Cognitive Poetics: from Interpreting to Experiencing what is Literary

David S. Miall

Paper prepared for Anglistentag 2007, 23-26 September, University of Münster, Germany

© David S. Miall (pre-print version; please do not cite from the text of this essay without permission)

1. Becoming empirical

A central claim of cognitive poetics is that it casts light on the process of reading. Thus Peter Stockwell in introducing cognitive poetics tells us that "It is all about reading literature" (2002, 11). Yet for Stockwell, like most other literary scholars, the reader throughout his book remains a theoretical one. Indeed, for most scholars in the cognitive paradigm their work is primarily analytical: what we find are theories of reading, based on cognitive constructs such as deixis, conceptual frame theory, or embodied metaphor. With a few exceptions (notably Gerard Steen, Yeshayahu Shen, and Reuven Tsur) what we don't encounter are actual readers or empirical studies of reading. Given that cognitive science, from which these scholars draw their inspiration, is fundamentally an empirical science involving experiment, this is perhaps rather strange. If I have a proposal about how, for example, deictic shifts require a certain reading process, shouldn't I investigate whether readers respond in the way I say they do? Shouldn't I try to find out if readers really do experience the process I have described? Perhaps I will find out that I am right; or perhaps I will find that other, competing processes make the process I have proposed impossible; or perhaps the process does occur, but I find out that there is a different explanation for it than the one I have proposed. This is the kind of problem that cognitive scientists typically try to resolve, by proposing hypotheses and setting out to test them empirically with real informants. Yet cognitive poetics has avoided this route, and as a result, I suggest, is in danger of falling back into the old hermeneutic model of text interpretation that it has proposed to supersede.

I should emphasize that the turn to empirical method is not for the purpose of testing interpretations of texts; it will, perhaps, assist us in understanding the psychological conditions that make interpretation possible. But even here I hesitate: there is a question whether interpretation is a necessary or typical outcome of literary reading - the kind of reading, that is, engaged in and enjoyed by the ordinary reader outside the academy. This is not something that we should take for granted; rather, it is in itself an issue for empirical study. If we set aside the assumption that interpretation is the aim of reading, this leaves the door open to the discovery of other possible correlates of reading - perhaps a range of experiential accompaniments or outcomes, some of which may have received little or no attention. This possibility may also influence the design of our empirical studies. We may want to avoid, for example, putting questions to our readers that probe overall comprehension of a text, since these are likely to evoke that sideways step into interpretation - the "Rule of Abstract Displacement," as Peter Rabinowitz calls it (1996, 139).

My aim in this paper, then, is to probe the purposes of cognitive poetics, and to outline what I see as a way forward that avoids some of the difficulties and limitations the discipline seems to have created for itself. I will discuss empirical studies of reading and draw some comparisons with the studies of reading to be found in cognitive poetics. My aim will be to show (in brief) that each discipline, that is, cognitive poetics and empirical studies, would benefit from the other, and to argue for a meeting of minds and methods.

One of the benefits I will point to later in this discussion, is that although the role of feeling has been theorized and studied in the empirical domain, it has been almost entirely ignored by cognitive poetics. As Meir Sternberg put it recently, cognitivists have made the mistake of setting "cognition against (at least above) emotion" (2003, 313), under the influence of "the party line that would put the cognitive before the affective (in temporal, causal, analytic, and/or scalar order)" (382). In our work, based on empirical findings (e.g., Miall & Kuiken 2002), we have suggested that feeling may be the vehicle for what is distinctively literary in literary response, thus to downplay or overlook feeling may seriously misrepresent the nature of literary reading.

The question of what is literary has, however, caused some notable ambivalence. Among cognitive poetics scholars Peter Stockwell (2002), for instance, regards the cognitive methods he imports as suitable for all purposes: since "there is nothing inherently different in the form of literary language, it is reasonable and safe to investigate the language of literature using approaches generated in the language system in general" (7). An eminent scholar of discourse processing takes the same approach: in Comprehension (1998) Walter Kintsch claims that "The comprehension processes, the basic strategies, the role of knowledge and experience, as well as the memory products generated, are the same for literary texts as for the simple narratives . . . used in our research. . . . The difference is in the 'what,' not the 'how'" (205). Similarly, the stylisticians Ronald Carter and Walter Nash (1990) refuse to separate the literary from the non-literary: "We do not consider such texts to be mutually exclusive and do not see why the same analytical procedures and schemas cannot be applied across a range of texts" (29). Almost the only cognitivist scholar to dissent from this is Reuven Tsur (1992), for whom literary principles are not reducible to cognition: "A major assumption of the present cognitive approach," he says, introducing his major book on the topic, "is that literature does have important operational principles that cannot be exhausted in terms of cognitive science" (3). In fact, as I remarked before, what is literary can be framed as an empirical question.

In order to provide a coherent focus to the discussion, I will consider the role of inference in reading. This will enable me to coordinate proposals about reading from four different disciplines, discourse processing, cognitive poetics, empirical studies of reading, and literary theory, and to raise some questions about the nature of reading. To anticipate my argument, I will suggest that the inference issue will enable us to reflect on what is literary, demonstrate a role for feeling that has been largely overlooked, shift emphasis away from interpreting texts to experiencing them, and demonstrate the value of the empirical method for literary studies.

2. Theories of Inference

Inferencing is clearly central to the process of reading, and central also to much scholarly work, whether on textual structures, interpretations, or reading processes. As Culpeper (2002) puts it, echoing a common observation, writers "can mean more than they say" (263). Hemingway (1932) makes this point in his well-known "iceberg" model: "If a writer of prose knows enough about what he is writing about he may omit things that he knows and the reader, if the writer is writing truly enough, will have a feeling of those things as strongly as though the writer had stated them. The dignity of movement of an iceberg is due to only one-eighth of it being above water" (192). But what are these things below the water-line that the reader can infallibly infer? In the literary domain, including cognitive poetics, we have no agreed upon model of the inferencing process, although a number of important proposals have been put forward. In discourse processing, in contrast, inference models have been proposed and validated empirically, and there is a greater measure of agreement in the field. I will give a brief account of a number of the models, drawing from different fields, and outline some of the principles that seem to be involved.

2.1 Discourse processing

Several models of inferencing have been put forward in discourse processing, focusing on the inferences said to be generated online by readers of narrative, that is, moment by moment during the reading process. I will not discuss the different models here or the reasoning behind their predictions, but outline one model, that of Graesser and his colleagues, which has been well developed in a number of papers and has strong empirical support. Moreoever, their model has been developed in the context of real - what they term "natural" - narratives, that is, short literary narratives, rather than the artificial, experimenter-produced narratives that are often used in these studies.

Graesser (in Graesser, Singer, & Trabasso 1994) considers thirteen classes of knowledge-based inferences that map onto the representation of the narrative in working memory. Of these, the first six types of inference are said to be generated online (a claim largely supported by empirical evidence). They are also said to occur automatically and invariantly within 650 msecs of passage onset (Magliano & Graesser 1991):

1. Referential, for instance, anaphoric references (e.g., who "he", or what "it" refers to)

2. Case structure role assignment (a noun phrase is assigned a particular role, such as agent, object, location, or time)

3. Causal antecedent (a cause is inferred linking the present proposition to what came before)

4. Superordinate goal (what motivates the character)

5. Thematic (the main point or issue of the passage being read, or the whole text)

6. Character emotional reaction (character's emotion in response to an event)

Their function can be summarized as follows: the first three establish local coherence at the sentence level; 3 and 4 provide explanations; while 4, 5, and 6 help establish global coherence. Other types of inference in their list that are not generated online include: causal consequence (at most points in a narrative there are too many possible consequences, so these are not usually computed); reader emotion; author intent, and several others.

Thus, according to Graesser, readers attempt to construct a representation of a narrative that is coherent both at the local and global levels. While Graesser considers what occurs when this is thwarted, he does not consider the literary potential of such moments. If global coherence is not achievable, he says, the reader settles for local coherence; but "If the reader believes that the text lacks local coherence, then the reader regards the text as incoherent" ( 378). But if an inference is not possible, or is either unclear or ambivalent, then literary processing may be initiated. An example is evident in the literary text that Graesser cites for his example inferences, Ambrose Bierce's little fable "How Leisure Came." In this very brief story, The Man to Whom Time Was Money accidentally sticks a pickle fork in his right eye, and pulls the eye out. In the next sentence we are told: "In buying spectacles the needless outlay for the right lens soon reduced him to poverty" (cited 375; from Fantastic Fables, 1899). No adequate causal antecedent can be inferred to account for why the man buys spectacles (although we are free to speculate); nor, since he no longer sees with the right eye, is it evident why he needs a right lens; especially odd is the dramatic effect of the "needless outlay" since it "soon reduced him to poverty." Of course, Bierce is playing here with his figure of a man whose orderly and obsessed life has been disrupted, and the sentence in question is evidently hyperbolic. If Graesser is right, however, and the reader attempts to generate inferences at this point, the regular inferential process will be thwarted; and within the first 650 msec an alternative processing strategy is likely to be invoked. As Graesser himself suggests, readers normally follow Grice's postulates, that whatever is mentioned by an author is relevant and important (379); thus, this sentence will not simply be dismissed as incoherent.

Problems with inferencing at this level, therefore, may launch literary processing - however we construe that. Frederick Bartlett, in his classic study Remembering (1932) showed that with a strange text from another culture, such as "The War of the Ghosts," readers recalling the story tended to rationalize it and bring it into line with the structure and types of explanation of familiar stories. Monika Fludernik's (1996) narrative theory makes important use of a similar point outlined by Jonathan Culler: in Culler's naturalization what is unfamiliar or strange "must be recuperated or naturalized, brought within our ken" (cited, 31). For Fludernik this strategy is central to the process of narrativization, making something a narrative, and it is located in the dynamics of the reading process, that is, it occurs online (34). Readers of Bierce, surprised by the comment on the right lens, will perhaps attempt to reconceptualize what they have understood, and find that a larger model of egotism and obsession is projected by these sentences. Graesser's account, then, argues that six types of automatic inferences are generated while reading narrative; but when normal inferencing is impossible or unsatisfactory we can consider its potential as a model for prompting the initial phases of literary processing.

Of the inference models I will mention, the one of Graesser, et al. (1994) is perhaps the best supported empirically. This is due to what the authors refer to as the "three-pronged method" (Magliano & Graesser 1991). First, readers of a narrative text, reading it sentence by sentence, are asked to think aloud about the meaning of each sentence or to ask questions; this generates a range of potential inferences. Second, theories of inferencing are elaborated based on alternative explanations for different classes of inference and whether a particular class is generated online. Third, experiments are conducted in which behavioural measures of readers are used to estimate whether a particular class of inferences is occurring during reading: for instance, given that forming an inference takes time, a regression analysis of reading times per sentence provides one measure; or if an inference takes the form of a particular word, the time taken by the reader to say the word when it is presented onscreen provides a measure of whether the inference was in the reader's mind or not.

Many other models of inferencing have been put forward. Here is a brief account of several others.

2.2 The minds of characters

Alan Palmer (in Herman's Narrative Theory and the Cognitive Sciences, 2003) puts forward a theory of the constructive activities required of the reader of narrative, in particular, those required to understand the minds of characters. He argues that "Because fictional beings are necessarily incomplete, frames, scripts, and preference rules are required to supply the defaults that fill the gaps in the discourse and provide the presuppositions that enable the reader to construct minds from the text" (325). In other words, literary fiction requires "the sets of instructions that relate to fictional minds" (324).

His first example is a sentence borrowed from Schank and Abelson (1977): "The policeman held up his hand and stopped the car" (325). What is required to understand it? According to Palmer, a number of intermediate inferences are required to create "the consciousness of both the policeman and the driver." Among the nine sentences he lists are: "the policeman perceived the car"; "the policeman came to the belief that he should stop the car"; "the driver perceived the policeman holding up his hand," and so on. Through this series of inferences "we create the driver's mind and the policeman's mind" (326). As this process is "effortless," according to Palmer, it is evident that this series of inferences must occur automatically and online. The inferences are all examples of what Graesser, et al. (1994) refer to as subordinate goal actions: they specify how an agent's action is achieved (here, most are enabling states of consciousness). According to Graesser, however, this class of elaborative inferences is not normally generated online; they are not required in order to generate a coherent account of what occurred. The meaning of this particular sentence is readily understood in relation to a common schema: that is, in the world in which most of us live, this is what policemen do; they frequently stop cars. We have no need to generate the consciousness of the policeman and the driver in order to understand it. This would occur only if something out of the ordinary were at stake: if the policeman were an impersonator, or the driver was running drugs, for instance.

2.3 Cognitive linguistics

Another class of inferences that has been proposed comes from cognitive linguistics. According to this theory, metaphoric mappings such as UNDERSTANDING IS SEEING, or ARGUMENT IS WAR occur at a fundamental level. As Yanna Popova (2002) puts it, "they incorporate universal correlations between sensory-motor experiences and cognitive responses to sensory input" (53), and thereby provide a (normally unconscious) basis for intepretation. The inferences we draw thus occur online as a part of the process of comprehension: "our intuitions and understanding are shaped by them automatically and unconsciously as we process textual information in short term memory" (66). Some important work has now been done towards validating this approach empirically (e.g., Gibbs 2006).

2.4 Deixis theory

Another source of inferences is indicated by deixis theory. In reading narrative we keep track of space and time, of the perspective of the main character or characters, and relations between characters; a shift in one or more of these elements requires inferences that re-establish the deictic centre and the significance of the new perspective - what Peter Stockwell has called "edgework" (2002, 49). For instance, Stockwell has shown that in Wuthering Heights a different set of thematic concerns is apparent when the narrator is Lockwood compared with when it is Nelly Dean: that is, either property issues, or the romance plot with its passionate and Gothic aspects (2002, 52).

2.5 Time in narrative

The role of time has been considered in detail in The Deconstruction of Time by David Wood (1989). While the distinction between story time and discourse time has been fundamental to narratology since the Russian Formalists, Wood's discussion implies that the reader of a narrative may keep track of time from a number of other perspectives. These include: the time of the reader (the reader's current situation, or memories of prior experiences of which they are reminded while reading); the time of the narrator; the time of the plot, in which conflicts or changes unfold; the time of actions (this is the discourse level), in contrast to the time of events (the story level); the time of characters, as they develop or the time in which they perceive or act or are aware of time; and the variations in time of the narrative discourse, including the scene/summary distinction. Apart from the time of actions, we have in our own work found empirical evidence for inferences in all these categories among readers who were asked to think aloud in response to one story. We cannot say, however, which of these types of inference might have been generated online.

2.6 Foregrounding

My research with Don Kuiken has been focused on foregrounding: we predicted that readers would be impelled to slow down when passages high in foregrounded features were encountered. In a number of studies with different literary stories we have found the correlation between foregrounded passages and reading times to be a replicable and robust one (Miall & Kuiken 1994). We also found that in rating passages from the stories, readers reported elevated levels of feeling in response to foregrounding, and higher levels of uncertainty. Thus we proposed that during the longer reading times that coincide with foregrounding, readers were initiating an inferencing process; since foregrounding was found defamiliarizing, readers were beginning the attempt to reconceptualize the meaning of the passage in question, and doing so through the vehicle of feeling, which may provide an avenue to relevant experiences, memories, concepts, or intertextual references. If this is correct, then such inferences would be generated online. We have no direct evidence for the role of inferencing in this context, however. (Given the high level of indeterminacy in the literary texts we used, the inferencing process is much less easy to track online than in the simpler narratives used in the Graesser studies.)

2.7 Character understanding

A number of studies focus on the inferences required to understand character in narrative. I will mention only five. Graesser, Bowers, Olde, White, & Person (1999) show that readers track what characters know, and that these inferences appear to be partly generated online. Catherine Emmott (2003) puts forward contextual frame theory, which argues that readers keep track of the social space in which characters are embedded, including the relative position and importance of characters; she proposes eight processes that readers are engaged in while reading a scene (305), several of which call for an online process of inference generation. Ralf Schneider (2001) develops a view of literary characters as mental models, incorporating psychological traits of characters, their emotions, and aims; inferences are generated both top-down, from readers' world knowledge, and bottom-up, such as readers' emotions about a character. A couple of studies have investigated whether readers track characters' emotions: Gernsbacher, Goldsmith, & Robertson (1992) demonstrated this in an empirical study that suggested emotions are inferred online; De Vega, Leon, & Diaz (1996) showed that readers track the shifting emotions of characters; in both studies emotions were not named in the narratives used, but had to be inferred by the reader. De Vega, et al., make the additional proposal that emotion representations provide a global coherence to narratives, in particular when a narrative is indeterminate (as literary narratives are likely to be).

3. An inference model?

This is just a small sampling of an extensive field, but it already shows how varied the range of inferencing processes might be: Graesser's knowledge-based inferences, Palmer's consciousness-creating inferences, Popova's metaphoric mappings, Stockwell's deictic edgework, David Wood's narrative times, the foregrounding model of Miall and Kuiken, and five different kinds of character inference: what characters know, their social space, Schneider's mental models, and the studies that show tracking of characters' emotions. Apart from Palmer's proposal, which I suggested was implausible, each of these proposals is well grounded in a coherent theoretical approach and several are supported empirically by experimental evidence. Are all these various inferences generated online during reading? The capacity limitations of working memory suggest that this may be unlikely. How should we assign priority to the models if we are seeking a coherent account of online narrative processing? Should we accept, for example, De Vega's view that character emotion provides global coherence? If so, what else forms a part of such a coherent mapping of a narrative?

I chose to consider online inferencing in this paper, because it is clearly fundamental to any cognitive account of reading narrative. If we fail to establish the grounds of reading at this level, then the rest of our cognitive theory is likely to be skewed. But the problem of inferencing is typical of the field of cognitive poetics as a whole: we have a range of proposals, each in itself often highly plausible and some (a few) supported by empirical evidence; but no overall theory of narrative processing that would enable us to assess the priority of the models. And most of the models, such as Palmer, Wood, or Emmott, are one-pronged: they consist only of theoretical proposals; although illuminated by example analyses of passages from specific narrative texts, they involve neither an attempt to map the field of inferences through think-aloud studies, nor any attempts to verify the theoretical proposals through empirical study of readers' behaviours. As Tony Jackson (2002) has observed, we are not in a position to modify or throw out such theories, since we have "no controlled experiments, no quantitative data" (177). Thus, he argues, we do not have in cognitive poetics a real interdiscipline (178).

I have no specific solution for this problem. As Jackson (2003) has also noted, although cognitive poetic theorists "find science-based concepts to be strongly appealing as a means of anchoring interpretation" (193), differences among literary scholars are unlikely to be resolved by importing cognitive constructs (195-6). What seems required is a serious commitment to empirical investigation where this is feasible. Thus I will return to Graesser's model and point out how we might develop it further in the cause of understanding literary reading. And here I will demonstrate a version of the three-pronged method based on our own research.

4. A three-pronged approach

Alongside the elaboration of a model of online inferences, Graesser and his colleagues (Zwaan, Magliano, & Graesser 1995) developed a theory of narrative processing based on the situation model concept (derived from earlier work by Van Dijk and Kintsch, 1983). This includes constructs for space, time, and causal connection, and textual features expected to influence reading: new arguments, and argument overlap. The segments of a short story (usually consisting of one sentence) are coded for these components, such that a 2, for instance, is used to indicate a shift in space at a given segment, and a 1 when there is no shift. Reading times per segment are then gathered from a group of readers who read the story at their normal speed on a computer screen. A multiple regression model is created in which the situation model components are used to predict reading times, on the assumption that a shift in space (for instance) requires additional processing time for the inferences that are generated online. Using data from readers of four short literary stories, their study provided support for the hypothesized situation model, showing that readers tracked temporal and causal dimensions in particular.

Building on this model, we (Miall & Kuiken 1999) selected one of the stories in the experiments of Zwaan, et al., Elizabeth Bowen's (1981) "The Demon Lover." Our earlier studies, as I described above, had shown that reading times can also be predicted by the occurrence of foregrounding. Thus we coded the Bowen story segments for foregrounding and, obtaining the situation model codes and the reading times of his participants from Rolf Zwaan, we set up another regression model that included both the situation model and the foregrounding codes. Our analysis showed that foregrounding was a stronger influence on reading times than any of the situation model components, and it provided an independent source of processing, since it correlated with none of the situation model measures. It is worth mentioning that readers probably had no distinct awareness of the presence of foregrounding in the story, yet the variations in their pace of reading clearly showed foregrounding to be a strong influence. Thus there are grounds for regarding the response to foregrounding as a component of literary reading.

With the regression model we thus have two prongs to the investigation: a theory of literary reading in the form of foregrounding; and an empirical test of its validity based on reader behaviour (i.e, reading times). The third prong in our research was to collect think-aloud protocols from readers, reading a literary story segment by segment on computer screen - a story for which we already have data on foregrounding and reading times ("The Trout," O'FaolŠin 1980-82). Think-aloud comments from 30 readers were analysed for naturally occurring categories (Kuiken & Miall 2001). The classes of commentary per segment were then compared with the story data, including our foregrounding codes, and readers' ratings of story segments. Given such an extensive database of commentaries, it becomes possible to confirm and extend some of the relationships already identified at the segment level. For example, in relation to foregrounded segments, we find that readers tend to quote these more often than non-foregrounded segments (r(82) = .369, p < .001); they also report experiencing feeling more often in response to foregrounded segments (r(82) = .247, p < .05). In addition, readers' ratings for uncertainty correlate at a significant level with the frequency of quotations (r(82) = .306, p < .01). Most interesting, perhaps, are the dynamic features of reading revealed when we examine the occurrence of comments in relation to segment position (i.e., where segments occur in the story).

The frequency of occurrence of a number of commentary types was influenced by the reader's position in the story. Correlating the segment numbers (that is, 1 through 84) with the frequencies of commentary types showed that earlier in the story readers had greater recourse to memory, in which personal or general experience was applied to assist in understanding the story (r(82) = -.267, p < .02); at the same time, they were also more likely to raise questions about the setting of the story (r(82) = -.229, p < .05); they felt less sure about the meaning of what they were reading (r(82) = -.467, p < .001); and early in the story they gave more thought to the motives of the characters (r(82) = -.243, p < .05). In contrast, later in the story readers were more likely to quote from the story (r(82) = .302, p < .01) and they were more actively engaged in anticipating where the story was headed (r(82) = .271, p < .02).

In this example, I have tried to show how research can build incrementally on what is already known. Here, in several stages, we have (1) acquired empirical information about the influence of foregrounding on reading, which is based on a theoretical model derived, ultimately, from the Russian Formalist critics; (2) extended the situation model proposal of Zwaan and his colleagues by building foregrounding on to it; and (3) extended the resulting model further by elaborating the basic correlational analysis with a rich database of readers' comments that confirm and extend the original findings.

5. The experience of literary reading

The research I have just been discussing was carried out specifically to help explore the role of feeling in literary response. One of our earliest findings was that the occurrence of foregrounding correlates with readers' ratings for feeling, feeling being more intense in response to such segments. The commentaries enable us to elaborate some of the experiences behind this finding: that is, we see readers responding to settings, to characters, or to story events whose descriptions are enriched with foregrounding. Feeling may play a role in other ways too: in an informal study I carried out recently in one of my classes, I asked students to choose two passages from a short story by Kate Chopin that they found striking, and to write a short commentary on their responses to the two passages. Analysis of the responses showed that 45 percent of the comments referred to readers' feelings (e.g., amusement, conflict, confusion, curiosity, empathy, excitement, irritation, etc.), while another 12 percent referred to characters' feelings. Thus over half of the comments involved feeling, suggesting that feeling is a primary component of readers' experience of a literary story. While we have speculated about the role that feeling may play in shaping readers' responses, this remains a largely uncharted area: it raises questions about what resources readers bring to bear through feeling, and how far a comprehension model based on feeling might depart from the standard situation model and the familiar mapping of cognitive processes to working memory and long-term memory.

Another finding of our protocol analysis was that, at least while reading the story or shortly afterwards, readers make very few comments about theme (Miall & Kuiken 1999), thus our studies seem to point us towards readers' experiences of the texts they read rather than to any impulse towards interpretation. Although several scholars of cognitive poetics have disclaimed an interest in interpretation, questioning "the standard academic practice of producing yet another intepretation of a text from the canon" (Steen & Gavins 2003, 2), the focus on analysing cognitive components of a text rather than readers' feelings ensures that interpretations tend to reappear rather often in their pages. While these may be valuable and insightful in their own right, they are tangential to the primary aim of cognitive poetics, i.e., a cognitive understanding of the processes at work during literary reading. Here is another indication that cognitivists should focus more explicitly, using empirical methods, on the experience of literary reading. By building on what has already been established, as I have suggested with the example of Graesser's inference model, this discipline would also offer the prospect of a more coherent approach to the issue of reading, rather than, as at present, providing only an unrelated set of proposed cognitive processes.

In this, finally, lies my appeal to cognitive poetics to redefine its aims, and to make it genuinely "three-pronged." Our research should coordinate our expertise in literary theory and analysis with a scientific model of reading based on cognitive science, given that this science not only provides directly relevant information about the mind's processes but also offers an empirical method for investigating theoretical claims. This would position cognitive poetics as a highly relevant component of any future literary studies.


Bartlett, Frederic (1932): Remembering: A Study in Experimental and Social Psychology. Cambridge: Cambridge University Press

Bowen, Elizabeth (1981): "The Demon Lover," in: Brown, C. (ed.): The Collected Stories of Elizabeth Bowen. New York: Alfred Knopf, 661-666

Carter, Ronald, & Nash, Walter (1990): Seeing Through Language. Oxford: Blackwell, 1990

Culpeper, Jonathan (2002): "A Cognitive Stylistic Approach to Characterisation," in: Semino, Elena; Culpeper Jonathan (eds.): Cognitive Stylistics: Language and Cognition in Text Analysis. Amsterdam and Philadelphia: John Benjamins, 251-277

De Vega, Manuel; Leon, Inmaculada; & Diaz, Jose M. (1996): "The Representation of Changing Emotions in Reading Comprehension," Cognition and Emotion, 10.3, 303-321

Emmott, Catherine (2003): "Constructing Social Space: Sociocognitive Factors in the Interpretation of Character Relations," in: Herman, David (ed.): Narrative Theory and the Cognitive Sciences. Stanford: CSLI Publications, 295-321

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

Gernsbacher, Morton A; Goldsmith, H. Hill; & Robertson, Rachel R. (1992): "Do Readers Represent Characters' Emotional States?" Cognition and Emotion, 6.2, 89-111

Gibbs, Raymond W., Jr. (2006): "Metaphor Interpretation As Embodied Simulation," Mind & Language, 21.3, 434-458

Graesser, Arthur C.; Bowers, Cheryl; Olde, Brent; White, Katherine; & Person, Natalie K. (1999): "Who Knows What? Propogation of Knowledge Among Agents in a Literary Story World," Poetics, 26.3, 143-175

Graesser, Arthur C.; Singer, Murray; & Trabasso, Tom (1994): "Constructing Inferences During Narrative Text Comprehension," Psychological Review, 101.3, 371-395

Hemingway, Ernest (1932): Death In the Afternoon. New York: Scribner's

Jackson, Tony E. (2002): "Issues and Problems in the Blending of Cognitive Science, Evolutionary Psychology, and Literary Study," Poetics Today, 23.1, 161-179

Jackson, Tony (2003): "'Literary Interpretation' and Cognitive Literary Studies," Poetics Today, 24.3, 191-205

Kintsch, Walter (1998): Comprehension: A Paradigm for Cognition. New York: Cambridge University Press

Kuiken, Don; & Miall, David S. (2001, February): "Numerically Aided Phenomenology: Procedures for Investigating Categories of Experience," FQS. Forum: Qualitative Social Research, 2.1

Magliano, Joseph P., & Graesser, Arthur C. (1991): "A three-pronged method for studying inference generation in literary text," Poetics, 20.3, 193-232

Miall, David S.; & Kuiken, Don (1994): "Foregrounding, Defamiliarization, and Affect: Response to Literary Stories," Poetics, 22.5, 389-407

Miall, David S.; & Kuiken, Don (1999): "What is Literariness? Three Components of Literary Reading," Discourse Processes, 28.2, 121-138

Miall, David S.; & Kuiken, Don (2002): "A Feeling for Fiction: Becoming What We Behold," Poetics, 30.4, 221-241

O'FaolŠin, S. (1980-82): "The Trout," in: The Collected Stories of SeŠn O'FaolŠin, Vol. I. London: Constable, 383-386

Palmer, Alan (2003): "The Mind Beyond the Skin," in: Herman, David (ed.): Narrative Theory and the Cognitive Sciences. Stanford: CSLI Publications, 322-348

Popova, Yanna (2002): "The Figure in the Carpet: Discovery or re-cognition," in: Semino, Elena; Culpeper, Jonathan (eds.): Cognitive Stylistics: Language and Cognition in Text Analysis. Amsterdam and Philadelphia: John Benjamins, 49-71

Rabinowitz, Peter J. (1996): "Reader Response, Reader Responsibility: Heart of Darkness and the Politics of Displacement," in: Murfin, R. C. (ed.): Joseph Conrad: Heart of Darkness, 2nd ed. Boston, New York: Bedford/St. Martin's, 131-147

Schank, Roger C.; & Abelson, Robert P. (1977): Scripts, Plans, Goals, and Understanding: An Inquiry into Human Knowledge Structures. Hillsdale, NJ: Erlbaum

Schneider, Ralf (2001): "Toward a Cognitive Theory of Literary Character: The Dynamics of Mental-Model Construction," Style, 35.4, 607-640

Steen, Gerard, & Gavins, Joanna (2003): "Contextualizing Cognitive Poetics," in: Gavins, Joanna; Steen, Gerard (eds.): Cognitive Poetics in Practice. London: Routledge, 1-12

Sternberg, Meir (2003): "Universals of Narrative and Their Cognitivist Fortunes (I)," Poetics Today, 24.2, 297-395

Stockwell, Peter (2002): Cognitive Poetics: An Introduction. London: Routledge

Tsur, Reuven (1992): Toward a Theory of Cognitive Poetics. Amsterdam: North-Holland

van Dijk, Teun A.; & Kintsch, Walter (1983): Strategies in discourse comprehension. New York: Academic Press

Wood, David (1989): The Deconstruction of Time. Atlantic Highlands, NJ: Humanities Press

Zwaan, Rolf A.; Magliano, Joseph P.; & Graesser, Arthur C. (1995): "Dimensions of Situation Model Construction in Narrative Comprehension," Journal of Experimental Psychology: Learning, Memory, and Cognition, 21.2, 386-397

return to reading

Document created October 12th 2007