lexicographical project. Fillmorean frame theory (see Cienki, this volume, chapter 7) has proved a highly stimulating framework for the description of lexical meaning, both theoretically and lexicographically. The FrameNet project that is currently being carried out in Berkeley is a large-scale attempt to build a digital lexicon according to the principles of frame theory. The theoretical background of the approach, as derived from frame theory as described in chapter 7 of this Handbook, was developed in publications such as Fillmore and Atkins (1992, 2000). Detailed information about the principles, purposes, and procedures within the FrameNet project are provided in a thematic issue of the International Journal of Lexicography (Fontenelle 2003). A full description of the project, together with the current release of the actual database, may also be found at the site of the Berkeley FrameNet project (http://www.icsi.berkeley.edu/~framenet). Basically, the information provided in the FrameNet database consists of three types of information. First, there is a description of the frames. This description in itself consists of different kinds of information. As an example, let us have a look at the Grant-permission frame. (The examples quoted here were derived from the FrameNet Web site in November 2003. It should be borne in mind that the Fram eNet project is far from being completed: no definitive value should be attributed to the examples.) The description begins with a global definition of the frame: In this frame a Grantor (either a person or an institution) grants permis- sion for a Grantee to perform an Action or for an Action to occur. This frame does not include situations where there is a state of permission granted by authority or rule of law. The permission for the Action may be limited to certain Conditions. The core FEs (frame elements) are then described separately: Action: The Action is the activity of the Grantee that the Grantor permits. Grantor: The Grantor is the individual or group of individuals vested with the authority to grant permission with respect to a specific domain and who grant permission to the Grantee. Grantee: The Grantee is the individual who is given permission to engage in a certain Action. Further, it is indicated how the frame in question relates to other frames. In the present case, for instance, Grant-permission is said to use the Communication frame, that is, it may take over elements from that frame. Finally, the LUs (lexical units) that exemplify the frame are listed: allow.v, greenlight.v, leave.n, let.v, okay.v, permission.n, permit.v, sanction.v, suffer.v, the go-ahead.n, the green light.n, the okay.n. The second basic type of information provided by FrameNet consists of the individual description of the lexical units. The description of the frames does not, in fact, exhaust the description of the lexical items that exemplify the frame. For instance, it is one of the basic insights of frame theory that words may vary indi- 1170 dirk geeraerts vidually in the patterns of frame element realization they allow. For example, given a description of the Commercial-transaction frame as ‘‘These are words that de- scribe basic commercial transactions involving a Buyer and a Seller who exchange Money and Goods,’’ the typical patterns for the verbs buy and sell differ in their configuration of frame patterns: buy: Buyer buys Goods from Seller for Money sell: Seller sells Goods to Buyer for Money. This implies that the description of the frames need to be complemented with a description of the individual valence patterns exhibited by the lexical items. This is all the more so because lexical units may feature in different frames. Take the example of suffer in the Grant-permission frame. It also occurs in the Catastrophe frame, defined as: The words in this frame involve an Undesirable_Event which affects the Undergoer negatively. No agent need be involved. Within the Catastrophe frame, suffer occurs in a number of valence patterns. In the following corpus-based examples, we successively find a passive construction with a by-phrase, a passive construction without a by-phrase, an intransitive use, a tran- sitive use, and a construction with a from-phrase expressing the frame element Undesirable_Event. The FrameNet project comprises an exhaustive description of these valence patterns. When I returned to the drawing room, a clergyman was talking about [the hardships] Undesirable_Event being suffered [by children] Undergoer in Berlin. The problems in applying the rule arise when [the damage] Undesirable_Event which has been suffered is not financial. [DNI] Undergoer And [social services] Undergoer suffered too, as they always will when wealth creation is despised. [DNI] Undesirable_Event [Both countries] Undergoer have suffered [prison violence, disorder and breakouts] Undesirable_Event . A year-long survey of Edinburgh has shown [much of the city] Undergoer to be suffering [from severe air pollution] Undesirable_Event . The set of examples just given illustrates the third main feature of the FrameNet database: for each of the LUs (lexical units) in the frames, the database provides a set of corpus-based examples illustrating the valence patterns and the configura- tions of frame elements that the LUs occur in. The examples are annotated, in the sense that the relevant frame elements are tagged. (In the examples, DNI stands for ‘‘definite null instantiation’’—a frame element that is missing but whose identity is understood from the context.) The FrameNet database mainly tries to serve two audiences (apart from the- oretical linguists): researchers in natural language processing, who may profit from lexicography 1171 the FrameNet information for any application that has to deal with word sense disambiguation, information extraction, and machine translation; and lexicogra- phers, who may profit from the semantic definitions, the combinatorial descrip- tions, and the corpus-based examples provided by FrameNet. 4. Further Prospects When dealing with the relationship between theoretical lexicology and practical lex- icography, one should definitely not assume that theoreticians in principle have the answers and that lexicographers simply have to follow: the relationship is not a one- sided one,and as I have stressed on an earlier occasion (Geeraerts 1997: 5), there should be a relationship of mutual inspiration between both disciplines. The present chapter has basically looked from lexicological theory as developed within Cognitive Lin- guistics to lexicographical practice rather than the other way round, but to round off, we may now reverse the perspective: to what extent is existing lexicography a source of inspiration for Cognitive Linguistics—or to what extent could it be? Although much of the work done in historical lexicology by cognitive linguists (see Grondelaers, Speelman, and Geeraerts, chapter 37, this volume) took inspira- tion from the great historical dictionaries like the Oxford English Dictionary, there are important aspects of current lexicographical practice that Cognitive Linguistics has only marginally touched upon, in spite of the fact that the lexicographical ap- proach has proved extremely rewarding for lexical analysis at large and for the study of polysemy in particular. Specifically, a collocational approach to lexical description, systematically singling out the constructional patterns that lexical items occur in and identifying different meanings through differences in collocational pat- terns, is a methodological focus for many current lexicographical projects (see, e.g., Moon 1998). However, collocational methods to get a grip on polysemy are not (yet) among the standard equipment of cognitive linguists. There are, in fact, two aspects to such a collocational approach: it is a corpus-based approach, and it generally uses a quantitative method for the analysis of the corpora. So to what extent does Cognitive Linguistics incorporate these two methodological approaches? First, the use of corpus materials was already part of early studies like Rudzka- Ostyn (1988), Schulze (1988), Geeraerts (1988), Dirven and Taylor (1988), or Goossens (1990). Typically, however, these were studies carried out by European cognitive linguists, while the early American studies were predominantly based on an intro- spective methodology. For the broader community of Cognitive Linguistics, the im- portance of corpus materials became a topic only since Barlow and Kemmer (2000), and as we saw in the description of the FrameNet project, using a corpus-based methodology is becoming more and more natural. 1172 dirk geeraerts Second, the use of quantitative data within Cognitive Linguistics is more recent than the use of corpus materials, but there are now various lines of research within Cognitive Linguistics using quantitative models and/or methods: Tomasello’s work on language acquisition (see chapter 41, this volume), Bybee’s work on linguistic change (see chapter 36), and the work by Grondelaers, Speelman, and Geeraerts on sociolinguistic variation (see chapter 37). The use of corpus materials, in short, is slightly more widespread in Cognitive Linguistics than the use of quantitative meth- ods, but neither is as yet a dominant approach. What Cognitive Linguistics could learn from current lexicography, then, is a combination of a corpus-based approach and a quantitative approach to tackle the collocational and combinatorial properties of lexical items. An interesting step in that direction may be found in the research program recently defined by Gries and Stefanowitsch (Gries 2003; Gries and Stefanowitsch 2003), connecting Construc- tion Grammar (see Croft, chapter 18, this volume) and quantitative corpus linguis- tics. To conclude, although Cognitive Linguistics appears to offer an exciting per- spective for the further development of lexicography and lexicographical theory, the interaction could be carried much further than is currently the case. REFERENCES Ayto, John. 1983. On specifying meaning. In Reinhard R. K. Hartmann, ed., Lexicography: Principles and practice 89–98. London: Academic Press. Barlow, Michael, and Suzanne Kemmer, eds. 2000. Usage-based models of language. Stan- ford, CA: CSLI Publications. Dirven, Rene ´ , and John R. Taylor. 1988. The conceptualisation of vertical space in English: The case of tall. In Brygida Rudzka-Ostyn, ed., Topics in cognitive linguistics 379–402. Amsterdam: John Benjamins. Fillmore, Charles J., and Beryl T. Atkins. 1992. Towards a frame-based lexicon: The semantics of RISK and its neighbours. In Adrienne Lehrer and Eva Feder Kittay, eds., Frames, fields, and contrasts: New essays in semantic and lexical organization 75–102. Hillsdale, NJ: Lawrence Erlbaum. Fillmore, Charles J., and Beryl T. Atkins. 2000. Describing polysemy: The case of ‘crawl.’ In Yael Ravin and Claudia Leacock, eds., Polysemy 91–110. Oxford: Oxford University Press. Fontenelle, Thierry, ed. 2003. FrameNet and frame semantics. Special issue of International Journal of Lexicography 16. Frawley, William. 1981. In defense of the dictionary: A response to Haiman. Lingua 55: 53–61. Geeraerts, Dirk. 1985. Les donne ´ es ste ´ re ´ otypiques, prototypiques et encyclope ´ diques dans le dictionnnaire. Cahiers de Lexicologie 46: 27–43. Geeraerts, Dirk. 1987. Types of semantic information in dictionaries. In Robert Ilson, ed., A spectrum of lexicography 1–10. Amsterdam: John Benjamins. Geeraerts, Dirk. 1988. Where does prototypicality come from? In Brygida Rudzka-Ostyn, ed., Topics in cognitive linguistics 207–30. Amsterdam: John Benjamins. lexicography 1173 Geeraerts, Dirk. 1990. The lexicographical treatment of prototypical polysemy. In Savas L. Tsohatzidis, ed., Meanings and prototypes: Studies in linguistic categorization 195–210. London: Routledge. Geeraerts, Dirk. 1997. Diachronic prototype semantics: A contribution to historical lexicology. Oxford: Clarendon Press. Goossens, Louis. 1990. Metaphtonymy: The interaction of metaphor and metonymy in expressions for linguistic actions. Cognitive Linguistics 1: 323–40. Gries, Stefan Th. 2003. Towards a corpus-based identification of prototypical instances of constructions. Annual Review of Cognitive Linguistics 1: 1–27. Gries, Stefan Th., and Anatol Stefanowitsch 2003. Collostructions: On the interaction between words and constructions. International Journal of Corpus Linguistics 8: 209–43. Haiman, John. 1980. Dictionaries and encyclopedias. Lingua 50: 329–57. Haiman, John. 1982. Dictionaries and encyclopedias again: Discussion. Lingua 56: 353–55. Hanks, Patrick. 1994. Linguistic norms and pragmatic exploitations, or why lexicographers need prototype theory, and vice versa. In Ferenc Kiefer, Gabor Kiss, and Julia Pajzs, eds., Papers in computational lexicography 89–113 . Budapest: Linguistics Institute. Kilgarriff, Adam. 1997. I don’t believe in word senses. Computers and the Humanities 31: 91–113. Lara, Luis Fernando. 1989. Dictionnaire de langue, encyclope ´ die et dictionnaire en- cyclope ´ dique: Le sens de leur distinction. In Franz Josef Hausmann, Oskar Reich- mann, Herbert Ernst Wiegand, and Ladislav Zgusta, eds., W € orterb € ucher/ Dictionaries/ Dictionnaires 1: 280–87. Berlin: Walter de Gruyter. Moon, Rosamund. 1998. Fixed expressions and idioms in English: A corpus-based approach. Oxford: Oxford University Press. Peeters, Bert, ed. 2000. The lexicon-encyclopedia interface. Oxford: Elsevier. Putnam, Hilary. 1975. The meaning of meaning. In Hilary Putnam, ed., Mind, language, and reality: Philosophical papers 2: 215–71. Cambridge: Cambridge University Press. Rey-Debove, Josette. 1971. E ´ tude linguistique et se ´ miotique des dictionnaires franc¸ais con- temporains. The Hague: Mouton. Rudzka-Ostyn, Brygida. 1988. Semantic extensions into the domain of verbal communi- cation. In Brygida Rudzka-Ostyn, ed., Topics in cognitive linguistics 507–54. Am- sterdam: John Benjamins. Schulze, Rainer. 1988. A short story of down. In Werner Hu ¨ llen and Rainer Schulze, eds., Understanding the lexicon: Meaning, sense and world knowledge in lexical semantics 394–410.Tu ¨ bingen, Germany: Max Niemeyer Verlag. Stock, Penelope F. 1983. Polysemy. In Reinhard R. K. Hartmann, ed., LEXeter ’83: Pro- ceedings of the Exeter lexicography conference 131–40.Tu ¨ bingen, Germany: Max Nie- meyer Verlag. Swanepoel, Piet. 1992. Linguistic motivation and its lexicographical application. South African Journal of Linguistics 10 : 49–60. Swanepoel, Piet. 1998. Back to basics: Prepositions, schema theory, and the explanatory function of the dictionary. In Thierry Fontenelle, Philippe Hiligsmann, Archibald Michiels, Andre ´ Moulin, and Siegfried Theissen, eds., Euralex ’98 Proceedings 655–66. Lie ` ge, Belgium: Universite ´ de Lie ` ge, De ´ partement d’anglais et de ne ´ erlandais. Temmerman, Rita. 2000. Towards new ways of terminology description: The sociocognitive approach. Amsterdam: John Benjamins. Van der Meer, Geart. 2000. Core, subsense, and the New Oxford Dictionary of English. In Ulrich Heid, Stefan Evert, Egbert Lehmann, and Christian Rohrer, eds., Euralex 2000 Proceedings 419–31. Stuttgart: Institut fu ¨ r Maschinelle Sprachverarbeitung. 1174 dirk geeraerts chapter 45 COGNITIVE LINGUISTIC APPROACHES TO LITERARY STUDIES: STATE OF THE ART IN COGNITIVE POETICS margaret h. freeman 1. Introduction In his statement, ‘‘Language is the child of the literary mind,’’ Turner (1991, 1996) reverses the traditional view that literature is a special, exotic subcategory of lan- guage by arguing that human language capabilities arose from the cognitive mapping projections of parable and story. Although Turner’s argument has not as yet received wide acceptance in either field of linguistics or literature (but see Modell 2003), the emergence of Cognitive Linguistics has encouraged the devel- opment of new relations between the two disciplines (see Geeraerts 1999 for a comprehensive survey of the historical development of linguistic semantics and literary theories). Just as literary texts may serve as legitimate data for under- standing the principles of language structure and use, linguistic analysis offers new perspectives on literary production, interpretation, reception, and evaluation (Bizup and Kintgen 1993; Hart 1995; Jahn 1997; Crane and Richardson 1999; Jackson 2000). Historically, a certain amount of tension has existed between the disciplines of linguistics and literature. For those of us engaged in bridging the two, the par- ticular form this tension takes—namely, that literary criticism contributes nothing to linguistic enquiry, and vice versa—has always seemed anomalous. However, this anomaly may have roots deeper than being simply a matter of turf wars. Recently, Burrows (2003) has characterized the split between scientific method and literary criticism as a comparison between Descartes’s retiring to his ‘stove’ to contemplate the foundations of knowledge and Montaigne’s retiring to his tower to write his Essays: Descartes’s stove and Montaigne’s library tower have given us two ways of liv- ing and thinking that are at root divergent. Stove people think that you can strip everything away and rebuild reality from precepts; tower people reckon that writing about and exploring or refining beliefs is the best you can do. For tower people, the process of writing and arguing is what thinking is; it is not concluding. (Burrows 2003: 21) Though the ways of the stove and the tower may appear fundamentally incom- patible, this chapter surveys recent work in applying cognitive linguistic approaches to literature that carry with them both the air of the tower and the heat of the stove. Literary critics have long been familiar with such topics as perspective, point of view, flashbacks, foreshadowing, and so on that cognitive linguists are just now exploring. One question that inevitably arises is what new insights Cognitive Linguistics provides in literary studies that literary criticism has not already dis- covered. The corollary, what literary criticism can contribute to Cognitive Lin- guistics, is almost always never asked (but see Brandt and Brandt 2005a). In its focus on the processes of literary creation, interpretation, and evaluation, Cog- nitive Linguistics contributes scientific explanations for the findings of literary critics and thus provides a means whereby their knowledge and insights might be seen in the context of a unified theory of human cognition and language. To this extent, the stove is not incompatible with the tower; to the contrary, neither functions completely or well without the other. Although ‘‘literature’’ in its broadest sense refers to all written texts, this chapter restricts its scope to the more narrowly focused term used to cover the literary genres of fiction, poetry, and drama, written instances of humor, multimedia forms such as film, and religious writings that display literary qualities, such as the Bible and mystic poetry. All these writings are oriented toward the expressive, the emotive, and the aesthetic; it is here that the more inclusive approach of Cog- nitive Poetics, particularly as practiced by Tsur (1992, 1998, 2003), may serve as a guide for further developments in the interdisciplinary area of linguistics and lit- erature. As Hamilton (2000: 3) notes, ‘‘Cognitive poetics can provide a sensible epistemology for the event of interpretation.’’ 1176 margaret h. freeman The past few years have seen an explosion in interest in cognitive approaches to literature. 1 These approaches include the development of methodologies for de- scribing both the production and reception of literary texts. Since the work pre- sented in this chapter describes a symbiotic relationship between literary and lin- guistic objectives, I have organized it according to challenges common to both. Each section highlights aspects important to literary and linguistic study and de- scribes work that suggests possible directions for future study. 2. Prototypicality and the Notion of Literature Several researchers have turned their attention to illuminating the nature of literature and its various genres through prototypicality theory as opposed to a classical, feature-based theory of categorization (Meyer 1997; De Geest and van Gorp 1999; G. Steen 1999b). From the perspective of literary criticism, the category ‘‘literature’’ has been so enlarged in the postmodern period as to include whatever a particular reader chooses to consider ‘‘text,’’ whether oral or written or even the nonlinguistic ‘‘signifiers’’ of culture. Under these circumstances, G. Steen wisely calls for an em- pirical research program to develop a taxonomy of discourse, in which literature may be positioned within the domain of discourse in general. He argues for a taxonomy in which ‘‘a prototypical approach emphasizes the hierarchical order of fuzzy concepts in a domain, using the same attributes for every level of conceptualization’’ (1999b: 116). The seven attributes he identifies are content, form, type, function, medium, domain, and language. The more abstract the level, the more certain attributes are unspecified. Thus, the basic level ‘‘novel’’ may be characterized by values of all seven attributes, whereas the superordinate term, ‘‘literature,’’ is characterized by domain (‘artistic’), content (‘fictional’), and function (‘positively affective’), but not by the other four. The advantage of G. Steen’s taxonomy is that it quickly identifies when theories of literature mix values belonging to different attributes, as Meyer (1997) does. In his analysis, the addition of the attributes of medium, language, and form to the term ‘‘literature’’ makes it less superordinate as a category and closer to the level of genre. In Meyer’s prototypical definition of literature, works that contain more features would be considered more literary or better examples of the category than those that contain less. In their focus on the basic level of literary genre, De Geest and van Gorp (1999) reveal the complexities of applying a prototype approach. They point out that identifying the ‘‘best’’ or more typical example of a literary category is not at all the same as an aesthetic evaluation: ‘‘the ‘best’ texts are almost by definition excep- tional cases which clearly are, at least in some aspects, atypical’’ (1999: 43). As the cognitive linguistic approaches to literary studies 1177 discussion of G. Steen’s taxonomy has noted, the greater the superordination of the category, the harder it is to establish prototypical instances; consider, for example, the difference between ‘‘poem’’ and ‘‘sonnet.’’ However, even the lower-level cat- egory is more problematic than it seems. Although the sonnet exists at a more subordinate level than the poem and thus might be more readily defined in pro- totypical terms, De Geest and van Gorp show that it is just as problematic; it would be strange, if not absurd, to consider a Petrarchan sonnet more prototypical than a Spenserian one, or a Spenserian than a Shakespearian. And then, what does one do with so-called sonnets whose rhyme, meter, structure, or number of lines vary from these established forms? Like G. Steen and Meyer, De Geest and van Gorp indicate that literary texts and genres must be considered along their evaluative and axiological components, considering norms, values, and models, as well as the author’s intentions and the reader’s expectations. They suggest that the concept of norm has to include not just what is proscribed but what is permitted. One possibility for achieving a proto- typical theory of literature would be to adopt De Geest and van Gorp’s (1999: 41) recognition that ‘‘the so-called ‘prototype’ need not exist in reality, since it is generally assumed to be a kind of hypothetical cognitive construction, a theoretical ‘fiction,’ ’’ much like Lakoff’s Idealized Cognitive Model that structures a con- ceptual domain. The ‘‘prototype’’ of a literary work would then include in its description an atypical example of its genre. This rather radical proposal—that the category of literary works needs to accommodate atypicality as prototypical—appears to undermine the very notion of prototypicality theory, so that literary critics might well question the relevance of applying it to literature in the first place. This is one example of the conflict between the stove and the tower. Understanding the nature of literature involves explaining its role in the workings of the embodied human mind. It might be argued that this begs the question: why should the methodology applied to understanding litera- ture (and the other arts) be necessarily a scientific one? Talmy’s (2000: 479–80) discussion of the parameter of protoypicality in the context of evaluation provides one answer: it is only by judging with respect to cultural norms that one can determine the relative status of a literary work as conforming to or challenging them. As Talmy notes, ‘‘Thus, it appears that certain long periods in Chinese art and literature maintained themselves with great conservatism, while this century in the West has rewarded authorial experimentation’’ (480). In this light, the ex- pectation that a literary work be atypical may be seen as the prototypical attitude to literature held by contemporary Western critics. Only by looking at literature using the same methodology that is applied to looking at other activities of the human mind can we fully comprehend the nature of the distinctions between creative and conventional expressions and trace the changes in their prototypical status through time. All the research surveyed in this chapter may be understood as examples of this principle. A case in point is Ravid and Hanauer’s (1998) study of how adult speakers of Hebrew show evidence of having a prototypical theory of rhyme. Their 1178 margaret h. freeman scientific analysis and empirical research confirm literary intuitions about the way readers respond to the kind of rhyme schemes that occur in a variety of poetic texts. One finding that ran counter to Ravid and Hanauer’s predictions—that Hebrew speakers tolerated contrasting coda consonants but not contrasting vowels in the post-stress syllable of modernistic rhymes—may possibly signify a dynamic shift of category boundary in process as Hebrew speakers grow more familiar with the rhyming practices of modernist poets. Whether Hebrew speakers in the future tolerate both post-stress consonants and vowels as members of the same rhyme category would be a hypothesis for such dynamic change and subject to further empirical research. A dynamic theory of prototypicality over time could explain how literary de- cisions as to what constitutes a literary text are made. For example, though Words- worth, in his second preface to the Lyrical Ballads, remarked that readers might question whether the poems included could be considered poetry at all, literary critics today perceive them to be classic examples of the genre of Romantic Poetry, a possible indication of category change over time. Evidence for a dynamic as opposed to static construal of prototypes is provided by two studies of prototypicality that involve literary texts. Głaz’s (2002) lexicological study of the concept domain of Earth looks at the use of the term in six novels by Kingsley Amis, alongside data collected from the 1995 editions of The T imes and The Sunday Times. Głaz combines Fuchs’s (1994) dynamic model of semantic space with Langacker’s (1987, 1991)net- work model to show how the use of a term opens a window onto its entire lexical network, with meaning construed by shifts in both intracategorial and extracate- gorial tensions set up by the context. Gibbs (2003: 38) recognizes that ‘‘prototypes are not abstract, pre-existing conceptual structures, but are better understood as prod- ucts of meaning construal.’’ These include interpreting context-sensitive meaning in literary texts, the judgment of novelty by skilled readers, and the fact that an ‘‘em- bodied view of meaning construal nicely captures at least some of what people see as poetic during their reading experiences’’ (39). Applying a dynamic view of prototype theory might well serve as a research agenda for understanding how prototypical judgments of literature change over time. 3. Conceptual Structure in Human Cognition and Narrative The aims of the tower are different from those of the stove. Literary critics focus on the emotional and aesthetic effects of literary works, cognitive linguists on ac- counting for the way language characterizes meaning. From a cognitive perspec- tive, literary critics are engaged in mapping the meanings of texts from various contextual domains. They are interested in the results of these mappings, not the cognitive linguistic approaches to literary studies 1179 . sanction.v, suffer.v, the go-ahead.n, the green light.n, the okay.n. The second basic type of information provided by FrameNet consists of the individual description of the lexical units. The description of the. digital lexicon according to the principles of frame theory. The theoretical background of the approach, as derived from frame theory as described in chapter 7 of this Handbook, was developed in. method for the analysis of the corpora. So to what extent does Cognitive Linguistics incorporate these two methodological approaches? First, the use of corpus materials was already part of early