at he is kept in the dark and a covert form (Jannidis, +2004: chap. 4 amp; 6, based on Emmott, 1997) .can be described as ascribing information to an agent in the text so as to provide a character in the storyworld with a certain property or properties, a process often referred to as ascribing a property to a character. In the 19th century, critics spoke of the difference between direct and indirect characterization and of the preference of contemporary writers and readers for the latter (Scherer, 1977). Until recently, characterization was understood as the text ascribing psychological or social traits to a character (eg Chatman, 1978), but in fact texts ascribe all manner of properties to characters, including physiological and locative (space-time location) properties. Yet some textually explicit ascriptions of properties to a character may turn out to be invalid, as when this information is attributable to an unreliable narrator or to a fellow-character (Narrator). Moreover, a textual ascription may turn out to be hypothetical or purely subjective. There are also texts and styles of writing (eg the psychological novel) which tend to avoid any explicit statements of characterization. The crucial issue in the process of characterization is thus what information, especially of a psychological nature, a reader is able to associate with any character as a member of the storyworld and where this information comes from. There are at least three sources of such information: (a) textually explicit ascription of properties to a character; (b) inferences that can be drawn from textual cues (eg she smiled nervously); (c) inferences based on information which is not associated with the character by the text itself but through reference to historically and culturally variable real-world conventions (eg the appearance of a room reveals something about the person living there or the weather expresses the feelings of the protagonist). A systematic description of such inferences employed in characterization is given by Margolin (1983). Inferences can be understood in terms of abductions (Keller, 1998: chap. 9, based on Peirce), so that the fundamental role of character models and of the character encyclopedia becomes obvious: the information derived from them is not included in the text, but is presupposed to a greater or lesser degree by it.key problem concerns the limits and underlying rules of such inferences when they are applied to fictional beings. Ryan (1980), noting that readers tend to assume that a storyworld resembles the real world unless explicitly stated otherwise, adopts the philosopher David Lewiss principle of minimal departure. In a thorough criticism of this and similar hypotheses, Walton points out that this would make an infinite number of inferences possible, and he comes to the conclusion: There is no particular reason why anyones beliefs about the real world should come into play. As far as implications are concerned, simple conventions to the effect that whenever such and such is fictional, so and so is as well, serve nicely [...] (Walton, 1990). This approach, in turn, increases the number of conventions without necessity and without providing any convincing argument as to how readers go about accessing these conventions, aside from drawing on their real-world knowledge, despite the fact that many conventions apply only to fictional worlds. Even so, this does not invalidate Waltons criticism, which can probably be refuted only by including another element: the fact that characters are part of storyworlds which are not self-contained, but communicated. Readers assumptions about what is relevant in the process of communication determine the scope and validity of inferences (Sperber amp; Wilson, 1986) .presentation of characters is a dynamic process, just as is the construction of characters in the readers mind. A powerful model for describing the psychological or cognitive dynamics coming into play here, based on the top-down and bottom-up processes observed during empirical studies on reading comprehension, has been proposed by Schneider (2001) building on concepts developed by Gerrig amp; Allbritton (1990). A top-down process occurs in the application of a category to a character, integrating the information given by the text into this category, while a bottom-up process results from the text information integrating a character into a type or building up an individualized representation. At the beginning of a character presentation, textual cues may trigger various types of categorization: social types (the teacher, the widow); literary types (the hero in a Bildungsroman); text-specific types (characters that do not change throughout the story). In contrast to the top-down processing that takes place in these forms of categorization is bottom-up processing. This occurs when the reader is unable to integrate the given information into an existing category, resulting...
