Contents
1. Ambiguity and context
2. Additional evidence on processing effects
3. The locus of context
4. Recontextualisation and the "shaping-eye" hypothesis
5. Conclusion: the constructivist model
Notes
List of Works Cited
Stanley Fish is on record for holding that literary theory in general and narratology in particular are "impossible" projects. Narratologists usually reciprocate by marking him down as counterproductive or by pointedly ignoring what he is saying. The present essay, however, argues that narratologists can learn much from Fish's notion of a "literature in the reader." It begins by reviewing the phenomenon of ambiguity, a topic that has attracted a certain amount of renewed attention in recent reading-oriented approaches, and by reorienting and supplementing Fish's examples with the test cases cited in cognitive linguistics, pragmatics, and artificial intelligence. Fish's process-oriented analysis of ambiguity is seen as paving the way toward a cognitive model of context on the one hand and to an account of "recontextualisation" heuristics on the other. Building on Fish's recontextualisation experiments, the essay makes an attempt to identify the axioms that characterise the new narratological projects.
One of the most serious flaws of classical narratology was that it allowed itself to be trapped in an atomistic-holistic doublebind. Seeing a story as a finished product, classical narratology crafted the tools, first, to take it to pieces, and then, to reassemble the pieces to explain the final state. Working both bottom-down from the finished product as well as bottom-up from seemingly independent small units (themselves deduced from the final product), classical narratologists not only accepted the paradoxical rationale of the hermeneutic circle but at the same time turned a blind eye to the dynamics of the text-reader interface that creates stories, plots, characters, and narrative situations in the first place.
The distinction between "text-as-product" and "discourse-as-process" (Brown and Yule 24), which highlights the aporias of the classical approach, is typically demonstrated by referring to the cognitive heuristics of disambiguation. Unfortunately, many of the strikingly contrarious positions on ambiguity tend to get sidetracked by a centripetal question -- the question whether "true" ambiguity exists at all. Consider the following selection of fairly recent views.
(1)
In real life -- that is, among real language users -- there is no such thing as ambiguity. (Mey, Pragmatics 7)
(2)
Narrowly defined, almost every utterance is ambiguous. In fact, almost every utterance is multiply ambiguous, with possible semantic interactions among its individual ambiguous constructions. [. . .] it is thus quite typical for an utterance to have dozens, or [End of p. 375] even hundreds, of possible propositional interpretations. However, speaker and hearer are normally able to select a single one of these interpretations without even realizing that they have made a choice. It is generally agreed that this choice is a function of the context; but to define the function, as opposed to claiming that it exists, is no easy task. (Sperber and Wilson, "Irony" 298)
(3)
[A] reader-oriented analysis of a text [. . .] can take the name "formalism" -- or "new formalism," if you like -- as a positive label. Its rhetorical purpose, then, is to call attention to the relative but crucial place of "good reading" [. . .] in the sense that, whatever its ambiguities, the word "cat" is a noun and not a verb, and refers to an animal, not an inanimate object, a mammal, not a fish, a singular not a plural, and the like; it is the opposite of obviously bad reading. [my] essays [. . .] try to formulate a framework for interpretations of narratives which bridges the gap between the text in its structured complexity and its effect upon its readers. (Bal, Story -Telling 11)
Presupposing a pragmatically rich environment of face-to-face communication, Jacob Mey considers ambiguity a distractive subject not meriting closer investigation. Against this, Dan Sperber and Deirdre Wilson stress (i) the pervasiveness of ambiguity in natural language, (ii) the cognitive relevance of ambiguity, particularly the computational effort involved in the resolution of ambiguity, (iii) the role of context as a disambiguating factor, and (iv) the extent of our ignorance in these matters. Lastly, quote (3) presents a view from a narratological perspective, tentatively suggesting a compromise position. Assuming that there are "good readings" that are free of, or uncontaminated by, ambiguity, Mieke Bal proposes to rescue the narratological project on the methodological basis of "formalism." Formalism, as illustrated in Bal's example of the word cat, can be put to work on both the nonambiguous (positive) meanings of the word as well as on its complement (i.e., non-cat) meanings. Hence formalism affords a decision procedure for distinguishing good from bad readings. Let us focus on good readings, Bal suggests, and narratology is in business, again.
Generally speaking, most modern narratologists will be happy with Bal's aims as expressed in (3). Like Bal, narratologists today believe in the teachability of narratology and the heuristic usefulness of models. Like Bal, they want to view texts in their "structured complexity," moving from a "reader-oriented analysis" to a systematic appreciation of "effects." Unlike Bal, however, many narratologists will be reluctant to use the term "formalism" for this kind of undertaking. They are also likely to question whether it is a good idea (i) to divide a text into ambiguous and unambiguous regions, and (ii) to let analysis focus on those parts that are unambiguous. In fact, Bal's notion of a safe area of non-ambiguity flies in the face of a widespread consensus that ambiguity is a central aspect of literary texts, possibly even a necessary condition. Moreover, the point so forcefully made by Sperber and Wilson in (2) is a point about natural language in general. Indeed, under scrutiny, Bal's seemingly commonsensical assumptions on the unambiguous aspects of the word cat are the most vulnerable part of her argument. Presumably, the ambiguity which Bal admits is that cat can refer to various types of felines (say, domestic and wild cats). However, on the evidence of dictionaries like the OED or Webster's Third, one finds that cat is exactly "multiply" ambiguous in Sperber and Wilson's sense, and that means ambiguous to the order not [End of p. 376] of merely two, but of fifteen to twenty meanings. On that basis, practically all of Bal's assertions about what a cat is not turn out to be simply wrong. Contrary to what Bal believes is obvious, the larger dictionaries advise us that cat can be used as a verb, can refer to a kind of fish (a catfish), can refer to a number of objects or persons, and (at least arguably) can be a plural as well as a singular.1 Even though it may seem unfair to challenge Bal's example by holding it against two encyclopaedic sources, one cannot, to save Bal's argument, exclude all specialist words, all trade names (cat: a caterpillar tractor), all nicknames ("the cat has done it again -- won in three straight sets"), all shortened forms (catfish; cat o' nine tails), all metonymic and metaphoric terms (a person resembling a cat), all politically incorrect and misogynist terms (to go catting -- to search for a sexual mate; cat -- a malicious woman). Paradoxically, then, Bal's decidedly bad reading of cat demonstrates better than any other argument that the realm of good readings is a horribly confined space and that the notion of restricted ambiguity (or, conversely, of partial univocity) is a dead end. Hence, the pervasiveness of ambiguity is best accepted without any qualification whatsoever. Rather than Mey's claim of the non-relevance of ambiguity or Bal's notion of a safe area of good readings the order of the day is Meir Sternberg's wholly uncompromising "Proteus Principle" -- "the many-to-many correspondences between linguistic form and representational function" (Sternberg, "Proteus" 112; see also Jahn, "Frames" 450).
Unfortunately, the Proteus Principle is an "anti-foundationalist" axiom, blocking the road rather than opening a door or levelling the ground. Whoever wants to rebuild narratology on such a negative foundation must accept ambiguity both as a stumbling block (something to be got rid of) and something whose processual dynamics and aesthetic significance is in urgent need of explanation. As it happens, this is where Stanley Fish enters the picture.
Oddly enough, the key "ambiguity" is missing from the indexes of Is There a Text in This Class and Doing What Comes Naturally, the two collections of essays that this paper will mainly refer to. Yet the essays contained in these two volumes are full of references to ambiguity, and an indexer could profitably arrange them under subdescriptors such as "in literature," "in natural language," "in puns," "local," "global," "illocutionary," etc. In fact, it is mainly the type of "temporary" or "resolvable" ambiguity that helps Fish argue his case for a "literature in the reader" (Is There ch. 1). One of the examples adduced by Fish is the following passage from a seventeenth-century sermon by Lancelot Andrewes.
(4)
He is found of them that seeke Him
not
but
of them that seeke Him [End of p. 377]
never
but found (qtd Is There 184)
Andrewes's text is clearly so difficult or -- in modern terms -- so uncooperative that an impatient reader may easily give up. Yet, as Fish shows, the passage can be construed in such a manner as to communicate a consistent and meaningful statement. The trick is to realize that the first two lines, taken jointly, formulate a first proposition which is then supplemented by a second proposition. To the first proposition He (God) is found by those who do not seek him, the rest of the passage adds the corollary and by those who do seek him he is always ("never but") found. Globally, then, the passage asserts that God is found by everybody, whether they seek him or not. As messages go, this is slightly anticlimactic, in fact it is as anticlimactic as a finished puzzle -- but here Fish's point is precisely that the passage is valuable not for its ultimate message but for the mental exercise that it affords as one labours over it. Recognition of this fact, Fish says, slips through the net of all product-oriented formalist approaches because "the only making of sense in a formalist reading is the last one." Fish, in contrast, holds that "everything a reader does, even if he later undoes it, is part of the 'meaning experience' and should not be discarded" (Is There 3-4).
The rift between product and process-oriented approaches is even stronger in disciplines like linguistics and artificial intelligence (AI). Consider standard examples such as the following:
(5)
Flying planes can be dangerous. (Chomsky 21)
(6)
Time flies like an arrow. (Brainerd 211)
In Chomskyan generative grammar, sentences like (5) are usually cited as evidence that an adequate competence grammar must be able to account for all of its virtual structural interpretations (whether an actual hearer in a specific situation is aware of them or not). Structurally ambiguous sentences are also the AI researchers' favourite examples for testing computational language-processing algorithms. To their delight, AI programmers quickly discovered the heuristics that enabled them to produce great numbers of virtual structural descriptions. As a matter of fact, their computers soon outperformed ordinary humans in deriving all possible parses of items like (6) (which actually admits of three readings). Then, to their dismay, the AI people found that what was needed for practical applications was not a host of virtual parses, but the human parser's "tunnel vision" (now considered a positive term) of seeing only whichever "most likely" parse was demanded by the pressure of the context. And oddly enough, even "zero" contexts apparently led to "most likely" parses. What was happening here was little accessible to rational explanation, let alone computational imitation. So intractable was the link between dismabiguation and context that many AI researchers and many formal linguists decided to shelf the project of a "theory of situations." Some AI theorists persevered, however, and in the mid-seventies, two basic but workable AI models of context became known as "frame" and "script" theories. Despite their [End of p. 378] obvious restrictions and shortcomings, the concepts postulated in these theories were quickly taken up in the cognitive and social sciences as well as in the humanities.2
One prominent linguist in the sixties took great pleasure in arguing against the grain of his interpretive community's product orientation. Like Fish, Charles F. Hockett was an inveterate inventor of intriguing test cases. (9), below, is one of them:
(7)
A man eating fish ...
... on Friday is not necessarily a Catholic.
... called the piranha is found in the tropical waters of Brazil.
... has an unbalanced diet. (Hockett, "Grammar" 226)
Hockett's test case is clearly only mildly ambiguous if enunciated carefully or spelled more helpfully, if listened to attentively or read in maximally conscious awareness. (It is questionable, however, whether such "ideal" scenarios ever come up in real life, or whether cognitive processing ever depends on them.) At any rate, what Hockett wants to focus on in (7) is the disambiguating function of subsequent context. As a sentence-initial fragment, a man eating fish is locally ambiguous because it may (be understood to) refer to a man who likes to eat fish or to some kind of predatory fish. If the sentence continues on Friday is not necessarily a Catholic then a man eating fish refers to a piscivorous man; if the sentence continues called the piranha is found in the tropical waters of Brazil the phrase is understood as referring to a hominivorous fish; and if it continues has an unbalanced diet then neither reading is suggested or confirmed and further disambiguating evidence is needed. The question raised by Hockett is this: when we hear a man eating fish, do we suspend processing of the text until some disambiguating clue comes up as further context rolls by; or do we settle for one of the readings immediately, making a stab at it and accepting the ill chance of going wrong, accepting also the possible necessity of having to backtrack and to revise?
Hockett's answer to this was that readings are always construed as early and quickly as possible, and this hypothesis is strikingly confirmed on the evidence of the so-called garden-path effect. The prototypical garden-path sentence is (8), an item invented by Thomas G. Bever; item (9) shows Hockett's awareness of the phenomenon, even though he does not identify it by the garden path term.
(8)
The horse raced past the barn fell. (Bever 316; Marcus ch. 9)
(9)
After John had started the car pulled up to the curb. (Hockett, "Where the Tongue Slips" 238)
As one can see, both (8) and (9) begin as seemingly ordinary sentences and then cause a sudden processing difficulty requiring backtracking and revision. Bever's example is [End of p. 379] rather a debilitating garden path -- one has (to be told) to re-interpret the initial segment as "the horse that was raced past the barn" to see that (8) is, after all, a perfectly grammatical sentence. Hockett's own example, (9), presents a rather milder garden path, easily "cured," in its written incarnation, by the insertion of a comma. (Note, however, that the common impulse to emend [8] to read ... past the barn and fell would have to count as a misreading.) As has been recognised since, the garden-path effect is quite a general phenomenon that can also occur in jokes, riddles, and other narrative genres, including fiction. To point up the narratological relevance of literary garden paths, consider the following passages from Ambrose Bierce's "An Occurrence at Owl Creek Bridge," the story of a man who is about to be hanged.
(10)
As Peyton Farquhar fell straight downward through the bridge he lost consciousness and was as one already dead. [. . .] Then all at once, with terrible suddenness the light about him shot upward with the noise of a loud splash; a frightful roaring was in his ears, and all was cold and dark. The power of thought was restored: he knew that the rope had broken and he had fallen into the stream. [. . .]
[Five pages of text follow, depicting Farquhar's
successful escape and his arrival at home, where he is greeted by his
wife.]
Ah, how beautiful she is! He
springs forward with extended arms. As he is about to clasp her he feels
a stunning blow upon the back of the neck; a blinding white light blazes
all about him with the sound like the shock of a cannon -- then all is
darkness and silence!
Peyton Farquhar was dead; his body, with a broken neck, swung gently from side to side beneath the timbers of the Owl Creek Bridge. (13, 18)
This closing passage is the story's surprise ending, and it produces the precise phenomenological shock that attends other garden-path constructions. Instructively, the point at which the reader is led astray can be pinpointed exactly -- it is the diegetic statement that Peyton Farquhar "knew that the rope had broken and he had fallen into the stream." For an authorial narrator to assert that Farquhar knew that p means to assert that p is factually true in the world of this fiction (this follows from the logic of so-called factive expressions such as know). Naturally, a reader prefers to believe that an authorial narrator would not assert what s/he knows to be false. It is only in retrospect that one can activate a fall-back frame of interpretation, recuperating good sense by assuming that the narrator picks this particular point in the story to delegate focalization to the focalizer, relinquishing his or her mimetic authority in the process.3
Targeting formalism at large, Fish derides all discovery procedures that are "incapable of finding value in temporal phenomena" (Is There 155). Narratology, in Fish's view, is a system of "formidable apparatuses" (Doing 567n2) in which the "practitioner [. . .] gives himself over to the theoretical machine, surrenders his judgement to it, in order to reach conclusions that in no way depend on his education, or point of view, or [End of p. 380] cultural situation" (Doing 319). This is all the more remarkable, Fish argues, when it can be easily shown that even meaning and truth are often (perhaps always) dependent on situational factors.
(11)
In the penultimate chapter of How To Do Things with Words, J. L. Austin presents a sentence and asks us to consider it. The sentence is "France is hexagonal," and the question he puts to it is a very familiar one in analytical philosophy: Is it true or false? The answer is not so familiar. It depends, says Austin: "I can see what you mean by saying that it is true for certain intents and purposes. It is good enough for a top-ranking general, perhaps, but not for a geographer" (p. 142). I[n] other words, the truth or falsehood of a sentence is a function of the circumstances within which it is uttered, and since it is always uttered within some set of circumstances or others, it is not in and of itself either true or false, accurate or inaccurate, precise or imprecise. (Is There 197)
Going one step further than Austin, Fish subsequently claims that it is not even the situation per se that is relevant, but the construction of the situation in the mind of the person hearing or articulating the true-or-false judgement. Context, Fish concludes, is in the mind. It is a position that seems to be gaining ground: as Sperber and Wilson argue at length more recently, context is never objectively given but always actively chosen (Relevance ch. 3.3).
The crucial strategic move in Fish's (and Austin's) explication of France is hexagonal is to "reconceive" the utterance in different situations. Recontextualisation is a "constructivist" procedure whose surprise potential is in many ways reminiscent of that of garden paths.4 In the realm of literary interpretation, the most radical example is Fish's famous Buffalo experiment, reported in "How To Recognize a Poem When You See One" (Is There ch. 14). In this experiment, Fish presented a list of names to a class of poetry students asking them to interpret it as a religious poem. Supposing that Fish's account is to be trusted (see Scholes for a critical comment) Fish's students acquitted themselves surprisingly well:
(12)
As soon as my students were aware that it was poetry they were seeing, they began to look with poetry-seeing eyes, that is, with eyes that saw everything in relation to the properties they knew poems to possess. [. . .] Skilled reading is usually thought to be a matter of discerning what is there, but if the example of my students can be generalized, it is a matter of how to produce what can thereafter be said to be there. Interpretation is not [End of p. 381] the art of construing but the art of constructing. Interpreters do not decode poems: they make them. (Is There 326-27)
It is at this point, precisely, that Fish executes a markedly constructivist turn. Unfortunately, he neither proceeds to embrace constructivism as a discipline nor does he enter into a dialogue with "radical" constructivists such as Heinz von Foerster, Paul Watzlawick, Humberto Maturana, or Siegfried J. Schmidt. Yet, to novices, his account is both more plausible and didactically useful than anything offered by the dyed-in-the-flesh constructivists themselves. Constructivists have an unfortunate habit of beginning their exposition by stating that the human mind is an "autopoietic system" -- a jargon that, in a classroom, comes across as nothing but the usual bit of intimidation. Fish's essay, in contrast, works well for students, arguing the case for constructivism in both entertaining and enlightening fashion.
Even though in his subsequent work Fish turns to what he considers anti-theories, embracing models like anti-foundationalism and deconstructivism, the questions raised by his account continue to be relevant to literary theory in general, and narratology in particular. Evidently, from Bal's formalist notion of "good readings" we must progress to Fish's notion of situated readings, readings that "produce what can [. . .] be said to be there" (Is There 327). To illustrate the impact of situatedness on a specifically narratological scenario, let us briefly conduct a recontextualisation experiment on a visual narrative. Consider the World War I propaganda poster shown in (13).
(13)
Daddy, what did you do in the great war? (Hillier 234).
A picture says more than a thousand words, and the spectator's gaze may variously focus on detail such as the pattern on the curtains, the characters' garments, the Alice-like appearance of the little girl, and so on. Apart from such descriptive detail, the narrative action that one constructs is that of a boy playing with a set of tin soldiers, a girl leafing through an illustrated journal, and a male adult (clearly, Daddy) sitting bemusedly in his chair. Proxemic relations (the characters' use of space) further suggest that it is the girl who asks the question quoted in the caption -- "Daddy, what did YOU do in the Great War?" Expecting Daddy's conversational turn, the viewer notes his troubled face, and the character's mind becomes as good as transparent: Nothing, Daddy thinks, I did nothing to help the war effort; how can I live with this shame (or something very much like it). The propaganda message to the public clearly is, do not become somebody like Daddy.[End of p. 382]
Recontextualising the picture is easy. To begin with, let the story-NOW be not post-World War I, but post-World War II; second, let the setting be not somewhere in the United Kingdom, but somewhere in Germany. Now the girl's question acquires a different slant, and Daddy may well think, Wish I had not done what I did! What suggests itself here, then, is a reading whose "cognitive payoff" (Jahn, "Speak" 177) easily supresses nonconfirmatory "facts" such as (i) that the original artist could not possibly have intended such a reading, or (ii) that the expression "Great War" more or less unambiguously refers to World War I. Clearly, the recontextualised reading is not a good reading in Bal's sense; however, the question at this point is not whether one [End of p. 383] reading is good, one bad; or one superior, and one inferior; or one correct, one incorrect -- rather, what is relevant is that, given the beliefs and memories of a culturally situated interpretive community, the second reading will come naturally and inescapably. Essentially, the first and "intended" interpretation of the picture is based on the same cognitive principles that produce other (but not necessarily "bad") readings in a different contexts.
Like all approaches, constructivism must define its conception of three basic concepts: expressions, mental models, and reality. While many semiotic-triangle type models make an attempt to establish a firm link between expression and reality (either by positing a signifier/signified relation or an expression/referent relation),5 the constructivist view denies the existence of "objective" correlatives or referents, and prefers to consider reality as a mere pattern of stimulations, "a succession of noises or marks," as Fish puts it (Doing 295). Secondly, the relationship between expressions and mental models is taken to be governed by the Proteus Principle. Thirdly, the relationship between reality and mental models is one of "seeing X (reality) as Y." Y, in this formula, is a Peircean "interpretant" (Morris 2-3; Peirce 2.õ228), a product of mental operations, interpretive strategies, and preference rule systems (Jackendoff, Semantics ch. 8). It is these operations and strategies that are the prime explicanda in a constructivist analysis.6
A constructivist conception of the expression/mental-models/reality triad informs most branches of what has come to be known as "postclassical narratologies" (Herman, "Narratologies"), especially the subdisciplines of historiographic narratology, postmodernist narratology, feminist narratology, possible-worlds narratology, constructivist narratology, and natural narratology. These postclassical approaches have all acquired a better awareness of the processual character of texts, they pursue more or less pronounced cognitive orientations, and they make a dedicated attempt to situate narrative in the cultural and historical contexts of interpretive communities. Five positions, in particular, stand out as postclassical narratological beliefs.
1. Proteus rules OK. Postclassical narratology prefers the Proteus Principle over the principle of univocity. Despite the fact that the Proteus Principle presents a negative and complicating condition, it enables the critic to recognise that texts materialise in a process of reversible decisions, and that text and world can only be interpreted in the frames and scripts of culturally inherited mental contexts.[End of p. 384]
2. Narratives are both products and processes. Classical narratological models arrive at their categories of events and existents by considering the text as a finished product and judging all textual detail from a global and retrospective view. Process-oriented analyses, in contrast, focus on the stepwise integration and combination of textual information, paying due attention to backtracking and revision. Both Sternberg (Expositional Modes) and Perry draw attention to literary uses of cognitive "primacy" and "recency" effects in story construction and characterisation; Perry, in particular, presents an account that establishes "a place for rejected meanings" (355). Ultimately, the main goal of the cognitive approach must be to develop a combined process-and-product model, one that builds complex conceptual structures by cumulatively integrating local interpretive decisions, including wrong turns (for an example, see the model proposed in van Dijk and Kintsch).
3. Narratives are interpretively situated. Rather than pursue the project of drawing up a timeless inventory of abstract laws and categories, postclassical narratology reconsiders and redefines its units in the shaping contexts of historical, cultural, and pragmatic parameters. One of the liberating consequences of situatedness is that texts and genres are no longer definable by sets of inherent qualities. It is a commonplace today that something can be read "either as literature or as history" or that "the same sentence can have different meanings in poetry and prose" (Culler, Structuralist 123).
4. Story is an interpretant; mimesis is an effect. While classical narratology strictly distinguishes between discourse as narrative's signifier and story as narrative's signified, the assumption that action is something that exists "prior to" and "independently of narrative presentation" is strongly challenged by Jonathan Culler, who identifies story "not as the reality reported by the discourse but as its product" ("Fabula" 28-9). Culler's position is clearly related to Fish's argument that readers "make" texts and that signifiers have no objective correlates in a world of signifieds or referents. Postclassical narratology today generally acknowledges that story, causality, and chronology are readerly constructs. Similarly, mimesis is no longer assumed to present an imitative picture of the world; indeed, mimesis does not even work for the narrative of verbal events because fictional speech, as Fludernik has shown, is an evocation of speech rather than an imitation or a reproduction (Fictions ch. 8).
5. Last but not least, cognitively oriented narratologies are likely to heed Wallace Chafe's advice that "the study of narratives can help us understand the workings of the mind" (96). According to Fish, an interpretive community supplies its members with the "standard stories" that situate "contextless" sentences and fill zero contexts with default settings. Using a variety of metaphors, theorists from various disciplines have suggested that a person's life plans are scripted on fairy-tales (Berne), that "everyone is a novelist" structuring his or her own life story (Dennett), that everyone has a "narrative identity" (Ricoeur), and that memory is a database of stories retrievable by a procedure called "reminding" (Schank). A number of mental narrativising devices were in fact already noted by Cohn in her analysis of "memory monologues" (ch. 6), and Mark Turner has recently demonstrated that "most of our experience, our knowledge, and our thinking is organized as stories" (i). Evidently, "mental narrative" [End of p. 385] is not just a subject in search of interdisciplinary treatment but also a challenge for a constructivist narratology that adapts and modifies itself as it grows.
1. Neither of the larger dictionaries explicitly lists a plural meaning of cat. I am assuming, however, that a fisherman talking to another fishermen would be more likely to say "Plenty of cat [=catfish] around" than "Plenty of cats around."
2. See Minsky ("Framework") and Schank and Abelson (Scripts) for the founding proposals of frame and script theory. For a narratological exploitation of these concepts, see Perry ("Literary Dynamics"), Fludernik's The Languages of Fictions and Towards a "Natural" Narratology, as well as the essays by Jahn listed in the references.
3. For further discussions of garden-path stories, see Perry ("Literary"), Mey ("Pragmatic"), and Jahn ("Speak").
4. As has been noted elsewhere, recontextualisation may serve both analytic and creative purposes. One of the most striking instances of the latter is William Carlos Williams's famous "plums poem" -- see Jonathan Culler (Structuralist 175) for a specific and Franz Stanzel for a more general account of what is now generally termed "found poetry." Recontextualisation also underlies many intertextual phenomena, especially the relationship between a text and its pretext(s).
5. See Whiteside for a useful survey of triangular semiotic models.
6. On a more general level, it is also legitimate to inquire into the "ecological viability" of interpretive strategies, including those that may lead to misinterpretations (Jackendoff, "Problem" 161). One recent view (actively pursued by AI researchers) is that our ability to perceive, think and speak the way we do is directly related to our being misled in certain circumstances (ambiguities, optical illusions, garden paths, and so on).