Toward a cognitive theory of literary character: the dynamics of mental-model construction

Ralf Schneider

Miss Bronte was struck by the force or peculiarity of the character of some one she knew; she studied it, and analyzed it with subtle power; and having traced it to its germ, she took that germ as the nucleus of an imaginary character, and worked outwards;--thus reversing the process of analyzation, and unconsciously reproducing the same external development.

Elizabeth Gaskell on Bronte's Shirley in The Life of Charlotte Bronte (1857)

1. A Cognitive Perspective on Literary Character

Mrs. Gaskell's statement about Charlotte Bronte's method of creating characters hints at the double nature of literary characters: on the one hand, they are based on real-life experiences with living persons; on the other, they are the result of processes of literary construction. (1) Whereas Gaskell looks to the author's contribution to construction, my aim is to look at literary character from the point of view of readers and to elucidate what effects this doubleness has on their experience of encountering characters in fiction. (2) It may be a truism to say that the reading of literary texts is a process in which textual information interacts with the reader's knowledge structures and cognitive procedures. (3) But in literary text-analysis the constraints on literary understanding that arise from the interactive nature of the reading process are rarely acknowledged. Whereas a number of theorists from Iser through to Perry and Phelan have paid attention to the dynamic aspects of narrative, such attention is by no means the rule, and categories for text analysis still tend to highlight the nondynamic, structural side. For the analysis of literary character, there exist some categories that at least show an awareness of the dynamics of reception. In a famous distinction between flat and round characters, implying such awareness, E. M. Forster defines flat characters as those who "are easily recognized," whereas round characters are "capable of surprising" (Aspects of the Novel 74, 81); for the experience of recognition and surprise, the reader must previously have established mental representations and expectations. Other categories, such as the well-known differentiation between static and dynamic characters, fail to account for these dynamics: to decide whether a character is static or dynamic, the reader would have to wait until he or she has read the whole book, since changes in the character's traits may occur late in the story. Of course, readers start forming impressions of characters from the very beginni ng of the reading process on, and from a reader-oriented point of view, the question is not whether a character is static or dynamic, but, rather, when and under which conditions he or she appears static or dynamic to the reader.

Even the categories aware of the temporal dimension of understanding seldom offer any detailed description of how the dynamic processing strategies of the reader interact with the successive presentation of information in the text. Drawing on results from cognitive psychology and cognitive social psychology, as well as from research in discourse processing, I attempt to capture the quality of this interaction more precisely than the more text-oriented, structuralist approaches have been able to do. My theory-building is similar to that of Richard Gerrig in his Experiencing Narrative Worlds (1993), in that my method attempts to align psychological models of the workings of cognition and emotion in text understanding with the description of textual properties. (4) In such an alignment, the interaction between reader and text appears, above all, as a dynamic process, for the framework of cognitive psychology affords a view not only on such general constraints on information processing and text-understanding as l imitations on working memory, but also on the interaction of bottom-up and top-down processing in using inference and forming hypotheses, activating schemas, and constructing categories. (5) More specifically, in my model, understanding literary characters requires our forming some kind of mental representation of them, attributing dispositions and motivations to them, understanding and explaining their actions, forming expectations about what they will do next and why, and, of course, reacting emotionally to them. All this happens through a complex interaction of what the text says about the characters and of what the reader knows about the world in general, specifically about people and, yet more specifically, about "people" in literature.

In contending that dynamic reading effects of character-reception can be explained more adequately if we describe literary characters from the reader's point of view and with the terminology of cognitive psychology, I follow a proposal made by Richard Gerrig and David Allbritton in an article published in Style in 1990. Their brief discussion of the general cognitive constraints of character-reception, however, requires considerable modification and elaboration if we want to arrive at a more detailed understanding of character-reception and at more adequate categories for the analysis of characters in novels, In a first step, I will therefore explicate my proposal to conceive of literary character as a mental model that the reader construes in the reading process through a combination of information from textual and mental sources. In this process of character construction, mechanisms of social cognition also play a crucial role, even though special conditions prevail due to the various textual sources of inf ormation on characters. In a second step, I will survey results from the study of discourse processing and social cognition, including emotional response, and will use them to describe the parameters of character-reception. In a third step, I will focus on the process rather than the structure of mental-model construction to examine dynamic reading effects. I will propose a flow-chart that tries to capture the most important dynamic stages of character-reception, and I will provide a set of categories for reader-oriented character analysis. While in my outline of a theory I will obviously need to make a number of simplifications, I intend neither to diminish the fullness of nor to restrict the range of experiences involved in understanding characters in narrative worlds. Instead, my theoretical model assumes that for a more refined theory of literary character as well as for empirical testing of such a theory we need first to formulate a heuristic framework that integrates results from empirical research--alb eit tentative--with narratological text analysis. Such a framework ought to be compatible with other observations, made in literary theory and empirical psychology, on the cognitive reality of narrative. My attempts to reformulate categories for character analysis in a way that is more sensitive to the mechanisms, dynamics and constraints of information processing in reading can therefore be understood to contribute to the development of a general "cognitive narratology." (6)

2. Literary Characters as Mental Models

The theory of mental models maintains that people construct some kind of holistic mental representation of their experience of the world and that, in tasks such as problem solving, such representations can provide a guideline for the operations of the mental apparatus. In this conception of how meaning is constructed, the approach of Johnson-Laird to mental models has been particularly influential. The notion of mental models has been applied to the description of text-understanding in various ways, (7) and many have become convinced that the construction of a mental model of the fictional world (cf. the terms "situation model" in van Dijk and Kintsch; "text world model" in de Beaugrande) involves a level of text-understanding deeper than that of the construction of hierarchical and cyclical propositional networks as described in the influential theory of Kintsch and van Dijk. Indeed, some claim that mental models capture "what the text is about, not the text itself" (Glenberg, Meyer and Lindem, "Mental Model s" 81), and that textual information serves as an instruction to construct mental models.

The theory of mental models offers new ways of looking at a whole range of aspects of text-understanding, and in such models the various dimensions of time, space, causation, intentionality, and agent have been widely researched empirically. While the mental constructions of fictional space and script-like or scenario-like situations with simple structures of goal-attainment or cause-and-event have received most attention, (8) I contend that our understanding of literary character can profit immensely from concentrating on the nature of characters in such mental representations. According to Sanford and Garrod, scenarios and scripts provide the mere framework in which character models are situated. They express this idea in terms of their differentiation between explicit and implicit representational focus:

At any point in reading, explicit focus contains tokens corresponding to the relevant individuals introduced into that world, whereas implicit focus contains a scenario or mental representation of the currently relevant aspects of the scenes portrayed, including the significance of each of the roles. (479)

Due to limitations of working memory, readers must be able to construct sub-models at various levels of representation, depending on the aspect of the fictional world on which they focus their attention. (9) Given that empirical evidence suggests that readers do not automatically construct elaborate spatial models (Zwaan and Radvansky 178), that in fact "protagonists and objects form the meat" on the backbone of situation models (173), and that literary works frequently describe characters in less eventful situations than the short texts used (and produced) for empirical testing, I would venture the hypothesis that readers of novels focus their attention predominantly on psychological traits, emotions, and aims of characters that are more abstract and less dependent on the immediate circumstantial conditions of individual situations. I would argue, then, that readers also construct mental models Of characters that serve to integrate all pieces of information on a character currently in explicit focus. Conside r, for instance, the way Adam Bede is introduced in George Eliot's eponymous novel of 1859:

Such a voice could only come from a broad chest, and the broad chest belonged to a large-boned muscular man nearly six feet high, with a back so flat and a head so well poised that when he drew himself up to take a more distant survey of his work, he had the air of a soldier standing at ease. The sleeve rolled up above the elbow showed an arm that was likely to win the prize for feats of strength; yet the long supple hand, with its broad fingertips, looked ready for works of skill. [...] The face was large and roughly hewn, and when in repose had no other beauty than such as belongs to an expression of good-humoured honest intelligence. (50)

At the moment a person is mentioned, described in terms of a social role, or referred to by a name or a personal pronoun, the reader must establish a mental token that remains in working memory as long as the text provides information on this entity, or, indeed, as long as the reader chooses to think about it. After that, depending on the stage of memory to which it has been relegated, it can be reactivated later for subsequent updating. (10) The example shows that character models must be multimodal, because information on characters can refer to all possible aspects of human life and present concrete physical traits (cf. Adam's height and stature), traits that are between concrete and abstract (cf. the hand that "looked ready for works of skill") as well as purely abstract ones, such as intellectual or emotional disposition (cf. the "good humoured honest intelligence"). It is difficult to describe the exact nature of representation in a mental model, for this question is linked with the controversial discus sion on mental imagery, which has too long a tradition to be subsumed here. (11) We can assume that mental representations such as mental models encompass both visually concrete and abstract information that may be digital or analogous (Garnham, Mental Models 28). Whenever triggered by the repeated use of the name, description, or pronoun, the character model will be reactivated and subjected to new information processing. Character models can be regarded as sub-models of the overall text-world model, and individual aspects of the character model, such as character traits, visual appearance, or single utterances, can come into representational focus.

3. Structural Components of Mental-Model Construction of Characters

Information from various sources, both textual and reader-centered, feed into the construction of mental character models. Text-understanding always combines top-down processing, in which the reader's pre-stored knowledge structures are directly activated to incorporate new items of information, and bottom up-processing, in which bits of textual information are kept in working memory separately and integrated into an overall representation at a later point in time. Top-down and bottom-up processing continually interact in the reading process on all levels: from the decoding of the graphic signs to the understanding of words, sentence structure and the contents of longer sections (see Lesgold and Perfetti; Sanford and Garrod 14-37; Adams; Kintsch). The differentiation between top-down and bottom-up processing is of major importance also for character-reception, as will be demonstrated below.

On the textual side, all direct or indirect sources of characterizing information can lead to the integration of new aspects into the model or to the modification of existing ones: (1) descriptions and presentations of a character's traits, verbal and nonverbal behavior, outer appearance, physiognomy and body language made by the narrator, by the character him- or herself or by other characters; (2) the presentation of consciousness and a character's mind-style; and (3) inferred character traits mapped metonymically from the presentation of fictional space to the character (see Margolin, "Characterization" and "Introducing"; RimmonKenan 59-70). On the reader's side, practically everything he or she knows about the world can be used in reception. As many have argued, including Tannen and Mandler, cognitive psychology has long since pointed out that knowledge is not found in the brain as a loose assembly of individual bits of information, but is stored in meaningful structures that arise from the individual's c ontact with the world. The organism constructs such structures either as categories according to the similarity of items, or as schemas (or frames or scripts) in accordance with the contiguity of the information encountered. (12) From the vast area of stored information about the world, schemas and categories situated in the domain of social and literary knowledge are of special relevance to character understanding. (13)

In social psychology, the tradition of cognitive research has described how social interaction leads to the formation of categorical and schematic structures of knowledge that create stability and reliability in dealing with others. Such structures allow us to understand situations and to attribute dispositions to others, but they may also create social stereotypes that can have negative effects on social life (see, for example, Cantor and Mischel; Wyer and Gordon). Every society, or group within a society, has a set of assumptions about human behavior that meet with a high degree of agreement and may lead to social stereotypes. Such assumptions function, in David Schneider's terms, as "implicit personality theories" in categorization and attribution processes. Personality theories provide both knowledge for efficient top-down processing and labels for the designation of person types or psychological dispositions. It must be emphasized that personality theories usually combine and popularize knowledge from va rious specialized discourses, that they entail not only descriptions but also evaluations of human behavior as socially acceptable or unacceptable, and that they are just as much subject to change as any other area of common knowledge in society. (14)

Readers can apply personality theories to the understanding of literary characters if they find that a character's traits agree with those of their social knowledge structures. (15) Furthermore, as Black, Galambos, and Read suggest, they can understand a fictional situation quickly and effortlessly if it resembles a stereotypical social scene or "script."

Literary knowledge is another important source for character-reception, because readers learn to form genre expectations both from their contact with books and in a formal literary education that is part of the overall socialization process. (16) Of course, literary and social knowledge can be intertwined, because in each period of literary history novels either reflect socially accepted structures of knowledge or respond to them critically. Readers who receive a training in literary analysis will acquire additional schemata that influence their understanding of texts. It has been pointed out by both literary theorists and empirical researchers that differences in the education of audiences make for different readings. Borrowing terms from Rabinowitz, Phelan, for instance, distinguishes "narrative audiences" from "authorial audiences." Narrative audiences concentrate more on the content level, the "mimetic" qualities of the novel, whereas authorial audiences are used to paying attention also to the "craftsman ship" of the literary work. A number of empirical studies have corroborated this difference between expert and nonexpert readers. Research shows that experienced readers and those who have received special training are able to activate schemata that help them describe the structural and linguistic construction principles of a text without much effort, and that they find it easier to make sense of apparently incoherent information (see, for instance, Graves and Frederiksen; Dijkstra; Peskin). We will return to the differences between models of character constructed by different types of readers later, but at this point we can assume that whereas the more readers have been trained to analyze novels in terms of structure and style, the more they will be able to activate such specific literary knowledge in top-down fashion also in the construction of a character model, nonexpert readers will tend to rely more on structures of social knowledge. But whether focused on the content level or enriched by an understandi ng of the techniques of character presentation, readers learn to expect certain characters with certain traits and functions to appear in certain types of texts.

Beside the more strictly knowledge-related structures, emotions play a crucial role in text-understanding, and especially in character-reception. (17) As many have shown, including Frijda and Lazarus, emotions are an irritatingly complex phenomenon that a number of different research traditions have tried to handle. In literary criticism, the term "identification" has been used dominantly to describe the way readers react emotionally to a character. I would like to suggest that instead of identification, the more adequate term "empathy" be used in a cognitive paradigm of literary studies to describe emotional reactions towards a characters' situations. As Zillman tells us, empathy results from the capacity of the reader to feel for the character because he or she can imagine a situation and its possible outcomes, anticipate what this must mean for the character, and at the same time evaluate this outcome as desirable or undesirable. Unlike identification, empathy does not require readers to share, or want to share, any number of traits with the character, nor does it require them to give up the position of an observer. (18)

Tan suggests that emotions occurring in the process of reception of fictional artefacts be differentiated into fiction-based emotions (f-emotions), which arise from the recipient's response to the occurrences in the fictional world, and artefact-based emotions (a-emotions), which arise from the recipient's response to the aesthetic qualities of a work of art. (19) F-emotions in character-reception are triggered when a character is in a situation the probable outcome of which the reader anticipates or by an already completed episode with whose outcome the reader is faced. In both cases, the quality of emotional response depends, in Zillman's view, on the previous formation of a positive or negative affective disposition towards the character or on what Gerrig calls a reader's "hopes and preferences" (69-77). If the reader has developed a liking for the character and the probable outcome of the situation is negative, the resulting empathic emotion is likely to be fear--consider, for instance, the situation of t he young Jane Eyre, in the first two chapters, where, in the household of the Reeds, she is continually ostracized, humiliated, and maltreated. If a positive outcome seems probable, the reader may feel hope--as is likely to be the case when Jane is allowed to leave the Reed family to go to school, and later, after her suffering at Lowood, when she decides to apply for a post as a governess. If a situation has already had negative effects on the character, the most likely emotion is pity, especially if the character is helpless against the forces of circumstance--a response that readers will be able to activate again and again when confronted with Jane Eyre's fate. Similar to pity, but perhaps requiring less empathic involvement, disappointment can result from a negative outcome of a situation--such as the realization that Jane's move to Lowood does not end her misery. Joy may result from a situation that has turned out positively--while, for some readers, the fact that Jane Eyre ends in a marriage will have s uch an effect, others will rejoice more in Jane's financial, physical, and moral superiority over Rochester at the end of the novel. Anger results from a situation that proves to be good for a disliked character, or from the negative effects of this character's actions on a likable character--e.g., Rochester's lie about his first marriage to Bertha Mason--or both.

The above description of fear, hope, pity, disappointment, joy, and anger schematizes the possible range of emotional response only very roughly, and it neglects the crucial question of how the reader establishes a positive or negative disposition towards the character in the first place. (20) In reading a novel, three major sources can contribute to establish a character as likable or unlikable: first, and most fundamentally, there is the reader's own value system, which allows him to pass moral judgments on the actions portrayed in the novel. This activity should not be seen as arbitrary or completely unrestricted, for when authors write they keep in mind a potential audience about whose "beliefs, knowledge and familiarity with conventions" they make assumptions. Readers, at the same time, will try to engage with the value systems, beliefs, and conventions put forth in a book and thus join this "authorial audience," at least for the duration of reading, even if their private, individual value systems as "ac tual audience" may differ from those of the "authorial audience." (21) One can argue that nineteenth-century authors were very much aware of their contemporary readers' standards of evaluation and were usually able to create immediate emotional involvement, for, according to Rabinowitz. this type of realist novel does not pose a significant distance between the actual and the "authorial audience," on the one hand, and, on the other, the "narrative audience" (i.e., the amount of "pretense" a reader has to invest in order to engage with the "reality" of a fictional world that may be at a distance from the reality of the actual audience). In portraying a character, authors will, if they want to achieve a certain disposition towards that character, try not to deviate too much from the standards of evaluation they expect their readers to apply. Later readers will of course have to make more effort to join the "authorial audience," but also contemporary readers may refuse to join the "authorial audience." If, say, nineteenth-century readers thought young Jane's rebellion against Aunt Reed impertinent, in spite of the apparent injustice Jane suffers, their empathic emotions towards Jane would have tended towards a negative disposition, and whether this reader's mental model of Jane would have maintained a negative stance throughout would have depended on whether the "primacy effect" of the first negative impression was superseded by a "recency effect," i.e., pity evoked by the negative implications of the ensuing events.

The second factor influencing the likability of a character and consequently the reader's willingness to empathize are the narrator's evaluative comments. If the narrator and a particular character are very close, i.e., if they share attitudes and beliefs or if they are situated in spatial and temporal vicinity to each other, readers may tend to empathize with that character more than with others. (22) Historically, the overt, heterodiegetic narrators of the nineteenth-century novel were traditionally invested with a high degree of credibility. (23) Because readers have little reason to mistrust the statements of this agency, they are likely to follow a narrator's invitation to empathize with a character, especially if, in addition, it presents the reader with the character's thought processes. But the narrator's stance can also create an emotional distance, frequently achieved by narratorial irony or a lack of insight into the character's thoughts or feelings, that may prevent empathic involvement. In Thacke ray's Pendennis of 1848-50, for instance, the narrator refuses to elaborate on young Pen's despair when his unhappy love relationship with Ms Fotheringay has ended: "As for Pen, he thought he should die. We are not going to describe his feelings or give a dreary journal of his despair and passion. Have not other gentlemen been balked in love besides Mr. Pen? Yes, indeed: but few die of the malady" (152). Even though Pen's emotional state is explicitly addressed here, the irony and the fact that there is no internal focalization at this point make it unlikely that the reader spends much effort empathizing with Pen in this situation. It is the effectiveness of such interventions by the narrator--engaging or distancing--that highlights a major difference between understanding characters in fiction and forming impressions in social cognition. In contrast to interpersonal understanding, the specific literariness of forming impressions of characters results from an additional stage of the narrator's mediation that is hardly ever unbiased. (24)

The third source of positive and negative disposition towards a character lies in other characters' judgments. These may work toward the same effects as the narrator's. Whether a reader accepts or rejects the innertextual evaluations depends on the reliability of the agency that utters them and on the status that the agency occupies in the innerfictional hierarchy of value systems. If a character passes a judgment on another character, it will have its effect only if the reader does not respond with dislike or suspicion to the character uttering the comment. The negative characterizations of Jane uttered by Aunt Reed and the headmaster of Lowood, Mr. Brocklehurst, are less likely to be credited by the reader than the comments made by her classmate Helen Burns and the teacher Miss Temple. Because the individual value systems of readers in a given period will always to some extent incorporate the common personality theories of that group, a certain amount of agreement on evaluation can be expected. Summing up, we can say that, in developing affective dispositions toward a character, readers always have to negotiate between their own frames of evaluation and any number of textual ones.

Theories of foregrounding, first formulated by the Russian Formalists and adapted to the empirical framework by Willie van Peer, David S. Miall, and Don Kuiken, can account for aesthetic response to cultural artifacts (Tan's a-emotions). Affect in response to the stylistic qualities of a literary text seems to arise from elements in the text that are foregrounded, i.e., show striking deviations from expectable patterns on the phonetic, grammatical, or semantic level. Readers' response to foregrounding lies, first, in a defamliarization--the slowing down of information processing, and second, in some kind of re-familiarization--attempts to integrate the foregrounded elements into a meaningful, coherent understanding of the text (see van Peer; Miall, "Anticipation"; Miall and Kuiken). Although empirical evidence suggests that readers both with and without training in literary analysis tend to describe the same passages of texts as foregrounded (Miall and Kuiken, "Foregrounding"), it is plausible to assume that it depends on a reader's previous experience with the construction of such higher-level structures of meaning whether response to aesthetic foregrounding results in enjoyment or displeasure. Support for this assumption comes from Berlyne's psychobiological approach to aesthetics. (25) According to this theory, too great an amount of new, unexpected, complex, or incongruent stimuli leads to an excess in physiological arousal of certain areas of the brain, and this excess is experienced as unpleasant. The organism experiences as pleasant only a moderate rise of arousal or a decreasing arousal level after an arousal peak. Aesthetic enjoyment is therefore crucially linked to the individual perceiver's level of tolerance for new, unexpected, and complex stimuli, in other words, to the reader's ability to resolve foregrounding and complexity in a way satisfactory to him or her. As mentioned above, experienced and expert readers encounter less difficulty in processing complex information. That aesthetic judgments of professional readers are therefore likely to address quite different phenomena from those that nonprofessionals pay attention to (cf. Dijkstra) will also influence their appreciation of a text's aesthetic qualities. Since personal aesthetic taste is even more subjective than social evaluation patterns of human behavior, this area seems to elude empirical testing more than other reading mechanisms. But as far as character-reception is concerned, we may at least establish two hypotheses. First, stylistic foregrounding in descriptions of characters adds to the complexity of the mental model of character, so that the reader is required to search for ways of making the style of the description fit in with the contents of the model. Likewise, foregrounding in character description can be used to heighten the reader's interest in that character and to separate a character model from others. Second, on the content level we may expect that characters whose traits and actions are too unfamiliar to the reader will inst igate as little aesthetic pleasure as those characters who simply repeat social or literary stereotypes and therefore present few arousal stimuli.

To sum up then, mental models of literary characters are able to integrate abstract and concrete information that proceeds from the reader's structures of social and literary knowledge, on the one hand, and from various textual sources, on the other. Emotional response to literary characters consists of empathic imagination of a character's situation and the meaning of its outcome for the character, and empathy depends on the previous evaluation of the character by readers according to their individual value system or their following innertextual evaluation cues. Finally, readers may pass aesthetic judgments on the presentation of characters. Figure 1 (adapted from Meutsch, "Mental Models" 324) presents, for easy survey, the constituent elements of mental-model construction of literary characters.

4. The Dynamics of Mental-Model Construction in Character-Reception and Categories for Character Analysis

Once a character model is established, it is continually updated to incorporate the latest information, as is the case with all mental representations in text comprehension (Collins, Brown, and Larkin, "Inference" 387; Glenberg, Meyer, and Lindem; Zwaan and Radvansky). Mental character models are therefore dynamic in that they adapt to new input of information. Richard Gerrig and David Allbritton have used a model that Marilynn Brewer proposed for the description of the dynamic social-cognitive processes of impression formation. According to this model, impression formation can take place either in the top-down or the bottom-up mode, depending on whether the subject is able to assimilate the target person into a structure of social knowledge stored in long-term memory. If this is the case (top-down processing), person perception takes place in the form of "categorization." If further information on the target person is encountered that requires a modification of the impression, a process of "individuation" ta kes place. If no social category is available or if the subject is especially interested in aspects of the target person other than category membership, impression formation proceeds bottom up and is called "personalization" (see Brewer). On the one hand, categorization guarantees efficient information processing, since it works automatically, top down, and does not demand too much of working memory's capacity. On the other, personalization requires conscious and careful observation of detail as well as successive integration of information, and it presupposes more tolerance for contradictory information. Person perception is thus subject to general constraints on information processing and follows a tendency to parsimonious use of processing capacities. Gerrig and Allbritton claim that these constraints are at work also in character-reception.

According to Gerrig and Allbritton, "the reader's act of constructing a literary character is initially one of trying to assimilate the character to some well-known category" ("Construction" 386). That is, the reader tries to apply top-down categorization as a preference rule. If this strategy fails, the reader establishes a "person-based" representation. Modifying the terminology used by Brewer and Gerrig and Allbritton, we can speak of "categorization" and "individuation," "decategorization" and "personalization" as distinct strategies of character-reception (see fig. 2). Expressed in terms of mental-model construction, this means that in categorization readers try to establish a holistic mental model of the character early on, one in which, at that point, they integrate all information available from text and memory. The model will possess a number of well-defined features from which expectations, hypotheses, and inferences as well as explanations concerning that character's behavior can be generated. If t he reader is unable (or unwilling) to categorize, the mental model of the character will be less specified and leave more space for input of additional information of all sorts. Few or no hypotheses and inferences can be drawn from such a personalized character model in its early stages. This distinction between categorized and personalized character models seems to be compatible with a distinction that, in Reading People, James Phelan makes between characters in which a "thematic" function dominates versus characters with a predominantly "mimetic" function. In his close readings of a number of narratives and in his analyses of other critical views on these, Phelan demonstrates that it is never the reader nor the text alone that "make" a character mimetic or thematic, but always a combination of the two. It is therefore worth inquiring further into the conditions--both textual and mental--of different constructions of character. Quite fundamentally, the first information presented about a character must be un derstood to be of prime importance for the dynamics of character-reception, because further inferences and hypotheses will be guided by the set-up of the initial model.

We need to specify further the assumption that readers choose a categorization strategy if they are able to activate a structure of knowledge that provides satisfactory explanations of a character, for there are different types of categorization. In accordance with the reader-related sources of mental-model construction mentioned above, we can distinguish between "social" and "literary categorization." In addition, a third strategy of categorization, one best termed "text-specific categorization," is frequently suggested by textual information.

4.1. Categorization, Individuation, Decategorization

A literary character can be introduced into the fictional world by explicit reference to a personality theory or social category. Textual cues for social categorization will therefore mainly be noun phrases naming professions and social roles, such as "the teacher," "the curate," or "the widow." Even less direct descriptions can trigger social categorization, if the reader finds that a character's traits match with those of a personality theory or social stereotype already available. If a nineteenth-century reader learned about a female character that she possessed "angelical sweetness and kindness," that "to love and to pray were the main occupations of this dear woman's life," and that she "was never tired of hearing the praises of her son" (all traits given Mrs. Pendennis in Thackeray's Pendennis (26)), the current gender model and the social role of motherhood must have been the most easily accessible structures of knowledge for mental-model construction. A special case of social categorization occurs if characters are modeled on real-life individuals whom the reader can recognize. Thus, for instance, the three sons of the archdeacon Dr Grantly in Anthony Trollope's The Warden (1855) are satirical copies of three bishops of three important dioceses in Trollope's time, whose personal idiosyncrasies or political actions would have made them famous to contemporary readers. The coincidence in the first names of the boys and the bishops serves as recognition cue (chapter 8).

"Literary categorization" arises when the reader recognizes in a character features of a literary stock character or when he or she can activate a genre schema with character slots to be filled. Thus, in a novel of development (or Bildungsroman), readers can expect at least one character to develop from childhood to adolescence and maturity in a number of stages, and at least one other character who provides counsel and guidance, namely the mentor figure. Readers do not have to rely solely on literary knowledge acquired in formal education, they can of course form their own literary categories of characters in reading out of school, if they encounter several characters in which they detect similarities. A special case of literary categorization can occur when a character is connected to a specific character in another literary work by way of intertextual reference. If someone has read Elizabeth Gaskell's North and South (1854-55), and then reads David Lodge's Nice Work (1988), they will be able to expect a lo ve relationship to develop against all odds between Vic Wilcox and Robyn Penrose and that Robyn will eventually give financial support to Vic. The expectation that Vic and Robyn might get married or enter a lasting relationship, however, will be disappointed.

The reader's categorization tendency may be successful even if he or she is not able to activate pre-stored social or literary knowledge. In many narrative texts, characters are introduced in a way suggesting that their dispositions and behavior are invariable, from which text-specific categorization can result. Reliable cues for such categorizations are words like "always," "never" and other expressions indicating habits or stable dispositions (see also the example of Mrs. Pendennis, above). Frequently, this kind of introduction is made by the narrator. If a character is introduced by the narrator stating that "[v]anity was the beginning and the end of Sir Walter Elliot's character" (in Austen's Persuasion, p. 36), the reader is not likely to expect anything but behavior supporting this trait. If, as is indeed the case in the pages following this characterization, the character's actions confirm the initial information, the reader's categorization tendency may turn into a fixedly categorized character model. Since text-specific categorization initially involves more bottom-up processing than the other categorization types, it may take the reader somewhat longer to form the model. Readers from a later period than the text's origin will have to rely heavily on text-specific information unless they have acquired socio-historical knowledge that turns them into well-informed readers who are able to activate quasi-social categorizations quickly.

Sometimes characters provide hints to the categorization of another character. In such a case, it is important that the categorizing characters stand higher in the reader's regard than the categorized character for the utterance to be credited and the mentioned trait to be integrated into the character model. Consider, for instance, how in Anne Bronte's The Tenant of Wildfell Hall (1848) the categorization judgments about Helen Graham are made by the community of the country-town (36-39) or, in a similar case, how the comments on young Maggie Tulliver are made by the Dodson sisters in book 1, chapter 7, of George Eliot's The Mill on the Floss (1860). In both cases, the categorizations prove to be mistaken and unjust, and will probably have no strong influence on the reader's conceptions of Helen and Maggie, because the characters uttering them are portrayed as narrow-minded, nosy and gossipy. In Bronte's novel, the other important source of information is the homodiegetic narrator Gilbert Markham, who is hims elf quite unable to make up his mind about this mysterious woman, so that the reader will have to rely on his/ her own judgment and observe Helen's actions and speeches more attentively. In The Mill on the Floss, not only does the heterodiegetic narrator clearly side with Maggie; but Maggie is also the dominant focalizer, so that it is easy for the reader to empathize with her and accept her value system rather than that of the Dodsons. In contrast, in Elizabeth Gaskell's Ruth (1853) Mr. Bradshaw is described in negative terms by Faith and Thurston Benson in a way that is likely to trigger categorization: "I tremble at the thought of his grim displeasure," and "He is so severe, so inflexible" (125). Since the Bensons have demonstrated their altruism and charity before these remarks, they are likely to range high in the estimation of the readers, and their categorization judgments are likely to be credited and used for the construction of the mental model of Bradshaw.

The presentation of Mr. Thornton in Gaskell's North and South is a more complicated case, and it shows that novels can thematize mechanisms of social cognition, especially impression formation blurred by social prejudice. In that novel, the protagonist and main focalizer, Margaret Hale, strongly disapproves of the new professions that had emerged with the industrialization of England, especially in the North. The characterization of the factory owner Thornton is at first based on Margaret's negative and simplified opinion of some professions: "I like all people whose occupations have to do with land; I like soldiers and sailors, and the three learned professions, as they call them. I'm sure you don't want me to admire butchers and bakers and candle-stick makers [...]?" (50). When Thornton and Margaret meet for the first time, the reader is presented with the internal focalization of both characters alternately, and their impressions of each other remain undecided between positive and negative (99ff.). What is more, his utterances do not provide information congruent with the category of factory owner. Although after the meeting Margaret confirms her prejudices by calling him a "tradesman" and "not quite a gentleman, but that was hardly to be expected" (102), Thornton displays positive traits that may lead the reader to question her categorization and deprecation of him. When directly after this situation Thornton undertakes to improve the living condition of the Hales--unknown to Margaret, but known to the reader--his actions contradict Margaret's negative characterizations and require readers to negotiate between the evaluative framework provided by Margaret and their own evaluation tendencies. Historically, readers belonging to the same social class as Margaret are likely to have followed her categorization of Thornton, especially since he characterizes himself as a supporter of the doctrines of self-help and laissez-faire associated with the trading classes (chapter 15). When, however, Thornton turns away from such value systems towards the end of the novel (chapter 50), those readers would either have had to decategorize him or to disregard the category-incongruent information. In contrast, readers unfamiliar with the evaluation tendencies towards the new industrial classes among the traditional professions, or those who were and are unwilling to follow Margaret's lead in forming an impression of Thornton, would have constructed a mental model of Thornton that was not prematurely fixed and would allow for the inclusion of category-incongruent information. Those readers would probably have appreciated the fact that when Thornton appears a second time in chapter 9, there are no further categorization cues for him, but the more for his mother, who is presented as a strong-willed, inflexible utilitarian tradeswoman and a Dissenter, and whose harshness and pride will elicit negative dispositions in most readers. Here, even those readers prejudiced against the trading classes would have been able to fix their negative attitudes on the mental model of Mrs. Thornton, so that the model of her son would appear in positive contrast.

The separation of categorization types is often less clear than the above description suggests, and there will be borderline cases: if a social stereotype is reproduced so often in literary texts that it becomes a literary stereotype as well, is the resulting categorization social or literary? Another example is characters who take names and traits from characters in the Bible, as is frequently the case in the Victorian novel. Is such a figure a literary category because it draws on the Bible as a textual source, or is it a social category because it is derived from a personality theory that has simply borrowed the name of the biblical prototype for labeling a specific sort of behavior? The distinction between categorization types is perhaps less important than the effects of categorization on mental-model construction, for these effects--quick completion of the mental model early on in the reading process, automatic and efficient inferencing, stable expectations of dispositions and behavior, and the impressi on of the character's explicability and reliability--all these tend to be the same regardless which type is activated.

There is one kind of literary categorization that will lead to less specific expectations and may, paradoxically, even support personalization cues: If a character is introduced as a potential protagonist in a love plot (Mr. Thornton in North and South) or in a Bildungsroman (Maggie Tulliver in The Mill on the Floss, Pen in Pendennis), this may evoke a number of expectations about the character's future actions in the reader, so that we may speak of literary categorization. At the same time, such recognition of literary structures may activate precisely the reader's interest in the character and make him expect rather unforeseeable, contradictory behavior, for which personalized processing would be more adequate. This paradox can be explained by conceiving of plot structures as relatively vague and abstract schemata that involve a number of functional categories to be filled by the characters. A character may thus fill in a certain slot in a plot schema, without the mental model of that character automaticall y being filled with specific information on a larger number of traits. Perhaps the continuing success of plots of love and of development in fiction and film is due to the fact that such structures present a comfortably familiar framework for engaging with literary characters, without however restricting the possible range of the characters' actions, thoughts and emotions too narrowly.

It is important to keep in mind that social categorization is likely to come easiest to most nonprofessional readers of novels, and that along with the knowledge structure, social categories activate evaluations that are transferred from the category to each member of that group (Brewer, "Dual Process Model" 19). By picking up a well-established mechanism of social categorization, authors need therefore use only a very small number of hints to achieve a certain disposition towards the character in the reader, (27) but in some cases that disposition may be evoked only to be proved unjustified later on. To lead readers to a categorized reception of a character and then urge them to execute a decategorization can be used as a strategy to hint at the dangers imminent in prejudiced social interaction.

Depending on the reader's knowledge, categorization effects can be increased if social, literary and text-specific categorization act in combination and accumulate their respective categorization potential. By giving cues for text-specific categorization on top of social and literary hints, some authors seem to want to make sure that even those readers who do not have the required social or literary knowledge at their disposal can arrive at a categorization of a character. In Dickens's Hard Times (1854), the notorious Mr. Bounderby, a caricature of the Victorian self-help man, not only professes the ideological maxims of self-help and political economy, but in addition is endowed with some idiosyncratic features that allow text-specific categorization. The quality and quantity of information presented about a character will support categorization tendencies. It is likely that readers feel they have learned enough about a character if expository information about that character is concentrated, rather than dis tributed over the text, if the information is not too extensive to be remembered nor contradictory according to common standards of social knowledge, and if the first descriptions or presentations of verbal or nonverbal behavior confirm categorization judgments.

As long as the text keeps. presenting more information about a categorized character without disturbing the initial set-up of the model, and without leading to a shift of focus away from the characteristic features, the mental model will be elaborated. But if subsequent information requires the reader to change some important aspects of the model, though leaving the initial category membership intact, the mental model undergoes some degree of modification and enters into a stage of individuation. Since a character who does not surmount the initial set of features at all would be experienced as boring--unless he or she is a comic character--, (28) individuation seems to be the norm rather than the exception. What is important in an individuated character model is that the original category membership is not given up.

An effect on the reader's model construction that is more significant than individuation is achieved if he or she encounters information that stands in direct opposition to the defining characteristics of the category, so that he must enter a process of decategorization. The impact of decategorization on the reader is considerable, because it is generally the case that information about events or people that does not match expectations leads to a rise in the level of awareness--a process that can be assumed to function in real life as well as literature (cf. Wyer and Gordon, "Cognitive Representation" 125; Black, Galambos and Read; Rimmon-Kenan 38). In addition to this deautomatization of perception, the reader has to revise his current model of the character completely and enter into a new process of model construction. This kind of model revision invalidates most previous implicit inferences, and the failure of model construction draws the reader's attention to the very construction process. It is of course the reader's prerogative to decide whether he or she wants to revise a model or not, but once textual information has started to contradict the categorization initially established, the reader will either have to disregard that information or try to adapt his model by extending the possibilities of category membership. The further presentation of information will then influence what sort of model the reader constructs next. At the end of Dickens's Hard Times, Mr. Gradgrind, the utilitarian, is forced into understanding that his system of education has produced nothing but misery for himself and his family. Gradgrind's complete renunciation of utilitarian values can give the reader occasion to decategorize his model of Gradgrind, but since he does not feature prominently in the remaining chapters, it is an open question whether a decategorization is really carried out and whether a new way of model-construction is chosen. As we have seen in the example of Mr. Thornton in North and South, social evaluation ten dencies may bias the reader's acceptance of category-incongruent information.

4.2 Personalization

The above considerations have all referred to the reader's tendency to use top-down processing in character-reception. The way of model construction we have termed personalization (bottom-up processing, which pays more attention to individual bits of incoming information), is by no means less important than categorization. One can even say that it is responsible for the more differentiated, more interesting and more effective cognitive and emotional responses in character-reception. In simple terms, personalization can occur whenever the reader does not categorize a character, i.e., when he or she is not able or willing to apply stored structures of knowledge for ad hoc impression formation. Even in that case the mental apparatus cannot entirely do without recourse to top-down processing, but according to Gerrig and Allbritton, the structures of knowledge that come into play are "specific recollections of the properties of specific individuals" ("Construction" 387) rather than abstract properties of whole gro ups of persons. Bottom-up processing requires conscious attention to incoming information and is therefore likely to consume more working-memory capacity than top-down processing.

The mere lack of knowledge for plausible categorization in the reader will not automatically give rise to personalization. Some textual strategies must support bottom-up processing. Thus, categorization as the preferred strategy of reception can be blocked by the distribution of expository information over longer stretches of text, so that there is never quite enough information available for fitting the character into a category. Authors often achieve this by introducing characters in action without previous narratorial commentary, by engaging them in dialogue on their first appearance, by having them described in contrasting terms by different other characters or by presenting the complex workings of a character's consciousness.

In the conceptions of both Brewer and Gerrig and Allbritton, emotional involvement with the target person or character is important for impression formation to be performed as personalization. In literature, characters whose traits do not allow easy categorization are in fact often presented in a way that also engages the reader's empathy from the beginning: a character may appear for the first time as being socially ostracized or otherwise emotionally isolated, suffering emotional or even physical distress (e.g., Maggie Tulliver in The Mill on the Floss, Sissy Jupe in Hard Times, the young Jane Eyre). If we keep in mind that personalization generally involves heightened awareness, it seems that this combination of cognitive attention and emotional involvement makes for extremely rich reading effects. If such characters are the main focalizer of a novel, their conceptions, dispositions, emotions, and motivations are easily accessible, and the reader's willingness to empathize is rewarded with much material fr om which to develop understanding and tolerance for the character. Personalized characters may, of course, also be members of a category, but characterization in their case is usually not focused on that fact, nor is category membership introduced at the beginning, and it appears as merely one attribute among many others.

It is possible that a reader overlooks categorization cues, tries for a personalization, and may nevertheless not find enough interesting information for a personalized mental model. We can then assume that some kind of depersonalization, takes place, after which Further categorizing information is sought. Because searching for information necessary to determine a more plausible processing strategy as well as processes of decategorization and depersonalization has a very disconcerting effect on the reader, it is therefore necessary to introduce phases of temporary openness and local instability into the description of mental-model construction, a phase that is not posited in either Brewer or Gerrig and Allbritton. If the text does not offer further categorization cues or any other information on that character after the unstable phase, the effect will surely be that readers will wish to rearrange their structures of knowledge in order to enlarge their power of assimilation for further processing of informatio n. The dynamics of mental-model construction in character-reception can be visualized in a flow-chart (fig. 2) according to the above explication.

One problematic aspect of the differentiation between personalized and categorized character models that has been left untackled so far is the question to what extent the training of readers in literary theory may influence their processes of character reception. Some theories tend to teach readers to look for categories of characters. Psychoanalytic theory provides search strategies that help readers detect when characters exhibit, for instance, an Oedipus Complex, and readers versed in feminist criticism or postcolonial theory may look out for the marginalized, the subaltern, or the subversive characters from the start. Whether aware or unaware, if readers look at a novel from the view-point of such theories, they apply top-down knowledge and structures of evaluation that may give their reading a quality different from that of other readers. Other theories, in contrast, may suggest a stance towards characters that supports forming personalized impressions. For example, in deconstruction a major reading stra tegy is precisely the search for contradictions and incongruities. It is an open question whether in their first readings of a novel even critics and theorists might follow nonacademic and nonprofessional reading strategies and apply their theories only afterwards, when they produce a published or publishable reading that makes sense within the discourse of the branch of criticism they represent. It would indeed be worth tracing how, in a number of critical readings of a single novel made from various critical and theoretical viewpoints, the modes of character reception are chosen to fit the critical framework. Perhaps in teaching literary theory, we must inevitably teach students to deviate considerably from their cognitive and emotional preferences in reading.

5. Conclusions

It is doubly unfortunate that, on the one hand, dynamic processes in reading have been neglected in narrative theory, and that, on the other, empirical research on the cognitive processes of text understanding has neglected character, for we can assume that most "common," i.e., nonacademic, readers focus their interest in the fictional world on the characters rather than on, for instance, fictional time or space or narrative situations. The above considerations have attempted to come to terms with the complexity of structures and processes in only this restricted area of literary understanding. The result was a set of hypotheses: Mental models of literary characters are complex in that they gather information about a character from many sources; they are flexible in that individual aspects can be shifted into representational focus; and they are dynamic in that they are successively refined--elaborated, modified or revised--in the reading process. Moreover, among character models, we may distinguish between t wo different types, the personalized and the categorized. These hypotheses ought to be empirically testable, since they are founded on other models and empirical results from cognitive psychology. At the same time, they can be used for more differentiated character analysis in literary studies. To analyze a character as personalized or categorized, and to capture the possible dynamic reading effects that ensue from these modes of model construction, requires considering the whole potential range of input sources in all imaginable combinations: textual cues, availability of readers' structures of knowledge, and their affective disposition towards the character.

The notion of different modes of character reception also sheds new light on the analysis of characterization in novels. Whereas it is possible to differentiate between direct characterization, in which character traits are explicitly named, and indirect characterization, in which character traits need to be deduced by the reader from the character's actions and appearance (see Rimmon-Kenan, chapter 5), this differentiation makes sense only if it pays attention to the kind of model a reader has constructed of the character. If a character model has been built as categorized, so that a set of expectations has been formed, and readers are then confronted later in the text with an indirect representation of a category-congruent trait, they will be able to process that information effortlessly in spite of its indirectness. If traits are presented that contradict the categorization, they are more likely to attract the reader's attention and suggest conscious decategorization if presented directly, since indirect p resentation is more easily disregarded or reinterpreted to fit the categorization. For personalized character models, in contrast, there will be more acceptance of incongruent character traits presented indirectly, and direct descriptions, especially those made by other characters, are less likely to be taken at face value uncritically.

The above remarks have tacitly implied that the cognitive theory of literary character ought to be applicable to any reader of any novel, even if I have applied it only to nineteenth-century English novels. To literary historians, the reconstruction of the structures of knowledge and habitual strategies of evaluation available to historical readers presents, of course, a formidable challenge. But if by a recontextualization of readers in cultural history we can single out discourses that were valid for certain groups, it should be possible to reach at least an approximation to historical reading processes, especially when the social structure and educational background of the reading public to whom a novel is addressed is fairly homogeneous. (29) When audiences grow increasingly varied and when social structures of knowledge lose general acceptance, this approximation will accordingly be more difficult. In each case the quality and quantity of parameters under scrutiny will determine how successful applicatio ns are,

The theory and graphs presented here simplify a number of the complexities of novel-reading. First, it is of course not enough to look at model constructions of individual characters, because the reactions of a reader to a novel will always involve the perception of characters in relation to each other. If readers use their processing strategies efficiently, it is plausible to suspect that they will try to group models of characters who share traits, especially if they appear in quick succession. At the same time, the isolation of characters who share no traits at all with any of the other activated models can be foregrounded against grouped characters. Second, I assume that readers do not form well-defined mental models of all characters, nor do they normally maintain each model throughout the reading process. It may well be that they establish models for every character only in the first pages or chapters and then decide to elaborate only on those that promise to be most interesting or on whom most informat ion is presented. Third, if categorization and personalization are dependent on such variable parameters as knowledge and a variety of textual cues, there may be strong and weak forms of both modes of construction. The more strongly one mode has been established, the greater the surprise, awareness, and effort if decategorization or depersonalization becomes necessary. Fourth, although there is a clear tendency for main characters to effect personalization and for minor characters to effect categorization, (30) this tendency is not necessarily a rule. Fifth, we need to inquire further into the nature of mental models of narrators. Whereas we may assume that narrator models are similar to character models in homodiegetic and autodiegetic narratives, this assumption is less secure in hetereodiegetic narrative situations. Although it is plausible to assume that there is a rule that readers prefer to form such a character-like model for an overt, obtrusive narrator, it is an open question whether readers also for m a mental model of a neutral, covert agency of narration at all, and, if so, how that would be related to the character models. Likewise, since the overtness of narrators can be described only with gradual rather than absolute categories, we must ask which degree of overtness and which textual cues would trigger the construction of a narrator model. There are sure to be more blind spots in the theory at this stage, and it is to be hoped that further developments in the study of text-understanding, social cognition, and of literary character will help to elucidate further aspects of the reception of character.

Notes

(1.) James Phelan (Reading People, Reading Plots) has captured this duality by describing literary character as a combination of a "mimetic" and a "synthetic" dimension. As will be shown below, although my approach shares a number of interests and assumptions with that of Phelan, he does not argue from the vantage point of cognitive psychology.

(2.) In this essay, I outline the theory I have put forth in greater detail in my Ph.D. thesis (Grundri[beta] zur kognitiven Theorie der Figurenrezeption). I first presented the notions it formulates in a paper at the conference of the International Association of Literary Semantics (IALS), held 1-4 September 1997 at Freiburg i. Br., Germany. I presented the contents of the present essay at the Centre de Recherche sur la Litterature et la Cognition (CRLC), University of Paris VIII, on May 15, 1999. A French translation of a shorter version of this essay has been published in TLE 17, edited by Yves Abrioux. I am indebted to the members of CRLC for a critical discussion, and to the editors of TLE for the permission to publish a version in English. I would also like to thank Manfred Jahn, Richard Aczel, Barbara Korte, and Julian Lethbridge for commenting on various earlier versions of this essay and two anonymous reviewers of Style for a number of helpful comments and suggestions.

(3.) Although this notion of interaction between reader and text is the common denominator of practically all reader-oriented theories, at least since the 1960s, from reception aesthetics to empirical theories they have varied considerably with regard to the dominance ascribed to either text or reader in this process.

(4.) My approach is narrower in scope than Gerrig' s because I focus on characters and no other elements of fictional worlds. At the same time, it aims at applicability to a wider range of literary texts, since Gerrig, although quoting actual novels as examples, is still quite in keeping with the tradition of empirical research at large, for he tends to use stories as examples that are strongly action-oriented (such as crime stories), feature episodes with a simple problem-solving structure (sequences from detective stories), and present rather drastic emotional states (usually, questions of life and death). It seems to me that such material leads to unwelcome simplifications, since it is likely to elicit mainly "story-driven" rather than "point-driven understanding," in the terminology of Vipond and Hunt ("Point-driven Understanding"), and will therefore hinder the analysis of readers' more subtle responses to characters in less action-oriented novels. Still, these texts can be used to test some general prin ciples of literary understanding, and my own approach would not have been possible without such research results.

(5.) In processing successive bits of textual information, readers continually draw backward and forward inferences, construct hypotheses they may or may not find confirmed, and anticipate and evaluate possible outcomes. On this basic principle of cognition in text understanding, see Collins, Brown, and Larkin; Gerrig; Graesser and Clark; Rickheit and Strohner.

(6.) Cf. the use of this term in a review article by Ibsch; although a number of contributions can be found on diverse aspects of cognition and narrative (see, e.g., Jahn, Fludernik, Cook), an integrative theoretical framework has not yet been developed. The contributions in van Oostendorp and Zwaan, in Naturalistic Text Comprehension, also attempt to bridge the gap between narratological text analysis and empirical research in discourse processing. Mary Crane and Alan Richardson's Web site (Literature, Cognition and the Brain), located at the English Department of Boston College, documents the impressive range of activities in the field of cognitive approaches to literature and provides abstracts and reviews as well as reports on work in progress.

(7.) See, e.g., Sanford and Garrod; van Dijk and Kintsch; Meutsch; Garnham; Schnotz. For a recent survey of empirical research on mental models, see Zwaan and Radvansky.

(8.) See Zwaan and Radvansky. For examples of studies in mental-model constructions of fictional space and situations, see the work by Sanford and Garrod; Glenberg, Meyer and Lindem; Bower and Morrow; Wilson et al.; Morrow; Zwaan and van Oostendorp.

(9.) For a general theory of the capacity of working memory in language comprehension, see Just and Carpenter. Graesser and Clark present a description of inference regulation that is also based on the observation that capacity limitations influence processing activities. This issue is important for mental-model construction in so far as inferencing procedures are responsible for connecting incoming information with knowledge structures stored in the model.

(10.) Ericsson and Kintsch's notion that working memory consists of two stages instead of one--the short-term working memory and the long-term working memory--seems to account plausibly for the accessibility of already processed information in discourse processing in a way that may be applicable also to the accessibility of information in mental character models. On this notion see also Zwaan and Radvansky, "Situation Models" 166-67.

(11.) The main debate is between, on the one hand, propositionalists, who argue for an all-propositional, digital basis of mental images (see Pylyshyn), and, on the other, analogists, who suggest that there exists a structural isomorphism between perceptions and representations and thus assume that images are basically analogous (see Kosslyn). For a more recent collection of studies in the field, see Cornoldi and McDaniel.

(12.) On categories, see the articles by Rosch and Mervis; on schemas, see Brewer and Nakamura; on frames, Minsky; on scripts, Schank and Abelson. The notion of schematic knowledge-structures has frequently been applied to descriptions of discourse processing; on scripts in text understanding, see, e.g., Bower, Black, and Turner; on frames, see Metzing, Jahn; compare de Beaugrande for a schematheoretic view of reading. The concept of schemas, however, has been criticized as being unsatisfactory for explaining the variety of structures of knowledge available for understanding text, as being too rigid for flexible and dynamic understanding, and as neglecting the emotional aspects of reading; for critical views, see Thorndyke and Yekovich, Spiro, and Miall ("Beyond the Schema Given").

(13.) On the relevance of real-life experiences with persons and situations for text understanding, see Black, Galambos, and Read; Rickheit and Strohner; Halasz; Pollard-Gott.

(14.) The nineteenth-century stereotype of the "fallen woman" is a pertinent example: as Mitchell attests, the discourses of religion, morality, economics and psychology (or medicine) came together to stigmatize a woman who had offended moral and religious standards of purity and undermined the popular medical assumption that women lack sexual enjoyment apart from that of reproduction, for which marriage was the socially and economically prescribed framework in Victorian bourgeois society.

(15.) As Phelan emphasizes when he speaks of the "thematic" function of characters, some of the "people" in fictional worlds can be used to represent classes of persons rather than individuals. The mechanisms of reception of such characters will be further discussed below.

(16.) Literary genre has therefore been reinterpreted as a cognitive rather than a merely textual phenomenon (see Fishelonv; Olson, Mack, and Duffy).

(17.) David Miall can be regarded as representative of the critics of purely cognitive research on text understanding and of the recognition of the emotional constituents of the construction of meaning. See his "Anticipation and Feeling" for a neuropsychological explanation of the connectedness of thinking and feeling. See also Gerrig's description of what he calls "participatory responses," i.e., responses such as hopes and preferences, suspense and mental reorganization of the events in a story ("replotting"); all these responses arise from a combination of emotional and cognitive information processing (see Gerrig, Experiencing 69-96).

(18.) Definitions of empathy as an emotional response to fictional characters have mostly been produced in film studies; see, for instance, Tan and Zillmann. Those can nevertheless be applied to the analysis of narrative prose if attention is paid to the differences in the communication structures of film and novel.

(19.) For a promising extension of this distinction, see Kneepkens and Zwaan, who emphasize that all emotional response to texts is ultimately based in the reader's own emotional disposition. See also Miall ("Affect and Narrative").

(20.) On the importance of evaluation, compare Vipond and Hunt: "literary reading ...] is not centrally a matter of transferring information, but of negotiating and sharing beliefs, values, and attitudes" ("Literary Processing and Response" 157).

(21.) On the differentiation of these audiences, see Rabinowitz, "Truth" 126, and his Before Reading 20-42 and 93-104. From the viewpoint of cognitive psychology, one has to note that Rabinowitz's audiences are ultimately situated within the text, that he is "not really concerned with the actual psychological processes by which a specific reader performs this act" (in this case, joining the narrative audience ["Truth" 128n. 15]), and that the idea of the reader "joining audiences" is metaphorical and needs to be checked against what we know about the actual processes of text understanding by thinking, for instance, of these audience stances as different sets of schemata activated at the beginning of the reading.

(22.) This assumption is supported by the results of an empirical study by Dixon and Bortolussi, who found that readers' willingness and ability to explain and justify a character's actions and motivations were greater when that character's utterances were presented in free indirect speech, a technique for presenting consciousness that indicates a closeness between character and narrator.

(23.) See Genette and Rimmon-Kenan for the terminology of narrative situations.

(24.) E. M. Forster has pointed to the advantages of presentations of fictional persons over real-life interactions of persons, since the latter depend on the quality and quantity of information presented about them. Forster says that "people in a novel can be understood completely by the reader, if the novelist wishes; their inner as well as their outer lives can be exposed. [... In a] novel we can know people perfectly, and, apart from the general pleasure of reading, we can find here a compensation for their dimness in life. [...] They are people whose secret lives are visible ...] they suggest a more manageable human race, they give us the illusion of perspicacity and power" (57, 69-70).

(25.) See also Cupchik and the modifications of this approach by Groeben and. Vorderer 155-65.

(26.) Quotations 8, 13, and 34, respectively.

(21.) Apart from ideological intentions, this strategy can be useful for efficient presentation of such information as was, for example, necessary in nineteenth-century serial publication and can be observed especially well in Charles Dickens's novels.

(28.) Compare Rimmon-Kenan: "habitual actions tend to reveal the character's unchanging or static aspect, often having a comic or ironic effect" (61).

(29.) I have aimed at such a recontextualized approximation to readings of ten Victorian novels in my thesis (Grundri[beta]) The discourses and personality theories that I took as parameters of Victorian thought of special relevance for character-reception were religion, gender attitudes, political economy and self-help, and physiognomy and phrenology. Furthermore, I tried to define the readership of novels in terms of financial resources, and I described reading expectations that arose from different publication formats (three-decker versus serial novel). The idea of this historical recontextualization was to demonstrate to what extent modern readings of the same novels must be different from the historical ones.

(30.) In Reading People, Phelan demonstrates that categorizable characters (those embodying the "thematic function," in Phelan's terminology) are frequently minor characters introduced in the story in order to highlight the individuality of the protagonists, whose "mimetic" aspects thereby come into view to make them more likely to be understood in a personalized way.

Works Cited

Adams, Marilyn Jager. "Models of Reading." Language and Comprehension. Ed. Jean-Francois Le Ny and Walter Kintsch. Amsterdam: North Holland, 1982. 193-206.

Austen, Jane. Persuasion. 1818. Ed. D. W. Harding. Harmondsworth: Penguin, 1985.

Beaugrande, Robert de. Text, Discourse and Process: Towards a Multidisciplinary Science of Texts. Norwood: Ablex, 1980.

Berlyne, D. E. Aesthetics and Psychobiology. New York: Appleton-Century-Crofts, 1971.

Black, John B., James A. Galambos, and Stephen J. Read. "Comprehending Stories and Social Situations." Handbook of Social Cognition. Vol. 3. Ed. Robert S. Wyer and Thomas K. Srull. London: Erlbaum, 1984. 45-86.

Bower, Gordon H., John B. Black, and Terrence J. Turner. "Scripts in Memory for Text." Cognitive Psychology 11(1979):177-220.

Bower, Gordon H., and Daniel G. Morrow. "Mental Models in Narrative Comprehension." Science 247 (1990):44-48.

Brewer, Marilynn B. "A Dual Process Model of Impression Formation." Advances in Social Cognition. Vol. 1: A Dual Process Model of Impression Formation. Ed. Thomas K. Srull and Robert S. Wyer. Hillsdale: Erlbaum, 1988. 1-36.

--, and G.V. Nakamura. "The Nature and Function of Schemas." Handbook of Social Cognition. Vol. 1. Ed. Robert S. Wyer and Thomas K. Srull. Hillsdale, London: Erlbaum, 1984. 119-60.

Bronte, Anne. The Tenant of Wildfell Hall. 1848. Ed. G. D. Hargreaves. Harmondsworth: Penguin, 1985.

Bronte, Charlotte. Jane Eyre. 1847. Ed. Q. D. Leavis. Harmondsworth: Penguin, 1985.

Cantor, Nancy, and Walter Mischel. "Prototypes in Person Perception." Advances in Experimental Social Psychology. Ed. Leonard H. Berkowitz. New York: Academic P, 1979. 4-51

Collins, Allan, John S. Brown, and Kathy M. Larkin. "Inference in Text Understanding." Theoretical Issues in Reading Comprehension: Perspectives from Cognitive Psychology. Ed. Rand J. Spiro, Bertram C. Bruce, and William F. Brewer. Hillsdale: Erlbaum, 1980. 385-407.

Cook, Guy. Discourse and Literature: The Interplay of Form and Mind. Oxford: Oxford UP, 1994.

Cornoldi, Cesare, and Marc A. McDaniel, eds. Imagery and Cognition. New York: Springer, 1991.

Crane, Mary, and Alan Richardson, eds. Literature, Cognition and the Brain (Web site). <http://www2.bc.edu/.~richarad/lcb/home.html>.

Cupchik, Gerald C. "A Decade after Berlyne: New Directions in Experimental Aesthetics." Poetics 15 (1986):345-69.

Dickens, Charles. Hard Times for These Times. 1854. Ed. David Craig. Harmondsworth: Penguin, 1985.

Dijkstra, Katinka. "How Professional Readers Differ from Leisure-Time Readers: An Empirical Investigation into Differences in Intrinsic and Social Aspects of Reading." Empirical Approaches to Literature and Aesthetics. Ed. Roger J. Kreuz and Mary Sue MacNealy. Norwood: Ablex, 1996. 539-77.

Dixon, Peter, and Marisa Bortolussi. "The Reader, the Narrator, and the Characters: A Cue-Interaction Model of Characterization." Empirical Approaches to Literature: Proceedings of the Fourth Biannual Conference of the International Society for the Empirical Study of Literature--IGEL. Ed. Gebhard Rusch. Siegen: LUMIS-Publications, Siegen U, 1995. 28-36.

Eliot, George. Adam Bede. 1859. Ed. Stephen Gill. Harmondsworth: Penguin, 1985.

--. The Mill on the Floss. 1860. Ed. A. S. Byatt. Harmondsworth: Penguin, 1985.

Ericsson, K. Anders, and Walter Kintsch. "Long-Term Working Memory." Psychological Review 102 (1995):211-45.

Fishelov, David. "Genre Theory and Family Resemblance--Revisited." Poetics 20(1991):123-38.

Fludernik, Monika. Towards a 'Natural' Narratology. London: Routledge, 1996.

Forster, E. M. Aspects of the Novel. 1927. Harmondsworth: Penguin, 1962.

Frijda, Nico. The Emotions. Cambridge: Cambridge UP, 1986.

Garnham, Alan. Mental Models as Representations of Discourse and Text. Chichester: Ellis Horwood, 1987.

Garrod, Simon C., and Anthony J. Sanford. "Referential Processes in Reading: Focusing on Roles and Individuals." Comprehension Processes in Reading. Ed. David A. Balota, Giovanni B. Flores d'Arcais, and Keith Rayner. London: Erlbaum, 1990. 465-85.

Gaskell, Elizabeth. North and South. 1854-55. Ed. Dorothy Collins. Harmondsworth: Penguin, 1985.

---. Ruth. 1853. Ed. A. Shelston. Oxford: Oxford UP, 1985.

Genette, Gerard. Narrative Discourse. Trans. Jane E. Lewin. Oxford: Blackwell, 1980.

Gerrig, Richard G. Experiencing Narrative Worlds: On the Psychological Activities of Reading. New Haven, London: Yale UP, 1993.

---. and David W. Allbritton. "The Construction of Literary Character: A View from Cognitive Psychology." Style 24 (1990): 380-91.

Glenberg, Arthur M., Marion Meyer, and Karen Lindem. "Mental Models Contribute to Foregrounding during Text Understanding." Journal of Memory and Language 26 (1987): 69-83.

Graesser, Arthur C., and Leslie F. Clark. Structures and Procedures of Implicit Knowledge. Norwood: Ablex, 1985.

Graves, Barbara, and Carl H. Frederiksen. "Literary Expertise in the Description of a Fictional Narrative." Poetics 20 (1991): 1-26.

Groeben, Norbert, and Peter Vorderer. Leserpsychologie: Lesemotivation-- Lekturewirkung. Munster: Aschendorff, 1988.

Halasz, Laszlo, ed. Literary Discourse: Aspects of Cognitive and Social Psychological Approaches. Berlin: de Gruyter, 1987.

Ibsch, Elrud. "The Cognitive Turn in Narratology." Poetics Today 1.2(1990): 411- 18.

Iser, Wolfgang. The Act of Reading: A Theory of Aesthetic Response. Baltimore: Johns Hopkins UP, 1978.

Jahn, Manfred. "Frames, Preferences, and the Reading of Third-Person Narratives: Towards a Cognitive Narratology." Poetics Today 18.4 (1997): 441-68.

Johnson-Laird, P. N. "Mental Models." Foundations of Cognitive Science. Ed. Michael I. Posner. Cambridge, MA: MIT P, 1989. 469-99.

---. Mental Models: Toward a Cognitive Science of Language, Inference and Consciousness. Cambridge: Cambridge UP, 1983.

Just, Marcel Adam, and Patricia A. Carpenter. "A Capacity Theory of Comprehension: Individual Differences in Working Memory." Psychological Review 99.1 (1992): 122-49.

Kintsch, Walter. "The Role of Knowledge in Discourse Comprehension: A Construction-Integration Model." Psychological Review 95 (1988): 163-82.

---, and Teun A. van Dijk. "Toward a Model of Text Comprehension and Production." Psychological Review 85.5 (1978): 363-95.

Kneepkens, E. W. E. M., and Rolf Zwaan. "Emotions and Literary Text Comprehension." Poetics 23 (1994): 125-38.

Kosslyn, Stephen Michael. Image and Mind. Cambridge, MA: Harvard UP, 1980.

Lazarus, Richard S. Emotion and Adaptation. New York: Oxford UP, 1991.

Lesgold, Alan M., and Charles A. Perfetti. "Interactive Processes in Reading Comprehension." Discourse Processes 1 (1978): 323-36.

Mandler, Jean M. Stories, Scripts, and Scenes: Aspects of Schema Theory. London: Erlbaum, 1984.

Margolin, Uri. "Characterization in Narrative: Some Theoretical Prolegomena." Neophilologus 67 (1983): 1-14.

---. "Introducing and Sustaining Characters in Literary Narrative: A Set of Conditions." Style 21(1987): 107-24.

---. "The What, When, Why and How of Being a Character in Literary Narrative." Style 24 (1990): 453-68.

Marshark, Marc, and Cesare Cornoldi. "Imagery and Verbal Memory." Imagery and Cognition. Ed. Cesare Cornoldi and Marc A. McDaniel. New York: Springer, 1991. 134-82.

Mervis, Carolyn B. "Category Structure and the Development of Categorization." Theoretical Issues in Reading Comprehension: Perspectives from Cognitive Psychology, Linguistics, Artificial Intelligence, and Education. Ed. Rand J. Spiro, Bertram C. Bruce, and William F. Brewer. London: Erlbaum, 1980. 279-307.

Metzing, Dieter, ed. Frame Conceptions and Text Understanding. Berlin: de Gruyter, 1980.

Meutsch, Dietrich. "Mental Models in Literary Discourse: Towards the Integration of Linguistic and Psychological Levels of Description." Poetics 15 (1986): 307-31.

Miall, David S. "Affect and Narrative: A Model of Response to Stories." Poetics 17 (1988): 259-72.

--- "Beyond the Schema Given: Affective Comprehension of Literary Narratives." Cognition and Emotion 3 (1989): 55-78.

---. "Foregrounding: Defamiliarization, and Affect: Response to Literary Stories." Poetics 22 (1994): 389-407.

---. and Don Kuiken. "Feeling and the Three Phases of Literary Response." Empirical Approaches to Literature: Proceedings of the Fourth Biannual Conference of the International Society for the Empirical Study of Literature-IGEL. Ed. Gebhard Rusch. Siegen: LUMIS-Publications, Siegen U, 1995. 282-90.

---. "Anticipation and Feeling in Literary Response: A Neuropsychological Perspective." Poetics 22 (1995): 275-98.

Minsky, Marvin. "A Framework for Representing Knowledge." The Psychology of Computer Vision. Ed. Peter Winston. New York: McGraw-Hill, 1975.211-77.

Mitchell, Sally. The Fallen Angel: Chastity, Class and Women's Reading, 1835-1880. Bowling Green/Ohio: Bowling Green U Popular P, 1981.

Morrow, Daniel G. "Spatial Models Created from Text." Naturalistic Text Comprehension. Ed. Herre van Oostendorp and Rolf A. Zwaan. Norwood: Ablex, 1994. 57-78.

Olson, Gary M., Robert L. Mack, and Susan A. Duffy. "Cognitive Aspects of Genre." Poetics 10 (1981): 283-315.

Paivio, Allan. Mental Representations: A Dual Coding Approach. New York: Oxford UP, 1986.

Perry, Menakhem. "Literary Dynamics: How the Order of a Text Creates its Meaning." Poetics Today 1 (1979): 35-62; 311-61.

Peskin, Joan. "Constructing Meaning When Reading Poetry: An Expert-Novice Study." Cognition and Instruction 16.3 (1998): 235-263.

Phelan, James. Reading People, Reading Plots: Character, Progression and the Interpretation of Narrative. Chicago: U of Chicago P, 1989.

Pollard-Gott, Lucy. "Attribution Theory and the Novel." Poetics 21 (1991): 499-524.

Pylyshyn, Zenon W. "What the Mind's Eye tells the Mind's Brain: A Critique of Mental Imagery." Psychological Bulletin 80.1 (1973): 1-24.

Rabinowitz, Peter. Before Reading: Narrative Conventions and the Politics of Interpretation. Ithaca: Cornell UP, 1987.

---."Truth in Fiction: A Reexamination of Audiences." Critical Inquiry 4 (1977): 121-41.

Rickheit, Gert, and Hans A. Strohner, eds. Inferences in Text Processing. Amsterdam: North Holland, 1985.

Rimmon-Kenan, Shlomith. Narrative Fiction: Contemporary Poetics. New York, London: Methuen, 1983.

Rosch, E. "Cognitive Representations of Semantic Categories." Journal of Experimental Psychology: General 104 (1975): 192-233.

Sanford, Anthony J., and Simon C. Garrod. Understanding Written Language: Explorations of Comprehension Beyond the Sentence. Chichester: Wiley and Sons, 1981.

Schank, Roger C., and Robert P. Abelson. Scripts, Plans, Goals and Understanding. London: Erlbaum, 1977.

Schneider, David. "Implicit Personality Theory: A Review." Psychological Bulletin 79.5 (1973): 294-309.

Schneider, Ralf. GrundriB zur kognitiven Theorie der Figurenrezeption am Beispiel des viktorianischen Romans. Tubingen: Stauffenburg, 2000.

_____. "Pour une theorie cognitive du personnage litteraire: la dynamique de construction d'un modele mental." Trans. Aline Jarousse. TLE: Theorie-Litterature--Enseignement 17 (1999): 119-45.

Schnotz, Wolfgang. "Textverstehen als Aufbau mentaler Modelle." Wissenspsychologie. Ed. Heinz Mandl and Hans Spada. Munchen: Pyschologie Verlags Union, 1988. 299-330.

Spiro, Rand J. "Long-Term Comprehension: Schema-based versus Experiential and Evaluative Understanding." Poetics 11 (1982): 77-86.

Tan, Ed S. Emotion and the Structure of Narrative Film: Film as an Emotion Machine. Mahwah: Erlbaum, 1996.

_____. "Film-Induced Affect as a Witness Emotion." Poetics 23 (1994): 7-32.

Tannen, Deborah. "What's in a Frame? Surface Evidence, for Underlying Expectations." New Directions in Discourse Processing. Ed. Roy O. Freedle. Norwood: Ablex, 1979. 137-81.

Thackeray, William Makepeace. The History of Pendennis. 1848-50. Ed. John Sutherland. Oxford: Oxford UP, 1994.

Thorndyke, Perry W., and Frank R. Yekovich. "A Critique of Schema-based Theories of Human Story Memory." Poetics 9 (1980): 23-49.

Trollope, Anthony. The Warden. 1855. Ed. Robin Gilmour. Harmondsworth: Penguin, 1986.

van Dijk, Teun A., and Walter Kintsch. Strategies of Discourse Comprehension. New York: Academic P, 1983.

van Oostendorp, Herre, and Rolf A. Zwaan, eds. Naturalistic Text Comprehension. Norwood: Ablex, 1994.

Van Peer, Willie. Stylistics and Psychology: Investigations of Foregrounding. London: Croom Helm, 1986.

Vipond, Douglas, and Russel A. Hunt. "Literary Processing and Response as Transaction: Evidence for the Contribution of Readers, Texts, and Situations." Comprehension of Literary Discourse: Results and Problems of Interdisciplinary Approaches. Ed. Dietrich Meutsch and Reinhold Viehoff. Berlin: de Gruyter, 1989. 155-74.

---. "Point-driven Understanding: Pragmatic and Cognitive Dimensions of Literary Reading." Poetics 13 (1984): 261-77.

Wilson, Stephanie Gray, et al. "Mental Models and Narrative Comprehension: Some Qualifications." Journal of Memory and Language 32(1993): 141-54.

Wyer, Robert S., and Sally E. Gordon. "The Cognitive Representation of Social Information." Handbook of Social Cognition. Vol. 2. Ed. Robert S. Wyer and Thomas K. Srull. London: Erlbaum, 1984. 73-150.

Wyer, Robert S., and Thomas K. Srull. "Human Cognition in Its Social Context." Psychological Review 39.3 (1986): 322-59. Zillmann, Dolf. "Empathy: Affect from Bearing Witness to the Emotions of Others." Responding to the Screen: Reception and Reaction Processes. Ed. J. Bryant and Dolf Zillmann. London: Erlbaum, 1991. 135-67.

Zwaan, Rolf A., and Herre van Oostendorp. "Spatial Information and Naturalistic Story Comprehension." Naturalistic Text Comprehension. Ed.

Herre van Oostendorp and Rolf A. Zwaan. Norwood: Ablex, 1994. 97-114.

---, and Gabriel A. Radvansky. "Situation Models in Language Comprehension and Memory." Psychological Bulletin 123.2 (1998): 162-85.

Ralf Schneider ([email protected]) received his Ph.D. from the University of Cologne, where he also taught. Since 1998 he has been lecturer in English literature at the English Seminar of Tubingen University. His areas of research include cognitive approaches to literature, the sociology of literature and culture, and the influences of the media on literature.

COPYRIGHT 2001 Northern Illinois University
COPYRIGHT 2003 Gale Group