the comprehension process if these notions are to be of theoretical interest (cf. Unger 2001). It is not enough to point out that people can categorize passages into these modes in order to claim their psychological reality – one needs to find effects on the comprehension process that are caused by them. Smith’s characterization of the modes contains two features: a typical type of situation entity, and a typical kind of temporal, spatial or metaphoric progression. However, the DRS rules for the modes take as input situation entities of the present and previous sentence or clause, compare them and give as result an instruction for temporal, spatial or metaphoric update. Mode recognition does not enter into the rules—only situation type recognition does. No other claim about causal relations between mode recognition and comprehension are made. Indeed it seems that no such claim can be made because of the way the notion of discourse passage is defined: “I use the term “passage” for text segments that realize a discourse mode. Passages must be long enough to establish the linguistic features that determine a mode. Two sentences suffice to do this” (p. 22). This means that a comparison of the interpretations of at least two adjacent sentences and their temporal or spatial relation is required as a precondition to see whether a certain discourse mode is realized. But then discourse modes cannot be claimed to have a causal role in the interpretation of the sentences in question.

The problem is further compounded by the fact that Smith defines the phenomenon of backgrounding in the following way: “Information is backgrounded if it fails to contribute to progression” (p. 34). This makes it in principle impossible to decide whether a given sequence of sentences realizes one passage containing backgrounding or whether it consists of two or more passages of different modes. Comparing Smith’s examples (1) of chapter two (p. 22) and (20) of chapter two (p. 36) illustrates this problem forcefully. Smith herself comments on this problem in the following way: “The examples show that with changes of entity there is a change from foreground to background in passages of a given mode. However, we tend to take one or two clauses merely as a shift from foreground to background” (p. 36). In other words, there is no non-arbitrary way to delimit discourse passages. It follows that discourse modes, which are realized in discourse passages, cannot be non-arbitrarily recognized either and hence cannot have a causal role in comprehension.

Questions of description

A strength of Smith’s book is that the approach is based on the investigation of passages from a corpus of natural texts (some of which are reproduced in an appendix). Therefore it should be possible to test the descriptive adequacy of her approach with natural texts. However, it seems that such a test reveals weaknesses in the descriptive validity of her approach.

Consider the distinction between description and narrative: on Smith’s account, the main difference is that these modes differ in terms of temporal progression. While the sentences in narratives are temporally related against each other (establishing a narrative time line through the relations between adjacent unbounded events), in descriptions each sentence is related deictically to speech time.

Now consider (3):

Is this text a description or a narration? It is clear that the sentences recount bounded events in temporal sequence, and the reader of this passage will understand it in this way. However, the linguistic forms employed (there were ...) suggest entities typical for description. There does not seem to be a clear-cut relation between these linguistic forms and the temporal relation of the sentences in this passage. Certainly Smith’s rules will give the wrong results for this example. The relevant construction rules are as follows (for informal paraphrases of these rules, see above):

In a clause n with temporal entities t_n: the situation entity in n is unbounded or atelic, of type e:

S[_Adv [Loc] + e + tense] → t_1n, t_2n, t_3n; A, B; adv, mt

In the context of temporal entities t_n of the DRS, if the situation entities associated with t_n-2 and/or t_n-1 are unbounded or atelic, and locating and temporal conditions obtain:

If the situation entities of n are unbounded and atelic entities of type e.

RT_x immediately preceding: RT_y = Rt_x

With RT_x immediately preceding; if e is a bounded event: RT_y > Rt_x

Applied to the first three clauses of example (3), none of the conditions of these rules are fulfilled, so no temporal advancement can be established with them. The first clause does not contain an adverb of location. The third clause (their leaves opened like a whole spring unfolding in a moment) does express a bounded event, but not the previous one. Thus, neither the rules for description or narration apply. Yet the situations expressed in these clauses intuitively follow each other in time, just as in narrative. Thus, Smith’s account of discourse modes (in particular that of temporal progression within the modes) is not able to account for (3). The rules stated are not general enough to explain the data.

Conclusion

This book applies the formal semantic theory of DRT to a wide variety of phenomena of discourse, all culminating in the discussion of discourse modes. The formal theory gives substance and structure to the approach and analyses, but avoids getting lost in highly formal theoretic detail. In this way it should be accessible to a wide audience, including non-specialists in formal semantics. It includes a very well-written introduction to DRT concentrating on those aspects that are important for the concerns of this book. The phenomena to be analyzed are always first discussed in an informal way so that the issues are explained well before they are analyzed in the formal theory. The breadth of issues discussed, the extensive review of the literature on them and the rich list of bibliographic references make the book a useful resource for students of discourse, including linguistic fieldworkers studying discourses in the lesser-studied languages of the world. However, it does not make a convincing case for its central claims, lacking both in theoretical and descriptive adequacy.