Listeners generally attempt to understand oral conversational stories by figuring out what the narrator is 'getting at'; their understanding is point-driven in this sense. Analogously, a form of reading in which readers expect to be able to impute motives to authors may also be called point-driven; it is a mode that seems especially useful for reading so-called 'literary' texts. Point-driven reading is conceptually distinguishable from story-driven and information-driven types. We argue that each type is associated with a number of cognitive strategies, with point-driven reading, specifically, characterized by coherence, narrative surface, and transactional strategies. Using a modern short story, we illustrate how point-driven readings might be differentiated from other kinds. An advantage of this conceptualization is that it enables one to generate empirically testable hypotheses about literary reading; we suggest a number of such hypotheses and methods of testing them.

Susan had been really worried about it, because she didn't know the other girls very well, and she couldn't . . . she didn't know how they were doing. So when she went in for it. she was really scared, and when she came out I guess she figured she'd probably failed or something. But when they posted the list, it turned out she got the highest mark.

Not surprisingly, however, the meaning a reader or listener might attribute to this story is affected by what he or she knows about the discourse in which it took place. Earlier in the conversation the narrator had told another anecdote, which also entailed a situation in which Susan had underestimated her academic ability. In light of this new information, we might begin to suspect that what the second story is really about is not so much Susan's getting a high mark as her tendency to be too self-critical.

It is also important to know that immediately preceding the 'mark' story the narrator had recounted an anecdote in which Susan had had a term paper (which had been duly submitted, marked, and graded) stolen from a box outside her instructor's office. The listener had then remarked that that must have been difficult but it wasn't as bad as losing the paper before it had been marked and graded. Knowing this broadens the meaning of the anecdote even further: apparently the topic of the entire conversation is the vagaries of academia, and perhaps not nearly so much the particular personality of Susan.

We might stop there and consider that the point -- that 'meaning' is profoundly influenced by context -- has been demonstrated. But the results of going a few steps further are equally dramatic, and suggest that a preoccupation with 'meanings' may not be the most useful or appropriate stance in an attempt to understand stories, tellers, and audiences. The following facts are surely not irrelevant for understanding this incident. The narrator of the story was a mechanic (not university-educated); the listener was a university English professor. The conversation took place while the mechanic was driving the professor to work, the professor's car being in for repairs. The two men knew each other only slightly. Susan was the mechanic's daughter; she attended not the professor's, but a neighboring institution.

Suddenly we are likely to become aware of an entire range of overtones and more or less covert purposes that were simply not available before. Now we wonder: why is the narrator telling this story? and the answer seems to lie in the fact that by telling it he can accomplish a number of things simultaneously. First, and most directly, he is establishing his daughter's identity as a student, and a very good student at that. By extension, he is affirming his own identity as a proper father, 'proper' both because his child is receiving a postsecondary education, and because he is taking appropriate interest in her progress. Nevertheless, the mechanic seems to be suggesting that universities can be cold, unfair, and apparently incomprehensible places. Depending on the previous relationship between the two men, one might suspect a covert and possibly non-conscious aggressiveness, a reflection of traditional 'town and gown rivalry. Or the narrator may be seeking clarification or understanding from his professorial listener as to what universities are really about.

How the professor responds to this story will depend, in turn, on what motives he imputes to the narrator. More generally, it seems clear that the listener's central task is not to infer the 'meaning' or 'gist' of a speaker's narrative, but rather to determine what the speaker might be 'getting at.' In a word, the listener tries to construct an appropriate point for the story. We call such listening 'point-driven.' Listening to oral conversational stories, we suggest, is normally of this kind.

Tellers, listeners, and points

To say that listeners attempt to construct points is not, however, to make clear just what sort of thing a 'point' actually is. Despite recent interest in the pragmatics of oral stories (Polanyi 1979, 1982; Robinson 1981), conversations (Schank et al. 1982), and narrative discourse generally (Prince 1983), definitions of point are hard to come by. Those that do exist are usually couched in negative terms: apparently it is easier to indicate what a point is not than to be clear about what it is. Perhaps the most memorable (negative) definition of point was that of Labov (1972: 366), who observed that a narrative without one is met with the "withering" rejoinder, "So what?"

It is, however, possible to define -- or at least to characterize -- more positively what it is that's missing in such a situation. Whether we term it an interpersonal move, a pragmatic gesture, or, as van Dijk and Kintsch (1983) do, a global speech act it is clear that a point involves not the exchange of information, but rather the sharing and comparing of values and beliefs, particularly by means of what Labov (1972: 366) has called evaluation: "the means used by the narrator to indicate the point of the narrative, its raison d'être: why it was told, and what the narrator is getting at." The kinds of things narrators ultimately 'get at,' as Polanyi (1979) shows, are socially and culturally shared values and beliefs. To make a point, then, or to construct one, is to engage in a fundamentally social, pragmatic, and interpersonal enterprise.

Points are not, however, characteristics only of face-to-face narrative situations. Returning to the story about Susan and her high mark, consider what happens when the 'frame' is extended another step, Suppose the conversation between the mechanic and the professor were a scene in a play or a chapter in a novel. Clearly, in such a case, an entirely new pattern of signification can come into effect. For example, our inferences about the mechanic's motives for telling the story may well be profoundly altered if we posit an author who has created that character in order that we should make such inferences. Consequently we are more likely to infer that the story is 'really about' the mechanic, rather than Susan or the university. When we read certain kinds of writing, then, we may have points to construct not only for narrators and other characters, but for authors as well. We are suggesting, in other words, that reading as well as listening can be 'point-driven.' Further, we argue that point-driven reading is an important -- arguably, an essential -- component of a type of reading commonly thought of as 'literary. '

In this paper we examine the pragmatics of literary reading by considering them as analogous to conversational story listening. Both reading and listening, then, can be seen as forms of a more general 'point-driven understanding.' Although other theorists have proposed analogies between conversational and literary stories (among them Pratt 1977 and Watts 1981), our approach is different in at least four ways. First, we are centrally concerned with an analogy between the texts' 'consumers' (the listener and the reader), and much less with one between the two texts themselves or between the texts' 'producers' (the teller and the author). Second, we maintain that an adequate account of literary reading must attend both to its social-pragmatic and to its psychological-cognitive dimensions (cf. John-Steiner and Tatter 1983). Third, we suggest that important distinctions can be made among various types or modes of reading; in particular, we distinguish among point-driven, story-driven, and information-driven types. Fourth, we believe that these distinctions can be used as a basis on which to generate predictions about specific reader behaviors. Accordingly, in the final sections of the paper we propose a number of hypotheses concerning strategies in literary reading, and then give some methodological suggestions as to how each hypothesis could be investigated experimentally.

Point-driven listening

When a story is told in an oral conversational setting, the narrator is granted the floor for one storytelling turn. The story recipients thereby undertake certain responsibilities: for instance, they are expected to indicate to the teller that they are understanding the story, and when it is finished they are expected to indicate whether or not they accept it. In return, listeners anticipate that they will be able to construct a point for the story. The likelihood that a listener will in fact be able to construct a point successfully, as well as the nature of that point, seems to depend on a number of simultaneous 'pressures' acting on the listener:

1. The listener's model of an author. The listener is more likely to construct a point successfully if he or she has a sense of an author who intends to make a point. Of course, in the oral situation it is not difficult for the listener to imagine an author -- quite simply because the author is physically present, in the person of the narrator. Further, point is influenced by what the listener knows about the narrator-author; for example, the kinds of points he or she usually makes, and whether, indeed, he or she usually intends to make (or succeeds at making) any point at all.

2. The listener's cultural expectations. Listeners are more likely to construct points that are socially and culturally salient. As Polanyi (1979: 207) argues, what can count as a point in a given culture is only material generally agreed upon by its members to be "self-evidently important and true." In North American culture, for instance, such basic cultural constructs would include THE INDIVIDUAL, FRIENDS, PROBLEMS, UNDERSTANDING. and many others (see Polanyi 1978 for an extensive list, and for examples of points appropriate to other cultures which would fail to render a story 'tellable' in North America). This helps explain why people often find stories from other cultural traditions difficult to understand, or even 'pointless.'

3. The listener's generic expectations. What is an appropriate point for one genre may be inappropriate for another. Therefore the point constructed by the listener will depend to some extent on the genre deemed to be in effect: personal narrative, tall tale, ghost story, anecdote, sermon, report, etc. For example, if the listener identifies the narrative as a 'parable, he or she will strongly expect there to be a point, but, equally strongly, expect it to be indirectly expressed.

4. Context. Context here refers to the physical and social setting in which the storytelling takes place; it, too, can affect what, if any, point is constructed by the listener. Physical setting includes such matters as environmental conditions, time of day, and so on; social setting includes the social identities of and role relationships between teller and recipient, as well as operative customs and mores. For example, in the story about Susan the listener may have felt more pressure to construct a point because he was the only story recipient (rather than a member of a larger audience); further, the nature of the point constructed was likely to have been strongly influenced by the social identities of the two men.

5. Cotext. Stories told in conversation do not come 'out of the blue' but instead are generally preceded by conversational interaction or even previous stories. Listeners expect that the current story will be somehow connectable to previous relevant discourse. (Such prior discourse may have occurred days or even weeks previously.) It may be that the story's topic can be related to a previous conversational topic; alternatively, the story may introduce a new topic but make a related point; other variations, of course, are possible. The mere availability of context makes it more likely that a listener will attempt to construct a point for the story.

6. Text. Not least of all, the point constructed by the listener will depend on the narrated text itself, which may include linguistic, paralinguistic, and kinesic levels of communication.

Collectively, the above pressures can be considered a 'pragmatic frame' which places limits on the nature of the point which may be constructed by a listener, and in some sense determines whether or not a point will be successfully constructed at all.

It is worth stressing here that we refer to the listener's 'construction' of point in order to underline our view that points aren't 'in' stories, waiting to be identified by perceptive listeners, but instead are constructed by listeners on the basis of various sources of information, only one of which is the text.

It is also important to note that the listening could fail for any number of reasons. For instance, the narrator may obviously intend to make a point, but the text may fail to support the point he or she intends to make. In that case narrator and listeners might still rescue the situation by 'negotiating' the point of the story; that is, by debating what the story may properly be said to be about (Polanyi 1979). A second cause of failure may be that the listener's and the narrator's sense of cultural or generic appropriateness differ too widely. For example, a 'story' may be told which according to its narrator is pointed but which the audience finds to be pointless. The audience then responds overtly or otherwise, with the question Labov (1972: 366) said "every good narrator is continually warding off,': "So what?" Narrators who tell pointless or irrelevant stories are seen as boring and inept and suffer a loss of face. And although listeners who miss points are less severely penalized in social terms, there are parallel consequences, particularly if there are other members of the audience who do 'get the point. '

Point-driven reading

Listening to an oral conversational story, then, is normally 'point-driven': recipients listen to the story in anticipation that they will be able to reconstruct it as a 'pragmatic gesture,' a 'global speech act.' Similarly, there is a type of reading which may be called 'point-driven,' because in it, too, the understander reads with the expectation that the text will enable the construction of a valid, pragmatic point. Point-driven reading is both similar to and different from point-driven listening. It is similar because in both types the construction of point is a function of the text, the comprehender's cultural, and the comprehender's generic expectations. Point-driven reading is different from listening' though, for several reasons. First, oral stories are generally immediately connectable to previous discourse, whereas written stories are relatively autonomous: the reader has less context available to constrain the nature of the point constructed. A second difference (which is quite obvious but which has not-so-obvious consequences), is that in listening but not in reading the producer and recipient of the text share the same socio-physical context. Consequently, listeners are able to see their authors whereas point-driven readers must imagine or 'invent' theirs. Furthermore, the chances of readers being able to construct points successfully are increased if they impute motives to authors: that is, if they imagine authors as beings who intend to make points.

By referring to the intentions of authors, however, we do not mean to argue (a la Hirsch 1967) either that what the author intended as a point is ascertainable, or that it can or should determine what the story's point will be. Just as the point of a conversational story is subject to negotiation among audience and producer, so what is taken to be a point for a written narrative will generally be the result of negotiation between the reader and the text.

It may seem odd to call this process negotiation,' since negotiation implies the possibility of change, and although readers may change, how can a text? The answer lies in a distinction between text' as marks-on-a-page and 'text' as that which is reconstructed by means of the transaction between the reader and the page (this is similar to Rosenblatt's 1978 distinction between 'text' and 'poem'). It is this reconstructed, virtual text that can, of course, change, as the reader reconstructs it again and again.

Nevertheless, this process of negotiation will fail if the text does not repay the attempt to construct a point from it. Here again point-driven listening and reading differ: whereas listeners can, by refusing to accept the story as narrated, influence the narrator to change his or her evaluation of the story or even the direction of the narrative itself, the ordinary reader has very little influence on an author. On the other hand, authors can anticipate this problem, and can therefore build into their stories the tools that a reader needs in order to assemble a pragmatic frame in which a point may be appropriately constructed.

A second reason point-driven reading may fail is that the person doesn't attempt to read this way in the first place. Presumably many people have rarely or never experienced success with this type of reading (even though they are routinely successful at point-driven listening), and therefore have not learned to use it -- or perhaps have learned not to use it, or not to use it in certain contexts. In any case, a reader's failure to assemble a pragmatic frame -whether because the story's built-in tool kit is inadequate, because the reader's and author's sense of generic appropriateness differ too greatly, or for some other reason -- will lead to the same withering "So what?' response that can occur in the face-to-face situation.

In fact, we have some evidence suggesting that this 'so what' response is a surprisingly frequent occurrence. In a series of studies, we had undergraduates (total N> 150) read a short story (John Updike's "A & P", 1962) under various task conditions, and then respond to a number of open-ended and leading questions. (Some of the students had, over a period of weeks, read the story as many as five times in all.) From their performance on these tasks, it appeared that a small number of the students -- approximately 5% -- were aware that it might be possible to impute motives to an intentional author. The majority of the readers, however, found the story to be incomplete and pointless -- just as if they were participating in a conversation in which the narrator was a known bore or social misfit who told stories that no one expected to have a point. It is of course true that the context in which this story was read made it less likely that a reader would expect it to have a pragmatic point; even so, the number who exhibited signs of attempting to read in such a way seemed startlingly low.

Point-driven reading, it should be stressed, is only one type of reading, appropriate for some situations but not others. It may be especially likely to occur in connection with 'literary' texts; however, as discussed further below, it is not the only way to read such texts. Nor does literary reading require literary texts [1]. On the contrary, what is important is that one reads a text -presumably any text will do (cf. 'found' poetry) -- in a literary way. Finally, we do not claim that point-driven reading is the best' or 'highest' type of reading. Or that it is, developmentally' a more advanced form. On the other hand, we would argue that it is a useful -- probably vitally useful -- addition to any mature reader's repertoire.

To further clarify the nature of point-driven reading, it may be helpful to distinguish it from two other types.

Information-driven reading

Very different from point-driven reading is a type of reading which we calls 'information-driven.' Information-driven reading is especially appropriate in learning-from-text situations where content is relevant. To take an extreme example, if one is reading a bus schedule it is hardly necessary to construct a 'model of an intentional author' or to see the text as some kind of 'pragmatic gesture.' Information-driven reading is most likely to occur in contextually-isolated situations, when the reader's task is to learn or remember the material, and when the text itself is fragmentary or inane (Beaugrande 1982) -- that is, doesn't repay the assumption of point. Since these conditions usually hold hi laboratory experiments on reading, it could be said that the resulting theories are essentially accounts of information acquisition, and may be only indirectly related to theories of literary reading (Dillon 1980).

Story-driven reading

It may be objected that the difference between 'point-driven' and 'information-driven' is merely another way of phrasing Rosenblatt's (1938, 1978) distinction between "aesthetic" and "efferent" reading. This is not the case, however. A way to make this clear is to describe a kind of reading which occupies a middle ground between point- and information-driven but which Rosenblatt's model would not distinguish from point-driven reading. According to Rosenblatt, aesthetic reading is concerned with " the lived-through experience" of reading the text, whereas efferent reading is concerned with what the reader "takes away" from the text. In our view, however, there is all important distinction to be made between kinds of reading which are concerned with "lived-through experience." While 'point-driven' reading is obviously concerned with that lived-through experience, so, eve would argue, is a quite different kind of reading, which we call 'story-driven.'

The term 'story-driven' is based on Chatman's (1978) distinction between 'story' and 'discourse' in narrative. Accordingly, story-driven readings tend to emphasize plot, character, and event, and to neglect the discourse' by which the events and characters are presented. Someone reading in a story-driven way will be looking for a 'good read' -- interesting, affectively-arousing events (Morgan and Sellner 1980), rounded characters, and the like -- but will not anticipate that the narrative will, in the way conversational narratives are expected to, invite and assist the construction of a valid 'point.' By the same token, a person reading in a story-driven way would not find it necessary to construct a model of the author: the story seems to exist, and can be enjoyed, quite independently of any implied author.

It is important to make clear that these three types or modes of reading -- point-, story-, and information-driven -- are not characteristics of readers, even though it seems probable that some readers only rarely or never engage in point-driven reading. Nor are they characteristic even of whole readings; it is likely that on any particular reading occasion, depending on his or her current goals, the fluent reader would have the flexibility to use a mixture of different strategies. In the following section we discuss some of these strategies in more detail.

Strategies in point-driven reading: an example

Up to now, we have been discussing the pragmatics of oral and written story comprehension. We suggested that people normally listen to, and sometimes read, stories in a 'point-driven' way, meaning that they attempt to assemble a pragmatic frame in which a point can be successfully constructed. However, this leaves open the question of what someone reading in a point-driven way actually does. That is, if reading a literary text is viewed from a cognitive, 'online' perspective, what precisely are the differences between point-driven and other types of reading? We find it useful, in this regard, to consider each type of reading as made up of a set of characteristic 'strategies' (van Dijk and Kintsch 1983), with the important qualification that 'strategy' does not imply deliberate or conscious use [2]. In particular, we claim that point-driven reading is characterized by three types of strategies: coherence, narrative surface, and transactional. To illustrate these (partly overlapping) strategies more clearly, consider how a point-driven understander might read a short story; for example, Updike's "A & P":

In walks these three girls in nothing but bathing suits. I'm in the third checkout slot, with my back to the door, so I don't see them until they're over by the bread. The one that caught my eye first was the one in the plaid green two-piece. She was a chunky kid, with a good tan and a sweet broad soft-looking can with those two crescents of white just under it, where the sun never seems to hit, at the top of the backs of her legs. I stood there with my hand on a box of HiHo crackers trying to remember if I rang it up or not. I ring it up again and the customer starts giving me hell. She's one of these cash-register-watchers, a witch about fifty with rouge on her cheekbones and no eyebrows, and I know it made her day to trip me up. She'd been watching cash registers for fifty years and probably never seen a mistake before.

in the big windows, over the bags of peat moss and aluminum lawn furniture stacked on the pavement, I could see Lengel in my place in the slot, checking the sheep through. His face was dark gray and his back stiff, as if he'd just had an injection of iron, and my stomach kind of fell as I felt how hard the world was going to be to me hereafter.

Consider, as an example, the incident involving the "cash-register-watcher" in the first paragraph of "A & P." According to a story-based analysis, this is at best a minor episode. Although it does help establish that Sammy is a cashier, is young, etc., it is hardly crucial in terms of plot -- it would not be part of a story summary, for instance. However, a reader who is reading in a point-driven, coherence-seeking way is likely to use even a throwaway episode such as this one. In this instance, a point-driven reading might well notice that Sammy's antipathy towards the cash-register-watcher connects with his attitude towards all regular customers (he later refers to them as "houseslaves," "sheep," etc.), and, furthermore, contrasts with his lovingly meticulous descriptions of the three girls [3].

Related to coherence is the question of when one evaluates the worth or quality of a text. We suggest that to read in an inescapably story-driven way is to suffer from premature evaluation: such readers may 'give up' early deciding after only a few episodes that the story isn't a good one, particularly if there seem to be a large number of unconnected elements or if the events aren't 'interesting' enough. By contrast, point-driven readers, who tend to view the text as a single, coherent speech act, are more likely to defer evaluation until all the evidence is in.

2. Narrative surface strategies. Narrative surface refers to the discourse aspects of narrative (e.g., point of view, tone, diction, 'style'), as distinguished from its story aspects (plot, setting, character) (Chatman 1978). Consider, for instance, the opening of "A & P": "In walks these three girls in nothing but bathing suits. I'm in the third checkout slot . . . ." For a story-based reading, what is important in these sentences is the setting information (the narrator is a cashier), and the initiating action information (enter three girls in bathing suits). However, the stylistic features -- Sammy's non-standard diction, grammar, and tense -- are relatively unimportant. We suggest that in a story-driven reading such features either would not be noticed at all, or, if they were noticed, would be taken as evidence of the author's carelessness or incompetence. A point-driven reading is likely to be different in two ways. First, the nonstandard elements are more likely to be noticed (again, it is important to stress that noticing, like other strategies, does not necessarily imply consciousness, but will have consequences). Second, the unusual features are more likely to be explained as a purposeful and deliberate contrivance: a dramatic device. A dramatic device, though, is used for some particular reason. Consequently someone who is reading in a point-driven way implicitly imputes a motive to the author and thus recognizes that there is a puzzle here to be resolved.

3. Transactional strategies. Where do texts come from? Our view is that people reading in a point-driven way implicitly realize that the text is an artifact, and therefore recognize the existence of an intentional being who is responsible for it. Because point-driven readers recognize intentionality behind. Or impute it to, the text, they realize that what is presented in the text -including the actions and beliefs of the characters -- cannot be taken at face value but instead must be interpreted in light of the fact that the characters are mere creations of an intentional being. In short, someone reading in a point-driven way is likely to notice discrepancies -- that is, ironies -- between the implied values and beliefs of the author as against the values and beliefs of the characters.

We claim that people reading in story- or information-driven ways do not adopt transactional strategies. Instead of seeing the text as an artifact they tend to see it as a natural phenomenon -- it's 'just there.' These readers take the actions and beliefs of the characters at face value, without considering the possibility that everything about the characters can be seen in relation to what an implied author might be getting at. In one sense, of course, what story- and information-driven readers do is reasonable: often a text can indeed be enjoyed (or learned) on its own terms, without making the assumption that it is the product of an intentional being. Then again, some texts simply don't repay the assumption of point. Nevertheless, story- and information-driven readings have the serious drawback that they tend not to recognize ironies -- a particularly serious problem when dealing with 'literary' texts.

What difference would transactional strategies make to a reading of "A & P"? Someone reading in a point-driven way is more likely to recognize that the text is an artifact, that it was created by an implied author named 'Updike' who is responsible for inventing a narrator-character named 'Sammy.' Once a distinction between implied author and narrator is made, the way is clear to recognize some possible differences between them. For example, people reading in a point-driven way are likely to infer that Updike probably does not share Sammy's blatantly sexist views of women ("do you really think it's a mind in there or just a little buzz like a bee in a glass jar?"). As well, the point-driven reader is more likely to accept the invitation to see 19-year old Sammy's quitting his job from a more mature viewpoint than Sammy himself does, and when Sammy comes to his realization of how hard the world is going to be to him, to see a bit of humor in that.

Making fine distinctions between an implied author and a characterized narrator is not so important to someone reading a text primarily for its information value or storyline. Thus it is not surprising that information- and story-driven readers tend to conflate 'author' and narrator': according to them, it is Updike who uses bad grammar and Updike who is sexist. It is also not surprising that when confronted with heavily ironic texts such as Nabokov's Pale Fire (1962) -- where there is no 'story' to speak of -- such readers tend to 'miss the point' entirely, and tend to feel dissatisfied with the reading experience. Blame, however, is usually attributed to the text -- it is as if they ask it, witheringly, "So what?"

Empirical questions

We have presented these three reading strategies in a rather informal and descriptive manner, and thus it may not be immediately apparent that they are all, in fact, empirically testable (although still largely untested) hypotheses. In this section we will try to give the hypothesized strategies of point-driven reading a firmer empirical base, first, by citing some relevant research findings in their support, and second, by proposing some methodologies and techniques by which they could be experimentally tested.

Coherence. We proposed that the coherence strategies of point-driven reading differ from those of the other types. There are some findings in the experimental literature that seem to support this claim. From sentence processing research, it is known that adult readers encode a sentence differently depending on whether they are trying to remember it or comprehend it (Aaronson and Scarborough 1976: Aaronson and Ferres 1984). In the memory task, extra processing occurs at phrase boundaries, but in the comprehension task, processing is dependent on semantic factors and is not related to structural divisions. Similarly, from discourse processing research, it is known that when a person is reading in order to remember a story, extra processing occurs at structural (episode) boundaries (Haberlandt 1980: Haberlandt et al. 1980): moreover, episodes tend to form cohesive chunks in memory (Black and Bower 1979).

We do not vet know whether, given tasks that promote point-driven comprehension rather than information-driven memory, discourse processing is less dependent on episode boundaries. This does, however, seem a reasonable inference. We therefore hypothesize that point-driven reading uses larger sized chunks than do information- or story-driven types. Two specific predictions follow: first, that point-driven readers should keep seemingly irrelevant textual elements in working memory longer: and second, that they should make more effort to integrate disparate and apparently unrelated details.

The length of time information is retained in working memory could be determined by an adaptation of Sternberg's (1966) "memory probe" technique (cf. Fletcher 1981). In this task, readers would be interrupted at various places and asked to verify whether or not a given probe (word, concept, sentence, etc., appeared earlier in the text. In a point-driven reading, information drawn from previous episodes should be more likely to be retained in working memory because the reader is attempting to create chunks that reach across episode borders; thus such readers should be faster than story-driven readers at verifying probes drawn from previous episodes. On the other hand, there should be no difference between the speed with which probes drawn from the current episode are verified by point-driven or story-driven readers.

The second prediction -- that people reading in a point-driven way make more effort to integrate apparently unrelated details -- could be investigated by means of 'process tracing' techniques. From protocol analysis, for instance (cf. Hayes and Flower 1980), it might be learned that point-driven readers are more persistent in attempting to integrate unexpected material. Similarly, eye movement patterns and inspection times (cf. Rothkopf and Billington 1979) may help determine how readers process unexpected text elements.

Narrative surface. Previous work in the area of sentence memory lends support to our hypothesis that point-driven readers should have especially accurate memory for narrative surface. Keenan and MacWhinney and their colleagues (Keenan et al. 1977; MacWhinney et al. 1982) have found that when conversational utterances have rich interactional (pragmatic) content, the surface form of such utterances is recognized better than equivalent utterances with neutral content. Analogously, we expect that readers who are reading in a point-driven (pragmatic) way will be more sensitive to, and thus have better memory for, narrative surface than will those reading in story- or information-driven ways.

We are currently investigating memory for narrative surface by having university students read a short story and then complete a recognition test containing verbatim as well as paraphrase and distractor items. We expect that students reading in a point-driven way will demonstrate better surface memory than other readers, and also that their errors will tend to be more consistent with the story's style (cf. Brewer and Hay 1984). We have also done some work in measuring sensitivity to narrative surface. Specifically, we have developed a 'branching text' exercise: this is an 'in process' task in which a reader is offered, at several places during the reading of a story, a set of possible alternate continuations, and asked to select the one most consistent with the rest of the story. The continuations are semantically similar but vary in point of view and tone: that is, they retain the same story but vary the discourse. We hypothesize that people reading in a point-driven way will be more sensitive to narrative surface, as indicated by better performance on this task relative to story- and information-driven readers.

Transactions. Social psychologists have known for some time that in contextually-isolated situations, people tend to take information at face value. For example, people are prone to make the 'fundamental attribution error' of imputing dispositional qualities to a person, even when the individual's behavior is clearly contaminated by situational constraints (Ross 1977). Recent evidence suggests, however, that this error is much reduced, or even eliminated, under conditions that are more ecologically -- that is, pragmatically -- valid: for example, when group discussion precedes the assessment of the person (Wright 1982). Similarly, it may be that the more pragmatically oriented, point-driven a reading is, the less likely a reader is to accept the information presented in a text at face value.

This could be investigated by having people read an ironically-structured, first-person literary text, then asking them to assess 'the narrator' and 'the author' in various ways. For instance, they could be asked to decide whether or not certain descriptors (e.g., Uses good grammar '; 'Is sexist') are true, false or indeterminate with respect to the author' and 'the narrator' (cf. Stephenson's [1953, in press] 'Q-sort' technique). The degree of similarity between the two sets of ratings could be statistically determined, and inferences then made as to whether the reader recognized two distinct entities, or a single blurred entity.

Experimental poetics

The concept of point-driven understanding appears to be useful in accounting for oral story comprehension, as well as in generating testable hypotheses concerning reading. It should be emphasized that we do not mean the methodological suggestions given above to be definitive, but rather to illustrate some of the ways that techniques from cognitive and experimental psychology could help answer questions in areas that traditionally have concerned only literary scholars. Unlike Updike's Sammy, whose stomach "kind of fell" when he contemplated the future, we are relatively optimistic that the future will bring important developments, both theoretical and empirical, in the study of literary reading.

References

Aaronson, Doris and Steven Ferres. 1984. Reading strategies for children and adults: Some empirical evidence. Journal of Verbal Learning and Verbal Behavior 23: 189-220.

Aaronson, Doris and Hollis S. Scarborough. 1976. Performance theories for sentence coding: Some quantitative evidence. Journal of Experimental Psychology: Human Perception and Performance 2: 56-70.

Beaugrande, Robert de. 1982. The story of grammars and the grammar of stories. Journal of Pragmatics 6: 383-422.

Black, John B. and Gordon H. Bower. 1979. Episodes as chunks in narrative memory. Journal of Verbal Learning and Verbal Behavior 18: 309-318.

Brewer, William F. and Anne E. Hay. 1984. Reconstructive recall of linguistic style. Journal of Verbal Learning and Verbal Behavior 23: 237-249.

Chatman, Seymour. 1978. Story and discourse. Ithaca: Cornell University Press.

van Dijk, Teun A. and Walter Kintsch. 1983. Strategies of discourse comprehension. New York Academic Press.

Dillon, George L. 1980. Discourse processing and the nature of literary narrative. Poetics 9: 163-180.

Fish, Stanley. 1970. Literature in the reader: Affective stylistics. New Literary History 2(1): 123-162.

Fletcher, Charles R. 1981. Short-term memory processes in text comprehension. Journal of Verbal Learning and Verbal Behavior 20: 564-574.

Greene, Graham. 1947. Nineteen stories. London: Heinemann.

Haberlandt, Karl. 1980. Story grammar and reading time of story constituents. Poetics 9: 99-116.

Haberlandt, Karl, Claire Berian and Jennifer Sandson. 1980. The episode schema in story processing. Journal of Verbal Learning and Verbal Behavior 19: 635-650.

Hayes, John R. and Linda S. Flower. 1980. 'Identifying the organization of writing processes'. In: Lee W. Gregg and Erwin R. Steinberg, eds., Cognitive processes in writing. Hillsdale. NJ: Erlbaum.

Hirsch, E. D. 1967. Validity in interpretation. New Haven: Yale University Press.

Hunt, Russell A. 1982. What happens when our students read, and what can we do about it? Paper presented at the Wyoming Conference on Freshman and Sophomore English, Laramie, Wyoming, 30 June.

Iser, Wolfgang. 1978. The act of reading. London: Routledge and Kegan Paul.

John-Steiner, Vera and Paul Tatter. 1983. 'An interactionist model of language development'. In: Bruce Bain, ed.,The sociogenesis of language and human conduct. New York: Plenum.

Keenan, Janice M., Brian MacWhinney and Deborah Mayhew. 1977. Pragmatics in memory: A study of natural conversation. Journal of Verbal Learning and Verbal Behavior 16: 549-560.

Labov, William. 1972. Language in the inner city. Philadelphia: University of Pennsylvania Press.

Levelt, W. J. M. 1974. Formal grammars in linguistics and psycholinguistics. Vol. III: Psycholinguistic applications. The Hague: Mouton.

MacWhinney, Brian, Janice M. Keenan and Peter Reinke. 1982. The role of arousal in memory for conversation. Memory & Cognition 10: 308-317.

Mandler, Jean M. and Marsha S. Goodman. 1982. On the psychological validity of story structure. Journal of Verbal Learning and Verbal Behavior 21: 507-523.

Morgan, Jerry L. and Manfred B. Sellner. 1980. 'Discourse and linguistic theory'. In: R. J. Spiro. B. C. Bruce and W. F. Brewer, eds., Theoretical issues in reading comprehension. Hillsdale. NJ: Erlbaum.

Nabokov, Vladimir. 1962. Pale fire. New York: G. P. Putnam's Sons.

Polanyi, Livia. 1978. The American story: Cultural constraints on the meaning and structure of stories in conversation. Unpublished doctoral dissertation, University of Michigan.

Polanyi, Livia. 1979. So what's the point? Semiotica 25 (3/4): 207-241.

Polanyi, Livia. 1982. Linguistic and social constraints on storytelling. Journal of Pragmatics 6: 509-524.

Pratt, Mary Louise. 1977. Toward a speech act theory of literary discourse. Bloomington: Indiana University Press.

Prince, Gerald. 1983. Narrative pragmatics, message, and point. Poetics 12: 527-536.

Robinson, John A. 1981. Personal narratives reconsidered. Journal of American Folklore 94: 58-85.

Rosch, Eleanor. 1973. 'On the internal structure of perceptual and semantic categories. In: T. E. Moore. ed., Cognitive development and the acquisition of language. New York: Academic Press.

Rosenblatt, Louise M. 1938. Literature as exploration. New York: Appleton-Century.

Rosenblatt, Louise M. 1978. The reader, the text, the poem. Carbondale: Southern Illinois University Press.

Ross, Lee. 1977. 'The intuitive psychologist and his shortcomings: Distortions in the attribution process. In: Leonard Berkowitz, ed., Advances in experimental social psychology. Vol. 10. New York: Academic Press.

Rothkopf, E. Z. and M. J. Billington. 1979. Goal-guided learning from text: Inferring a descriptive processing model from inspection times and eye movements. Journal of Educational Psychology 71: 310-327.

Schank, Roger C., Gregg C. Collins, Ernest Davis, Peter N. Johnson, Steve Lytinen and Brian J. Reiser. 1982. What's the point? Cognitive Science 6: 255-275.

Stephenson, William. 1953. The study of behavior: Q-technique and its methodology. Chicago: University of Chicago Press.

Stephenson, William. In press. Q-methodology and English literature. In: Charles R. Cooper, ed., Researching response to literature and the teaching of literature. Norwood, NJ: Ablex.

Sternberg, Saul. 1966. High-speed scanning in human memory. Science 153: 652-654.

Updike, John. 1962. Pigeon feathers and other stories. New York: Knopf.

Watts, Richard J. 1981. The pragmalinguistic analysis of narrative texts. Tübingen: G. Narr.

Wright, Edward F. 1982. Group discussion effects on the dispositional bias in attitude attributions. Unpublished doctoral dissertation. University of Alberta.

Douglas Vipond received his doctorate in psychology from York University (Toronto). In 1976-77 he did a postdoctoral year at the University of Colorado, working with Walter Kintsch on text comprehension.

Russell A. Hunt completed a doctorate in eighteenth century English literature and criticism at Northwestern University in 1969. He spent 1982-83 on sabbatical leave as research associate at Indiana University, working on language development, pragmatics, semiotics, and reading.

[2] Similarly, reader-response critics such as Stanley Fish (1970) have observed that many reading events take place below the level of consciousness, but nonetheless are 'real' and do have consequences. A distinction can be made (Levelt 1974; Mandler and Goodman 1982) between psychological validity -- phenomena that have demonstrable effects on processing with or without awareness, and psychological reality -- phenomena that we can think and talk about. In this sense strategies are psychologically valid but not necessarily psychologically real.

[3] Actually, "A & P" does not provide an ideal example of this strategy. Even the cash-register-watcher incident is somewhat problematic: although structurally unimportant, it seems to be quite well recalled, possibly due to its imagibility or favorable serial position. Still, our point is not that people reading in a story-driven way will forget this episode but that they will tend not to connect it to anything else. A better example of a text where the use of this strategy might be tested can be found in Hunt (1982). This study reports that in readings of the Graham Greene short story "The Second Death" (1947), in which the salient detail is a caterpillar crawling across a leaf, only a very small number of readers connected the caterpillar with the story's other concerns.

*Authors' address: both authors are at St. Thomas University, Fredericton, New Brunswick, Canada E3B 5G3, Douglas Vipond in the Department of Psychology and Russell A. Hunt in the Department of English. This work was supported by NSERC grant A6401 to D. Vipond, and by SSHRC grant 410-84-0105 to R. A. Hunt.

We thank James A. Reither for insightful comments on an earlier version of this paper.