Kurt W. Fischer and Daniel Bullock
What is the nature of children's knowledge? How does their knowledge change with development? In pursuing these fundamental questions in the study of cognitive development, researchers often expand their focus to include a range of children's behaviors extending far beyond the standard meaning of knowledge.
In the two primary cognitive-developmental traditions, the questions typically take different forms. In the structuralist tradition, influenced strongly by the work of Jean Piaget, Heinz Werner, and others, the questions are: How is behavior organized, and how does the organization change with development? In the functionalist tradition, influenced strongly by behaviorism and information processing, the question is: What are the processes that produce or underlie behavioral change? In this chapter we review major conclusions from both traditions about cognitive development in school-age children.
The study of cognitive development, especially in school-age children, has been one of the central focuses of developmental research over the last 25 years. There is an enormous research literature, with thousands of studies investigating cognitive change from scores of specific perspectives. Despite this diversity, there does seem to be a consensus emerging about (1) the conclusions to be reached from research to date and (2) the directions new research and theory should take. A major part of this consensus grows from an orientation that seems to be pervading the field: It is time to move beyond the opposition of structuralism and functionalism and begin to build a broader, more integrated approach to cognitive development (see Case, 1980; Catania, 1973; Fischer, 1980; Flavell, 1982a). Indeed, we argue that without such an integration attempts to explain the development of behavior are doomed.
The general orientations or investigations of cognitive development are similar for all age groups—infancy, childhood, and adulthood. The vast majority of investigations, however, involve children of school age and for those children a number of specific issues arise, including in particular the relationship between schooling and cognitive development.
This chapter first describes the emerging consensus about the patterns of cognitive development in school-age children. A description of this consensus leads naturally to a set of core issues that must be dealt with if developmental scientists are to build a more adequate explanation of developmental structure and process. How do the child and the environment collaborate in development? How does the pattern of development vary across traditional categories of behavior, such as cognition, emotion, and social behavior? And what methods are available for addressing these issues in research?
Under the framework provided by these broad issues, there are a number of different directions research could take. Four that seem especially promising to us involve the relationship between cognitive development and emotional dynamics, the relationship between brain changes and cognitive development, the role of informal teaching and other modes of social interaction in cognitive development, and the nature and effects of schooling and literacy. These four directions are taken up in a later section.
Patterns Of Developmental Change
One of the central focuses in the controversies between structuralist and functionalist approaches has been whether children develop through stages. Much of this controversy has been obscured by fuzzy criteria for what counts as a stage, but significant advances have been made in pinning down criteria (e.g., Fischer and Bullock, 1981; Flavell, 1971; McCall, 1983; Wohlwill, 1973). In addition, developmentalists seem to be moving away from pitting structuralism and functionalism against each other toward viewing them as complementary; psychological development can at the same time be stagelike in some ways and not at all stagelike in other ways. As a result of these recent advances in the field, it is now possible to sketch a general portrait of the status of stages in the development of children.
The General Status Of Stages
Children do not develop in stages as traditionally defined. That is, (1) their behavior changes gradually not abruptly, (2) they develop at different rates in different domains rather than showing synchronous change across domains, and (3) different children develop in different ways (Feldman, 1980; Flavell, 1982b).
Cognitive development does show, however, a number of weaker stagelike characteristics. First, within a domain, development occurs in orderly sequences of steps for relatively homogeneous populations of children (Flavell, 1972). That is, for a given population of children, development in a domain can be described in terms of a specific sequence, in which behavior a develops first, then behavior b, and so forth. For example, with Piaget and Inhelder's (1941/1974) conservation tasks involving two balls or lumps of clay, there seems to be a systematic three-step sequence (see Hooper et al., 1971; Uzgiris, 1964): (1) conservation of the amount of clay (Is there more clay in one of the balls, even though they are different shapes, or do they both have the same amount of clay?), (2) conservation of the weight of clay (Does one of the balls weigh more?), and (3) conservation of the volume of clay (Does one of the balls displace more water?). The explanation and prediction of such sequences is not always easy, but there do seem to be many instances of orderly sequences in particular domains.
Second, these steps often mark major qualitative changes in behavior—changes in behavioral organization. That is, in addition to developing more of the abilities they already have, children also seem to develop new types of abilities. This fact is reflected in the appearance of behaviors that were not previously present for some particular context or task. For example, in pretend play the understanding of concrete social roles, such as that of a doctor interacting with a patient, emerges at a certain point in a developmental sequence for social categories and is usually present by the age at which children begin school (Watson, 1981). Likewise, the understanding of conservation of amount of clay develops at a certain point in a developmental sequence for conservation.
More generally, there appear to be times of large-scale reorganization of behaviors across many (but not all) domains. At these times, children show more than the ordinary small qualitative changes that occur every day. They demonstrate major qualitative changes, and these changes seem to be characterized by large, rapid change across a number of domains (Case, 1980; Fischer et al., in press; Kenny, 1983; McCall, 1983). Indeed, the speed of change is emerging as a promising general measure for the degree of reorganization. We refer to these large-scale reorganizations as levels. We use the term steps to designate any qualitative change that can be described in terms of a developmental sequence, regardless of whether it involves a new level.
Third, there seem to be some universal steps in cognitive development, but their universality appears to depend on the way they are defined. When steps are defined abstractly and in broad terms or when large groups of skills are considered, developmental sequences seem to show universality across domains and across children in different social groups. When skills of any specificity are considered, however, the numbers and types of developmental steps seem to change as a function of both the context and the individual child (Bullock, 1981; Feldman and Toulmin, 1975; Fischer and Corrigan, 1981; Roberts, 1981; Silvern, 1984). For large-scale (macrodevelopmental) changes, then, there seem to be some universals, but for small-scale (microdevelopmental) changes, individual differences appear to be the norm. The nature of individual differences seems to be especially important for school-age children and is discussed in greater depth in a later section.
Large-Scale Developmental Reorganizations
In macrodevelopment there seem to be several candidates for universal large-scale reorganizations—times when major new types of skills are emerging and development is occurring relatively fast. Different structuralist frameworks share a surprising consensus about most of these levels, although opinions are not unanimous (Kenny, 1983). The exact characterizations of each level also vary somewhat across frameworks. Our descriptions of each level, including the age of emergence, are intended to capture the consensus.
Between 4 and 18 years of age—the time when many children spend long periods of time in a school setting—there seem to be four levels. The first major reorganization, apparently beginning at approximately age 4 in middle-class children in Western cultures, is characterized by the ability to deal with simple relations of representations (Bickhard, 1978; Biggs and Collis, 1982; Case and Khanna, 1981; Fischer, 1980; Isaac and O'Connor, 1975; Siegler, 1978; Wallon, 1970). Children acquire the ability to perform many tasks that involve coordinating two or more ideas. For example, they can do elementary perspective-taking, in which they relate a representation of someone else's perceptual viewpoint with a representation of their own (Flavell, 1977; Gelman, 1978). Similarly, they can relate two social categories, e.g., understanding how a doctor relates to a patient or how a mother relates to a father (Fischer et al., in press).
The term representation here follows the usage of Piaget (1936/1952; 1946/1951), not the meaning that is common in information-processing models (e.g., Bobrow and Collins, 1975). Piaget hypothesized that late in the second year children develop representation, which is the capacity to think about things that are not present in their immediate experience, such as an object that has disappeared. He suggested that, starting with these initial representations, children show a gradual increase in the complexity of representations throughout the preschool years, culminating in a new stage of equilibrium called ''concrete operations'' beginning at age 6 or 7.
Research has demonstrated that children acquire more sophisticated abilities during the preschool years than Piaget had originally described (Gelman, 1978), and theorists have hypothesized the emergence of an additional developmental level between ages 2 and 6—one involving simple relations of representations. The major controversy among the various structural theories seems to be whether this level is in fact the beginning of Piagetian concrete operations or a separate reorganization distinct from concrete operations. Many of the structural approaches recasting Piaget's concepts in information-processing terms have treated this level as the beginning of concrete operations (Case, 1980; Halford and Wilson, 1980; Pascual-Leone, 1970).
For Piaget (1970), the second level, that of concrete operations, first appears at age 6-7 in middle-class children. In many of the new structural theories, concrete operations constitute an independent level, not merely an elaboration of the level involving simple relations of representations (Biggs and Collis, 1982; Fischer, 1980; Flavell, 1977). The child comes to be able to deal systematically with the complexities of representations and so can understand what Piaget described as the logic of concrete objects and events. For example, conservation of amount of clay first develops at this level. In social cognition the child develops the capacity to deal with complex problems about perspectives (Flavell, 1977) and to coordinate multiple social categories, understanding, for example, role intersections, such as that a man can simultaneously be a doctor and a father to a girl who is both his patient and his daughter (Watson, 1981).
The third level, usually called formal operations (Inhelder and Piaget, 1955/1958), first emerges at age 10-12 in middle-class children in Western cultures. Children develop a new ability to generalize across concrete instances and to handle the complexities of some tasks requiring hypothetical reasoning. Preadolescents, for example, can understand and use a general definition for a concept such as addition or noun (Fischer et al., 1983), and they can construct all possible combinations of four types of colored blocks (Martarano, 1977). Some theories treat this level as the culmination of concrete operations, because it involves generalizations about concrete objects and events (Biggs and Collis, 1982). Others consider it to be the start of something different—the ability to abstract or to think hypothetically (Case, 1980; Fischer, 1980; Gruber and Voneche, 1976; Halford and Wilson, 1980; Jacques et al., 1978; Richards and Commons, 1983; Selman, 1980).
Recent research indicates that cognitive development does not stop with the level that emerges at age 10-12. Indeed, performance on Piaget's formal operations tasks even continues to develop throughout adolescence (Martarano, 1977; Neimark, 1975). A number of theorists have suggested that a fourth level develops after the beginning of formal operations—the ability to relate abstractions or hypotheses, emerging at age 14-16 in middle-class Western children (Biggs and Collis, 1982; Case, 1980; Fischer et al., in press; Gruber and Voneche, 1976; Jacques et al., 1978; Richards and Commons, 1983; Selman, 1980; Tomlinson-Keasey, 1982). At this fourth level, adolescents can generate new hypotheses rather than merely test old ones (Arlin, 1975); they can deal with relational concepts, such as liberal and conservative in politics (Adelson, 1975); and they coordinate and combine abstractions in a wide range of domains.
Additional levels may also develop in late adolescence and early adulthood (Biggs and Collis, 1980; Case, 1980; Fischer et al., 1983; Richards and Commons, 1983). At these levels, individuals may able to deal with complex relations among abstractions and hypotheses and to formulate general principles integrating systems of abstractions.
Unfortunately, criteria for testing the reality of the four school-age levels have not been clearly explicated in most cognitive-developmental investigations. There seems to be little question that some kind of significant qualitative change in behavior occurs during each of the four specified age intervals, but researchers have not generally explicated what sort of qualitative change is substantial enough to be counted as a new level or stage. Learning a new concept, such as addition, can produce a qualitative change in behavior; but by itself such a qualitative change hardly seems to warrant designation as a level. Thus, clearer specification is required of what counts as a developmental level.
Research on cognitive development in infancy can provide some guidelines in this regard. For infant development, investigators have described several patterns of data that index emergence of a new level. Two of the most promising indexes are (1) a spurt in developmental change measured on some continuous scale (e.g., Emde et al., 1976; Kagan, 1982; Seibert et al., in press; Zelazo and Leonard, 1983) and (2) a transient drop in the stability of behaviors across a sample of tasks (e.g., McCall, 1983). Research on cognitive development in school-age children would be substantially strengthened if investigators specified such patterns for hypothesized developmental levels and tested for them. Available evidence suggests that these patterns may index levels in childhood as well as they do in infancy (see Fischer et al., in press; Kenny, 1983; Peters and Zaidel, 1981; Tabor and Kendler, 1981).
In summary, there seem to be four major developmental reorganizations, commonly called levels, between ages 4 and 18. Apparently, the levels do not exist in a strong form such as that hypothesized by Piaget (1949, 1975) and others (Pinard and Laurendeau, 1969). Consequently, the strong stage hypothesis has been abandoned by many cognitive-developmental researchers, including some Piagetians (e.g., Kohlberg and Colby, 1983). Yet the evidence suggests that developmental levels fitting a weaker concept of stages probably do exist.
Relativity And Universality Of Developmental Sequences
One of the best-established facts in cognitive development is that performance does not strictly adhere to stages. On the contrary, developmental stages vary widely with manipulations of virtually every environmental factor studied (Flavell, 1971, 1982b). Developmental unevenness, also called horizontal decalage (Piaget, 1941), seems to be the rule for development in general (Biggs and Collis, 1982; Fischer, 1980). During the school years it may well become even more common than in earlier years. By the time children reach school age they seem to begin to specialize on distinct developmental paths based on their differential abilities and experiences (Gardner, 1983; Horn, 1976; McCall, 1981). Some weak forms of developmental stages—what we have called levels—probably exist, as we have noted, but they occur in the face of wide variations in performance.
Since developmental unevenness has been shown to be pervasive, it seems inevitable that developmental sequences will vary among children and across contexts. Unfortunately, there have been few investigations testing for variations in sequence. Most of the studies documenting the prevalence of decalage are designed in such a way that they can detect only variations in the speed of development on a fixed sequence, not variations in the sequence itself. The dearth of studies testing for individual differences in sequence, apparently arises from the fact that cognitive developmentalists have been searching for commonalities in sequence, not differences.
Nevertheless, a few studies have expressly assessed individual differences, and their results indicate that different children and different situations do in fact produce different sequences (Knight, 1982; McCall et al., 1977; Roberts, 1981). A plausible hypothesis is that developmental sequences are relative, changing with the child, the immediate situation, and the culture.
To examine this hypothesis researchers must face an important hidden issue—the nature and generality of the classifications used to code successive levels or steps of behavioral organization. Indeed, when issues of classification are brought into the analysis, it becomes clear that universality and relativity of sequence are not opposed. With a general mode of analysis, children can all show the same developmental sequence in some domain, while with a more specific mode of analysis they can all demonstrate different sequences in the same domain.
Figure 3-1 helps show why. The arrows and solid boxes depict developmental paths taken by two children, boy X on the left and girl Y on the right. The letters in the boxes indicate the specific content of the behaviors at each step, and the hyphens connecting letters indicate that two contents have been coordinated or related. The word step is used to describe a specific point in a sequence without implying how that step relates to developmental levels such as those described above.
Two developmental sequences demonstrating both commonalities and individual differences.
Depending on how these sequences are analyzed, they can demonstrate either commonalities or individual differences—that is, that both children move through the same sequences or that each child moves through a different sequence. When viewed in terms of the specific steps each child traverses, the figure shows different developmental sequences. At step 1, child X can control skill or behavior F, and at step 2 he can control skills F and M separately but prefers F. Finally he reaches step 3, where he can relate F to M. Child Y at step 1 can control skill M, and at step 2 she can control both M and F but prefers M. Finally she reaches step 3, where she can relate M to F. For example, in social play, F might represent the social category for father, M the social category for mother, F-M an interaction in which the father dominates, controlling what the mother does, and M-F an interaction in which the mother dominates, controlling what the father does. Thus, all three steps clearly differ for the two children.
Such plurality would seem to contradict the idea of a universal developmental sequence, since the two children are demonstrating different sequences for similar content. Yet when the specific steps are characterized more generally, it is possible to see these different paths as variations on a common theme. Analysis in terms of the social categories present, for instance, leads to the conclusion that steps 2 and 3 are the same in the two children: At step 2 both children comprehend the two separate categories of mother and father, and at step 3 they both understand how a mother and a father can interact.
In a still more general classification, the steps can be defined in terms of social category structure rather than the particular categories. Then, steps 2 and 3 remain equivalent for the children, and, in addition, step 1 becomes equivalent, since both children control similar structures, a single category (mother or father). In addition, skills that deal with markedly different contents can also be considered equivalent. An interaction between a doctor and a patient is equivalent structurally to the interaction between mother and father at step 3, since both interactions involve a social role relation between two categories.
When cognitive-developmental theorists posit general developmental levels, they are defining developmental sequences even more abstractly—in terms of highly general, structural classes of behaviors. For the level of concrete operations, for example, the conservation of amount of clay can be considered structurally equivalent to the intersection of social categories (Fischer, 1980). Conservation of clay involves the coordination of two dimensions (length and width) in two balls of clay, and the intersection of categories involves the coordination of two social categories for two people (such as doctor/father with patient/daughter).
These considerations lead to a reconceptualization of the controversy over whether developmental sequences are relative or universal. For highly specific classes of behavior, universality would seem impossible, relativity inevitable. At the extreme, even the social category of mother is not the same for the two children, since the behaviors and characteristics that each child includes in the category undoubtedly differ. As a result of such variations, no two randomly chosen children could be expected to show the same specific developmental sequences. Even identical twins exposed to, say, a common mathematics curriculum would follow developmental paths for mathematics that differed in detail. Thus, a useful analysis must distinguish irrelevant from relevant detail and generalize over the latter.
Of course, what counts as relevant detail depends on the researcher's purpose. And care must be taken to avoid trivialization of the issue of universality in a second way—by using overly general or ill-defined classes. It is important that what counts as an equivalent structure be specified with some precision. For example, all instances of two units of something cannot be counted as equivalent unless there is a clear rationale for classifying the units as equivalent. With social categories, it would seem unwise to treat "mother" as structurally equivalent to "corporation president." One of the primary tasks for cognitive developmentalists is to devise a system for analyzing structural equivalences across domains (Flavell, 1972, 1982a; Wohlwill, 1973).
Assuming an opposition between relativity and universality, then, is too simple, because at times individual differences may usefully be seen as variations on a common theme. Many of the current disagreements among researchers about universality and relativity in sequences could be clarified by consideration of the nature of the structural classifications being used. In practice, investigators can use a straightforward rule of thumb: They can construct their classes at an intermediate degree of abstraction—neither so specific as to miss valid generalization nor so general that they serve only the purpose of imposing order.
How the controversy about relativity and universality will be resolved rests in part on whether the structures and processes of developmental reorganization can be usefully regarded as similar across different domains of cognition and across children who differ in their achievements within domains. Can the growth of linguistic skill be usefully described in the same terms as the growth of mathematical skill? Or are there distinct linguistic and mathematical faculties whose development remains fundamentally dissimilar in any useful system for classifying sequences (Gardner, 1983)? Is the difference between a retarded child and a prodigy a difference of sequence or a difference in the speed of mastering what can usefully be considered the same sequence (Feldman, 1980)? These questions are just beginning to be addressed in a sophisticated manner.
Processes Of Development
Many of the questions about the nature of developmental stages, their universality, and the extent of individual differences would be substantially clarified by a solid analysis of the processes underlying cognitive development. However, the best way to conceptualize the results of the extensive research literature on developmental processes is very much an open question. No emerging consensus is evident here, except perhaps that none of the traditional explanations is adequate. Three main types of models have dominated research to date.
The first type of model grows out of Piaget's approach. The developing organization of behavior is said to be based fundamentally in logic (Piaget, 1957, 1975). Developmental change results from the push toward logical consistency. Stages are defined by the occurrence of an equilibrium based on logical reversibility, and two such equilibria develop during the school years—one at concrete operations and one at formal operations.
Tests of this process model have proved to be remarkably unsuccessful. The primary empirical requirement of the model is that, when a logical equilibrium is reached, individuals must demonstrate high synchrony across domains. The prediction of synchrony arises from the fact that at equilibrium a logical structure of the whole (structure d'ensemble) emerges and quickly pervades the mind, catalyzing change in most or all of the child's schemes. Consequently, when a 6-year-old girl develops her first concrete operational scheme, such as conservation of number, the logical structure of concrete operations should pervade her intelligence in a short time, according to Piaget's model. Her other schemes should quickly be transformed into concrete operations.
Such synchrony across diverse domains has never been found. Instead, synchrony is typically low, even for closely related schemes such as different types of conservation (e.g., number, amount of clay, and length). Even if one allows that several concrete operational schemes might have to be constructed before the rapid transformation occurs, the evidence does not support the predicted synchrony (Biggs and Collis, 1982; Fischer and Bullock, 1981; Flavell, 1982b).
Efforts to study other implications of the logic model also have failed to support it (e.g., Braine and Rumain, 1983; Ennis, 1976; Osherson, 1974). Several attempts have been made to build alternative models based on some different kind of logic (e.g., Halford and Wilson, 1980; Jacques et al., 1978). But thus far there have been only a few studies testing these models, and it is therefore not yet possible to evaluate their success.
The second type of process model in cognitive-developmental theories is based on the information-processing approach. The child is analyzed as an information-processing system with a limited short-term memory capacity. In general, the numbers of items that can be maintained in short-term memory are hypothesized to increase with age, thereby enabling construction of more complex skills. The exact form of the capacity limitation is a matter of controversy, but in all existing models it involves an increase in the number of items that can be processed in short-term or working memory. The increase is conceptualized as a monotonic numerical increment from 1 to 2 to 3, and so forth, until some upper limit is reached.
This memory model has been influential and has generated a large amount of interesting research, although it has not yet produced any consensus about the exact form of the hypothesized memory process (Dempster, 1981; Siegler, 1978, 1983). One of the primary problems with the model seems to be the difficulty of using changes in the number of items in short-term memory to explain changes in the organization of complex behavior. Although analysis of behavioral organization is always difficult, the distance between the minimal structure in short-term memory and the complex structure of a behavior such as conservation or perspective-taking seems to be particularly difficult to bridge. How can a linear numerical growth in memory be transformed into a change from, for example, concrete operational to formal operational perspective-taking skills (Elkind, 1974)? Although such a transformation may be possible, its nature has not proved to be transparent or simple (Flavell, 1984).
Moreover, how to conceptualize working memory is itself a controversial issue. Various investigators have challenged the traditional conceptualization that there is an increase in the size of the short-term memory store (Chi, 1978; Dempster, 1981; see also Grossberg, 1982: chs. 11 and 13). Fortunately, ever richer developmental models involving ideas about working memory capacity have continued to appear since Pascual-Leone's (1970) ground-breaking work (see Case, 1980; Halford and Wilson, 1980), and perhaps one of these will be successful in overcoming the problems mentioned.
The third common type of model assumes that development involves continuous change instead of general reorganizations of behavior like those predicted by the logic and limited-memory models. The fundamental nature of intelligence is laid down early in life, and development involves the accumulation of more and more learning experiences. Behaviorist analyses of cognitive development constitute one of the best-known forms of this functionalist model. A small set of processes defines learning capacity, such as conditioning and observational learning, and all skills—ranging from the reflexes of the newborn infant to the creative problem solving of the artist, scientist, or statesman—are said to arise from these same processes (Bandura and Walters, 1963; Skinner, 1969). Any behavioral reorganizations that might occur are local, involving the learning of a new skill that happens to be useful in several contexts.
Some information-processing approaches also assume that the nature of intelligence is laid down early and that development results from a continuous accumulation of many learning experiences: The child builds and revises a large number of cognitive "programs," often called production systems (Gelman and Baillargeon, 1983; Klahr and Wallace, 1976). Children construct many such systems, such as one for conservation of amount of clay and one for conservation of amount of water in a beaker. At times they can combine several systems into a more general one, as when conservation of clay and conservation of water are combined to form a system for conservation of continuous quantities. These reorganizations remain local, however. There are no general levels or stages in cognitive development—no all-encompassing logical reorganizations and no general increments in working memory capacity.
Researchers who believe in the continuous-change model tend to investigate the effects of specific types of processes or content domains on the development of particular skills. One of the processes emphasized within the continuous change framework has been automatization, the movement from laborious execution of a skill or production system to execution that is smooth and without deliberation. Several studies have demonstrated that automatization can produce what seem to be developmental anomalies. When school-age children are experts in some domain, such as chess, they can perform better than adults who are not experts (Chi, 1978). More generally, many types of tasks produce no differences between the performances of children and adults (Brown et al., 1983; Goodman, 1980).
In research on specific content domains, the general question is typically how the nature of a domain affects a range of developing behaviors. For example, the nature of language, mathematics, or morality is said to produce "constraints" on the form of development in relevant behaviors (Keil, 1981; Turiel, 1977). Development in domains that involve self-monitoring, such as knowledge about one's own memory processes (metamemory), is hypothesized to have general effects on many aspects of cognitive development (Brown et al., 1983; Flavell and Wellman, 1977).
Within the continuous-change, functionalist framework, investigators often assume that there is some intrinsic incompatibility between general cognitive-developmental reorganizations and effects of specific domains or processes. Yet it is far from obvious that any such incompatibility exists. The process of automatization can have powerful effects on developing behaviors, and at the same time children can show general reorganizations in those behaviors (Case, 1980). The domain of mathematics can produce constraints on the types of behaviors children can demonstrate, and at the same time those behaviors can be affected by general reorganizations. The reason for the assumption of incompatibility seems to be that developmentalists view the logic and limited-memory models as incompatible with the continuous-change model.
The assumption of incompatibility between reorganization and continuous change seems to stem from the fundamental starting points of the models: The logic and short-term memory models focus primarily on the organism as the locus of developmental change, whereas the continuous models focus on environmental factors. Several recent theoretical efforts have sought to move beyond this limit of the three standard models by providing a more genuinely interactional analysis, with major roles for both organismic and environmental influences (Fischer, 1980; Halford and Wilson, 1980; Silvern, 1984). Approaches that explicitly include both organism and environment in the working constructs for explaining developmental processes may provide the most promise for future research.
The Central Issues In The Field Today
The differences among the traditional approaches to development are important to understand, but they seem much less significant today than they did 10 years ago. A pervasive change in orientation seems to be taking place among behavioral scientists—a shift away from emphases on competing theories toward integrating whatever tools are available to explain behavior in the whole person, in all of his or her complexity. The present era seems to be a time of integrating rather than splitting. Structuralism and functionalism, for example, are seen not as competing approaches but as complementary ones, emphasizing different aspects of behavior and development. This new orientation is evident throughout this volume.
In the study of cognitive development, this change in the field appears to be associated with attempts to go beyond certain fundamental limitations of previous approaches and to move toward a more comprehensive framework for characterizing and explaining cognitive development. At least three basic questions have arisen as part of this movement toward a new, integrative framework. All three involve efforts to avoid conceptual orientations that have proved problematic in past research. The most fundamental of the three questions is: How do child and environment jointly contribute to cognitive development? The other two questions involve elaborations of this question: How do developing behaviors in different contexts and domains relate to each other? What methods are appropriate for analyzing cognitive development? In a general way the answers to these questions apply to development at any age, but the answers apply in particular ways to school-age children.
The Collaboration Of Child And Environment
The central unresolved issue in the study of cognitive development today seems to be the manner in which child and environment collaborate in development. As a result of the cognitive revolution, it is generally accepted that the child is an active organism striving to control his or her world. But this emphasis on the active child often seems to lead to a neglect of the environment. Contrary to the structural approaches of such theorists as Piaget (1975) and Chomsky (1965), it appears to be impossible to explain developing behavior without giving a central role to the specific contexts of the child's action, including those in the school environment (see Scribner and Cole, 1981; Flavell, 1982b).
Giving context a central role does not mean merely demonstrating once again that environmental factors affect assessments of developmental steps. Researchers have documented these effects in thousands of studies, thus pointing out the inadequacies of the Piagetian approach to explaining the unevenness of development. Surely Piaget, Kohlberg (1969, 1978), and other traditional structural theorists have failed to deal adequately with the environment. It is also true, however, that the functionalists have not produced a satisfactory alternative—an approach that both deals with the environment's roles in development and treats children as active contributors to their own development (Lerner and Busch-Rossnagel, 1981). An analysis of the collaboration of child and environment in development is just as unlikely to arise from a functionalist emphasis on the environment as from a structuralist emphasis on the child.
Why has the study of cognitive development repeatedly fallen back on approaches that focus primarily on either the child or the environment? Why have developmentalists failed to build approaches based on the collaboration of child with environment?
Historically, developmental psychology has been plagued by repeated failures to accept what should be one of its central tasks: to explain the emergence of new organization or structure. These failures have most commonly taken either of two complementary forms. In one form, nativism, the structures evident in the adult are seen as already preformed in the infant. These structures need only be expressed when they are somehow stimulated or nourished at the appropriate time in development. In the second form, environmentalism, the structures in the adult are treated as already preformed in the environment. These structures need only be internalized by some acquisition process, such as conditioning or imitation. Typically, structuralist approaches assume some form of nativism, and functionalist approaches assume some type of environmentalism.
Although it is common to focus on the difference between nativism and environmentalism, there is a fundamental similarity, a common preformism.
Both approaches reduce the phenomena of development to the realization of preformed structures. The mechanisms by which the structures are realized are clearly different, but in both cases the structures are present somewhere from the start—either in the child or in the world (Feffer, 1982; Fischer, 1980; Sameroff, 1975; Silvern, 1984; Westerman, 1980).
A mature developmental theory, we believe, must move beyond explanation by reduction to preexisting forms. It must build constructs that explain how child and environment collaborate in development, and one of the primary tasks of such constructs must be to explain how new structures emerge in development (Bullock, 1981; Dennett, 1975; Haroutunian, 1983).
If the future is not to be a reenactment of the past, it is important to ask why it has been so difficult to avoid drifting toward one or another type of preformism. Why has no well-articulated, compelling alternative to preformism been devised? Any compelling alternative to preformism must describe how child and environment collaborate to produce new structures during development. Constructing such a framework is an immensely difficult task. At the very least, the framework must make reference to cognitive structure, environmental structure, the interaction of the two, and mechanisms for change in structure. The scope of these issues makes such a framework difficult to formulate and difficult to communicate once formulated.
Unfortunately, even approaches that have explicitly attempted to move beyond preformist views have typically failed to do so. Piaget provides a case in point. He set out expressly to build an interactionist position, an approach that would deal with both child and environment and thus avoid the pitfalls of nativism and environmentalism (Piaget, 1947/1950). Yet the theory he eventually built placed most of its explanatory weight on the child and neglected the environment.
Consider, for example, his famous digestive metaphor for cognitive development. Just as the digestive system assimilates food to the body and accommodates to the characteristics of the particular type of food, so children assimilate an object or event to one of their schemes and accommodate the scheme to the object or event. Piaget seems to have chosen this metaphor expressly as a device to avoid preformist thinking, yet he still drifted back toward preformism. In practice, the focus for applications of the metaphor was the assimilation of experience to preexisting schemes. The other side of the metaphor—accommodation to experience—was systematically neglected. For example, Piaget (1936/1952, 1975) differentiated many different types of assimilation but generally spoke of accommodation in only global, undifferentiated terms.
Similarly, the structures behind Piaget's developmental stages—concrete operations and formal operations in school-age children—were treated as static characteristics of the child. The environment was granted an ill-defined role in supporting the emergence of the structures, but the structures themselves were treated as if they came to be fixed characteristics of the child's mind (Piaget and Inhelder, 1966/1969). In a genuinely interactionist position, these structures would have been attributed to the collaboration of the mind with particular contexts. Piaget's neglect of the environment became particularly evident when he was faced with a host of environmentally induced cases of developmental unevenness (termed horizontal decal-age). His response was that it was simply impossible to explain them (Piaget, 1971:11). Because of Piaget's neglect of the environment, even supporters of his position have argued that it is essentially nativist (Beilin, 1971; Broughton, 1981; Flavell, 1971).
Toward A Remedy
If the foregoing diagnosis is accurate, any remedy must explicitly counteract the tendency to drift toward attributing cognitive structures to either the child or the environment. What is needed seems to be a framework providing constructs and methods that force researchers to explicitly deal with both child and environment when they characterize how new structures emerge in development.
What might such a framework look like? Many would recommend general systems theory, because it views the child as an active component in a larger-scale dynamic system that includes the environment. To date, however, systems theory does not seem to have been successful in promoting research explicating the interaction between child and environment in development. Many investigators appear simply to have learned the vocabulary of the approach without changing the way they study development. Apparently, the concepts of systems theory lack the definiteness needed to guide empirical research in cognitive development toward a new interactional paradigm. A few provocative approaches based on general systems concepts have begun to appear in the developmental literature (e.g., Sameroff, 1983; Silvern, 1984), but they seem to bring to bear additional tools that specifically promote interactional analyses.
It is in such practical tools that the proposed remedy lies. To promote interactional analyses, a framework needs to affect the actual practice of cognitive-developmental research. We would like to suggest that the concept of collaboration may provide the basis for such a framework.
The Collaborative Cycle
Human beings are social creatures, who commonly work together for shared goals. That is, people collaborate. Often when two people collaborate to solve a problem, neither one possesses all the elements that will eventually appear in the solution. During their collaboration, a social system (Kaye, 1982) emerges in which each person's behavior supports the other's behavior and thought in directions that would not have been taken by the individuals alone. Eventually a solution—a new cognitive structure—emerges. It bears some mark of each individual, yet it did not exist in either person prior to the collaboration, nor would it have developed in either one without the collaboration. Indeed, even after the structure has developed, the individuals may be able to access it only by reconstituting the collaboration. Of course, besides having the same two people collaborate again, it is also possible for one of them to collaborate with a different partner (Bereiter and Scardamalia, 1982; Brown et al., 1983; Maccoby and Hartup, in this volume).
Figure 3-2 shows this developmental process as a collaborative cycle. The two left circles represent, respectively, structures that are external and internal to an individual. Consider a girl engaged in solving a puzzle with her father. The father provides external structures to support or scaffold her puzzle solving by stating the goal of the task, lining up a puzzle piece to highlight how it fits in its particular place, providing verbal hints, and so forth (Brown, 1980; Kaye, 1982; Wertsch, 1979; Wood, 1980). The child's knowledge and skills for solving the puzzle constitute the core of the developing internal structures.
Development schematized as a collaborative cycle.
The collaboration of external and internal structures produces the behavioral episodes represented in the right circle. The girl and her father work at solving the puzzle, and, as a result of the collaboration, she can achieve a scaffolded mental state, which she could not achieve by herself as quickly or in the same form.
The feedback arrows running from the right circle to the left ones in Figure 3-2 show the dependence of developmental change on collaboration. By performing the task in a scaffolded interaction, the girl learns the goal of the puzzle and how to go about solving it without her father's help. She develops more sophisticated internal structures so that she is less dependent on the complex external structures provided by her father. Of course, the development of this ability takes many steps: The father constantly updates his scaffolding to fit the child's present knowledge and skill. In this way, developmental change occurs both inside the child and outside her—an often overlooked fact to which we will return.
In much human behavior there is indeed a collaboration between two or more individuals. Recent socially oriented analyses of development have emphasized this process. Sometimes the emphasis is on the joint contributions of collaborating individuals, and the process is called coregulation or something similar (see Feldman, 1980; Markus and Nurius, Maccoby, and Weisner, in this volume; Westerman and Fischman-Havstad, 1982). Sometimes the emphasis is on the role of the parent or older child in supporting and advancing the child's behavior, and the process is called scaffolding or something similar, as in Figure 3-2 (Brunet, 1982; Kaye, 1982; Laboratory of Comparative Human Cognition, 1983; Lock, 1980; Vygotsky, 1934/1978; Wertsch, 1979; Wood et al., 1976; Wood, 1980).
Even when a child is acting alone, collaboration can occur because the nonpersonal environment can play the role of collaborator. Because environments have structures, every environment supports some behaviors more than others. For example, a tree that has strong branches far down on its trunk provides strong support for climbing, a tree with only high branches provides less support, and a pole with no branches provides little support.
Of course, much about human environments is socially constructed. Consequently, the collaboration between child and environment often involves other people even when no other person is immediately present, because people have constructed the physical environment to correspond with mental structures that organize their activity. Good examples include a library with a spatial/topical organization of its many books and a classroom with its desks, chalkboards, and wall displays all designed to facilitate the types of interactions needed for schooling.
Implications For Research
Although the collaboration approach has not yet been fully articulated, it already seems to have straightforward implications for research practice. If child and environment are always collaborating to produce a behavior, explanations of that behavior must invoke characteristics of both. As a practical procedure to encourage such explanations, investigators can use research designs that vary important characteristics of both the child and the environment. With such designs, variations in both child and environment are likely to affect behavior (Fischer et al., in press; Hand, 1981).
A series of studies on the development of understanding social categories illustrates how this type of research design can lead to analyses of the collaboration between child and environment in cognitive development (Hand, 1982; Van Parys, 1983; Watson and Fischer, 1977, 1980). The studies were designed to test several predicted sequences for the development of social categories such as the social roles of doctor and patient and the social-interaction categories of ''nice'' and "mean." Each study was designed to include variations in both the child and the environment.
The main variable involving child characteristics was age. A wide age range was included in each study to ensure substantial variation in children's capacities to understand the social categories. Ages ranged from 1 to 12 and thus included the relevant periods for the major developmental reorganizations in preadolescent school-age children.
To determine the contribution of environmental characteristics, behavior was assessed under three different conditions, which were designed to provide varying degrees of support for advanced performance. In a structured condition—the elicited-imitation assessment—a separate task was administered to test each predicted step in the developmental sequence. The subject was shown a story embodying the skill required for that step and was asked to act out the story. Thus this condition provided high environmental support for performance at every step. The other two conditions provided less support and thus assessed more spontaneous behavior. In the free-play condition, each child played alone with the toys, acting out his or her own stories. In the best-story condition the experimenter returned to the testing room and asked the child to make up the best story he or she could.
The results showed a systematic effect of environmental support on the child's performance, but the effect varied as a function of the developmental level of the child's best performance. For the first several steps in the developmental sequence, virtually all children showed the same highest step in all three conditions. However, a major change occurred beginning with the first step testing the developmental level of simple relations of representations (which typically emerges at approximately age 4). At this step most children performed at a higher step in the structured assessment than in the two more spontaneous conditions, and that gap grew systematically in the later steps in the sequence. Figure 3-3 shows these results for the studies of the social roles of doctor and patient, and parallel results were obtained in studies of the social interaction categories of nice and mean (Hand, 1982) and the self-related categories of gender and age (Van Parys, 1983).
A systematic change in the proportion of children showing the same step in elicited imitation and free play. Adapted with permission from Watson & Fischer (1980). Copyright © American Psychological Association.
A similar design and method was used to test for an analogous phenomenon in adolescents. The developmental sequence involved the moral concepts of intention and responsibility. It was predicted that at the cognitive-developmental level of formal operations (also called "single abstractions") subjects would show the same highest step in a structured assessment and in a spontaneous condition. However, when they became capable of performing at the next developmental level, relations of abstractions, a major gap would appear between performance in the structured and spontaneous conditions. The prediction was supported. Once again, the highest developmental step that the individual demonstrated varied systematically as a function of both the individual's capacity and the environmental condition (Fischer et al., 1983).
In analyzing results of this sort a proponent of a noncollaborative approach would ask which condition provides the best assessment of the child's true competence. The collaboration theorist replies, "You've missed the point. Competence as traditionally assessed is a joint function of child and environment." The child does not have any true competence independent of particular environmental conditions. Competence varies with degree of support.
Even for an individual child research can be designed to investigate variations in both the child and the environment. Cole and Traupman (1983), for example, assessed a learning disabled child's capabilities using a range of cognitive tests and examined his performance in settings outside the classroom. They found that, in settings involving social interactions with other people, his disabilities were hardly noticeable because he used his social skills to compensate for them. Thus, the portrait of the child in a standard testing situation was vastly different from the portrait in a real-life social setting.
It is surprising how few cognitive-developmental studies have systematically varied characteristics of both child and environment. Typically, studies examine either changes with age and ability or changes resulting from environmental factors. In the infrequent studies that include variations in both child and environment, the interpretations often neglect the interaction and instead focus on the child and the environment separately. For example, many studies criticizing Piaget's work demonstrate that variations in environmental conditions produce developmental unevenness (decalage), but they seldom deal with the variations as a function of children's ages or abilities. Fortunately, there are a growing number of exceptions to this characterization—studies that seriously consider the effects of both child and environment on performance. The results of these studies are already beginning to transform explanations of cognitive development (see O'Brien and Overton, 1982; Rubin et al., 1983; Tabor and Kendler, 1981).
The Transformation Of Concepts Of Ability And Competence
As these research examples illustrate, analyzing development as a collaborative process leads to a reconceptualization of many basic cognitive-developmental concepts. Since every behavior can be seen to depend on a collaboration between child and environment, it becomes impossible to analyze any behavior without including both organismic and environmental factors.
Cognitive developmentalists and psychometricians commonly speak of children's ability, or capacity, or competence, as if a child possessed a set of static characteristics that could be defined independently of any context: One child has the ability to understand conservation of water, and another child does not. As soon as the collaborative role of the environment is introduced, these concepts must be radically changed. Competence is not a fixed characteristic of the child but an emergent characteristic of the child in a specific context. It is not enough to make a distinction between competence and performance, because in standard usage this distinction begs the question. The assumption is made that children really do possess a set of competences, but they are somehow prevented from demonstrating them in their performance (Overton and Newman, 1982). If concepts such as ability and competence are to be consonant with a collaboration approach, they must be redefined in terms of the interaction of child with environment.
Within a collaboration approach, concepts of ability and competence retain their utility, because the child is part of the analysis, too. In certain contexts, children perform up to a certain level of complexity and not beyond it, thus demonstrating a certain competence for those contexts. At times children show partial knowledge of what is needed for a particular task (Brown et al., 1983; Feffer, 1982) and so demonstrate the competence for collaboration with a more knowledgeable partner. Also, children evidence large individual differences in the facility with which they can generalize an ability to new contexts, thus demonstrating variations in the competence to generalize. Upon the emergence of formal operation, for example, very bright children seem to be able to use their new capacity quickly in a wide range of tasks, whereas children of normal intelligence take much longer to extend the capacity to many tasks (Fischer and Pipp, 1984; Webb, 1974).
The collaboration orientation poses many new questions for the study of cognitive development. It is not enough to ask questions such as: How does the child's behavior change with age, or how does the child's behavior change as a function of experience? Instead, questions like the following need to be asked: Why do children often perform below capacity? How does context support or fail to support high level performances that are known to be within the child's reach? How do specific collaborative systems support the acquisition of particular skills in different ways at different developmental levels? How is the nature of the child's experience jointly regulated by the child and by resources (human and other) available in the child's environment? Later, we examine several lines of research that show promise of contributing answers to such questions.
Integrating Across Traditional Research Categories
In the same way that scholars are coming to treat child and environment as collaborators in development they are recognizing the need to integrate the traditional categories for categorizing behavior. Cognition and emotion, for example, are not separate in the developing child. There seem to be at least three reasons for this changing orientation.
First, after decades of research, developmentalists have found that a child's behavior does not fit neatly into separate boxes labeled cognition, emotion, motivation, social skills, personality, and physical development (see, for example, Harter, 1982, 1983; Selman, 1980). Indeed, even behavior in more restricted, intuitively appealing categories such as perspective taking and conservation does not fit together coherently (see Hooper et al., 1971; Rubin, 1973; Uzgiris, 1964). Behavioral development has not proved to follow the "obvious" categories devised by developmentalists.
Second, the general movement toward integrating diverse approaches and dealing with the whole child leads not only to an emphasis on the collaboration of child and environment but also to the consideration of relations between behaviors in the traditional categories: How does emotional development relate to cognitive development? How does social development relate to cognitive development? Instead of one set of researchers studying a cognitive child, while another set studies a social child, and still another set studies an emotional child, the field is moving toward viewing the child as a whole—a cognitive, social, emotional, motivated, personal, biological child.
Third, during the last 20 years the cognitive-developmental orientation has become a dominant influence in the study of development, and it has provided a major impetus toward integration. The central questions in the study of cognitive development involve the organization of behavior and the processes underlying behavioral change. Because these questions are so general and fundamental, their applicability is not limited to the traditional domain of cognitive development—increments in knowledge about "cold" topics, such as objects, space, and scientific principles. All behavior, including that involving "hot" topics, such as emotions and social interaction, is organized in some way and undergoes developmental change.
The movement toward integration across behavioral categories has been promising, and many interesting results have come from research in this new tradition. But thus far progress has been limited by several conceptual difficulties.
Overcoming The Obstacles
One of the central conceptual problems has been the tendency to reify the traditional behavioral categories despite the lack of evidence that children's behavior fits the categories. Thus, the most common hypotheses about the relationship between, for example, cognitive development and social development have assumed the validity of cognition and social skills as separate categories. This assumption is especially clear when cognitive development is postulated as a prerequisite for social development.
One such hypothesis that has received much attention involves the relation between cognition and morality: Cognitive development is hypothesized to be a prerequisite for moral development (see Kohlberg, 1969). In practice, this proposition has been taken to mean that performance on Piagetian tasks is a prerequisite for performance on Kohlberg's moral dilemmas. Why should conservation of amount of clay, for instance, be a prerequisite for moral reasoning based on normative concepts of good and bad (Kohlberg's stage 3)? Is there any sense in which conservation is included in the concepts of good and bad? Or is there any way that conservation is more fundamental to mental functioning than concepts of good and bad? Isn't it just as reasonable (or unreasonable) to suggest that concepts of good and bad may be a prerequisite for conservation? If evidence does not support the division of behavior into separate categories of cognition about science problems and moral reasoning, it cannot be meaningful to suggest that such cognition is a prerequisite for moral reasoning (Rest, 1979, 1983).
A similar problem arises when investigators assume that the behaviors captured by the traditional categories are totally separate, showing no relation to each other at all. One of the most neglected topics for school-age children is emotional development, which is sometimes treated as if it is not related at all to cognitive development. Perhaps this assumption helps explain why cognitive developmentalists have omitted emotions from their research agenda. In a later section we suggest some guidelines for stimulating the study of emotional development in school-age children, especially as it relates to cognitive development.
A third, related conceptual problem has been the assumption that one variable can capture an entire behavioral category. Self-esteem as assessed by a questionnaire is treated as measuring the core of the developing self (Hatter, 1983; Markus and Nurius, in this volume; Wylie, 1979). The stage of moral judgment, as assessed by reasoning about a set of moral dilemmas, is believed to assess the fundamental nature of moral development (Rest, 1983).
This mistaken assumption is at the heart of a recent controversy about the nature of brain-behavior relations. Several investigators have used measurements of the growth rate of children's heads as indexes of changes in the children's ability to learn (Epstein, 1978; Toepfer, 1979). Although no measures of learning were used, conclusions were drawn from the head-growth data about what children of different ages were able to learn. The relationship between brain growth and cognitive development is an exciting topic worthy of research, as we discuss later. It is important, however, that researchers differentiate what they are measuring from other developmental changes. Relationships between developments in different domains cannot be assumed; they must be assessed.
Implications For Research
Since the traditional categories for categorizing behavior do not seem to capture either the way behavior is organized or how its organization develops, it makes sense to analyze development across categories. More generally, the concern for explaining development in the whole child and for building a framework that emphasizes the collaboration of child and environment demands that researchers assess behavior in multiple contexts and with various methods. In doing such research, however, developmentalists need (1) to avoid allowing the categories to limit their thinking, as when cold cognition is considered to be a prerequisite for moral reasoning, and (2) to avoid assuming that a single variable will provide a valid index of overall cognitive functioning, as when head growth is treated as if it directly reflects cognitive changes.
In practice, doing research on development across traditional categories is closely related to doing research on the collaboration of child and environment in development. In both cases a number of variables must be measured in several settings, and the investigator must analyze not only each variable itself but also the relations among variables. Consider, for example, research on the effects of divorce on the school-age child. It would appear to be wise to assess (1) the child's understanding of family roles and the effects of divorce on that understanding, (2) the child's emotional reactions to the divorce, (3) the types of social interactions between parents and child and the changes in those interactions that resulted from the divorce, (4) the child's attitudes toward the parents, and so forth. On the basis of the collaboration argument, it may also be important to measure each of these factors under several different degrees of environmental support. Obviously, such research is difficult because it can quickly become unmanageably complex.
Despite this complexity it is possible to do research on patterns of development across categories without either being overwhelmed by complexity or becoming entangled in the conceptual problems that have plagued much past research. At least two helpful guidelines can be articulated: First, development should be analyzed in what promises to be a coherent domain of personal functioning. For example, an investigator might study the mastery of early skills involved in learning to read words (for example, Knight, 1982) or the relationship of divorce to a child's understanding and use of social roles in the family. Within such domains the investigator can examine development in different contexts while still keeping the project within a manageable scope. In addition, the coherence of the domain itself will often provide environmental support to guide the investigator's efforts.
Second, the researcher needs to use methods and measures appropriate to the questions being addressed. Of course, this admonition has been made often. In cognitive-developmental research, however, inadequate methods have been used repeatedly even when appropriate methods were available. In addition, recent innovations in developmental methodology have provided powerful methods for studying many fundamental developmental issues, including relationships between development in different contexts.
Methods Of Assessing Development Change And Continuity
Cognitive-developmental research has not generally been distinguished by the sophistication of its methodology. One of the primary reasons has been that the traditional methods used in the behavioral sciences are not appropriate for studying such issues as developmental change and continuity (Wohlwill, 1973). Analysis of variance, for example, was originally constructed to test whether one or more factors made a difference in the outcomes of independent, equivalent groups. It was not constructed to examine questions about cognitive-developmental issues such as changes in the organization of behavior.
Children almost invariably become smarter as they grow older, and so it has been a simple matter in cognitive-developmental research to use analysis of variance to demonstrate differences between age groups and to use correlations to demonstrate relations between development and age. By themselves, such differences and relations can be uninteresting unless they help answer important questions such as the following: Do children show a systematic developmental sequence in a given domain? Does that sequence demonstrate reorganization of behavior? Are there differences in the speed of developmental change at different times in that domain? Across domains or contexts, are there systematic relations among sequences, reorganizations, changes in speed, or other developmental patterns?
Fortunately, there has recently been substantial progress in constructing designs, measures, and statistics for asking developmental questions (Applebaum and McCall, 1983; Bart and Krus, 1973; Coombs and Smith, 1973; Fischer et al., in press; Krus, 1977; Siegler, 1981; Wohlwill, 1973). Although we do not review all these methods here, we do sketch some of the important concepts behind them.
Systematic change is clearly one of the fundamental concerns of developmental science in general. In cognitive development the tool used most often to describe and analyze systematic change has been the developmental sequence—a series of steps, levels, or stages that portray how behavior gradually changes from some starting point to some endpoint (Flavell, 1972). As a descriptive tool the sequence has been at the center of cognitive-developmental research, providing the core set of observations on which most cognitive-developmental theories are based, ranging from classical approaches (for example, Piaget and Inhelder, 1966/1969; Werner, 1957) to more recent ones (for example, Case, 1980; Siegler, 1981).
Developmental sequences demonstrate not only developmental change but also a form of developmental continuity. They describe how one type of behavior gradually changes into another, and scales based on sequences can be used to examine when change is relatively gradual and continuous and when it is relatively abrupt and discontinuous.
Since the developmental sequence is so important to the study of cognitive development, scaling should clearly be a central concern in research. Documenting that a description of a series of steps in fact forms a scale would seem to be integral to the research enterprise, yet very few investigations of cognitive development in school-age children demonstrate a basic concern with scaling.
The most common type of study in published cognitive-developmental research fits the following description. Children from a few different age groups are tested on several tasks. For example, 5-, 8-, and 11-year-olds are tested on three tasks: one task for conservation of number of plastic chips, one for conservation of amount of clay in a ball, and one for conservation of amount of water in a beaker. Performance on each task is scored on a three-step hypothesized sequence. Step 1 reflects a clear nonconservation response, such as a statement that the amount changes when the array is transformed. Step 2 indicates a transitional or ambiguous response, as when a child states that the amount stays the same but gives no satisfactory elaboration or explanation. Step 3 indicates an answer showing full conservation. An analysis of variance is then performed on the results, which demonstrate that, for each of the three tasks, performance improved across the three age groups and that performance for one or two tasks was significantly better than that for the other tasks. For example, children had significantly more advanced scores for conservation of number than for the other two tasks.
These analyses clearly demonstrate that the older groups performed better than the younger ones—hardly a surprise. The results document little else of interest, failing even to test directly for any developmental sequences. They do not adequately test the hypothesized three-step sequence, nor do they demonstrate that the three conservation tasks form a two-step sequence, with conservation of number developing before the other two.
To test a developmental sequence an independent assessment is required of each step in the hypothesized scale (Fischer and Bullock, 1981). With such an assessment it is possible to test directly whether one step comes consistently before or after another. Performance on the independent assessments should form a Guttman (1944) scale, in which every child passes all the steps prior to his or her highest step passed (and fails all the steps after the lowest step failed). Table 3-1 shows the possible performance profiles that are consistent with a simple eight-step Guttman scale. Scales can also be more complex, with two or more tasks at a single step, as for step 2 in Table 3-2. Indeed, methods are available for tracing highly complex scales, such as those that branch into multiple parallel paths (Bart and Krus, 1973; Coombs and Smith, 1973; Krus, 1977).
Strong Scalogram Method: Profiles for an 8-Step Developmental Sequence.
Profiles for a Measure With Two Tasks at Step 2.
The design of the hypothetical study of conservation allows only one such direct test for sequence. Because of the independent assessment of the three types of conservation, a sequence involving those types can be tested. For example, consider a two-step sequence in which the first step is full understanding of conservation of number and the second is full understanding of either conservation of clay, water, or both. With that sequence every child should show one of the profiles for steps 0, 1, and 2 in Table 3-2. However, it is not possible to test directly the hypothesized three-step developmental sequence (from nonconservation to conservation with explanation) within each type of conservation, because with the specified design of the study the steps are not assessed independently.
There is another method that provides independent assessments without requiring a separate task for each step—the longitudinal design traditionally espoused for developmental research (Wohlwill, 1973). Longitudinal testing of children on the three conservation tasks would make it possible to determine whether for each task and every child, steps always occurred in the predicted order. From one testing to the next, children should either move to a higher step or remain at the same step. This design has been used very effectively in research on moral development to demonstrate that the stages hypothesized by Kohlberg do in fact form a developmental sequence (Colby et al., 1983; Kuhn, 1976; Rest, 1983). The use of scalogram assessments in longitudinal research would provide even greater power and precision, however. With separate tasks to assess each step, individual children's development could be traced in detail. We know of no studies of cognitive development in school-age children using scalograms with a longitudinal design.
Of course, longitudinal research is not needed to test a developmental sequence. With a cross-sectional design, powerful methods are available for rigorously testing a predicted developmental sequence, as suggested by Tables 3-1 and 3-2. Scalogram statistics can be used to test how well the data fit the predicted scale (Green, 1956), and measures approximating a developmental scale can be devised when a specific sequence cannot be predicted. A strong scalogram measure, in which a different task is constructed a priori to assess each predicted step in a sequence, can be especially useful because the theoretical interpretation of each task can be specified unambiguously. For the most part, however, researchers have not taken advantage of the obvious virtues of scalogram methods for testing sequences or other hypotheses about development.
In most published studies, scalability tests are not reported even when the design allows them. The apparent reason for the neglect of scalogram methods is that, when they were used to test some of the detailed developmental sequences inferred by Piaget from mean age differences between tasks, the scalability of the sequences was poor (Hooper et al., 1979; Kofsky, 1966; Wohlwill and Lowe, 1962). Instead of concluding that Piaget's sequences were incorrect, developmentalists seem to have shot the messenger that brought the bad news: They discarded neglect of a powerful method appears to be coming to an end. scalogram methods, for the most part. Fortunately, this unwarranted
The cognitive-developmental issues that can be addressed with scalogram methods include the following: (1) With independent assessments of each steps, the parallels and differences between developments in different contexts can be traced precisely (Corrigan, 1983). (2) Individual differences in developmental sequences can be directly tested, especially when separate assessments are used to detect hypothesized differences (Knight, 1982). (3) Changes in the speed of development can be detected.
The particular method will vary with the hypothesis, of course. For instance, to test for changes in the speed of development, such as spurts and plateaus, it is essential that subjects be sampled such that their ages are distributed evenly (Fischer et al., in press). If a developmental spurt is predicted at age 10, for example, it is necessary to sample children evenly throughout the age range between 9 and 11. If all children tested are at a few restricted ages, such as within a few months of age 9 or 11, it will be impossible to determine whether a difference between 9- and 11-year-olds reflects a developmental spurt, since the distribution of ages alone will produce a bunching of subjects at certain steps in the scale.
Several studies using appropriate designs to assess speed of development have found that speed does seem to accelerate at certain ages during the school years and to slow down at other ages (Jacques et al., 1978; Kenny, 1983; Tabor and Kendler, 1981). That is, there may be periods of discontinuity and periods of continuity as assessed by speed of development. Current data are consistent with the hypothesis that spurts are associated with the large-scale reorganizations or levels described earlier (Fischer and Pipp, 1984), although more research is necessary to fully test this hypothesis.
In general, research with infants and young children has used much more sophisticated scaling methods than has research with school-age children. For example, Seibert and Hogan (1983), Uzgiris and Hunt (1975), and others have devised a number of scales for infant cognitive development in which each step in a predicted sequence is assessed independently. These scales have been used by various investigators to examine developmental change with some precision (Hunt et al., 1976; Seibert et al., in press). Using methods that approximate a Guttman scale, McCall et al. (1977) analyzed a longitudinal study of performance on infant intelligence tests to assess both changes in the speed of development and individual differences in developmental sequences. We know of no large-scale research projects on school-age children that have used such sophisticated methods to assess developmental change.
Developmental sequences are a central concern in cognitive research, but an emphasis on the relations of behavior across contexts highlights the centrality of a second, related issue: the generality or breadth of applicability of a skill or scheme. A full analysis of the skill underlying a behavior should predict not only where that behavior will fall in a developmental sequence but also how the skill will be evident across a range of contexts.
In recent years several investigators have elaborated a set of methods for assessing the rules underlying a behavior and explaining how those rules apply across contexts (Klahr and Wallace, 1976; MacWhinney, 1978; Siegler, 1981). Siegler (1983) provides an especially clear statement of the logic of rule assessment and focuses on school-age children (as does most rule-assessment research).
Typically, ''rule'' refers to a mental procedure whose operation affects performance on many problems within a task domain. Virtually all of the various approaches to specifying rules derive from the theory of production systems (Newell and Simon, 1972), which analyzes human behavior in terms of systems of rules for generating actions. A rule is defined in terms of a condition-action pair, in which the condition for taking some action is specified abstractly. For example, in simple arithmetic tasks involving division, such as 13 divided by 3, a sequence of rules can be used to describe the division procedure. After an estimate has been made of the whole number required in the quotient, a rule applies for dealing with what is left over, the remainder: If the remainder is less than the divisor, a fraction is made, with the remainder as the numerator and the divisor as the denominator. The "if" clause specifies the condition, and the "then" clause gives the action to be followed. For 13 divided by 3 the estimated whole number is 4. Application of the rule leads to the following procedure: The remainder of 1 is less than the divisor of 3, and therefore the remainder is made into a fraction of 1/3.
To use this rule across division problems, the child must check the current situation to see whether it meets the condition specified in the rule. Such checking can be done only if the rule is represented in some general format. To start with, the child must be able to distinguish which number is currently serving as remainder and which as divisor. Neither remainder nor divisor can be specified in the rule in terms of particular numerical values, such as 1 and 3, respectively, because across problems all numbers can be in both categories.
Researchers can determine whether a child is using such a rule in some set of problems by testing him or her on a number of division problems. The child is said to be using the rule whenever the pattern of behaviors (answers or methods of solution) on some set of the problems fits the rule. The child does not have to state the rule explicitly.
Though the concept of "rule" was controversial two decades ago, today it provides a basis for one of the most promising approaches for exact specification of the cognitive structures underlying child performance. Indeed, it also promises more generally to provide a powerful tool for describing change and continuity in cognitive organization.
In practice, research based on the rule-assessment approach has been characterized by two prominent features. First, it has provided highly differentiated models of regularities in behavior across contexts, including not only correct performances but also errors. This research has articulated the Piagetian hypothesis that errors form coherent patterns that derive from developmentally immature procedures (see Roberts, 1981; Siegler, 1981, 1983). Thus, both errors and correct performances can serve as indexes of the current state of a child's rule system for a particular task domain.
Second, the rule-assessment approach has fostered what might be called a "particulate" view of the child's mind. The methods are designed to detect rules in specified, interrelated tasks, in which the rules are described in terms closely tied to the tasks. Changes in performance are typically explained in terms of modifications, additions, or deletions of particular rules. Just as the philosopher Hume was criticized for depicting the mind as a "bundle of perceptions," some researchers who use rule-assessment techniques might be criticized for depicting the mind as a bundle of rules. Although such localism avoids the postulation of global, vague cognitive metamorphoses, it is in danger of treating the child too narrowly—as merely a solver of division problems, for example.
Early Childhood ConclusionAngela Oswalt, MSW
As this article has shown, children continue to grow and to change in amazing ways as they move from toddlerhood to school-age. During the Preoperational stage, between ages 2 to 7 years, young children continue to grow taller as their bodies take on more adult proportions. They gain the ability to run and to climb stairs independently, as well as to cut with scissors and to grip a writing tool. Cognitively, young children learn how to think symbolically, which leads to make-believe play, and their language explodes and matures. Emotionally, children learn how to express their own feelings and to feel reflective empathy. Socially, they begin to cultivate relationships with peers as well as deepen family relationships. Morally, they begin to understand "right" versus "wrong," and to understand they have the choice about which way to go. Sexually, young children continue to form their gender identity and begin to understand what it means to be male or female.
Beyond just understanding how young children are growing and developing during these early years, parents also need practical, everyday knowledge on how to care for their children and how to meet their needs. More information on Parenting can be found in our corresponding article on Parenting Skills for the Preoperational Level.