Edited by D. Olson & N. Torrance (pp. 123- 140) New York: Cambridge University Press

Скачать 73.83 Kb.
НазваниеEdited by D. Olson & N. Torrance (pp. 123- 140) New York: Cambridge University Press
Размер73.83 Kb.
  1   2   3

In Modes of Thought, Edited by D. Olson & N. Torrance (pp. 123- 140) New York: Cambridge University Press.

Inference in Narrative and Science

Keith Oatley

Centre for Applied Cognitive Science, Ontario Institute for Studies in Education

and Department of Psychology, University of Toronto

Introduction: Narrative thinking

“Happy families are all alike; every unhappy family is unhappy in its own way.” So wrote Tolstoy (1877) in the famous opening sentence of Anna Karenina. We are prompted to think: are we going to read about a happy family? No — if the author has written a sentence with a first part about happy families and a second part about every unhappy family this story must be about a family that is unhappy in a distinctive way. So: with this sentence Tolstoy prompts our minds into motion. Further thoughts may occur: perhaps there is a brief frisson of emotion, or a flash of memory. “Will this unhappy family be unlike mine?”

Another famous first sentence is by I.A. Richards (1925): “A book is a machine to think with” — what kind of thinking do books enable us to do? In particular, what do books of fiction enable us to do as compared with books of science? Bruner’s (1986) proposal that there are two modes of thought, a narrative mode for thinking about human action and a paradigmatic mode for thinking about mechanisms and natural science is a productive one (in his chapter in this volume Bruner calls them agentive and epistemic). Here I suggest that the narrative mode is useful in many domains, not just in fiction. It is based on distinctive psychological processes, and it is used widely in explanation, including scientific explanation.

One of the properties of the narrative mode is that objects expressed in this mode — that is to say stories about agents — slip easily into the mind. The story of Anna Karenina is an example. The mind is more resistant to objects based on the paradigmatic mode. At least such objects need elaborate cultural assistance to allow them to enter the mind, for example knowledge about how to reason mathematically, or how to understand statistical data presented in tables or diagrams, or how to draw inferences validly from scientific experiments.

Scientists do not restrict themselves to mathematics, diagrams, and experiments. They use narrative too. Here are a few sentences from Richard Feynman’s famous textbook of physics (1963). They are from his introduction to Newton’s third law of motion which is that: “For every action there is an equal and opposite reaction.” Feynman introduces his discussion in narrative mode, with a story about two agents called particles:

Suppose we have two small bodies, say particles, and suppose that the first one exerts a force on the second one, pushing it with a certain force. Then, simultaneously, according to Newton’s Third Law, the second particle will push on the first with an equal force, in the opposite direction . . . (p. 10-2)

A little further down the page, Feynman switches to the paradigmatic mode, with an equation for these equal and opposite forces:

dp1/dt = - dp2/dt.

In this equation, the momentum of Particle 1 is p1 and the momentum of Particle 2 is p2, force is the rate of change of momentum (dp/dt), and the equation expresses the idea that these two forces are equal and opposite, but the equation or indeed the reason why one might wish to write it are incomprehensible without knowledge of calculus, a Western cultural product. What Feynman is doing, of course, is to begin his explanations in narrative mode in order to connect with our ordinary human intuitions, which will then be formalized by means of the cognitive prostheses of differential equations, from which a whole set of new inferences can be made that are unavailable to people who are naive in mathematics.

Let us now take an example from fiction: Sherlock Holmes has just examined two ears sent in a cardboard box to a lady in Croydon. Lestrade, the inspector of police, thinks that the ears must have been sent as a practical joke by some medical students. Holmes speaks first, then Lestrade:

“. . . this is not a practical joke.”

“You are sure of it?”

“The presumption is strongly against it. Bodies in the dissecting rooms are injected with preservative fluid. These ears bear no sign of this. They are fresh, too. They have been cut off with a blunt instrument, which would hardly happen if a student had done it . . . We are investigating a serious crime (Doyle, 1981, p. 892).”

Soon after this speech Holmes notices that the ears of the recipient of the gruesome package have similar conformations to those of one of the ears in the box, and infers that this ear belonged to a close relative or hers. He explains to Watson that he noticed this ear's characteristics because had studied ears, and written two articles in the previous year’s Anthropological Journal on their idiosyncratic shapes.

Having given examples from science and from fiction, let me add one further example: from autobiography. In 1907 the founder of American pragmatism and semiotics, C.S. Peirce wrote an article for the Atlantic Monthly in which he recounted the following incident. The article was rejected, but an account of the incident, published posthumously, was reproduced in an article by Sebeok and Umiker-Sebeok (1983) which has served as inspiration for my chapter, also both a source and a key for some of Peirce’s writings. (They use Peirce, 1935-1966, as their source, for quotations from Peirce here, I give the page numbers in their article.)

In 1879 Peirce had sailed in a steam ship from Boston to New York to attend a conference. After disembarking he discovered he had left behind a valuable watch that had been been given him by the US Coast and Geodetic Survey, as well as his overcoat. He went immediately back to the ship to find his things gone. He arranged to have all the ship's waiters lined up before him:

I went from one end of the row to the other, and talked a little to each one, in as dégagé a manner as I could, about whatever he could talk about with interest, but would least expect me to bring forward, hoping that I might seem such a fool that I should be able to detect some slight symptom of his being the thief. When I had gone through the row I turned and walked from them, though not away, and said to myself, “Not the least scintilla of light have I got to go upon.” But thereupon my other self (for our communings are always in dialogues) said to me, “But you simply must put your finger on the man. No matter if you have no reason, you must say whom you will think to be the thief.” I made a little loop in my walk, which had not taken a minute, and as I turned toward them, all shadow of a doubt had vanished (Sebeok & Umiker-Sebeok, 1983 p. 11-12).

Peirce took one man aside, but could not persuade him by reason, threat, or the offer of $50, to give up the lost objects. He had the man followed . . . After a complex pursuit, Peirce regained his possessions. The man he picked out had indeed stolen them.

In “The Adventure of the Cardboard box” shortly after inferring that the parcel containing the two ears could only be explained by a crime, Sherlock Holmes gave one of the best definitions of this kind of inference that Peirce had made to identify the thief: “to reason backward from effects to causes” (Doyle, 1981, p. 895). Though Holmes claimed that his inferences were “as infallible as so many propositions of Euclid” (Doyle, 1981, p. 23, Study in Scarlet), Peirce was more candid: such an inference is always a guess, “a singular salad . . . whose chief elements are its groundlessness, its ubiquity, and its trustworthiness” (Sebeok & Umiker-Sebeok, 1983, p. 16). Peirce saw the ability to make such inferences, which he called “abductions,” as part of our given mental equipment: the human mind “having been developed under the influence of the laws of nature, for that reason naturally thinks somewhat after nature’s pattern” (Sebeok & Umiker-Sebeok, 1983, p. 17). Abductive guesses are not always correct, but they are correct far more frequently that would occur by chance.

In 1878 Peirce had proposed that there are only three forms of inference. Abduction is one of these. He expressed the idea in terms of syllogistic figures about drawing beans from a bag, like this (Sebeok, 1983, p. 8).


Rule: All the beans from this bag are white

Case: These beans are from this bag

Result: These beans are white


Case: These beans are from this bag

Result: These beans are white

Rule: All the beans from this bag are white


Rule: All the beans from this bag are white

Result: These beans are white

Case: These beans are from this bag

Deduction is inference in which we reason from a rule, generalization, or theory, to some particular instance, or “case” as Peirce calls it. Johnson-Laird (1993) has written that in deduction there is no gain of semantic information. When we reason in the opposite direction there is a gain of semantic information. One form of semantically increasing information is induction: reasoning from several instances (such as the observed cases of beans from the bag) to a generalization (about all the beans). The other form is abduction, to an explanation of how something came about (for instance how a particular set of beans was chosen). Peirce also referred to abduction as “hypothesis,” or “retroduction,” and sometimes as “speculative modelling.” Abduction is always guessing but, as Peirce said, “we must conquer the truth by guessing, or not at all” (Sebeok & Umiker-Sebeok, 1983, p.11).

We can, however, perhaps make a more specific proposal about the grounds of abduction than Peirce’s “singular salad.” Generally abductive inferences are made from two elements: an observation and a relevant base of knowledge. In “The Adventure of the Cardboard Box” Holmes inferred that the severed ears were not from a medical dissecting room because there was no preservative other than rough salt (observation), and because he had general knowledge of what goes on in medical schools.

Of course Holmes is mythological rather than a realistic character, and Conan Doyle is a crafty story teller. His stories are heroic, allowing us to identify with Holmes. Though there is a representation of a potential critic in Dr Watson, his commentary is always supportive of Holmes’s inferences, allowing extra links to be added to chains of reasoning, rather than questioning any of them. In detective stories of the kind that Edgar Allan Poe invented and Conan Doyle made popular, certainty of such inferences can be claimed — and of course in the story the inferences do turn out to be correct. We have to exert ourselves and step outside the frame of “The Adventure of the Cardboard Box” to make inferences other than Holmes’s. Holmes infers the ears were cut from dead people, but there is nothing in the first part of the story to prevent an inference that the ears were cut from living people and that a ransom note would shortly be delivered. To make such a different inference would be to start writing a different story. So we stay in Conan Doyle’s story, and Holmes’s abduction that the old lady’s sister had been murdered slips into our minds with ease and satisfaction. The story is about Holmes’s intentions and actions with which we easily identify. We follow it as if we were Holmes, as if these inferences were popping into our own minds.

Generally, in order to understand any narrative we do make a range of inferences (Graesser, Singer, & Trabasso, 1994). Principal among these are inferences from actions and events in the story to explanations in terms of the goals and plans of characters (Wilensky, 1978).

Human life, as many social scientists have observed, is founded on being able to act purposefully in the world, using information in planning processes to work out how to act. In understanding a story, we run these planning processes, as it were, backwards: finding explanations in terms an agent’s goals and plans for the actions and events that we read about. As Wilensky (1978) has put it, when we read a story we are not trying to work out what will happen next. Rather, we summon up a range of explanations for what does happen. The genre of the detective story, or mystery, involves us in focussing specifically on explanations of the actions of one character — the murderer.

The effect of prompting explanations of action is not confined to narratives of detection. It is characteristic of all narrative, from the recounting of one’s own actions as Peirce did, to the highest art. In Anna Karenina, for instance, Anna and Vronsky fall in love near the beginning of the novel. Why? Tolstoy goes to considerable lengths to provide the reader with the background information that they had both led divided lives. He shows us Vronsky going to the station to meet his mother, not because he wanted to, but because he would always act in obedience to her. Thus, we are prepared for him to notice Anna getting off the train, because he is not fully engaged in what he is doing. As to Anna, she acts respectably although married to someone whom she does not love. When Anna and Vronsky fall in love, then, this does not come exactly as a surprise. Rather it grows out of their characters, goals, and life situation. Without being able to make the inferences about the roots of their actions in their characters and life situations we would probably not enjoy the novel. As Henry James (1884) asked: “What is character but the determination of incident? What is incident but the illustration of character?” In the nineteenth-century novel, then, what we abduce in explaining action is character. By contrast in scientific exposition, narrative typically gives way to other forms of discourse, and to inferences about how things work.

What individual human minds can’t do

Suppose that Wilensky (1978) is right in proposing that a basic process of reading narrative is summoning up explanations — that is to say making abductive inferences — about what is going on. This may happen either because, in a detective story, the protagonist with whom we identify makes inferences that will later be corroborated by events, or more subtly because characters’ actions allow us to infer character, as in the novels of Tolstoy and Henry James.

Perhaps, indeed, abduction is a kind of inference that is characteristic of fiction. Readers are easily led into abductive inferences. Perhaps a distinctive attribute of fiction — what makes it fiction rather than science — is that the inferences seldom stand realistic scrutiny, let alone scientific scrutiny. Abduction is nowadays defined as reasoning towards the best explanation. In a story, because the writer constructs a closed world he or she can make sure an inference is indeed the best explanation. But in real life, looking forward from any event into the future we usually do not knowwhich explanation will be the best one: a guess may be better than a random shot as Peirce argued, but it will be a guess. Perhaps abduction is only for fiction, and other domains that need no firm anchor in fact. Perhaps climbing through a web of abductions allows us merely to escape into fanciful but non-existent universes. Perhaps science depends on different forms of inferencing. Perhaps it depends as Bacon proposed, on inductions from sets of cognate instances to generalizations and laws. Perhaps, when information is to be increased scientifically, induction is the heart of the paradigmatic (epistemic) mode of thought.

Brown and Clement (1987) have shown, however, that induction to a scientific generalization is very difficult. Their subjects were high-school students who might subsequently take physics, but had not yet done so. There was a pretest with three questions about Newton’s third law (the one mentioned above in discussing Feynman’s textbook). One question was about a book on a table: does the table exert an upwards force on the book? In a post-test, after instruction on Newton’s third law, the same three questions were asked again, together with two more. From 14 students who could not correctly answer the question about the book on the table in the pretest, the experimenters randomly selected two groups of seven students each.

One group was instructed using the best available high-school physics textbook. It was an innovative and well written book. The section used for instruction on Newton’s third law was an exposition that dealt explicitly with whether a table on which a book is resting exerts an upwards force on the book. It started straight out by saying that the table does exert a force on the book. Then came a set of other examples of action and reaction, including a finger pressing on a stone (Newton’s own example), a rifle kick, an athlete running. Then the text returned to the example of the book on the table. The message to this group of students was that they should, under guidance of the text, make an inductive inference from a set of examples to Newton’s third law.

The second group of students was not encouraged to do induction. Instead they were offered a bridging concept that offered an explanation of why Newton’s third law worked — a concept that would allow them to do abduction. Here is the concept they were offered: first think of pushing down with your hand on a spring resting on a table; notice how the spring would push back on your hand. Then think of a book resting on a long springy plank that has been laid across a gap between two sawhorses. Imagine the plank’s springiness pushing back on the book. Next imagine the plank becoming thicker and thicker. The springiness of the plank has not gone away, but as the plank gets thicker it needs more force to let us see it bending. The thicker plank’s springiness is stiffer. All materials have springiness, dependent on the bonds between the molecules, so that even when no deformation of a plank or any other material is visible, this springiness is still there. So it is in a table; its springiness pushes back on a book resting on it.

So: the students in the second group were offered an explanation that could enter their minds. It was the missing half, as it were, of the abduction: observation — the book rests on the table, explanation — its downward force is exactly opposed by an equivalent force of the springiness of the table. In the post-test, seven out of seven of the students who were given the explanation about the springiness of materials answered correctly the question about the book on the table, significantly more than the two out of seven students from the induction group. The abduction group also gave correct answers for the other examples in the post test, and in this also did significantly better than the induction group.

There is a growing number of indications that induction to productive generalizations is difficult for humans. Brown & Clement’s study is one such. Others are shown from Case’s (1991) extensive studies of how children who have skills in one domain, though they involve comparable procedures to problems in adjoining domains, do not transfer these skills. They do not on their own make the generalizing induction, though they can be instructed to acquire central conceptual structures that do generalize across domains.

In an early and important paper on artificial intelligence Newell (1972) made a related point about psychology, which had tied itself to an inductive model of science. Newell argued that in psychology one can not hope to perform a set of experiments, each yielding one small piece of empirically established knowledge, and then by induction from these observations arrive at general principles of mental functioning.

  1   2   3


Edited by D. Olson & N. Torrance (pp. 123- 140) New York: Cambridge University Press iconGenesee, Fred & John Upshur. (1996) Classroom-based evaluation in second language education. Cambridge: Cambridge University Press

Edited by D. Olson & N. Torrance (pp. 123- 140) New York: Cambridge University Press iconReal Analysis and Probability, R. M. Dudley, Cambridge Studies in Advanced Mathematics, Vol 74, Cambridge University Press. This is the main textbook. I have

Edited by D. Olson & N. Torrance (pp. 123- 140) New York: Cambridge University Press iconА. Л. Осипян Рецензия на книгу P. D. Curtin “Cross-Cultural Trade in World History”. Cambridge: Cambridge University Press, 1984
Рецензия на книгу P. D. Curtin “Cross-Cultural Trade in World History”. Cambridge: Cambridge University Press, 1984
Edited by D. Olson & N. Torrance (pp. 123- 140) New York: Cambridge University Press iconCambridge University Press

Edited by D. Olson & N. Torrance (pp. 123- 140) New York: Cambridge University Press iconCambridge University Press

Edited by D. Olson & N. Torrance (pp. 123- 140) New York: Cambridge University Press iconHarvard university press, cambridge, massachusetts

Edited by D. Olson & N. Torrance (pp. 123- 140) New York: Cambridge University Press iconReferences Lothaire, M. (1983), Combinatorics on Words. Addison-Wesley, Reading, ma. Lothaire, M. (2002), Algebraic Combinatorics of Words. Cambridge University Press, Cambridge. Regular grammars

Edited by D. Olson & N. Torrance (pp. 123- 140) New York: Cambridge University Press iconContext and Culture in Language Teaching [Текст] / Claire Kramsch. Oxford; New York : Oxford University Press, 2004, 2006. VIII, 295 p IL. Oxford
Текст] / Claire Kramsch. Oxford; New York : Oxford University Press, 2004, 2006. VIII, 295 p IL. (Oxford Applied Linguistics). Bibliogr.:...
Edited by D. Olson & N. Torrance (pp. 123- 140) New York: Cambridge University Press iconGoldreich, O., Computational Complexity: a conceptual Perspective. 2008: Cambridge University Press

Edited by D. Olson & N. Torrance (pp. 123- 140) New York: Cambridge University Press icon1) Dickin ap, 1995. Radiogenic Isotope Geology. Cambridge University Press (reprinted in 1997, 2000, 2002)

Разместите кнопку на своём сайте:

База данных защищена авторским правом ©lib.znate.ru 2014
обратиться к администрации
Главная страница