Concept Formation

views updated

Concept Formation


Our world is filled with sets of objects, events, and ideas that share some common quality while differing in other characteristics. An organism that learns to respond to the common quality of a given set has learned a concept. The child, for example, learns the concept of blueness when he is able to select a blue balloon or a blue ball or a blue wagon from similar objects of different colors. When he makes this conceptual response, he is abstracting the blueness from other properties of the objects, like size, shape, or function. Similarly the child learns the concept of triangle when he distinguishes triangles of various shapes (right-angle, isosceles, scalene, equilateral, and obtuse) from other geo-metrical forms. Or he learns the concepts of right and wrong when he discriminates between socially approved and disapproved behavior.

In experimental psychology the field of concept formation has achieved neither the methodological nor the conceptual unity characterizing such areas as sensation, perception, learning, and motivation. Traditionally, psychologists have not looked upon conceptual behavior as reflecting any unique or fundamental psychological process. Instead the common view has largely been that conceptual behavior is a function of the interaction between basic psychological processes. As a result, psychologists have for the most part sought to extend theories from other areas (e.g., learning, perception) to conceptual behavior instead of formulating hypotheses concerned exclusively with the facts of concept formation.

Historical introduction. Experimental efforts to understand concepts were made initially by psychologists who by training or intellectual kinship were related to Wilhelm Wundt (1832–1920). For Wundt, introspection, the method of observing one's own conscious experience, was the primary method of the psychological laboratory. The goal in investigating concepts was, first, to identify their basic mental elements and, second, to discover the principles that govern their combination—a goal previously set in investigations of sensory and perceptual experience. As a step toward understanding the conscious correlates of a concept, Titchener described it as a “symbol which holds a large number of particular ideas together” ([1896] 1921, p. 312).

Although the goals of introspective psychology seemed clear, the results were not. Introspection, as a general method of psychology and as a specific technique for revealing the mysteries of thought, possessed severe limitations. Of paramount importance was its apparent unreliability. The content of the mind seemed to be influenced by how one was trained to observe it. Elements of consciousness observed in one laboratory could not be detected in another.

Introspective psychology was also handicapped by two related characteristics of phenomenal experience that apparently interfered with the analysis of the higher mental processes. First, mental experience seemed to be in a state of constant flux, incapable of being analyzed into static elements. William James, in his discussion of the conscious experiences associated with conceptual thinking (1892), recognized the inadequacy of formal introspection as a technique for describing subjective life. Second, introspective analysis of thought by H. J. Watt and N. Ach at Wiirzburg, Germany, sometimes revealed an extreme paucity of conscious experience and at other times, none at all.

Dissatisfaction with the introspective method encouraged some psychologists to propose a new approach to the science of psychology. Behaviorists argued that psychology would remain methodologically unreliable, and empirically restricted, if it persisted in identifying its subject matter as conscious experience. Instead, it was suggested, psychology should be conceived as the objective experimental science of behavior (Watson 1913).

Although the philosophical issues were more complicated than Watson understood, his vigorous and convincing arguments redirected experimental psychology in the United States from the study of consciousness to the study of behavior. The examination of the experiential correlates of concepts was discarded in favor of the objective investigation of their acquisition and utilization.

Hull (1920), in a doctoral thesis, illustrated how behavioristic methodology could be applied to the study of concepts. He trained college students to associate each of 36 Chinese characters with one of six sounds. These characters belonged to six different groups, each of which had a common part known as a radical. The appropriate sound (e.g., oo yer) was the same for all the characters having the same radical. The subjects were under the impression that their task was to memorize the response (the appropriate sound) to each stimulus (the Chinese character). After a subject was able to give the correct response for all 36 characters, he was tested with an entirely new set of characters that had the same radicals. Thus it was possible to discover whether the subject learned to respond to the common element (the radical) or whether he learned the appropriate response to each individual character. If he had learned to respond to the common element, he should now respond appropriately to all the new characters which had that common element. If, however, he had only learned to respond to the individual character, he would be unable to identify the new characters.

The results indicated that the subjects as a group behaved in a manner falling somewhere between the two extremes. On the very first trial 27 per cent of the responses were correct, indicating that the subjects had learned, to some extent, to respond conceptually, that is, to react to the common element characteristic of all the instances of that concept. It is interesting to note that Hull found the most effective method of teaching the six concepts was by presenting each series of characters with the common element in red, thus facilitating the discrimination between the essential cues.

Discrimination learning as a model. Hull's study set the course of research in conceptual behavior in the United States for decades to come. First, it offered an experimental method that could determine objectively the influence of numerous variables on concept formation. Second, the use of stimulus-response language in analyzing conceptual behavior was encouraged. The implication of this technical language system, a system that is widely used by neobehaviorists (the present-day descendants of Watson), is that there are at least three fundamental sets of variables in any psychological event: environmental (S), behavioral (ft), and the association (–) between them. Within this language system a concept becomes “a common response to a set of dissimilar stimuli” (see T. S. Kendler 1961, p. 447). Finally, the results of Hull's study suggested that concept formation could be conceptualized as a case of discrimination learning based upon principles of conditioning. Pavlov explained the learning of a discrimination (conditioned differentiation) in terms of three fundamental principles: conditioning, stimulus generalization, extinction. When training a dog to discriminate between a white and a gray circle, it proved necessary not only to condition him to respond to the white circle but also to train him not to respond to the gray one. A dog who has been conditioned to salivate to a white circle will, without any additional training, salivate to similar stimuli, such as gray circles. Differential responses to gray and white circles can be learned only if the effects of stimulus generalization are neutralized. This can be done by extinguishing the generalized response to the gray circle, i.e., by not giving the dog food in the presence of the gray circle. Thus a discrimination can be established by the joint action of conditioning the response to the appropriate stimulus and extinguishing the generalized response to the inappropriate stimulus [seeLearning, article onDISCRIMINATION LEARNING; and the biography ofHull].

This discrimination model, based upon the mechanisms of conditioning, generalization, and extinction, seemed to many psychologists obviously applicable to concept formation. The child who is learning the concept of red must not only acquire a common response to all red stimuli but must also extinguish the response to similar stimuli, such as orange objects.

The S-R discrimination model of concept formation instigated a wide variety of research studies, far too numerous to summarize here. Of the more systematic research programs some were primarily concerned with investigating the influence of a specific set of variables; others were more interested in analyzing the discrimination process underlying concept formation.

Order of difficulty. Heidbreder (1948) in a classic series of studies, with college students serving as subjects, investigated the order of difficulty of a ttainins different kinds of concepts: those representing objects (e.g., tree), form (e.g., circle), and number (e.g., five). Under varying conditions, the order of difficulty—concrete objects, form, and number—remained constant. This order was interpreted as being positively correlated with the “thing character,” the concreteness, of the critical feature of the concept; a tree possesses more “thing character” than does a circle, which in turn is more concrete than five (see Vinacke 1952, chapter 7, for discussion of concrete-abstract dimension of conceptual behavior). The order of difficulty Heidbreder initially found was not, however, always replicated. By modifying the critical cues on which the concepts were based, the hierarchy of dominance could be changed. In another set of studies under somewhat different experimental conditions (e.g., Grant & Curran 1952), attaining concepts based on color proved to be more difficult than identifying either concepts of form or number. This simple ordering of difficulty again does not tell the whole story. With practice, concepts of number become relatively easy to learn. Also number concepts, once learned, tend to perseverate more than do concepts of form or color. The problem of the hierarchy of dominance of various concepts becomes even more complicated when the behavior of children is considered (see Russell 1956, chapter 8; Thompson [1952] 1962, chapter 9). Some data suggest that children under age three find form concepts easier to acquire than color concepts, but from three to five the relationship is reversed. Perhaps a fundamental order of difficulty will never be found, because the basic mechanisms responsible for concept formation will favor one order under one set of experimental conditions and an entirely different order under a different set of conditions.

Mediating processes. Rats have been reported capable of learning to choose a triangular form when paired with other forms despite variations in the size, shading, position, and amount of outline of the triangles. But is this “truly” conceptual behavior? Osgood believes not, since the rat probably does not understand the abstract concept of triangularity and therefore would not respond to “three dots in a triangular arrangement … or to three places on a map, a three cornered block …” (1953, p. 667). He suggests that there are two kinds of classificatory behavior. One kind occurs when an organism responds to a common specific element in different stimulus patterns (e.g., the specific Chinese radicals in Hull's study). The other kind of classificatory behavior, the “true” conceptual type, occurs when varying stimulus patterns, not necessarily containing any features in common, elicit a common mediating abstract response that serves as a cue for conceptual behavior. Although the qualification “not necessarily containing any features in common” can be a source of semantic confusion (see Kendler 1964, p. 219), Osgood's distinction directs attention to mediating processes that many investigators assume are of crucial importance to the discrimination mechanism underlying conceptual behavior. Experiments with simple concept-formation tasks with children of normal intelligence ranging from three to ten years of age (Kendler & Kendler 1962) suggest that the behavior of very young children is consistent with the principles of discrimination learning based on lower animals in that their responses are directly under the control of environmental stimuli. The behavior of older children, in contrast, is dependent upon responses, usually implicit, that intercede between the external stimulus and the overt response to provide stimulation that influences the course of overt behavior. As children grow older, the proportion of those who behave in a manner that reflects mediating processes increases in a stable fashion. The transition from behavior based on a direct connection between external stimulus and overt response to mediated behavior enables children to learn concepts more rapidly and utilize them under a wider range of conditions.

What are these mediational events underlying conceptual behavior and how can they be controlled? Instructing children to verbalize the relevant cues in a simple concept-learning task encourages mediational behavior. But whether these instructions to name provide the essential cue (e.g., the word “black”) for conceptual behavior or instead direct attention to the essential cue (e.g., the stimulus black) or function in both ways simultaneously is not yet known.

Second signaling system. Closely related to the work of the American mediational theorists are the efforts of the Soviet psychologists who investigate what Pavlov described late in his career as the second signaling system. The first signaling system dealt with stimulation from the external world. But man, Pavlov noted, is a verbal organism that can produce words that can function as stimuli. Such stimuli, which constitute the second signal system, differ qualitatively from others because they can represent in an abstract manner the multitude of separate stimuli of the first signal system. According to Liublinskaya (1957) a word becomes a signal of the second signaling system only when it becomes a concept.

Natural development—Jean Piaget. There is practically no limit to the number of models of behavior that have been or can be used to investigate conceptual behavior. One of the distinguishing characteristics of both the S–R and the second signaling system orientations is the belief that classificatory behavior could, at least in large measure, be reduced to principles of behavior observed in conditioning. Other investigators of conceptual processes do not feel compelled to commit themselves to any psychological reductionism, while some actually oppose it. Jean Piaget has spent over forty years investigating the development of thought, or as he sometimes refers to it, “conceptual operations,” in children. The fruit of his efforts, and those of his co-workers, is a theory that seeks to describe intellectual development from infancy to adolescence. The formulation is based on seminaturalistic observations of behavior of children in response to ingeniously designed problems and adroit questioning. In general the findings suggest to Piaget and his co-workers that intellectual development can be characterized as a series of stages in which each stage lays the foundation for the successor. The concepts Piaget deals with are quite different from those investigated by the S-JR psychologists, who for the most part investigate nominal concepts, e.g., forms, shapes, vegetables. Piaget deals with concepts of relationships, such as those involved in logical operations, which have relevance for the way the child perceives and organizes his world. For example, in the first stage of intellectual development (from birth to two years of age) the child learns the concept of permanency of objects; a doll still exists after it is removed from the child's sight. In the second stage, which covers from two to seven years, the child can use language, and the internalization of actions becomes possible. Symbolic function appears, but there is an absence of both “reversible operations” and of the concepts of conservation of quantity, size, etc. (e.g., the child is apt to say that four checkers placed far apart are more than four checkers placed close together). In the third stage (ages 7–11) the conceptual operations of the child include the ability to perform concrete operations that belong to the logic of classes and relations but do not take into account their combinatorial possibilities. The fourth stage (11–15), which leads to adult logic, is marked by the ability to reason by hypothesis. The logic is now concerned with propositions as well as objects. Movement from one stage to another is determined by a number of experiential and maturational variables, related to, but not exclusively determined by, age.

Piaget's impact on behavioristically oriented psychologists was initially not very great because of their strong methodological commitments to rigorous experimental procedures. In recent years, with an increase of interest in cognitive processes and the publication in English of a book (Flavell 1963) that attempts to both streamline and clarify Piaget's conceptions, a meeting of interests, if not of minds, has become more probable. Actually the interests of Piaget and the S–R psychologists supplement each other, the former being primarily concerned with changes that take place as children mature, while the latter are interested in discovering general principles of behavior that are applicable to all age groups [seeDevelopmental psychology, article onA THEORY OF DEVELOPMENT].

Information and decision theories. There are psychologists whose orientations and conceptualizations are at odds with the neobehaviorists. Beginning with a disenchantment with the capability of a model of behavior based upon principles of conditioning and learning to represent the richness of true-life cognitive processes (e.g., Miller et al. 1960), these psychologists have turned to the recent developments of information theory, decision theory, and computers for both assistance and inspiration. Hovland (1952) offered a “communication analysis” of concept learning in which the information transmitted by specific instances of concepts could be measured. Such an analysis provided a frame of reference against which to evaluate the efficiency of the information-processing ability of human subjects. This communication-analysis orientation led to the planning of computer programs designed to simulate human behavior as well as experiments aimed at contributing information on which programs could be based. The task of simulating conceptual behavior has proved to be more difficult than initially anticipated (see Hunt 1962). Success will not be achieved by isolated demonstrations of the similarity between the behavior of computers and humans. What are needed, assuming that the primary interest is in the behavior of organisms and not the artificial intelligence of computers, are general principles that will generate programs capable of simulating human conceptual behavior exhibited in a wide variety of experimental studies. Hovland concluded that computers can function as potential aids in understanding concept formation by “sharpening our formulations concerning mental processes and phenomena,” in encouraging “theories that have both descriptive and predictive power,” and for coping with the complex problem of dealing with a multitude of interacting variables (Hovland 1960, pp. 691–692).

Bruner, Goodnow, and Austin (1956) approach the problem of concept attainment in a manner similar to the method of computer simulation of cognitive processes—but without the use of computers. Using the term “strategy”—borrowed from the mathematical theory of games and functioning like a computer program—they analyze the behavior of subjects categorizing a set of cards containing geometrical figures. In different experiments the subjects were instructed to form categories that were conjunctive, disjunctive, or relational. Interest was centered on the kind of strategies used to attain these concepts; a strategy being defined as “a pattern of decisions in the acquisition, retention, and utilization of information that serves to meet certain objectives …” (1956, p. 54). Their objectives were to maximize the information obtained from each instance, to reduce “cognitive strain,” and to regulate the risk. The authors analyzed the results primarily in terms of four ideal strategies, and in terms of the advantages and disadvantages of each. Although the term “strategy” is rich in connotations, the questions of how they develop and influence behavior are far from clear. The value of the construct strategy would seem to depend on additional theoretical refinement and empirical data [seeDecision theory; Information theory].

In closing this article it may be appropriate to offer a brief summary of the crucial problems confronting investigators of concept formation. There is little doubt that the discrimination process is of primary importance in concept formation. The best method of teaching a concept would be to arrange the optimal conditions for discriminating between instances that belong to a concept and those that do not. Although such a principle would be generally accepted, there would be much disagreement about its specific interpretation. Whether optimal conditions for discrimination could be best arranged by reinforcing correct habits and not reinforcing incorrect ones, by encouraging suitable mediational responses, by training the organism to perceive crucial differences, by developing appropriate cognitive systems, or by some favorable combination of all of these factors—all these issues would be open to dispute. Basic to this disagreement are two related questions: Do these apparent differences always represent real differences? If so, does their resolution depend upon their being cast in precise mathematical language?

These theoretical issues are influencing specific experimental problems and shaping general research programs. A frank recognition among most investigators is that, to some extent, future theoretical progress must await advances in our understanding of the relationship between conceptual behavior and verbal and developmental processes as well as the design of new experimental techniques to tap the wide variety of concepts humans do learn and use.

Howard H. Kendler and

Tracy S. Kendler

[Directly related are the entriesReasoning and logicandThinking.]


Bruner, Jerome S.; Goodnow, J. J.; and Austin, G. A. 1956 A Study of Thinking. New York: Wiley.

Flavell, John H. 1963 The Developmental Psychology of Jean Piaget. Princeton, N.J.: Van Nostrand.

Grant, David A.; and Curran, Joan F. 1952 Relative Difficulty of Number, Form, and Color Concepts of a Weigl-type Problem Using Unsystematic Number Cards. Journal of Experimental Psychology 43:408–413.

Heidbreder, Edna 1948 The Attainment of Concepts: VI. Exploratory Experiments on Conceptualization at Perceptual Levels. Journal of Psychology 26:193–216.

Hovland, Carl I. 1952 A “Communication Analysis” of

Concept Learning. Psychological Review 59:461–472.

Hovland, Carl I. 1960 Computer Simulation of Thinking. American Psychologist 15:687–693.

Hull, Clark L. 1920 Quantitative Aspects of the Evolution of Concepts: An Experimental Study. Psychological Monographs 28, no. 1.

Hunt, Earl B. 1962 Concept Learning: An Information Processing Problem. New York: Wiley.

James, William (1892) 1948 Psychology. Cleveland, Ohio: World.

Kendler, Howard H. 1964 The Concept of the Concept. Pages 211–236 in Symposium on the Psychology of Human Learning, University of Michigan, 1962, Categories of Human Learning. New York: Academic Press.

Kendler, Howard H.; and Kendler, Tracy S. 1962 Vertical and Horizontal Processes in Problem Solving. Psychological Review 69:1–16.

Kendler, Tracy S. 1961 Concept Formation. Annual Review of Psychology 12:447–472.

Liublinskaya, A. A. 1957 The Development of Children's Speech and Thought. Pages 197–204 in Brian Simon (editor), Psychology in the Soviet Union. Stanford Univ. Press.

Miller, George A.; Galanter, E.; and Pribram, K. H. 1960 Plans and the Structure of Behavior. New York: Holt.

Osgood, Charles E. (1953) 1959 Method and Theory in Experimental Psychology. New York: Oxford University Press.

Russell, David H. 1956 Children's Thinking. Boston: Ginn.

Thompson, George G. (1952) 1962 Child Psychology. Boston: Houghton Mifflin.

Titchener, Edward B. (1896) 1921 An Outline of Psychology. 3d ed. New York: Macmillan.

Vinacke, W. Edgar 1952 The Psychology of Thinking. New York: McGraw-Hill.

Watson, John B. 1913 Psychology as a Behaviorist Views It. Psychological Review 20:158–177.

Concept Formation

views updated

Concept formation

Learning process by which items are categorized and related to each other.

A concept is a generalization that helps to organize information into categories. For example, the concept "square" is used to describe those things that have four equal sides and four right angles. Thus, the concept categorize things whose properties meet the set requirements. The way young children learn concepts has been studied in experimental situations using so-called artificial concepts such as "square." In contrast, real-life, or natural, concepts have characteristic rather than defining features. For example, a robin would be a prototypical or "good" example of the concept "bird." A penguin lacks an important defining feature of this categoryflight, and thus is not as strong an example of a "bird." Similarly, for many children the concept "house" represents a squarish structure with walls, windows, and a chimney that provides shelter. In later development, the child's concept of house would be expanded to include nontypical examples, such as "teepee" or "igloo," both of which have some but not all of the prototypical characteristics that the children have learned for this concept.

Natural concepts are often learned through the use of prototypes, highly typical examples of a category like the robin cited above. The other major method of concept learning is through the trial-and-error method of testing hypotheses. People will guess or assume that a certain item is an instance of a particular concept; they then learn more about the concept when they see whether their hypothesis is correct or not.

People learn simple concepts more readily than complex ones. For example, the easiest concept to learn is one with only a single defining feature. The next easiest is one with multiple features, all of which must be present in every case, known as the conjunctive concept. In conjunctive concepts, and links all the required attributes. For example, the concept square is defined by four sides and four 90-degree angles. It is more difficult to master a so-called disjunctive concept, when either one feature or another must be present. People also learn concepts more easily when they are given positive rather than negative examples of a concept (e.g., shown what it is rather than what it is not).

Further Reading

Bruner, Jerome S. Studies in Cognitive Growth: A Collaboration at the Center for Cognitive Studies. New York: Wiley, 1966.

Ginsburg, Herbert, and Sylvia Opper. Piaget's Theory of Intellectual Development. 3rd ed. Englewood Cliffs, NJ: Prentice-Hall, 1988.

Lee, Victor, and Prajna Das Gupta. (eds.) Children's Cognitive and Language Development. Cambridge, MA: Blackwell Publishers, 1995.

McShane, John. Cognitive Development: An Information Processing Approach. Oxford, Eng.: B. Blackwell, 1991.

Piaget, Jean, and Barbel Inhelder. The Growth of Logical Thinking from Childhood to Adolescence. New York: Basic Books, 1958.

Sameroff, Arnold J., and Marshall M. Haith. (eds.) The Five to Seven Year Shift: The Age of Reason and Responsibility. Chicago: University of Chicago Press, 1991.