Chapter 4
Idealized Cognitive Models
Sources of Prototype Effects
The main thesis of this book is that we organize our knowledge by means of structures called idealized cognitive models, or ICMs, and that category structures and prototype effects are by-products of that organization. The ideas about cognitive models that we will be making u of have developed within cognitive linguistics and come from four sources: Fillmore's frame mantics (Fillmore 1982b), Lakoff and Johnson's theory of metaphor and metonymy (Lakoff and Johnson 1980), Langacker's cognitive grammar (Langacker 1986), and Fauconnier's theory of mental spaces (Fauconnier 1985). Fillmore's frame mantics is similar in many ways to schema theory (Rumelhart 1975), scripts (Schank and Abelson 1977), and frames with defaults (Minsky 1975). Each ICM is a complex structured whole, a gestalt, which us four kinds of structuring principles:
- propositional structure, as in Fillmore's frames
- image-schematic structure, as in Langacker's cognitive grammar
- metaphoric mappings, as described by Lakoff and Johnson
- metonymic mappings, as described by Lakoff and Johnson
Each ICM, as ud, structures a mental space, as described by Fauconnier.
维新派的代表人物
Probably the best way to provide an idea of what ICMs are and how they work m categorization is to go through examples. Let us begin with Fillmore's concept of a frame. Take the English word Tuesday. Tuesday can be defined only relative to an idealized model that includes the natural cycle defined by the movement of the sun, the standard m
eans of characterizing the end of one day and the beginning of the next, and a larger ven-day calendric cycle--the week. In the idealized model, the week is a whole with ven parts organized in a linear quence, each part is called a day, and the third is Tuesday. Similarly, the concept weekend requires a notion of a work week of five days followed by a break of two days, superimpod on the ven-day calendar.
Our model of a week is idealized. Seven-day weeks do not exist objectively in nature. They are created by human beings. In fact, not all cultures have the same kinds of weeks. Consider, for example, the Baline calendric system:
The two calendars which the Baline employ are a lunar-solar one and one built around the interaction of independent cycles of day-names, which I shall call "permutational." The permutational calendar is by far the most important. It consists of ten different cycles of day-names, following one another in a fixed order, after which the first day-name appears and the cycle starts over. Similarly, there are nine, eight, ven, six, five, four, thr
ee, two, and even--the ultimate of a "contemporized" view of time--one day-name cycles. The names in each cycle are also different, and the cycles run concurrently. That is to say, any given day has, at least in theory, ten different names simultaneously applied to it, one from each of the ten cycles. Of the ten cycles, only tho containing five, six, and ven day-names are of major The outcome of all this wheels-within-wheels computation is a view of time as consisting of ordered ts of thirty, thirty-five, forty-two and two hundred and ten quantum units ("days").... To identify a day in the forty-two-day t--and thus asss its practical and/or religious significance-one needs to determine its place, that is, its name in the six-name cycle (say Ariang) and in the ven-day cycle (say Boda): the day is Boda-Ariang, and one shapes one's actions accordingly. To identify a day in the thirty-five day t, one needs its place and name in the five-name cycle (for example, Klion) and in the ven-: for example, For the two-hundred-and-ten-day t, unique determination demands names from all three weeks: for example, Boda-Ariang-Klion, which, it so happens, is the day on which the most important Baline holiday, Galungan, is celebrated. (Geertz 1973, pp. 392-93)
Thus, a characterization of Galungan in Baline requires a complex ICM which superimpos three week-structures--one five-day, one six-day, and one ven-day. In the cultures of the world, such idealized cognitive models can be quite complex.
The Simplest Prototype Effects
国库集中支付中心In general, any element of a cognitive model can correspond to a conceptual category. To be more specific, suppo schema theory in the n of Rumelhart (1975) were taken as characterizing propositional models. Each schema is a network of nodes and links. Every node in a schema would then correspond to a conceptual category. The properties of the category would depend on many factors: the role of that node in the given schema, its relationship to other nodes in the schema, the relationship of that schema to other schemas, and the overall interaction of that schema with other aspects of the conceptual system. As we will e, there is more to ICMs than can be reprented in schema theory. But at least tho complexities do ari. What is particularly interesting is that even if one
t up schema theory as one's theory of ICMs, and even if the categories defined in tho schemas were classical categories, there would still be prototype effects-effects that would ari from the interaction of the given schema with other schemas in the system.
A clear example of this has been given by Fillmore (1982a). The example is a classic: the category defined by the English word bachelor.
The noun bachelor can be defined as an unmarried adult man, but the noun clearly exists as a motivated device for categorizing people only in the context of a human society in which certain expectations about marriage and marriageable age obtain. Male participants in long-term unmarried couplings would not ordinarily be described as bachelors; a boy abandoned in the jungle and grown to maturity away from contact with human society would not be called a bachelor; John Paul II is not properly thought of as a bachelor.
北湖公园
In other words, bachelor is defined with respect to an ICM in which there is a human society with (typically monogamous) marriage, and a typical marriageable age. The idealized model says nothing about the existence of priests, "long-term unmarried couplings," homoxuality, Moslems who are permitted four wives and only have three, etc. With respect to this idealized cognitive model, a bachelor is simply an unmarried adult man.
This idealized model, however, does not fit the world very precily. It is oversimplified in its background assumptions. There are some gments of society where the idealized model fits reasonably well, and when an unmarried adult man might well be called a bachelor. But the ICM does not fit the ca of the pope or people abandoned in the jungle, like Tarzan. In such cas, unmarried adult males are certainly not reprentative members of the category of bachelors.
The theory of ICMs would account for such prototype effects of the category bachelor in the following way: An idealized cognitive model may fit one's understanding of the world either perfectly, very well, pretty well, somewhat well, pretty badly, badly, or not at all. If the ICM in which bachelor is defined fits a situation perfectly and the person referred to by the term is unequivocally an unmarried adult male, then he qualifies as a member of the category bachelor. The person referred to deviates from prototypical bachelorhood if either the ICM fails to fit the world perfectly or the person referred to deviates from being an unmarried adult male.
情商提高的表现是Under this account bachelor is not a graded category. It is an all-or-none concept relative to the appropriate ICM. The ICM characterizes reprentative bachelors. One kind of gradience aris from the degree to which the ungraded ICM fits our knowledge (or assumptions) about the world .
I his account is irreducibly cognitive. It depends on being able to take two cognitive model
s--one for bachelor and one characterizing one's knowledge about an individual, say the pope--and compare them, noting the ways in which they overlap and the ways in which they differ. One needs the concept of "fitting" one's ICMs to one's understanding of a given situation and keeping track of the respects in which the fit is imperfect.
This kind of explanation cannot be given in a noncognitive theory-- one in which a concept either fits the world as it is or not. The background conditions of the bachelor ICM rarely make a perfect amless fit with the world as we know it. Still we can apply the concept with some degree of accuracy to situations where the background conditions don't quite mesh with our knowledge. And the wor the fit between the background conditions of the ICM and our knowledge, the less appropriate it is for us to apply the concept. The result is a gradience--a simple kind of prototype effect.
Lie
A ca similar to Fillmore's bachelor example, but considerably more complex, has been
discusd by Sweetr (1984). It is the category defined by the English word lie. Sweetr's analysis is bad on experimental results by Coleman and Kay (1981) on the u of the verb lie. Coleman and Kay found that their informants did not appear to have necessary and sufficient conditions characterizing the meaning of lie. Instead they found a cluster of three conditions, no one of which was necessary and all of which varied in relative importance:
A consistent pattern was found: falsity of belief is the most important element of the prototype of lie, intended deception the next most important element, and factual falsity is the least important. Informants fairly easily and reliably assign the word lie to reported speech acts in a more-or-less, rather than all-or-none, fashion, . . . [and] . . . informants agree fairly generally on the relative weights of the elements in the mantic prototype of lie.
Thus, there is agreement that if you steal something and then claim you didn't, that's a go
od example of a lie. A less reprentative example of a lie is when you tell the hostess "That was a great party!" when you were bored stiff. Or if you say something true but irrelevant, like "I'm going to the candy store, Ma" when you're really going to the pool hall, but will be stopping by the candy store on the way.
An important anomaly did, however, turn up in the Coleman-Kay study. When informants were asked to define a lie, they consistently said it was a fal statement, even though actual falsity turned out consistently to be the least important element by far in the cluster of conditions. Sweetr has obrved that the theory of ICMs provides an elegant way out of this anomaly. She points out that, in most everyday language u, we take for granted an idealized cognitive model of social and linguistic interaction. Here is my revid and somewhat oversimplified version of the ICM Sweetr propos:
THE MAXIM OF HELPFULNESS
People intend to help one another.
This is a version of Grice's cooperative principle.
THE ICM OF ORDINARY COMMUNICATION
牵挂的拼音(a) If people say something, they're intending to help if and only if they believe it.
(b) People intend to deceive if and only if they don't intend to help.
THE ICM OF JUSTIFIED BELIEF
(c) People have adequate reasons for their beliefs.
(d) What people have adequate reason to believe is true.
The two ICMs and the maxim of helpfulness govern a great deal of what we consider ordinary conversation, that is, conversation not constrained by special circumstances. For example, if I told you I just saw a mutual friend, under ordinary circumstances you'd probably assume I was being helpful, that I wasn't trying to deceive you, that I believed I had en the friend, and that I did in fact e the friend. That is, unless you have reason to believe that the maxim of helpfulness is not applying or that one of the idealized models is not applicable, you would simply take them for granted.
The ICMs provide an explanation of why speakers will define a lie as a fal statement, when falsity is by far the least important of the three factors discovered by the Kay-Coleman study. The two ICMs each have an internal logic and when they are taken together, they yield some interesting inferences. For example, it follows from (c) and (d) that if a person believes something, he has adequate reasons for his beliefs, and if he has adequate reasons for believing the proposition, then it is true. Thus, in the idealized
world of the ICMs if X believes a proposition P. then P is true. Converly, if P is fal, then X doesn't believe P. Thus, falsity entails lack of belief.
In this idealized situation, falsity also entails an intent to deceive. As we have en, falsity entails a lack of belief. By (a), someone who says something is intending to help if and only if he believes it. If he doesn't believe it, then he isn't intending to help. And by (b), someone who isn't intending to help in giving information is intending to deceive. Thus, in the ICMs, falsity entails both lack of belief and intent to deceive. Thus, from the definition of a lie as a fal statement, the other properties of lying follow as conquences. Thus, the definition of lie does not need to list all the attributes. If lie is defined relative to the ICMs, then lack of belief and intent to deceive follow from falsity.
As Sweetr points out, the relative importance of the conditions is a conquence of their logical relations given the ICMs. Belief follows from a lack of intent to deceive and truth follows from belief. Truth is of the least concern since it is a conquence of the othe
r conditions. Converly, falsity is the most informative of the conditions in the idealized model, since falsity entails both intent to deceive and lack of belief. It is thus falsity that is the defining characteristic of a lie.张澜简介
Sweetr's analysis provides both a simple, intuitive definition of lie and an explanation of all of the Coleman-Kay findings. The ICMs ud are not made up just to account for lie. Rather they govern our everyday common n reasoning. The results are possible becau the ICMs have an internal logic. It is the structure of the ICMs that explains the Coleman-Kay findings.烧牛腩
Coleman and Kay discovered prototype effects for the category lie-- situations where subjects gave uniform rankings of how good an example of a lie a given statement was. Sweetr's analysis explains the rankings on the basis of her ICM analysis, even though her ICM fits the classical theory! Nonprototypical cas are accounted for by imperfect fits of the lying ICM to knowledge about the situation at hand. For example, whit
e lies and social lies occur in situations where condition (b) does not hold. A white lie is a ca where deceit is not harmful, and a social lie is a ca where deceit is helpful. In general, expressions such as social lie, white lie, exaggeration, joke, kidding, oversimplification, tall tale, fiction, fib, mistake, etc. can be accounted for in terms of systematic deviations from the above ICMs.
Although neither Sweetr nor anyone el has attempted to give a theory of complex concepts in terms of the theory of ICMs, it is worth considering what would be involved in doing so. As should be obvious, adjective-noun expressions like social lie do not work according to traditional theories. The category of social lies is not the interction of the t of social things and the t of lies. The term social places one in a domain of experience characterized by an ICM that says that being polite is more important than telling the truth. This conflicts with condition (b), that intent to deceive is not helpful, and it overrides this condition. Saying "That was a great party!" when you were bored stiff is a ca where deception is helpful to all concerned. It is a prototypical social lie, though it is
not a prototypical lie. The concept social lie is therefore reprented by an ICM that overlaps in some respects with the lying ICM, but is different in an important way. The question that needs to be answered is whether the addition of the modifier social can account for this difference systematically. Any general account of complex concepts like social lie in terms of ICMs will have to indicate how the ICM evoked by social can cancel one condition of the ICM evoked by lie. while retaining the other conditions. An obvious suggestion would be that in conflicts between modifiers and heads, the modifiers win out. This would follow from the general cognitive principle that special cas take precedence over general cas.
Cluster Models: A Second Source of Prototype Effects
It commonly happens that a number of cognitive models combine to form a complex cluster that is psychologically more basic than the models taken individually. We will refer to the as cluster models.
Mother
An example is the concept mother According to the classical theory, it should be possible to give clear necessary and sufficient conditions for mother that will fit all the cas and apply equally to all of them. Such a definition might be something like: a woman who has given birth to a child. But as we will e, no such definition will cover the full range of cas. Mother is a concept that is bad on a complex model in which a number of individual cognitive models combine, forming a cluster model. The models in the cluster are:
- The birth model: The person who gives birth is the mother.
The birth model is usually accompanied by a genetic model, although since the development of egg and embryo implants, they do not always coincide.
- The genetic model: The female who contributes the genetic material is the mother.
- The nurturance model: The female adult who nurtures and rais a
. . ... ..
child is the mother of that child.
- The marital model: The wife of the father is the mother.
- The genealogical model: The clost female ancestor is the mother.
The concept mother normally involves a complex model in which all of the individual models combine to form a cluster model. There have always been divergences from this cluster; stepmothers have been around for a long time. But becau of the complexities of modern life, the models in the cluster have come to diverge more and more. Still, many
people feel the pressure to pick one model as being the right one, the one that "really" defines what a mother is. But although one might try to argue that only one of the characterizes the "real" concept of mother, the linguistic evidence does not bear this out. As the following ntences indicate, there is more than one criterion for "real" motherhood:
- I was adopted and I don't know who my real mother is.
- I am not a nurturant person, so I don't think I could ever be a real mother to any child.
- My real mother died when I was an embryo, and I was frozen and later implanted in the womb of the woman who gave birth to me.
- I had a genetic mother who contributed the egg that was planted in the womb of my real mother, who gave birth to me and raid me.
- By genetic engineering, the genes in the egg my father's sperm fertilized were spliced together from genes in the eggs of twenty different women. I wouldn't call any of them my real mother. My real mother is the woman who bore and raid me, even though I don't have any single genetic mother.
In short, more than one of the models contributes to the characterization of a real mother, and any one of them may be abnt from such a characterization. Still, the very idea that there is such a thing as a real mother ems to require a choice among models where they diverge. It would be bizarre for someone to say:
- I have four real mothers: the woman who contributed my genes, the woman who gave birth to me, the woman who raid me, and my father's current wife.
When the cluster of models that jointly characterize a concept diverge, there is still a stro
ng pull to view one as the most important. This is reflected in the institution of dictionaries. Each dictionary, by historical convention, must list a primary meaning when a word has more than one. Not surprisingly, the human beings who write dictionaries vary in their choices. Dr. Johnson cho the birth model as primary, and many of the applied linguists who work for the publishers of dictionaries, as is so often the ca, have simply played it safe and copied him. But not all. Funk and Wagnall's Standard cho the nurturance model as primary, while the American College Dictionary cho the genealogical model. Though choices made by dictionary-makers are of no scientific importance, they do reflect the fact that, even among people who construct definitions for a living, there is no single, generally accepted cognitive model for such a common concept as "mother."
When the situation is such that the models for mother do not pick out a single individual, we get compound expressions like stepmother, surrogate mother, adoptive mother, foster mother, biological mother, donor mother, etc. Such compounds, of cour, do not repre
nt simple subcategories, that is, kinds of ordinary mothers. Rather, they describe cas where there is a lack of convergence of the various models.
And, not surprisingly, different models are ud as the basis of different extended ns of mother. For example, the birth model is the basis of the metaphorical n in
- Necessity is the mother of invention.
while the nurturance model is basis for the derived verb in
郑州人事- He wants his girlfriend to mother him.
The genealogical model is the basis for the metaphorical extension of mother and daughter ud in the description of the tree diagrams that linguists u to describe ntence structure. If node A is immediately above node B in a tree, A is called the mothe
r and B. the daughter. Even in the ca of metaphorical extensions, there is no single privileged model for mother on which the extensions are bad. This accords with the evidence cited above which indicates that the concept mother is defined by a cluster model .
This phenomenon is beyond the scope of the classical theory. The concept mother is not clearly defined, once and for all, in terms of common necessary and sufficient conditions. There need be no necessary and sufficient conditions for motherhood shared by normal biological mothers, donor mothers (who donate an egg), surrogate mothers (who bear the child, but may not have donated the egg), adoptive mothers, unwed mothers who give their children up for adoption, and stepmothers. They are all mothers by virtue of their relation to the ideal ca, where the models converge. That ideal ca is one of the many kinds of cas that give ri to prototype effects.
