literary communication: effects of reader-narrator cooperation

26
ELSEVIER Poetics 23 (1996) 405-430 P()ETICS Literary communication: Effects of reader-narrator cooperation Peter Dixon a,*, Marisa Bortolussi b a Department of Psychology, University of Alberta, Edmonton, Alberta, Canada T6G 2E9 b Department of Modern Languages and Comparative Studtes, Unwersity of Alberta, Edmonton, Alberta, Canada T6G 2E6 Abstract We propose that readers process text as if they were communicating with the narrator. This entails, first of all, that readers construct a mental representation of the narrator's knowledge, perspective, and goals, and second, that they cooperate with the narrator by interpreting the characters and events of the described world in a way that makes the narrator's stance rational and justified. It is hypothesized that both of these processes depend on the location that is assigned to the narrator in the described world. We present a model of how the narrator's location is constrained by dialogue style and how the assigned location affects the evaluation of the characters and the perceived attributes of the narrator. The model was tested by systematically manipulating dialogue style in the story 'Rope' by Katherine Ann Porter. Readers' evaluations of the characters and inferences about the narrator closely matched the predictions of the model. I. Introduction An enduring controversy in literary theory is one concerning the nature of communication in literature. As readers, we have a strong intuition that a text 'communicates' some message or intent to us. This intuition permeates much of critical literary scholarship; for example, textual analyses commonly entail infer- ences about the author's intended message or implicit world view. In the present article, we develop a related, but critically different, analysis. Specifically, we propose that readers often process a text as if it were communication, even though there may be only a limited sense in which veridical communication occurs * Corresponding author. Tel.: +1 403-492-2318; Fax: +1 403-492-1768; E-mail: pdixon~ psych.ualberta.ca 0304-422X/96/$15.00 © 1996 Elsevier Science B.V. All rights reserved SSDI 0304-422X(95)00007-0

Upload: peter-dixon

Post on 21-Jun-2016

214 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Literary communication: Effects of reader-narrator cooperation

ELSEVIER Poetics 23 (1996) 405-430

P()ETICS

Literary communication: Effects of reader-narrator cooperation

P e t e r D i x o n a,*, M a r i s a Bor to lu s s i b

a Department of Psychology, University of Alberta, Edmonton, Alberta, Canada T6G 2E9 b Department of Modern Languages and Comparative Studtes, Unwersity of Alberta, Edmonton,

Alberta, Canada T6G 2E6

Abstract

We propose that readers process text as if they were communicating with the narrator. This entails, first of all, that readers construct a mental representation of the narrator's knowledge, perspective, and goals, and second, that they cooperate with the narrator by interpreting the characters and events of the described world in a way that makes the narrator's stance rational and justified. It is hypothesized that both of these processes depend on the location that is assigned to the narrator in the described world. We present a model of how the narrator's location is constrained by dialogue style and how the assigned location affects the evaluation of the characters and the perceived attributes of the narrator. The model was tested by systematically manipulating dialogue style in the story 'Rope' by Katherine Ann Porter. Readers' evaluations of the characters and inferences about the narrator closely matched the predictions of the model.

I. Introduct ion

A n endur ing controversy in literary theory is one concerning the nature of communica t ion in literature. As readers, we have a strong intuition that a text ' communica tes ' some message or intent to us. This intuition permeates much of critical literary scholarship; for example, textual analyses commonly entail infer- ences about the author ' s in tended message or implicit world view. In the present article, we develop a related, but critically different, analysis. Specifically, we propose that readers of ten process a text as i f it were communicat ion, even though there may be only a limited sense in which veridical communica t ion occurs

* Corresponding author. Tel.: +1 403-492-2318; Fax: +1 403-492-1768; E-mail: pdixon~ psych.ualberta.ca

0304-422X/96/$15.00 © 1996 Elsevier Science B.V. All rights reserved SSDI 0 3 0 4 - 4 2 2 X ( 9 5 ) 0 0 0 0 7 - 0

Page 2: Literary communication: Effects of reader-narrator cooperation

406 P. Dixon, M. Bortolussi / Poetics 23 (1996) 405-430

between the historical author and actual readers. This kind of reading strategy has a clear implication: If readers treat text as communication, they must construct a mental representation of the individual with whom they are communicating. Moreover, this individual is not the author, but rather the implied speaker of the words of the text, that is, the narrator. Following from this approach, we present here a theoretical account of one essential aspect of the narrator's representation, namely, the position of the narrator in the described world. The account is supported by the results of an experimental investigation of narratorial technique.

1.1. Literature as communication

Inspired by linguistic theory (e.g., Lyons, 1977), early theoreticians such as Jakobson (1960) attempted to describe literature building on the theoretical framework adopted for communication in conversational discourse. On this ap- proach, the author of a work is conceptualized as the sender of an intended message. This message is coded in the text, and the task for readers is to decode the intended message based on their knowledge of both linguistic and literary conventions and the social and cultural context in which the text is embedded. This framework accounts for why literary works may be poorly received when removed from their original cultural context, and it provides insight into the role of readers' knowledge of literary form and genre. Moreover, the general approach has provided the basis for theoretical developments in narratology (e.g., Lanser, 1981) and literary reception (e.g., Iser, 1974, 1978).

However, this view of literature has a number of fundamental drawbacks. First, because the author and the reader are removed in space and time, they do not share a common perceptual and pragmatic context. They may not even share common linguistic, literary, cultural, or social knowledge. This means that there is only a limited sense in which author and reader can develop a mutual understand- ing of their shared knowledge, or common ground (Stalnaker, 1978), and cannot engage in a variety of processes fundamental to conversational communication, such as collaborating on references (e.g., Clark and Wilkes-Gibbs, 1986) and repairing misunderstandings (e.g., Schegloff et al., 1977). Yet literary works can be successfully appreciated without this common ground: For example, Garcla- M~rquez' One Hundred Years of Solitude became an international best-seller despite the fact that most of its readers knew little of its author or the political and historical situation to which it refers.

Second, it is often unreasonable to ascribe a single, coherent intention to the author of a literary work. To begin with, extended literary works are generally produced over a period of time, and it is unlikely that the complex structure of goals and intentions that motivate the work remain perfectly consistent throughout this process. Moreover, creating literature may involve a certain amount of experimentation, in which an author introduces variations in structure, content, and style without explicit expectations about how those variations will affect readers. More difficult still are works (such as Faulkner's As I Lay Dying) in which there are multiple narrators presenting divergent points of view. Furthermore, one

Page 3: Literary communication: Effects of reader-narrator cooperation

P. Dtxon, M. Bortolussi / Poetics 23 (1996) 405-430 407

cannot overlook the importance of characters' dialogues; they too can represent the author's ideas to differing degrees. However, since characters often contradict each other, the authorial point of view becomes, to use Bakhtin's (1981) terminol- ogy, ' refracted' and virtually impossible to reconstruct. As Bakhtin has argued, to posit an authorial intention as the basis for literary communication is to presup- pose, falsely, that narrative discourse is 'monologic', whereas it is clearly 'dialogic' and polyphonic.

Third, the view that readers decode an intended message from a text trivializes the variability of literary reception. For example, it invites the inference that there is a single, 'correct ' interpretation of a literary work (the intended interpretation) and that other interpretations should be dismissed as inappropriate or invalid. Moreover, some aspects of literary works are inherently ambiguous, and although the author may have had a particular intention in mind, that intention is essentially unknowable and cannot be recovered reliably by readers. Thus, interpretations of literature necessarily must be idiosyncratic to a certain extent, and it is unreason- able to require that readers find a 'correct', intended message in order to process literature appropriately. In sum, authors and readers do not share common ground as required for communication; literary works are unlikely to represent a single coherent message; and readers' interpretations of literature are variable and idiosyncratic.

Scholars have responded to these problems in a variety of ways. For example, it has been suggested that in constructing a text, authors make special allowances for the lack of common ground with readers (e.g., Winograd, 1977). Lanser (1981) has suggested that communication in literature is special and does not follow the rules derived from conversational discourse. Similarly. Schmidt (1980) has argued that there are essential conventions that apply specifically to the processing of litera- ture. Still another approach is to distinguish the communicative level of processing from other, lower levels at which these problems may not apply (e.g., Schram and Steen, 1992). In our view, these various perspectives serve to underscore the fact that literature is not communication in any simple sense and that one cannot assume that the mechanisms of conversational communication apply unaltered to the processing of literary texts. In particular, communication theory by itself cannot reconcile the intuition that texts 'communicate' with the pragmatics of the actual reading process.

1.2. Communication as reading process

Our solution to this dilemna is to propose that readers generally use communi- cation as a reading strategy (cf. Van Dijk and Kintsch, 1983). This approach allows us to build on the concepts and distinctions from linguistic communication theory while minimizing its limitations when applied to literature. In a communicative reading strategy, readers would process the words of the text in much the same way they would if the words were encountered in normal conversation, even though they cannot be veridically communicating with the historical author. This view implies, first of all, that readers are 'communicating' with the implied speaker

Page 4: Literary communication: Effects of reader-narrator cooperation

408 P. Dixon, M. Bortolussi / Poetics 23 (1996) 405-430

of the words of the text. In other words, communication is with the narrator, not the author. This does not preclude the possibility of making other, additional inferences about a text and how it was created. For instance, sophisticated readers may construct a representation of the implied author whose views may or may not be the same as those of the narrator. However, our proposal is that these additional inferences and representations typically do not supplant the more basic process of communicating with the narrator. Processing the narrator's voice as if it were communication is likely to be a natural, automatic process, and consequently it probably forms the basis of more sophisticated inferences about the implied authorial intentions. In the present research, though, we are concerned primarily with the first-order communication with the narrator, rather than with higher-order communication with an implied author. More elaborate theories of literary pro- cessing and interpretation are likely to depend on an adequate understanding of this basic process.

A further implication of a communicative reading strategy is that readers construct a mental representation of the narrator. It is generally argued that in order to communicate, each participant in a conversation must generate a mental representation of the other participant's perspective, knowledge, and goals (Winograd, 1977; Clark and Clark, 1977). For example, in order to answer the question, "Is it snowing outside?" in a sensible way, one must have information about what the speaker knows and why he or she is asking the question. "No, it never snows in the summer" might be a sensible response to a toddler under one set of circumstances, while "Yes, the roads are very slick" might be an appropriate response to an adult under other circumstances. We believe a similar representa- tion of the perspective, knowledge, and goals of the narrator is also required when reading a literary text as if it were communication. Such a representation is also likely to include information about the narrator's attitudes and belief about characters and events in the story. In short, processing the text as communication requires a mental representation of the stance of the narrator. Moreover, because this representation is constructed by readers as a natural part of comprehension, it would be constructed regardless of whether or not it is supported by explicit references to the narrator in the text. In particular, we assume that a representa- tion of the narrator and his or her stance is constructed even for texts in which the narrator is absent and is not explicitly part of the described world.

Linguists such as Grice (1975) have argued that communication in conversation relies on a principle of cooperation: One assumes that the other participant is rational and is providing only necessary and sufficient information given their knowledge, perspective, and goals. We believe that similar cooperation occurs with readers processing the narrator. Readers assume that the narrator is rational and that the narration suffices to understand the story events and the narrator's stance. To cooperate with the narrator in this communication process, then, readers must use their own knowledge and experience to generate 'implicatures', inferences about the described world and the story events that allow the narrative to make sense. In particular, this means that readers will attempt to rationalize and justify the stance of the narrator towards the characters and events of the story. Our view

Page 5: Literary communication: Effects of reader-narrator cooperation

P. Dixon, M. Bortolusst / Poetics 23 (1996) 405-430 409

is that such justification is part of the process of cooperating with the narrator in order to foster communication.

1.3. A model o f narratorial location

A central prediction of this analysis is that the reader's representation of the narrator must include information about the location of the narrator in the described world. In order to communicate information about the described world, the narrator must exist in that world. Moreover, information about the narrator's location is a prerequisite for making a variety of inferences necessary for communi- cation. For example, location determines the perspective from which events are described and is essential for understanding focalisation, changes in perspective, and the knowledge of actions and events that can be attributed to the narrator. Further, the narrator's location could have a central role in determining the scope and constituents of narrative sequences, and movements of the narrator may be correlated with spatial and temporal shifts of the narrative. It seems inevitable, then, that readers depend on the location of the narrator as a natural part of the communication process.

In the present article, we also propose that the location of the narrator plays a central role in the process of characterization: In the absence of information to the contrary, readers assume that the narrator's position is consonant with other aspects of narrator's stance towards characters and events in the story. In particu- lar, if the narrator is spatially proximal or adjacent to a particular character, the narrator will be presumed to share attitudes and beliefs with that character. In part, we believe this mechanism is motivated by the requirements of communicat- ing with the narrator: Because readers need to have a representation of the narrator's attitudes and beliefs, they will use the physical location of the narrator as a cue to those attitudes and beliefs when other, more explicit cues are unavailable. However, we argued above that readers attempt to use their own knowledge and experience to justify attitudes and beliefs of the narrator. Conse- quently, if the narrator shares attitudes and beliefs of an adjacent character in the story, the net result is that the thoughts and actions of that character are likely to be more justified and rationalized in the mind of the reader than those of other characters. In effect, readers may 'identify' with that character, interpret that character's actions in terms of their own experiences, and react to that character in a more sympathetic, positive manner. Thus, the position of the narrator is critical in characterization because it can affect the way in which readers' own experiences are used in understanding characters' thoughts and actions.

It seems likely that this aspect of narratorial processing can be manipulated in order to direct readers' involvement in the story. For example, if the author's intention is served by having readers understand a character's actions in terms of their own knowledge and experience, the author may omit explicit justification for that character's behaviour and instead describe events from the character's per- spective. Our analysis of this kind of text is that the reader is likely to locate the narrator adjacent to the character, infer that the narrator shares attitudes and

Page 6: Literary communication: Effects of reader-narrator cooperation

410 P. Docon, M. Bortolussi / Poettcs 23 (1996) 405-430

beliefs with that character, and then cooperate with the narrator by attempting to infer what circumstances and interpretation would be required for those attitudes and beliefs to be reasonable. Similarly, movements of the narrator may be used to invite certain kinds of inferences by the reader. For example, changing the focalisation of a narrative from one character to that of another may signal to the reader that the narrator has moved to be adjacent to the second character and that the reader should attempt to rationalize and explain the new character's attitudes and beliefs in order to foster communication. In our view, these kinds of processes and effects are mediated by the location the reader assigns to the narrator.

1.4. Narratorial location and dialogue style

Although a variety of text features contribute to the location that is assigned to the narrator, in the present work we focused on dialogue style. We hypothesize that different styles of dialogue place various degrees of constraint on the location readers can be expected to assign to the narrator. For example, consider the dialogue fragments in Table 1. Indirect speech (the first example) places the least constraint on the narrator's position. Although the fragment implies that the narrator knows the content of the utterance, there is no logical requirement that he or she know the exact wording and no requirement that he or she actually heard the utterance. For instance, if this fragment were conveyed during a conversation, one could easily suppose that the conversational participant learned of the utterance second-hand.

Direct speech (the second example) provides an exact transcription of the utterance. In this kind of construction, the voice of the narrator is temporarily suspended in favour of the (transcribed) voice of the speaking character. However, texts generally provide ample cues that it is the narrator who has provided the transcription. For example, the narrator often provides tags or comments about the transcription (e.g., 'he said', 'she argued forcefully'). Thus, direct speech implies that the narrator must have witnessed the utterance in person, and, as a consequence, that the location of the narrator is constrained to be within hearing distance of the speaker.

Table 1 Dialogue styles

Indirect speech He remembered that he 'd forgotten the coffee and said that he would have to go back to town to get it.

Direct speech "Gosh, I've forgotten the coffee," he remembered. "Lord, now I'll have to go back to town to get it."

Free indirect speech Gosh, he 'd forgotten the coffee. Lord, now he'd have to go back to town to get it.

Page 7: Literary communication: Effects of reader-narrator cooperation

P. Dixon. M. Bortolusst / Poetics 23 (1996) 405-430 411

The third example in Table 1 is of free indirect speech. We believe that this style of dialogue places the greatest amount of constraint on the presumed location of the narrator. Although scholars differ in terms of the hypothesized effect of free indirect speech on readers (e.g., McHale, 1978), there seems to be little ambiguity about its fundamental characteristics. In particular, free indirect speech is marked by the presence of two narrative voices in the same sentence: the character's, as indicated by the use of vocabulary and sentence structure that would only be appropriate in transcribed conversation, and the narrator's, as indicated by the use of pronouns and verb tense that would only be appropriate in indirect speech. Our view is that this intermingling of the two voices in the same sentence suggests to the reader that the narrator and the speaking character are close or adjacent in the described world. In a sense, the narrator and the character are so close that one cannot be sure who is speaking. Thus, free indirect speech places the greatest degree of constraint on the location of the narrator of these three dialogue styles.

This analysis of free indirect speech can be illustrated by considering the narratorial technique used in the story 'Rope' by Katherine Ann Porter. The story relates an argument between a husband and wife that ensues when the man returns from town with a large coil of rope. Almost all of the story consists of dialogue, so there are few cues that can be used to locate the narrator in the described world. The reader is forced to rely instead on the way in which the dialogue is related in order to locate the narrator. However, in this text the speech of both characters is conveyed in free indirect speech. According to our analysis, this should constrain the narrator to be located adjacent to each character while he or she is speaking. Initially, the reader may locate the narrator near the female character when she begins speaking in free indirect style, but then move the narrator to the male character when he provides a rejoinder also in free indirect style. Subsequently, the reader would have to move the narrator back to the woman when she speaks again, back to the man when he replies, and so on. We suspect that the net effect of this technique is that the narrator will take on attributes and attitudes of not just one character, but of both. And because of communicative cooperation, the reader will attempt to rationalize and justify the attitudes of each character in turn. This kind of effect on the reader may have been intentionally created by the author as an essential part of the story.

2. Experiment 1

In order to test this analysis of communication and narratorial technique, we conducted two experiments in which we systematically manipulated the dialogue style used in 'Rope ' and observed the concomitant changes in readers' evaluation of the characters and their perception of the narrator. We compared the original version of the story to three modified versions. In one version, the female character's speech was changed to direct, quoted speech, but the speech of the male character was left as in the original. In a second version, the male character's speech was changed to the direct style and the female character was left un-

Page 8: Literary communication: Effects of reader-narrator cooperation

412 P. Dixon, M. Bortolussi / Poetics 23 (1996) 405-430

changed. And in a third version, the speech of both characters was changed to direct speech. According to our model of narratorial location, these different versions should place various degrees of constraint on where readers may be expected to locate the narrator. When both characters' speech is related in the free indirect style (as in the original story), the narrator should be constrained to be adjacent to each character while he or she is speaking; when one character's speech is in free indirect and the other's is in direct, the location of the narrator should be constrained only during the speech of the first character and may shift at other times; and when the speech of both characters is in direct style, the narrator's location should not be constrained to be adjacent to either character.

The effect of these constraints on the location of the narrator and readers' reaction to the characters was assessed after reading the story. Readers were asked a series of questions concerning their attitudes towards the male and female characters and their impression of the narrator. According to our analysis of narratorial location and communicative cooperation, we predicted that readers would be more likely to rationalize and justify the attitudes and actions of a character when the dialogue style constrained the narrator to be adjacent to that character. For example, the evaluation of the male character relative to the female character should be more positive when the speech of the male character is in free indirect and the speech of the female character is in direct. This is because the dialogue style should constrain the narrator to be adjacent to the male character rather than the female character, and as a consequence the reader should be more likely to rationalize and justify the male character's attitudes and beliefs.

Similar predictions follow for the kinds of characteristics readers are likely to attribute to the narrator. If the dialogue style constrains the narrator to be adjacent to one character but not adjacent to the other character, one may expect the narrator to take on attributes of the nearby character. In 'Rope' , a salient characteristic that might be attributed to the narrator is gender. For example, if only the speech of the male character is in free indirect style, we predict that the narrator will be located adjacent to the male character for a significant portion of the story and as a consequence is more likely to be seen as male. The narrator may be less likely to be seen as male when the speech of both characters is in free indirect style (because the narrator is constrained to be adjacent to both charac- ters), and the narrator may be less likely to be seen as male when neither character's speech is in free indirect (because the narrator is not constrained to be near either character).

2.1. Method

2.1.1. Materials In the experiment, subjects were asked to read one of four versions of the story.

In the original, the speech of both the male and female characters was almost entirely in free indirect speech. In the Male Direct version, the speech of the male character was changed to direct speech; in the Female Direct version, the speech of the female character was changed to direct speech; and in the Both Direct

Page 9: Literary communication: Effects of reader-narrator cooperation

P. Dixon, M. Bortolussi / Poetics 23 (1996) 405-430 413

version, all of the dialogue was changed to direct speech. Samples of the original and modified texts are shown in Table 2. Generally, the new versions of the story required minimal changes to the text: The words attributed to a given character in the original were quoted, tags were added to identify the speaker at a number of locations in the text, new paragraphs were inserted to introduce a new direct speaker, and pronouns and verb tenses were changed as appropriate. The story versions were 2939, 2965, 2979, and 2990 words in length for the Original, Female Direct, Male Direct, and Both Direct versions respectively.

2.1.2. Procedure After reading the story, subjects answered a series of questions concerning their

overall evaluation of the story, their perception of the two main characters, and their impression of the narrator. Subjects then completed a questionnaire concern- ing their knowledge about and attitude towards reading and literature. Following the questionnaire, subjects then reread the story and again answered the story questions. There were few systematic differences between the responses after first and second reading and the results reported below are the average of the two.

2.1.3. Measurements The present report focuses on the responses to five questions pertaining to

readers' reactions to the two characters: 'How much did you like the character in the story?', 'Is the character in the story reasonable and rational?', 'Is the character tolerant, sympathetic, and understanding?', 'Does the character feel love and affection for his /her partner?', and 'Is the character responsible for the argument?' These questions were asked of both the male and female character; half of the subjects were asked first about the male character and then about the female character, while the other half were asked about the female character first. Subjects responded to the questions by selecting a number on a nine-point scale labeled with polar opposites appropriate to the question. The responses to the questions were added to form a composite index of readers' evaluation of each of the characters. In addition, readers were asked to rate how strong their impression was that the narrator was a man or a woman, also on a nine-point scale.

2.1.4. Subjects Subjects were undergraduates at the University of Alberta recruited through

announcements in psychology classes. There were 83 men and 79 women in the study, divided into four groups: 39 read the original version (23 men and 16 women), 31 read the Male Direct version (12 men and 19 women), 48 read the Female Direct version (24 men and 24 women), and 44 read the Both Direct version (24 men and 20 women). The experiment was run in small groups of 5-20.

2.2. Character evaluations

Fig. 1 shows a summary of the overall evaluation of the male and female characters as a function of story version. The three story versions constructed for

Page 10: Literary communication: Effects of reader-narrator cooperation

414 P. Dixon, M. Bortolussi / Poetics 23 (1996) 405-430

the study are laid out along the horizontal axis according to the hypothesized constraints on the location of the narrator. The Male Direct version is plotted on the left of the graph. In this case, we believe that the narrator is constrained to be adjacent to the female character for a significant portion of the narration because the free indirect style used for her speech suggests proximity. The direct speech

Table 2 Excerpts from 'Rope'

Original Had he brought the coffee? She had been waiting all day long for

coffee. They had forgot it when they ordered at the store the first day.

Gosh, no, he hadn't. Lord, now he'd have to go back. Yes, he would if it killed him. He thought, though, he had everything else. She reminded him it was only because he didn't drink coffee himself. If he did he would remember it quick enough. Suppose they ran out of cigarettes? Then she saw the rope. What was that for? Well, he thought it might do to hang clothes on, or something. Naturally she asked him if he thought they were going to run a laundry? They already had a fifty-foot line hanging right before his eyes? Why, hadn't he noticed it, really? It was a blot on the landscape to her.

Male direct Had he brought the coffee? She had been waiting all day long for

coffee. They had forgot it when they ordered at the store the first day.

"Gosh, no," he answered, "I didn't. Lord, now I'll have to go back. Yes, I will if it kills me. I thought, though, that I had everything else." She reminded him it was only because he didn't drink coffee himself. If he did he would remember it quick enough. Suppose they ran out of cigarettes? Then she saw the rope. What was that for?

"Well," he answered, "I thought it might do to hang clothes on, or something." Naturally she asked him if be thought they were going to run a laundry? They already had a fifty-foot line hanging right before his eyes? Why, hadn't he noticed it, really? It was a blot on the landscape to her.

Female dtrect "Did you bring the coffee?" she asked him. "I've been waiting

all day long for coffee." They had forgot it when they ordered at the store the first day.

Gosh, no, he hadn't. Lord, now he'd have to go back. Yes, he would if it killed him. He thought, though, he had everything else.

"It's only because you don't drink coffee yourself," she reminded him. "If you did you would remember it quick enough. Suppose we ran out of cigarettes?" Then she saw the rope. "What is that for?" she asked. Well, he thought it might do to hang clothes on, or something.

Page 11: Literary communication: Effects of reader-narrator cooperation

P. Dixon, M. Bortolussi / Poetics 23 (1996) 405-430 415

Table 2 (continued)

Naturally she asked him, "Do you think we're going to run a laundry? We already have a fifty-foot line hanging right before your eyes? Why, didn't you notice it, really? It's a blot on the landscape to me."

Both direct She asked him, "Did you bring the coffee? I've been waiting all

day long for coffee." They had forgot it when they ordered at the store the first day.

"Gosh, no," he answered. "I didn't. Lord, now I'll have to go back. Yes, I will if it kills me. I thought, though, that I had everything else."

"'It's only because you don't drink coffee yourself," she reminded him. " I f you did you would remember it quick enough. Suppose we ran out of cigarettes?" Then she saw the rope. "What is that for?" she asked.

"'Well," he said, "I thought it might do to hang clothes on. or something."

Naturally she asked him, "Do you think we're going to run a laundry? We already have a fifty-foot line hanging right before your eyes. Why, didn't you notice it, really? It's a blot on the landscape to me."

used for the male character in this condition places much less constraint on the location of the narrator. The Female Direct version is plotted on the right of the graph. In this case, the direct speech used for the female character is not constraining, while the free indirect speech used for the male character constrains the narrator to be adjacent to the male character while he is speaking. The Both Direct version is plotted in the middle because the direct speech style does not constrain the narrator to be near the male or female character. The vertical axis shows the mean composite evaluation of the characters, with larger values indicat- ing more positive evaluations. The error bars around each point indicate 95% confidence intervals for the population mean, adjusted for pairwise comparisons. Thus, when error bars for two points fail to overlap one can be reasonably confident that the population means for those conditions are different. In particu- lar, noting that two error bars do not overlap is equivalent to rejecting the hypothesis that the two means are equal with an alpha-level of.05. In the following presentation, we do not report inferential statistics for various contrasts since this information is redundant with and less interesting than the information provided by the figures (cf. Cohen, 1994; Loftus, 1993).

The results indicate that narratorial technique has a substantial effect on the evaluation of the male character. When the male character's speech is direct, the male character is evaluated less positively; when the female character's speech is direct the male character is evaluated more positively. In the version in which both characters' speech is direct, the evaluation is intermediate. The original version is plotted as a separate point in the middle of the graph. Like the Both Direct

Page 12: Literary communication: Effects of reader-narrator cooperation

416 P. Dixon, M. Bortolussi / Poetics 23 (1996) 405-430

15

Male Character

~ g i n a l

Female Character

Original

I I I I I

Male Both Female Male Both Female Direct Direct Direct Direct Direct Direct

Fig. 1. Evaluations of the male and female characters in Experiment 1.

version, the evaluation of the male character with this version was intermediate between the Male Direct and Female Direct versions. A similar, weaker trend is observed for the female character, but in this case the direction of the effects is exactly opposite that observed for the male character. For example, the evaluation is highest when the female character's speech is in free indirect style and the male's is in direct, and the evaluation is lowest when the female's speech is in direct style and the male's is in free indirect. In addition, evaluations of the female character were generally lower than those of the male character. The evaluation of the male and female characters can be summarized succinctly by considering their relative evaluation. Relative evaluation was computed by subtracting the composite evaluation for the female character from that for the male character. Thus, relative evaluation is greater than zero when readers prefer the male character to the female, and is less than zero when they prefer the female character. Relative evaluation, shown in Fig. 2, summarizes the pattern of results found separately for the male and female evaluations shown in Fig. 1: There is an overall preference for the male character that increases from the Male Direct version to the Female Direct version, with the Original and Both Direct versions falling in between.

The general pattern of results is consistent with the proposed model of narrato- rial location and communicative cooperation. We assume that the use of free

Page 13: Literary communication: Effects of reader-narrator cooperation

P. Dixon, M. Bortolussi / Poetics 23 (1996) 405-430

4

417

~ 2

' 1

_= t~

-3

- 4 i i i

Male Direct Both Direct Female Direct Fig. 2. Evaluations of the male character relative to the female character in Experiment 1.

indirect speech for a character invites the reader to rationalize and justify those statements. This assumption follows from the presumed nature of communication in literature and the process of locating the narrator: Free indirect speech constrains the narrator to be located adjacent to the speaking character, and because of this proximity the narrator is presumed to share attitudes and attributes of that character. Further, in order to cooperate with the narrator 's effort at communication, the reader will attempt to justify and explain those attitudes the narrator is presumed to share. The net result is that statements made using free indirect speech will be more justified in the mind of the reader than one would otherwise expect. However, the results also show an unexpected preference for the male character. We suspect that this preference is unrelated to narratorial tech- nique and instead arises from the content of the story. One possibility is that readers find the actions and statements of the male character in the story to be more easily understood and rationalized than those of the female character. For example, it may be easy for readers to believe that the male character's action - bringing home the coil of rope - should not be a problem and that therefore the female character is at fault for initiating the argument. In contrast, sympathizing with the female character requires interpreting the rope as symbolic of more extensive problems in the relationship, and justifying this position may be more difficult for readers because these problems are largely unstated in the text. This

Page 14: Literary communication: Effects of reader-narrator cooperation

418 P. Dixon, M. Bortolussi / Poettcs 23 (1990) 405-430

Table 3 Predictions for relative character evaluation in Experiment 1

Story version Male character Female character Relative evaluation

Male direct Content Free indirect 0 Female d i rec t Content + Free indirect 2 Both direct Content 1 Original Content + Free indirect Free indirect 1

analysis suggests that the male character should be preferred when all other things are equal.

Table 3 provides a detailed account of how these assumptions account for the relative evaluation of the characters shown in Fig. 2. Each entry under the 'Male Character ' and 'Female Character ' headings indicates a basis for justifying the characters ' actions and statements. The entry 'Content ' appears for the male character in each of the four conditions following the argument above; we assume that the statements and actions of the male character are likely to be more readily justified than those of the female character. The entry 'Free indirect' appears in each case where a character 's dialogue is conveyed in free indirect speech; we assume that this dialogue style invites the reader to justify and rationalize that character 's actions. For example, 'F ree indirect' is entered for the female charac- ter in the Male Direct condition because in that condition her speech is conveyed in free indirect style. The last column indicates the predicted relative evaluation of the two characters, calculated by simply subtracting the number of entries for the female character from that for the male character. As can be seen, the pat tern of predicted results is exactly the same as that found in Fig. 2.

2.3. At tr ibutes o f the narrator

Another test of our approach to narratorial location is to assess readers ' perception of the narrator directly. We assume that when the narrator is located adjacent to a character, readers may infer that the narrator shares attitudes and attributes with that character. So in the present study, we asked whether readers perceived the narrator as a man or a woman; when the narrator is located adjacent to the female character readers should tend to see the narrator as a woman and when the narrator is located adjacent to the male character readers should tend to see the narrator as a man. Fig. 3 shows the mean response, graphed in the same way as Fig. 1. In this figure, the responses have been scaled so that 0 indicates a neutral impression of gender, positive values indicate male gender and negative values indicate female gender. Many aspects of these results are consistent with our interpretation of the effect of speech style. In the Original version, the narrator is constrained to be adjacent to both the male and female characters while they are speaking. Thus, the location of the narrator would suggest that the narrator shares gender with both (or neither) character and readers perceive the narrator as neutral, neither clearly male nor female. The contrast between the

Page 15: Literary communication: Effects of reader-narrator cooperation

P. Dixon, M. Bortolusst / Poetics 23 (1996) 405-430

Male 4

419

Z -1

-2

-3

Female -4 I j I

Male Both Female Direct Direct Direct

Fig. 3. Perceived gender of the narrator in Experiment 1.

Male Direct version and the Female Direct version is also clearly consistent with our analysis: The narrator is seen as male in the Female Direct version because the narrator is constrained to be adjacent to the male character while the narrator is less likely to be seen as male in the Male Direct condition because the narrator is constrained to be adjacent to the female character instead.

The one aspect of the results that was unexpected in this interpretation is the overall impression that the narrator is male; even in the Male Direct version, the narrator is perceived as only neutral gender rather than female as might have been expected. In this respect the pattern of results is similar to that found for relative character evaluation. The male character is preferred overall (Fig. 2) and the narrator is seen as male overall (Fig. 3). However, the perceived gender of the narrator is unlike relative character evaluations in the Original version in which both characters' speech is related in free indirect style. For relative evaluations, the male character is preferred in this condition, while for perceived gender, the narrator is seen as neither male nor female. This result suggests that different mechanisms are operating for relative evaluation and perceived gender. We hypothesize that perceived gender is sensitive to weak cues in 'Rope' for the location of the narrator that operate when the narrator is not otherwise con- strained. In particular, near the end of the story, the woman leaves the scene while

Page 16: Literary communication: Effects of reader-narrator cooperation

420 P. Dixon, M. Bortolussi / Poettcs 23 (1996) 405-430

Table 4 Predictions for perceived gender of the narrator in Experiment 1

Story version Male character Female character Perceived gender

Male direct (Scene) Free indirect - ½ Female direct Free indirect 1 Both direct (Scene) ½ Original Free indirect Free indirect 0

the narration continues with the speech and actions of the male character. This aspect of the story may provide some association between the narrator and the male character, even when the male character's speech is not related in free indirect style.

Table 4 summarizes how this assumption, combined with constraints arising from the use of free indirect speech, predict the pattern of gender attributions shown in Fig. 3. In each of the four conditions, an entry is made for a character whenever the story suggests the narrator should be located adjacent to that character. In the Male Direct version, the speech of the female character is in free indirect style, and so the narrator is constrained to be adjacent to the female character while she is speaking. In this condition, the speech of the male character is in direct style, and so there is little constraint on the narrator's position while he is speaking. However, as argued above, the reader may locate the narrator adjacent to the male character near the end of the story while the female character is absent from the scene. The fact that this is a relatively weak tendency is shown in Table 4 by putting the entry 'scene' in parentheses. Similar considerations serve to fill out the balance of the table. In the Female Direct version, the male character's speech is in free indirect style, so 'Free indirect' is entered for the male character in that condition; the fact that the female character leaves the scene near the end of the story is unimportant in this condition because the narrator is already constrained by the dialogue style of the male character to be located nearby. In the Both Direct version, dialogue style does not constrain the location of the narrator and the only hypothesized effect is the weak association between the narrator and the male character encountered at the end of the story. Finally, in the original version, free indirect speech is used throughout and the narrator is constrained to be adjacent to the male character while he is speaking and the female character while she is speaking.

The predicted judgments of gender are shown in the final column. We have simply subtracted the number of entries found for the female character from that for the male character, counting the parenthesized 'scene' entries at half value. Consequently, positive values would count as an overall perception of the narrator as male, while negative values would count as an overall perception of a female narrator. As can be seen, the pattern of predictions is similar to that obtained in Fig. 3: The Male Direct, Both Direct, and Female Direct versions are ordered in terms of increasing certainty that the narrator is male, while the narrator in the

Page 17: Literary communication: Effects of reader-narrator cooperation

P. Dixon, M. Bortolussi / Poetics 23 (1996) 405-430 421

original version is seen as neutral, neither male nor female. The only deviation between the predicted pattern and the obtained pattern is that the narrator in the Male Direct version was predicted to be seen as somewhat female; the obtained result is that the narrator is seen as neutral. However, to anticipate, a comparable manipulation of dialogue style in Experiment 2 produced precisely the predicted pattern of gender judgments. Thus, we are inclined to view the apparent deviation here as experimental error.

3. Experiment 2

In the second experiment, we investigated the unexpected overall preference for the male character in 'Rope' . As is apparent from Figs. 1 and 2, the male character is perceived as more rational, reasonable, and likable than the female character. In our interpretation of these results, we attributed this preference to the content of the story. That is, readers may perceive the actions and statements of the female character to be more argumentative and more difficult to justify than those of the male character. Crucially, this tendency would have to be independent of the inferred location of the narrator, because dialogue style has virtually identical effects on the evaluation of the male and female characters. For example in Fig. 1, the evaluation of the female character shown is almost a precise mirror image of the evaluation of the male character. However, another possibility is that the observed differences between the male and female characters is related to the expectations and biases of the readers rather than story content. For example, readers may expect men to be more rational and reasonable than women, particu- larly in the context of an argument, or they may be more likely to expect the protagonist in a story to be male. These two possible interpretations are tested in Experiment 2.

In order to distinguish these two explanations for the overall preference for the male character, we manipulated the gender of the two story characters. New versions of the story were created, parallel to those used in Experiment 1, in which the male character was changed to a woman and the female character was changed to a man. Thus, in the new versions, the female character returns from town while the male character stays home. On the hypothesis that the male character was evaluated more positively because of readers' biases or expectations, one would expect to find little difference in the evaluations of the characters in these new story versions: Readers should still evaluate the male character more highly, even though the roles of the two characters in the story are reversed. On the other hand, if the preference for the male character is related to the content of the story, one should now find that readers prefer the female character. That is because in the new versions, the female character performs the same actions and utters the same statements that presumably produced the preference for the male character in Experiment 1.

This manipulation of gender also provides a test of our account of the perceived gender of the narrator. In most of the conditions in Experiment 1, there was an

Page 18: Literary communication: Effects of reader-narrator cooperation

422 P. Dixon, M. Bortolussi / Poetics 23 (1996) 405-430

overall tendency to see the narrator as male. We argued that was due to the absence of the female character from the narrative scene for a short period near the end of the story. However, an alternative interpretation is that readers expect narrators to be male in the absence of other information, and that this bias combines with the cues garnered from dialogue style to produce the effects shown in Fig. 3. According to this account, the manipulation of gender in Experiment 2 should have little effect, and there should still be a strong tendency to see the

Table 5 Excerpts from 'Twine'

Original Had she brought the tobacco? He had been waiting all day long for

a cigarette. They had forgot it when they ordered at the store the first day.

Gosh, no, she hadn't. Lord, now she'd have to go back. Yes, she would if it killed her. She thought, though, she had everything else. He reminded her it was only because she didn't smoke herself. If she did she would remember it quick enough. Suppose they ran out of coffee? Then he saw the twine. What was that for? Well, she thought it might come in handy, maybe for hanging plants or for tying up the bushes in the fall. Naturally he asked her if she thought they were going to open a plant nursery? They already had a large spool of rope sitting on the back porch. Why, hadn't she noticed it, really? It was a pain in the neck to him.

Male direct "Did you bring the tobacco?" he asked her. "I've been waiting

all day long for a cigarette." They had forgot it when they ordered at the store the first day.

Gosh, no, she hadn't. Lord, now she'd have to go back. Yes, she would if it killed her. She thought, though, she had everything else.

"You know, it's only because you don't smoke yourself," he chided. "If you did you would remember it quick enough. Suppose we ran out of coffee?" Then he saw the twine. "What is that for?" he asked. Well, she thought it might come in handy, maybe for hanging plants or for tying up the bushes in the fall.

He asked her, "Do you think we're going to open a plant nursery? We already have a large spool of rope sitting on the back porch. You mean you've never noticed it? It's a real pain in the neck to me."

Female direct He asked her if she had brought the tobacco? He had been waiting

all day long for a cigarette. They had forgot it when they ordered at the store the first day.

"Gosh, no," she answered. "I didn't. Lord, now I'll have to go back. Yes, I will if it kills me. I think, though, that I have everything else." He reminded her it was only because she didn't smoke herself. If she did she would remember it quick enough. Suppose they ran out of coffee? Then he saw the twine. What was that for?

Page 19: Literary communication: Effects of reader-narrator cooperation

P. Dixon, M. Bortolussi / Poetics 23 (1996) 405-430 423

Table 5 (continued)

"Well," she answered, "I thought it might come in handy, maybe for hanging plants or for tying up the bushes in the fall." Naturally he asked her if she thought they were going to open a plant nursery? They already had a large spool of rope sitting on the back porch. Why, hadn't she noticed it, really? It was a pain in the neck to him.

Both dtrect "Did you bring the tobacco?" he asked her. "I 've been waiting

all day long for a cigarette." They had forgot it when they ordered at the store the first day.

"Gosh, no, I didn't," she answered. "Lord, now I'll have to go back. Yes, I will if it kills me. I think, though, I got everything else."

"It 's only because you don't smoke yourself," he reminded her. " I f you did you would remember it quick enough. Suppose we ran out of coffee?" Then he saw the twine. "What ' s that for?" he asked.

"Well," she replied, "I thought it might come in handy, maybe for hanging plants or for tying up the bushes in the fall."

Naturally he asked her, "Do you think we are going to open a plant nursery? We already have a large spool of rope sitting on the back porch. Have you really not noticed it? It's a pain in the neck to me."

narrator as male. However, if this tendency is caused instead by the short absence of the female character from the narrative, the perceived gender of the narrator in Experiment 2 should be reversed. With the gender of the characters switched, it is the male character who would be absent from the scene, and this should produce a concomitant tendency to see the narrator as female. Further, on this account, the reversal of the character's genders should have little effect on the original version in which the speech of both characters is in free indirect style. We argued previously that the absence of one character from the scene has little effect in this condition because the location of the narrator is already constrained by the use of free indirect speech for both characters. Thus, in Experiment 2 we expect that the perceived gender of the narrator should be neutral in this condition, just as in Experiment 1.

3.1. Method

3.1.1. Materials All of the story versions used in this experiment were based on a new version of

'Rope ' which we entitled 'Twine'. The new story was virtually identical to the original except that a man was used in place of the female character and a woman was used in place of the male character. A few additional changes were made to keep the male and female actions and roles from seeming unusual or atypical. For example, in the original 'Rope' , the argument ensues when the man returns from

Page 20: Literary communication: Effects of reader-narrator cooperation

424 P. Dixon, hi. Bortolussi / Poetics 23 (1996) 405-430

town with a coil of rope; in 'Twine', the argument is precipitated by the woman bringing home a large ball of twine. However, every effort was made to retain the language and essential content of the original story.

Four versions of 'Twine' were created that paralleled the four versions of 'Rope' in Experiment 1. The first version was similar to the original version of 'Rope' in that the speech of both characters was related in the free indirect style. For consistency with the nomenclature adopted in Experiment 1, we refer to this version as the 'original' version of 'Twine'. Three additional versions were created like those used in Experiment 1: In the Male Direct version, the speech of the male character was changed to direct speech; in the Female Direct version, the speech of the female character was changed to direct speech; and in the Both Direct version, all of the dialogue was related as direct speech. As in Experiment 1, these additional versions incorporated minor variations in vocabulary and typography. Excerpts from the four versions of 'Twine' are shown in Table 5. The stories were 3111, 3132, 3134, and 3137 words in length for the Original, Male Direct, Female Direct, and Both Direct versions respectively.

3.1.2. Procedure In general, the procedure was identical to that used in Experiment 1. Readers

read one of the four versions of the story, answered the same questions used in Experiment 1 pertaining to the story and the characters, filled out a questionnaire on their experience and knowledge pertaining to literature, and then reread the story and answered the questions again. The only essential difference in the two experiments was the nature of the story. Composite character evaluations were computed for the male and female character as in Experiment 1.

3.1.3. Subjects Fifty-one women and 39 men from introductory psychology courses participated

in the study for course credit. Subjects were assigned randomly to condition, resulting in 9 women and 13 men in the Male Direct version, 13 women and 11 men in the Female Direct version, 18 women and 6 men in the Both Direct version, and 11 women and 9 men in the Original version.

3.2. Character evaluations

The composite character evaluations (shown in Fig. 4) display exactly the same pattern of results found in Experiment 1. In particular, when the narratorial technique constrains the narrator to be adjacent to a particular character, that character is evaluated more positively than when the narrator may be more distant. This result obtains for both the male and female character, just as in Experiment 1. However, what is different in the present results is the overall evaluation of the male and female characters, independent of speech style. This is shown clearly by the relative evaluations shown in Fig. 5. In Experiment 1, the male character was preferred overall and the relative evaluations were all positive; in the present experiment, this preference is reversed, the female character is evaluated more

Page 21: Literary communication: Effects of reader-narrator cooperation

P. Dixon, M. Bortolussi / Poetics 23 (1996) 405-430 425

O

~5 "d

Male Character

Original

T I

Female Character

g" inal

I I I l I I

Male Both Female Male Both Female Direct Direct Direct Direct Direct Direct

Fig. 4. Evaluations of the male and female characters in Experiment 2.

highly, and the relative evaluations are negative. This would seem to be conclusive evidence that the content of the story, rather than a priori expectations and biases, lead the reader to prefer one character over the other.

Predictions based on our model of narratorial location are shown in Table 6. The model predicts these results in exactly the same way as those of Experiment 1 (cf. Table 3), except that in this case, the content of the story leads to an overall preference for the female relative to the male character. In fact, if Figs. 2 and 5 were replotted in terms of story roles rather than gender roles, they would look virtually identical. In both experiments, readers overwhelmingly prefer the charac- ter that goes to town and returns with the rope or twine over the character that stays at home, regardless of whether the character is male or female.

3.3. Attributes of the narrator

The perceived gender of the narrator is shown in Fig. 6. In many respects the pattern shown in this figure is comparable to that shown in Fig. 3 for Experiment 1. When the dialogue style constrains the narrator to be adjacent to a character, the narrator is more likely to be seen as sharing the gender of that character. For example, Fig. 6 shows that the narrator is more likely to be seen as male in the

Page 22: Literary communication: Effects of reader-narrator cooperation

426 P. Dixon, M. Bortolussi / Poetics 23 (1996) 405-430

4

~ 2

' 1

° ~

-3

-4

Original ~ / ~

I I I

Male Direct Both Direct Female Direct

Fig. 5. Evaluations of the male character relative to the female character in Experiment 2.

Female Direct version (in which the narrator is constrained to be adjacent to the male character) than in the Male Direct version (in which the narrator is con- strained to be adjacent to the female character). The same qualitative pat tern of results was found in Experiment 1. However, when the actual responses are compared, it is clear that the two experiments produced a large and systematic difference in perceived gender of the narrator: The narrator in 'Rope ' (Experi- ment 1) was generally perceived as male, while the narrator in 'Twine ' (Experi- ment 2) was generally perceived as female. On the other hand, the variation of gender roles across experiments had little effect on the perceived gender of the narrator when the dialogue of both characters was in free indirect style. In the original version of 'Rope ' , the narrator was seen as neither strongly male nor

Table 6 Predictions for relative character evaluation in Experiment 2

Story version Male character Female character Relative evaluation

Male direct Content + Free indirect - 2 Female direct Free indirect Content 0 Both direct Content - 1 Original Free indirect Content + Free indirect - 1

Page 23: Literary communication: Effects of reader-narrator cooperation

P. Dixon, M. Bortolussi / Poetics 23 (1996) 405-430

Male 4

427

-~ 0

Z -1

-2

-3

Female -4 I I I

Male Both Female Direct Direct Direct

Fig. 6. Perceived gender of the narrator in Experiment 2.

strongly female, despite the fact that in the other modified versions the narrator was likely to be seen as male. The same result holds for the original version of 'Twine': This pattern of results is exactly what would be predicted by the narrato- rial location model: The use of free indirect speech for both characters constrains the narrator to be adjacent to a character while he or she is speaking and the narrator is not strongly associated with either character. However, when a charac- ter's speech is in direct style, the narrator's location is relatively unconstrained and may be affected by other, weaker cues in the story.

These predictions are detailed in Table 7. As for the predictions for Experiment 1 (Table 4), an entry is made for each character when his or her speech is in free indirect style; in these conditions, the narrator is constrained to be adjacent to that

Table 7 Predictions for perceived gender of the narrator in Experiment 2

Story version Male character Female character Perceived gender

Male direct Free indirect - 1 Female direct Free indirect (Scene) ½ Both direct (Scene) - ½ Original Free indirect Free indirect 0

Page 24: Literary communication: Effects of reader-narrator cooperation

428 P. Dixon, M. Bortolussi / Poetics 23 (1996) 405-430

character and there may be a tendency to attribute the gender of that character to the narrator. When a character's speech is related in direct style, the narrator's location is unconstrained and may be affected by other aspects of the narration. The difference between these predictions and those made for Experiment 1 is that in 'Twine' it is the male character who leaves the scene briefly rather than the female character. The result is that in this experiment the weak scene cues favour the female character rather than the male character, and overall readers are expected to see the narrator as female. The predicted pattern of results (shown in the last column of Table 7) match the obtained results precisely.

4. General discussion

The results of these studies provide support for our view of the communicative reading strategy and the role of dialogue style in constraining the location of the narrator. When readers treat text as communication, they of necessity must construct a representation of the narrator that includes his or her location in the described world. Dialogue style constrains the location assigned to the narrator: With direct speech the narrator is constrained only to be within hearing distance, while with free indirect speech the narrator is constrained to be adjacent to the speaking character. The link between the location of the narrator and readers' interpretation of the story depends on communicative cooperation. In the absence of information to the contrary, readers attribute attitudes and beliefs of adjacent characters to the narrator, and then cooperate with the narrator by interpreting the described world so as to make those attitudes and beliefs rational and sensible. Thus, when free indirect speech style constrains the location of the narrator to be adjacent to the speaking character, the net result is that the attitudes and beliefs of that character are likely to be seen as more justified and understandable than would otherwise be the case.

This theoretical approach provides a detailed account of the effect of manipu- lating dialogue style in the two experiments reported here. In both experiments, a character is evaluated more positively when his or her speech is conveyed in free indirect style and the other character is conveyed in direct dialogue style. Our interpretation of this result is that the free indirect dialogue style provides a cue that the narrator is adjacent to the speaking character, and by virtue of that proximity, the attitudes and beliefs of that character become more rationalized and justified in the mind of the reader. Experiments 1 and 2 together show that this result is obtained for both characters in the story and is obtained regardless of the character's gender. The pattern of relative evaluations of the two characters in both experiments can be predicted precisely by assuming that the location of the narrator determines the narrator's attitudes, and that the evaluation of a character depends on both narrator's stance towards that character as well as the content of the story.

The experiments also confirm our predictions about how the location assigned to the narrator affects readers' interpretation of the narrator. We assume that in

Page 25: Literary communication: Effects of reader-narrator cooperation

P. Dixon, M. Bortolussi / Poetics 23 (1996) 405-430 429

the absence of other information, the narrator is assumed to be similar to the adjacent character. In 'Rope ' there are few explicit cues as to the attributes of the narrator. Thus, readers infer that the narrator is male when the narrator is constrained to be adjacent to the male character and female when the narrator is constrained to be adjacent to the female character. When the narrator is con- strained to be adjacent to both characters at different times, the reader interprets the narrator as neither clearly male nor female. Thus, the results on the narrator's perceived gender provide converging evidence for our model of narratorial loca- tion and communicative cooperation.

We believe that our approach represents an advance over standard scholarship on narratology in that we are concerned primarily with the way in which the narrator is processed by the reader, rather than with systematic descriptions of features of the text associated with narrative technique. In particular, we suggest that effects of narratorial technique are mediated by the nature and content of readers' representation of the narrator, and that this representation is determined primarily by the requirements of the communicative reading strategy. In the present research, we investigated the effect of one element of narratorial tech- nique, dialogue style, on one aspect of the narrator's representation, location. Clearly, further research is required to understand more fully how readers process the narrator and to determine precisely how other elements of narratorial tech- nique affect this process. However, we believe that viewing the representation of the narrator as an essential part of a communicative reading strategy provides a crucial insight into the role of the narrator in literary processing.

The results of this research also provide interesting insight into the technique used in the original 'Rope' . The use of free indirect speech for both characters provides a paradoxical cue for the position of the narrator: The technique intertwines the voice of the narrator with the voices of both characters in the story and suggests that the narrator is proximal to each character in succession. Accord- ing to our analysis, this technique invites readers to cooperate with each character in turn and to see his or her side of the argument. Moreover, the constant shifting of narratorial location may produce some confusion in the mind of the reader as to what attitudes and beliefs should be attributed to the narrator and what stance they should be attempting to rationalize overall. In both cases, the net effect of this particular narratorial technique may be to undermine the tendency to see the argument in black-and-white terms in which one character is right and the other wrong. Thus, the reader may be led to appreciate the larger circumstances and societal conventions that constrain the interpersonal roles these two individuals must adopt. It seems likely that the title of the story is a metaphor for these constraints.

References

Bakhtin, M.M., 1981. The dialogic imagination. Austin, TX: University of Texas Press. Clark, Herbert H. and Eve V. Clark, 1977. Psychology and language. New York: Harcourt Brace

Jovanovich.

Page 26: Literary communication: Effects of reader-narrator cooperation

430 P. Dixon, M. Bortolussi / Poetics 23 (1996) 405-430

Clark, Herbert H. and D. Wilkes-Gibbs, 1986. Referring as a collaborative process. Cognition 22, 1-39. Cohen, Jacob, 1994. The earth is round (p < 0.05). American Psychologist 49, 997-1003. Coste, Didier, 1989. Narrative as communication. Minneapolis, MN: University of Minnesota Press. Grice, H.P., 1975. Logic and conversation. In: P. Cole and J.L. Morgan (eds.), Studies in syntax, Vol. 3,

41-58. New York: Seminar. Iser, Wolfgang, 1974. The implied reader: Patterns of communication in prose fiction from Bunyan to

Beckett. Baltimore, MD: John Hopkins University Press. Iser, Wolfgang, 1978. The act of reading: A theory of aesthetic response. Baltimore, MD: John Hopkins

University Press. Jakobson, Roman, 1960. Closing statements. In: T.A. Sebeok, Style and language. New York: Technol-

ogy Press of Massachusetts Institute of Technology and John Wiley and Sons. Lanser, Susan, 1981. The narrative act: Point of view in prose fiction. Princeton, NJ: Princeton

University Press. Loftus, Geoffrey R., 1993. A picture is worth a thousand p values: On the irrelevance of hypothesis

testing in the microcomputer age. Behavior Research Methods, Instruments, and Computers 25, 250-256.

Lyons, John, 1977. Semantics. Cambridge: Cambridge University Press. McHale, B., 1978. Free indirect discourse: A survey of recent accounts. PTL: A Journal for Descriptive

Poetics and Theory of Literature 3, 249-287. Stalnaker, R.C., 1978. Assertion. In: P. Cole (ed.), Syntax and semantics, Vol. 9: Pragmatics, 315-332.

New York: Academic. Schegloff, E.A., G. Jefferson and H. Sacks, 1977. The preference for self-correction in the organization

of repair in conversation. Language 53, 361-382. Schmidt, S.J., 1980. Grundriss der Empirischen Literaturwissenschaft I. Der gesellschaftliche Hand-

lungsbereich Literatur. Braunschweig-Wiesbaden: Vieweg. [Reprinted 1991, Suhrkamp, Frankfurt a.M.]

Schram, Dick H. and Gerard J. Steen, 1992. 'But what /s literature?' A programmatic answer from the empirical study of literature. SPIEL 11,239-258.

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

Winograd, Terry, 1977. A framework for understanding discourse. In: Marcel Adam Just and Patricia A. Carpenter, Cognitive processes in comprehension, 63-88. Hillsdale, NJ: Erlbaum.