VIII

 

Conversation and Narration

 

1. CONVERSATION

 

1.1 According to Peter Hartmann (1970: 91), every text essentially functions as a contribution to a dialogue (see also Coulthard 1977: 100). For various text types, the dialogue between producer and receiver is carried out with a greater or lesser MEDIATION in regard to SITUATIONALITY (I.4.11.5; cf. Beaugrande & Dressler 1970: Ch. VIII). In conversation,. mediation is not extensive, due to mutual awareness of participants, usually (except via telephone or the like) supported by physical presence. The immediacy of the communicative situation leads to heavy reliance on INTERTEXTUALITY (1.4.11.6; Beaugrande & Dressler 1978: Ch. IX), the principle whereby the textuality of any one text arises from interaction with other texts. What is cohesive, coherent, and acceptable in conversation may be quite different from what meets those standards in other modes of communication.

1.2 In VI.4.2 I noted the dual status of texts in discourse as both action and meta-action, i.e. verbal monitoring of actions and situations. These two outlooks can lead to different research methods for the study of conversation:

1.2.1 The action-oriented perspective began with the behaviorist definition of conversation as a pairing of stimulus and response (Ruesch 1957: 189). This narrow approach was replaced by the investigation of TURN-TAKING, in which a discourse action and reaction are seen as constituents of a “speech exchange system” (Sacks, Schegloff, & Jefferson 1974: 696). Here, sociologists have undertaken to define the ways people select or delegate speaking turns in a conversation. Most recently, conversational actions have been probed from the standpoint of how people plan to attain goals (Winograd 1977a; Allen & Perrault 1978; Cohen 1978; McCalla 1978a, 1978b; Allen 1979; cf. VI.A).

1.2.2 The meta-action-oriented perspective is obliged to deal with content and topic, issues pursued at first with hesitation (Sacks 1968, cited in Coulthard 1977: 75) in absence of general methods to deal with meaning. Little can be gained by the common linguistic procedure of assigning structural descriptions to abstract sentences. However, advances have been forthcoming from disciplines besides linguistics: sociology (Sacks, Schlegloff), discourse analysis (Sinclair, Coulthard),1  [1. Many early the papers on conversational analysis were circulated as mimeographs and were difficult to obtain. The situation worsened after the sudden death of Harvey Sacks in 1975, whose collected papers were published only many years later.]

  and artificial intelligence (Grosz 1977; Schank 1977; Lehnert 1978),” so that topic and content are gradually becoming more thoroughly explored in this domain.

1.3 Earlier work on conversation understandably preferred to address relatively restricted domains. The study of comparatively stabilized communicative situations, such as rituals (Salmond 1974), verbal duels (Dundes, Leach, & Ozkõk 1972; Labov 1972a, 1972c), litigation (Frake 1972; Leodolter 1975), and chanting (Scherzer 1974), is concemed with a limited range of conventional topics and actions. The study of  “registers” (characteristic styles of text production in typical situations or among certain groups, e.g. dialect options, (cf. Blom & Gumperz 1972; Ervin-Tripp 1972), focuses more on the variations within the virtual systems of sound and grammar/syntax than on topic or action.

1.4 Harvey Sacks (cited in Coulthard 1977: 75) states as a “general rule about convetsation that it is your business not to tell people what you can suppose they know.” It might be more accurate to state that a great deal of conversational material is indeed already known to all participants, but that the particular configuration of text-world models in discourse is not known as such or in the conscious mind, due to new combinations, limitations, modifications, or directions (see IV.3.14). We could say that the DISCOURSE MODEL which conversational participants co-operate in building (cf. Reichman 1978; Rubin 1978b; Webber 1978) often fails to provide a complete or exact match with the stored knowledge of those who enter the receiver role. This partial match is especially attributable to the enormous diversity of sources for conversational materials, such as the following:

1.4.1 Typical and determinate concepts and relations in world knowledge can safely be taken to be accessible to conversational participants at large, e.g. that the sky is blue, water boils and freezes, humans live in houses, and so on. Such knowledge is therefore too trite for conversing even with strangers.

1.4.2 Cultural and social attitudes, such as conventions of politeness or standards of desirability and value (cf. VI.A.10), can be presupposed to apply in most situations, unless signals are given to the contrary.

1.4.3 Conventional scripts and goals serve to alert conversationalists to what people are expected to say, and why, in familiar situations. In the enactment of scripts, e.g., going to a restaurant, participants need not provide explanation of why they are speaking, as long as they are conforming to expectations.

1.4.4 Apperceivable traits of the current situation are presumably known to all participants present or can be pointed out with little difficulty. The countless conversations about weather fall under this heading, aided by the social convention that weather is a universally acceptable topic.

1.4.5 Episodic knowledge of shared experiences among participants applies when some past situation involved the presence of people in the current conversation. This knowledge contributes to the speaker’s internal model of the hearer (cf. I.6.1), and extends the store of shared situations  (cf. Clark & Marshall 1978). The speaker may assign as defaults and preferences the components of his or her own self-model to the hearer-model (Cohen 1978: 93; cf. VIII.1.14).

1.5 Conversation differs from other text types especially in its greater reliance on SITUATIONALITY (VIII.1.1 ), whether current or shared in the past. This factor allows rich UPDATING of expectations and steady FEEDBACK about the effects of utterances (cf. Rubin 1978b). The participants’ plans and goals will be more directly evident or firmly established in prior knowledge (cf. the notion of “life themes” in Schank & Abelson 1977).

1.6 This immediate situationality lends conversation an enormous range and flexibility. Paul Grice (1975, 1978) has undertaken to systemize conversation somewhat by formulating some “conversational maxims” with the status of preferences or defaults. In comparing his maxims to my own model, I noticed what seemed to be some unclearness and overlap in his definitions. I accordingly asked Grice himself about the disputed points, and my discussion below incorporates the explanations he kindly afforded (for a more detailed treatment, cf. Beaugrande & Dressler 1980: Ch. VI).

1.6.1 The principle of CO-OPERATION is cited as: “Make your conversational contribution such as is required, at the state at which it occurs, by the accepted purpose or direction of the talk exchange in which you are engaged” (Grice 1975: 45; cf. Clark & C1ark 1977: 122ff.). My own criteria of INTENTIONALITY and ACCEPTABILITY (I.4.11.3f.) seem applicable here in regard to participant attitudes, and that of SITUATIONALITY in regard to “direction and purpose.” I cited some examples of deliberately unco-operative utterances in II.1.8 and IV.3.7.

1.6.2 The principle of QUANTITY is cited as: “Make your contribution as informative as (but not more informative than) is required” (Grice 1975: 45). This principle concerns the amount of content presented, and seems related to my notion of RELEVANCE to communicative plans (cf. I.4.14; VII.2.8). I would not equate the principle with my own notion of informativity as expounded in Chapter IV, because I am concerned more with knownness and expectedness than with volume. A text which is “more informative than is required” would, in my model, be too discontinuous or discrepant (cf. IV.l.12); for Grice, such a text could be merely too extensive.

1.6.3 The principle of QUALITY is concerned with truthfulness: “Do not say what you believe to be false, or that for which you lack adequate evidence” (Grice 1975: 46). The expectation that a textual world ought to match at least the determinate elements and configurations of a corresponding world ­knowledge pattern is, I suspect, stronger for text types other than conversation (e.g. science texts, cf. VII.1.8.6). Indeed, conversation often demands false assertions for the sake of social conventions, e.g., insincere praise of other people’s appearance or possessions. Also, the pursuit of a plan may require false assertions for motives such as we saw in the stage play in VI.4. Still, these usages are probably parasitic on a principle such as Grice has proposed, or else they would not be effective.

1.6.4 The principle of RELATION is cited simply as “be relevant” (Grice 1975: 46). Grice’s notion subsumes at least some factors of my notions of relevance (oriented toward a plan or goal) and knowledge access (what kind of contents are in principle related to each other). One could devise cases where these two notions of mine come into conflict, for example, when someone’s plan calls for a sudden change of content to a topic not accessible from the previous one (e.g. Mrs. Haggett’s attempt to stall in (86) of sample (188); or the mother’s change of song in (15) of sample (247). But these cases could plausibly be seen as violations of normal conversation.

1.6.5 The principle of MANNER is rather diverse: (1) “be perspicuous”; (2) “avoid obscurity of expression “; (3) “avoid ambiguity”; (4) “be brief”; and (5) “be orderly” (Grice 1975: 46). In a recent presentation at the University of Bielefeld Symposium on Theories of Language Use (June 1979), Grice also proposed to state this maxim as “Frame whatever you say such as to be appropriate to a reply.” These notions can be clarified as follows (I rely here again on Grice’s explanations). Be perspicuous would yield this maxim: “Be such that the intentions you have for what you say are plainly served.” Avoid obscurity could be stated as: “Do not be difficult to understand.” Avoid ambiguity can be rephrased as: “Do not express yourself such that your audience will take meanings other than what you intend.” Be brief could be: “Do not use more time than necessary to make your contribution”; hence, “quantity” concerns how much you say, and “brevity” concerns how much you take to say it. Be orderly could be restated as: “Present your materials in the order in which they are required.” Grice’s illustration of appropriateness to a rep/y was the use of formats suitable for a denial.

1.6.6 Grice also has introduced the notion of IMPLICATURE (not a maxim) for utterances whose intended utilization is recoverable from their conceptual content only via social conventions (cf. also McCawley 1978; Sadock 1978). A well-known illustration is the utterance:

 

      (208) Can you pass the salt?

 

where a request is presented like an inquiry about someone’s abilities. The need for implicature arises from the ASYMMETRY between the connectivities of concepts/relations and those of planning. My notion of RELEVANCE might account for how implicatures are created and recovered.

1.7 The principles suggested by Grice and others exert powerful controls on expectations, defaults, and preferences in conversation. Their violation may elicit regulative utterances such as:  

 

(209) So what? [violation: co-operation]

(210) Big deal! [violation: quantity]

(211) Why are you telling me this? [violation: perspicuity]

(212) I don’t know what you’re talking about! [violation: obscurity]  

 

These signals cannot be used freely on participants with pronounced social dominance over the speaker. Also, some social situations require people to converse in absence of materials needed to live up to Grice’s principles. In many discussions of the weather, (209) through (212) would not be usabIe responses.

1.8 We can distinguish between the DISCOURSE ACTIONS of INVOKING: calling up material presumed to be known to participants; and INFORMING: modifying known material or presenting new material (cf. VI.4.14). The distinction is one of degree rather than opposition, and exists not in the material as such, but in the participants’ outlook on material and on each other. Invoking is a good means for maintaining the desirable social states of acceptance and solidarity (VI.4.10), and thus accounts for many conversations of low informativity. Invoking can serve for exploring the attitudes of certain other people whose co-operation is needed for the speaker’s plans.

1.9 I have proposed to define TOPIC with reference to the density of conceptual-relational configurations in text-world models (e.g. in III.4.27). A single utterance in conversation might not have its own topic, but might rather present material which would become topical if developed in follow-up utterances (Schank 1977: 424). Hence, topic is a dynamic aspect of the flow and shift of knowledge drawn from the various sources enumerated in VIII.1.4ff. Topic shifts/are especialIy pronounced among participants with rich knowledge stores regarding each other’s personal histories. An illustration would be this dialogue I overheard on the University of Florida campus: .

 

(213.1) Hey, what’s happening?

(213.2) Keeping busy. You going to the game Saturday?

(213.3) If I get the physics paper done. Your brother back yet?

(213.4) Sometime next week. Got hung up somehow.

(213.5) Sounds just like him.

 

Without a context of shared experience, these utterances would hardly occur together. Conversely, unco-operative speakers can deflect interaction by deliberately discontinuous shifts, as in this recent campus exchange:

 

(214.1) BIBLE EVANGELIST: It’s a fearful thing to meet with God the  King!

(214.2) STUDENT: Like when Godzilla meets King Kong?

 

The student used superficial similarities among expressions to move the topic from religion to a monster movie made some years ago.

1.10 If one participant has the initiative, others can restrict their contributions to simple feedback-”commentation” in the sense of Roland Posner (1972) (cf. IV.3.8). Consider this exchange between Sam Weller and another servant in the Pickwick Papers (Dickens 1836-37: 547):2 [2. For the sake of illustration, I omit the cues such as ‘said so-and-so; in these passages, unless—as in (237)—they are relevant to the discussion. I have normalized a few dialect spellings for the ease of non-English readers.]

 

(215.1) I’m afraid I’ve been dissipating.

(215.2) That’s a very bad complaint that.

(215.3) And yet the temptation, you see, Mr. Weller.

(215.4) Ah, to be sure.

(215.5) Plunged into the very vortex of society, you know, Mr. Weller.

(215.6) Dreadful indeed.

 

Sam’s (ironic) contributions display his solidarity without affecting topic flow. This type of response can extend back further than to a previous utterance, though clarification may be needed (Dickens 1836-37: 552):

 

(216.1) What a lucky fellow you are!

(216.2) How do you mean?

(216.3) That there young lady. She knows what’s what, she does. [Mr. Weller closed one eye, and shook his head from side to side.]

(216.4) I’m afraid you’re a cunning fellow, Mr. Weller.

 

Sam’s first remark looks back to a topic raised some time before — the youngest daughter of the footman’s master “leaning heavily on his shoulder’ — so that its motivation is not at once obvious. The final remark (216.4) of the footman takes Sam’s feedback as grounds for inferring his mental abilities.

       1.11 It should be possible to state some strategies for generating feedback. Like many other issues, topic flow depends on how knowledge is acquired, stored, and utilized (cf. III.3.7). First, participants should know what elements of knowledge are connected to each other, for example, by appealing to global patterns like frames, schemas, plans, and scripts. Second, participants should distinguish what elements are INTERESTING because they involve possible PROBLEMS, that is, uncertainty of access or variability of node content in either real-world events and situations, or internal knowledge stores, e.g., how something is to be obtained or achieved, or how improbable or infrequent something is. By focusing on such problems, one can respond to many conversational contributions with follow-up questions via the LINK TYPES proposed in III.4.7ff.:

 

(217) Why did you do that? [reason-of]

(218) What happened then? [proximity in time-to]

(219) How did you manage that? [enablement-of, instrument-of]

(220) Why did you do that? [purpose-of]

(221) What brought that on? [reason-of]

(222) When did that happen? [time-of]

(223) Where did that happen? [location-of]

(224) What’s it made of? [substance-of]

(225) How did you find out? [apperception-of]

(226) How did you think that up? [cognition-of]

(227) Where did you get it from? [entry-into-possession-of]  

 

1.12 To use these questions, participants would search the applicable knowledge pattern for problematic entries (cf. examples in Schank 1977). The acceptability of a question rises with the uncertainty of the elements in its focus, and with their relevance to an event, action, object, or situation involved. A statement like:

 

(228) I fell in love last night.

 

could sensibly be responded to with (218), (221), (222), or (223), but hardly with (217), (219), (220), and (224) would be less likely: falling in love is popularly supposed not to be guided by reason or purpose, nor to have substance. A good test for problematic and interesting aspects of some assertion is to try the effects of these various follow-up questions upon it. Put into a quasi­-Gricean maxim, but with my terms, we obtain: “Select an active node of the discourse world and pursue from it a pathway whose linkage or goal node is problematic or variable.”

1.13 The relevance of such linkages directs the course of conversation in many ways. Here Mr. Pickwick is hearing one of Sam WelIer’ piquant stories (Dickens 1836-37: 651):

 

(229.1) ‘Next morning he gets up, has a fire lit, orders in three shillings worth of crumpets, toasts them all, eats them all, and blows his brains out’.

(229.2) ‘What did he do that for?, inquired Mr. Pickwick abruptly; for he was considerably startled by this tragical termination of the narrative.

 

Although the pro-form ‘that’ is itself non-determinate, there is no doubt that Mr. Pickwick’s follow-up question (229.2) is directed only toward the final action of ‘blowing one’s brains out’, normally requiring an unusually powerful “reason-of’ linkage. People ‘get up’, ‘light fires’, ‘eat crumpets’, etc. in the normal course of life.

 1.14 Actions of some participant that appear unique or without reason are likely topics for the conversation (Dickens 1836-37: 651):  

 

 (230.1) Will you allow me to inquire why you make up your bed under  that there deal table?

 (230.2) Cause I was always used to a four-poster afore I came here, and I find the legs of the table answer as well.  

 

A speaker can encourage conversation by withholding some problematic knowledge and allowing others to make guesses. A cobbler Sam meets in debtors’ prison turns out to have a very informative reason for his arrest (Dickens 1836-37: 653) (231.1-5) This reason is so non-expected that a regulatory interchange concerning believability ensues (231.6-7):

 

(231.1) What do you suppose ruined me now?

(231.2) Why, I suppose the beginning was that you got into debt.

(231.3) Never owed a farthing. Try again.

(231.4) You didn’t go to law, I hope?

(231.5) Never in my life. The fact is, I was ruined by having money left me.

(231.6) Come, come, that won’t do.

(231.7) Oh I daresay you don’t believe it. I wouldn’t if I was you; but it’s true all the same.  

 

The final remark (231.7) clearly signals how participants project their own knowledge and beliefs onto others in conversation (VIII.1.4.5).

1.15 Topic flow can also move along links of class inclusion. The conversation may be directed toward a class or superclass from which an instance or subclass has been mentioned. In the following exchange between Sam WelIer and his father, the discussion of the situation drifts onto a SUPERTOPIC (cf. Schank 1977) about a prototype of the class of ‘prophets’ (Dickens 1836-37: 641) (232.1-2) :3 . To the extent that the elder Mr. Weller is himseIf included under the heading ‘prophet’ by METACLASS INCLUSION (cf. IlI.3.20), we might have here an example of a METATOPIC (cf. Schank 1977).

 

(232.1) WelI now, you ‘ve been a-prophesying away very fine, like a red­faced Nixon as the sixpenny books gives pictures on.

(232.2) Who was he, Sammy?

(232.3) This here gentleman was a prophet.

(232.4) What’s a prophet?

(232.5) Why, a man as tell what’s a-going to happen.

(232.6) I wish I’d known him, Sammy. Perhaps he might have throwed a small light on that here liver complaint as we was a-speaking of just now. Howsoever, if he’s dead and ain’t left the business to nobody, there’s an end on it. Go on, Sammy.

 

The elder Mr. Weller steers the topic back to an earlier one (the ‘liver complaint’ of his wife’s pastor) and then signals, with ‘go on’, a return to the previous point (232.1), where the digression occurred  (232.6).

1.16 The flow of topic can also be guided by traits or objects of the current situation (VIII.l.4.4). When Mr. Pickwick is driven by the elder Weller to ‘the turnpike at Mile-End’, this topic shifts to the ‘pike-keeper’ (i.e. toll taker) (Dickens 1836-37: 318):

 

(233.1) Very queer life is a pike-keeper’s, sir.

(233.2) Yes; very curious life--very uncomfortable.

(233.3) There alI on ‘em men as has met with some disappointment in life.

(233.4) Ay, ay?

(233.5) Yes. Consequence of which they retires from the world and shuts themselves up in pikes; partly with the view of being solitary, and partly to revenge themselves on mankind by taking tolls.

 

1.17 These examples illustrate conversations whose components are less goal-directed than those we looked at in the stage play in VI.4. They are typical of situations where people are motivated by social factors to uphold a continuity of communication. That continuity requires a corresponding continuity of topic, but shifts may pursue a wide variety of links. Situational settings and participants’ episodic apperception and know ledge can readily be used, since they rest upon experiential continuity.

1.18 The question of how SPEAKING TURNS are allotted is correlated both with topic flow and with participant roles. Sacks, Schegloff, and Jefferson (1974) distinguish two groups of conventions: (1) the current speaker selects the next one, for instance, by direct address; and (2) the next speaker “self-selects” by beginning to speak at an available utterance boundary. Conversations contain remarkably few long silences, and little overlap among utterances. Participants must have powerful and efficient strategies for introducing their contributions at the opportune instant. Sacks et al. (1974: 709) note the heuristic role of the sentence as an indicator of completing utterances (cf. III.4.26), though sentences provide options for continuation in many cases (cf. Coulthard 1977: 59). To retain the turn and discourage interruptions, people can insert cues of incompleteness, such as ‘however’, ‘and then too’, so that the intention to continue is evident (cf. Sacks’ notion of “utterance incompletor” cited in Coulthard 1977: 57). The speaker can also announce a forthcoming series of contributions, e.g. these utterances at a (yawn) recent university meeting:

 

(234) I’d like to say three things about that.

(235) There are several points we’re overlooking here.

 

The speaker hopes that no reply will come except at most an encouragement to continue. Anything else would be considered as clear an interruption as breaking into the midst of a half-uttered sentence.

1.19 The assignment of turns can also depend on the distribution of knowledge among participants. Someone reputed to be very knowledgeable on a topic has a right to be heard if the topic comes up in conversation. Such is the case with Mr.Pickwick’s remark made ‘hoping to start a subject which all the company could take a part in discussing’ (Dickens 1836-37: 293):

 

(236.1) Curious little nooks in a great place like London these old inns are.

(236.2) By Jove, you have hit upon something that one of us at least would talk upon forever. You’ll draw old Jack Bamber out.

 

And Jack Bamber was indeed ‘the figure that now started forward and burst into an animated torrent of words’. Similarly, people encourage conversation by searching memory for a special topic that a given participant, especially a socially important personage, should know about (Dickens 1836-37: 530):

 

(237.1) “Have you seen his lordship’s mail cart, Bantam?” inquired the Honourable Mr. Crushton after a short pause, during which […] Mr. Crushton had been reflecting upon what subject his lordship could talk about best.

(237.2) “Dear me, no,” replied the M.C .; “a mail-cart! What an excellent  ideal Remarkable!”

(237.3) “Gwacious heavens!” said his lordship, “I though evewebody had seen the new mail cart; it’s the neatest, pwettiest, gwacefulIest tbing that ever wan upon wheels. Painted wed, with cweam piebald. [etc.]”4 [4. His lordship has an infantile speech habit of replacing [r] with [w].]

 

This ‘gwacious’ conversationalist was ‘the richest young man in Bath’.

1.20 Prior knowledge among participants also plays a part in QUESTION ANSWERING. Truthful answers that fail to account for the questioner’s  purpose may not be appropriate, as in Lehnert’s (1978: 5) samples:

 

(238.1) Do you drink?

(238.2) Of course. AlI humans drink.

(239.1) Who wasn’t in class today?

(239.2) George Washington and Moby Dick.

(240.1) Would you like to dance?

(240.2) Sure. You know anyone who wants to?

 

The questioner of (238.1) and (240.1) presumably wants to offer drinking and dancing to the addressee, and the answer violates the principle of co­operation (VIII. 1.6. 1). The questioner of (239.1) doubtless wishes to learn the identity of those members of the class who should have been present, but weren’t, so that the answer violates the principle of quantity (VIII.I.6.2).

 1.21 Shared procedures for maintaining coherence allow considerable economy in exchanges like this (Ortony 1978b):

 

 (241.1) Would like a piece of cake?

 (241.2) I’m on a diet.

 

Ortony argues that the coherence of the answer rests on an underlying chain of inferences such as the following:

 

(241.2a) People on diets ought not to eat fattening things.

(241.2b) Cake is fattening.

(241.2c) I ought not to eat any cake.

(241.2d) I will not eat any cake.

 

He points out (1978b: 76) that any of these steps in the chain of reasoning could also be mapped onto a surface utterance instead of (241.2). These steps must all be available anyway for the discourse action of (241.2) to take effect.

 1.22 Participants may be under a social obligation to correct inferable prior knowledge. In an exchange like the following (cf. Kaplan 1978: 204):

 

 (242.1) Which students got a grade of F in CIS 500 in Spring 1977?

 (242.2a) None.

 (242.2b) CIS 500 was not given in Spring 1977.

 

the response (242.2a) is literally true if the computer science course was not given, but it is misleading, whereas (242.2b) is helpful in correcting a wrong presupposition. Answerers might, of course, have reasons for encouraging wrong presuppositions, as in the soldiers’ reply (Carrol 1860: 110):

 

(243.1) “You shan’t be beheaded!” said Alice, and she put them (the gardeners] into a large flower-pot that stood near. The three soldiers wandered about for a minute or two, looking for them, and then quietly marched off after the others.

(243.2) “Are their heads off?” shouted the Queen.

(243.3) “Their heads are gone, if it please your majesty!” the soldiers shouted in reply.

 

1.23 By the same token, questioners can influence the answerer’s state by deploying strategic phrasing. Since the question action indicates that reason for doubt exists, positives encourage negative responses and vice-versa:

 

(244a) Do you think you ought to go?

(244b) Don’t you think you ought to go?

 

From (244a), the hearer could infer the speaker’s belief that ‘going’ is inadvisable, and from (244b) just the opposite (cf. Fillenbaum 1968). The question format interacts with the conventions for negation. If negation is typically used for material that might otherwise be believed (ef. IV.1.25), the question countermands that setting by implying that the addressee has doubtful grounds for disbelief. Elizabeth Loftus (1975) explores a number of ways in which question formats set up expectations in eliciting eyewitness reports. When test subjects were asked (246) after (245a), 53% said yes, but only 35% said yes after (245b):

 

(245a) How fast was Car A going when it ran the stop sign?

(245b) How fast was Car A going when it turned right?

(246) Did you see a stop sign?

 

yet there was no stop sign shown in the film of the accident!

1.24 Rachael Reichman (1978) has undertaken to describe the mechanisms of topic flow and shift across the various turns of entire conversations: using actually recorded sample conversations, she argues that coherence relations can obtain between chunks of discourse in which topic appears to change over considerable distance. She proposes a distinction between issues spaces (“a general issue of concern” plus the agents, affected entities, and times, etc. involved) and event spaces (“a particular episode and the events that occurred therein” plus agents, affected entities, times, locations, etc. involved) (Reichman 1978: 291f.). The coherence of conversation depends on how these space types are related. An illustrative or restatement relation obtains when an event space is adduced to demonstrate or clarify what has been asserted in an issue space. Conversely, a generalization relation obtains if an event space is followed up with a discussion of the “general activity” to which the event belongs. If an issue space or event space is temporarily abandoned in favor of an unrelated one and then resumed, we have interruption and return relations. If an event space is used to show that two issue spaces are contingent upon each other (e.g., via causality), we have a subissue relation; if the two issue spaces are merged into “one composite issue,” we have a joining relation. A respecification relation obtains if an event or issue already fully discussed is rediscussed in a different perspective. A total shift relation obtains if the new discourse chunk is not at all related to its predecessor. Reichman shows that these various discourse relations are frequently accompanied by surface signals such as ‘like’ and ‘like when’ (illustrative), ‘by the way’ (interruption), ‘anyway’ (return), and so forth.

1.25 The mechanisms of conversation are undeniably complex. Nonetheless, the work I have reviewed in this section promises to reveal at least some of the major factors worthy of exploration. There is a pronounced interaction among sources of knowledge, organization of topics, participant roles, and criteria for considering what is interesting and worth talking about. An entire discourse must have textuality, even when the textuality of its component texts is not obvious in isolation (cf. VIII.1.1; Beaugrande & Dressler 1980). Clearly, the study of conversation must be carried out with co­operation among the various disciplines—linguistics, psychology, sociology, anthropology, computation—that will profit from insights into this versatile and vital domain.

 

2. NARRATION

 

2.1 The investigation of stories prefigures the general trend to which this book also belongs. Early methods inspired by linguistic structuralism sought to isolate standard units in chains (cf. Propp 1928; Dundes 1962; Bremond 1964). Later on, transformational grammar became the source of inspiration (e.g. Greimas 1967; Zolkovskij & Sceglov 1967). Recently, however, attention has been directed away from abstract units and forms toward cognitive processes in the comprehension of stories (e.g. Charniak 1972; Kintsch 1974, 1977b, 1979a; Rumelhart 1975, 1977b, 1978; Mandler & Johnson 1977; Schank & Abelson 1977; Thorndyke 1977; Cullingford 1978; Rieger 1978; Wilensky 1978; Beaugrande & Colby 1979; Beaugrande & Miller 1980). The trend is thus away from abstractions upon surface artefacts and their features toward human activities of utilizing texts. Whereas the former are often specific to a language, a topic, or a cultural and historical domain, the latter may be UNIVERSAL (cf. IV.3.17.ff.).

2.2 One major consequence has been the realization of how much prior knowledge is deployed by the understander. The effects of SCHEMAS as global knowledge patterns applied to stories have been irrefutably demonstrated. Readers can put scrambled stories back into the proper order (Kintsch 1977b; Kintsch, Mandel, & Kozminsky 1977; Stein & Nezworski 1978). The removal of material needed to match important schema elements interferes with comprehension and recall (Thorndyke 1977). Stories in which events of different sequences are so interlaced that concurrent schemas must be maintained for each sequence are rearranged so as to separate the schemas (J. Mandler 1978).

2.3 Despite its recognized importance, the schema appears in very diverse formats in research. Some investigators envision a set of REWRITE RULES of the familiar transformational type, in which large story components are “rewritten” as smaller ones (e.g. Rumelhart 1975; Mandler & Johnson 1977; Simmons 1978). Others make use of TREES in which story constituents are arranged in a hierarchy of size, containment, or importance (cf. Bower 1976; Rumelhart 1977b; Thorndyke 1977). These two formats are essentially equivalent, because the rewriting, in effect, acts as a parent node descending to offspring nodes (hence Mandler & Johnson use both formats). However, the cognitive implications of formatting have often been glossed over. Where do the story components actually come from? Are they (1) segments of supersegments, (2) instances of a class, (3) elements of an unordered set, or (4) products of transformational derivation? These relationships would have significantly different impacts on actual processing.

2.4 Ideally, hierarchical structuring ought to reflect cognitive priorities. The higher-up components should be noticed and recalled better than the lower-down ones (Meyer 1975, 1977). However, the data I reviewed for the ‘rocket’ text suggest that recall is more diffuse and topographical in nature as documented by the contrast between our protocols in VII.3.32ff. and the idealized hierarchical summaries generated by Simmons’ computer simulation in the Appendix. People apparently retain quite a lot of material that would figure as lower-down components in a hierarchy. Their recall manifests the priority of connectivity and continuity more than that of height in tree structures.

2.5 To clarify issues of topography vs. hierarchy, we could explore the effects of BOTTOM-UP input on the TOP-DOWN input during story comprehension (cf. I.6.5). During the PROCEDURAL ATTACHMENT of a story schema to an actual story text, the schema is evidently specified and modified as occasion arises (Beaugrande & Miller 1980). The enduring qualities of great folktales must depend upon the processing of their inherent structures in interaction with schemas. By the standards of informativity and interestingness proposed in chapter IV, it follows that these famous tales cannot be a perfect match for the stored schema pattern: some uncertainties, alternatives, and surprises are, I suspect, virtually obligatory for the actualization of interesting and enduring stories. Indeed, one might want to insert such a requirement into the story schemas themselves (cf. Beaugrande & Colby 1979).

2.6 A minimal STORY-WORLD must contain at least a pair of states linked by an action or event. But to be interesting, the story-world needs a structure in which the progression from the INITIAL to the FINAL state is not so obvious that it would happen on its own in the normal course of things. For a story-world fraught with alternative pathways, the narrator and the readers engage in joint PROBLEM-SOLVING in which the narrator’s solution eludes that of the readers at least some of the time.

2.7 Narrators can create uncertainty by using CHARACTERS (story-world persons) with opposing PERSPECTIVES. A given character is assigned a particular goal to seek in the course of events (cf. the notions of “objective” and “achievement” in Bremond 1973). If the reader audience sees that goal with positive values, the character will be a PROTAGONIST; for negatively valued goals, the role is that of ANTAGONIST.5  [5. Like language regularities in general (cf. note 14 to Chapter I), this one can be turned around for special effect, e.g. in the “picaresque” narrative where the protagonist’s goals violate official standards of conduct, though readers may still find them positive in context. A completely goalless neutral protagonist such as that in Camus’ L’étranger is both hard to present and not especially convincing, to me at least.] The interaction of characters appears as a pursuit and mutual blocking of goals (cf. Wilensky 1978; compare the notion of “polemics” in Greimas 1970). Goal-blocking readily upholds uncertainty, especially when the narrator creates a powerful and resourceful antagonist. The event or action which makes a main goal decisively attainable or non-attainable is a TURNING POINT. In terms of the drama, a positive turning point for the protagonist is the conventional mark of “comedy,” and a negative one the mark of “tragedy.”

2.8 These considerations might be used to formulate some STORY­TELLING STRATEGIES (rather than abstract rewrite roles) such as the following (cf. Beaugrande & Colby 1979: 45f.):

2.8.1 Create a STORY-WORLD with at least one CHARACTER.

2.8.2 Identify an INITIAL STATE, a PROBLEM, and a GOAL STATE for the character.

2.8.3 Initiate a pathway that attempts to resolve the problem and attain the goal state.

2.8.4 Block or postpone the attainment of the goal state.

2.8.5 Mark one event or action as a TURNING POINT.

2.8.6 Create a FINAL ST ATE identified as matching or not matching the goal state.

2.9 These strategies can be applied recursively, generating STORY EPISODES of varying complexity or number. My own definition of “episode” is that of a space in a story-world with an initial state, a problem, a turning point, and a goal state (but compare the definitions in Rumelhart 1975, 1977b; Kintsch 1977b; Simmons 1978). A frequent demand for recursion arises from having a story-world with multiple main characters, each of them assigned actions and goals. The story-world with PROTAGONIST and ANTAGONIST could be governed by a role set like this (cf. Beaugrande & Colby 1979: 46):

2.9.1 Create a story-world with two characters, the PROTAGONIST P and the ANTAGONIST A.

2.9.2 Create a PROBLEM for P that is caused or desired by A, and a goal state desired by P and opposed by A.

2.9.3 Initiate a pathway that attempts to resolve P’s problem and attain P’s goal state.

2.9.4 Create actions of A to block P’s solution and goaI.

2.9.5 Mark one action or event as a turning point in which either P’s or A’s plans and values win out.

 2.9.6 Create a final state identifiable as matching or being relevant to either P’s or A’s goal state.

 2.10 If the narrator makes the antagonist extremely powerful, a compensatory strategy may be required:

 2.10.1 Introduce one or more HELPING CHARACTERS to create ENABLEMENTS or to block DISENABLEMENTS of P’s actions and goals.

2.11 The traditional categories of the narrative (cited in Kintseh 1977b) could be viewed as clusterings of realizations for these various strategies. The EXPOSITION would include 2.8.1 through 2.8.3; the COMPLICATlON would be in 2.8.4; and the RESOLUTION in 2.8.5 and 2.8.6. The realization of the strategies is flexible in many ways. By adding more characters, the narrator has the options of making their goals in COMPETITION or in CONCORD (Wilensky 1978). For competition, Wilensky (1978) discusses various means of “anti-planning” including sabotage, concealment, distraction, removing enablements, and overpowering. Also, a single character may have different goals in conflict with each other (cf. VI.4.11; Wilensky 1978: ch. 6). Even just one goal may raise formidable problems if its attainment is difficult enough. Wilensky (1978: 253) cites typical circumstances of difficult plans:

2.11.1 if the plan requires exceptionally great RESOURCES, such as are probably not available;

2.11.2 if the plan leads to UNSTABLE GOALS (cf. VIII.2.22).

2.12 All of the participants in narrating-narrator, audience, and the story-world characters are engaged in activities of planning and predicting. The narrator must: (1) plan out coherent tracks of states and actions for each character; (2) relate narrated actions to recoverable plans of the agent character; and (3) anticipate and monitor how the audience recovers or reconstructs characters’ plans and predicts upcoming actions and events. The narrator needs to outplan the audience at least sometimes to keep the story interesting. The narrator can achieve this effect in several ways:

 2.12.1 by selecting a rather improbable pathway to follow in the story line, e.g. by having characters make bad or unreasonable selections and decisions;

2.12.2 by introducing unforeseeable interactions among events, e.g. by having independent characters suddenly happen to come into contact or conflict;

2.12.3 by purposefully withholding knowledge that would otherwise render upcoming events predictable, e.g., by failin