The Epigenesis of Mind

7	Constraining Nativist Inferences about Cognitive Capacities

Kurt W. Fischer

Thomas Bidell

Harvard University

Biological factors clearly play a major role in cognitive development, setting constraints on behavior that provide directions to development. As Peter Marler (this volume) has argued persuasively, even the capacity to learn is itself a species-specific characteristic, determined in part by genetic inheritance. For too long, researchers in cognitive development have merely assumed that such constraints are at work and avoided the difficult business of ferreting out the specific nature of the genetic constraints and their relation to the development of cognitive skills. This gap is now beginning to be closed, thanks largely to a newly emerging research tradition variously referred to as structural-constraint theory, rational constructivism, or neo-nativism.

The purpose of this chapter is to offer a critical appraisal of some of the theoretical claims and research methods of this new tradition. We believe that the study of biological constraints on cognitive development is a timely and important new trend in the field; yet in any new approach there is always a risk that new discoveries will be overgeneralized. New, and sometimes startling, information about infants’ and children’s seemingly precocious skills promises to illuminate the relations between biological constraints and cognitive development. At the same time, inferences about innate knowledge or concepts have been drawn that are overgeneralizations not warranted by the evidence. To avoid such overgeneralizations, we suggest theoretical and methodological guidelines for constraining these inferences, placing the evidence within an epigenetic framework that both emphasizes the importance of the early behaviors and specifies their limitations.

GENETIC CONSTRAINTS, GENETIC DETERMINISM, AND EPIGENESIS

Although there is not yet a consensus among neo-nativists as to the defining characteristics of their program, there does seem to be at least one central shared premise, as is evident in this book: The behavioral abilities with which human beings are genetically endowed are far richer and more complex than traditional accounts of cognitive development imply. New research seems to have revealed rich sets of perceptual and cognitive abilities in infants and young children. A key neo-nativist argument appears to be that these early abilities show the starting points from which cognitive development must emerge. As starting points, they set limits or constraints on what is possible and thereby help to channel the direction of development. Thus stated, the notion of genetic constraints complements other contemporary perspectives and helps broaden our overall view of cognitive developmental processes.

However, there seems to be a tension within the neo-nativist perspective between two possible versions of the position. Some interpretations go beyond the general notion of genetic constraints to a strong form of genetic determinism. These interpretations not only posit constraints on the process of development; they seem to imply that certain ideas, like object or number concepts, are in themselves innate.

Consideration of the relation between the human hand and the structures people build with it provides an informative analogy. With its opposable thumb, the human hand is a genetically determined tool with which people both learn about and transform the world. In an important sense, the structure of the human hand anticipates the nature of objects and the possibilities of what people can build. The genetically determined structure of the hand clearly constrains the ways in which people can work and thus sets limits on what we can build. But it does not predetermine or specify the forms of the constructions themselves. Likewise, the genetically determined structure of the brain (plus the nervous system and body) does not predetermine or specify the forms of ideas or concepts. That is the role of the human agent using the hand and the brain working in environmental settings.

The Role of Genetic Constraints in Epigenesis

The broader notion of genetic constraints, as opposed to strong genetic determination, is clearly consistent with the interactionist position from developmental biology called epigenesis (Fischer & Silvern, 1985; Gottlieb, 1983). In epigenesis, the developing organism is often portrayed as moving through an epigenetic landscape that defines potential developmental pathways. A single behavior or state is interpreted in terms of where it fits in the landscape along a pathway moving from earlier (usually simpler) to later (usually more powerful or differentiated) characteristics. Three types of factors contribute to (constrain) the shape of the landscape and the pathway through it—collaborating to mold development: (a) genetic endowment, (b) environment (both social and physical), and (c) self-regulation of the organism. In epigenesis, development is not reduced to any one of these factors but is analyzed in terms of the collaboration among them.

Within the epigenetic perspective there is plenty of latitude for variation in specific theories about the nature of interactions among these factors. For instance, Piagetian theory has tended to emphasize more strongly the role of self-regulation, whereas Vygotskian theory has put stronger emphasis on the social environment. Indeed, because of the gap left historically by an overemphasis on self-regulatory and environmental factors, cognitive developmental theory in general stands to be enriched by a careful examination of the role of innate constraints in relation to the other developmental factors.

However, interpretations based on genetic determination of ideas do not hold much promise of filling this gap in our knowledge, because they effectively reduce the explanation of development to a single factor, the genes, while assigning inconsequential roles to self-regulation and the environment. Genetic determinism cannot effectively illuminate the relations between the genome, self-regulation, and the environment, because it does not address those relations. Just as behaviorism presented a distorted picture of cognitive development by overemphasizing environmental factors, genetic determinism distorts the epigenetic landscape by eliminating the role of the person and privileging the genetic contribution.

Questioning Genetic Determinist Tendencies

Most of the body of neo-nativist research and interpretation appears to point in the direction of a broad genetic constraint model, compatible both with the epigenetic framework and with most conventional theories of cognitive development, including Piagetian constructivism. This work is bringing our knowledge of infant and early childhood cognition to a new level of specificity and thus providing a basis for a new articulation of the relations between genetic, self-regulatory, and environmental factors.

However, some interpretations tend in the direction of genetic determinism. These interpretations suggest that the surprisingly rich repertoire of cognitive abilities being uncovered in infants and young children are appearing so early and are so incompatible with standard developmental theory that they must represent innate concepts.

There are good scientific reasons to examine such claims carefully. Such strong innatist interpretations of behavior can be misleading. They are too easy. They skip the difficult job of teasing out the relations among the multiple factors involved in development and simply assign a genetic origin to the observed behavior.

There are also social reasons to be wary about simple innatist interpretations. Scientific claims about the genetic origins of human intelligence and culture have often been misused to justify discriminatory social policies (Lewontin, Rose, & Kamin, 1984). Scientists therefore have a responsibility to proceed with great caution in developing and disseminating theories of this kind.

In the following pages we examine some examples of interpretations that claim or strongly imply innate origins of concepts. We argue that such claims are unwarranted because they are based in conceptual and methodological flaws. On the conceptual side, these problematic interpretations misrepresent Piagetian theory or provide overly simple interpretations of behaviors that develop early. On the methodological side, this false impression is supported by the failure to properly analyze relevant developmental factors. Finally, we suggest a methodological framework for investigating and interpreting apparently precocious behaviors—a framework that promotes epigenetic analysis and makes such conceptual errors much less likely.

Based on these considerations, we believe that there is no compelling basis for inferring innate determination of ideas or concepts. There is, however, a great deal of valuable new information about the earliest periods of cognitive development that can provide new insight into the role of genetic constraints in human development.

CONCEPTUAL DIFFICULTIES WITH INFERENCES OF INNATISM

Inferences about innate ideas seem to be based on two problematic approaches to interpretation of evidence. The first approach involves the unjustified rejection of the Piagetian framework for interpreting infant behavior based on inaccurate representations of Piaget’s predictions or the confounding of Piaget’s views with other positions. If these erroneous arguments are accepted, a strong nativist position appears to be the default choice.

The second problematic approach is the argument from precocity: If a behavior tied to a particular concept is found precociously—far earlier than in previous research—then the concept associated with that behavior must be innate. This argument is potentially much more powerful than the positions based on misinterpretations of Piaget, because it seems to stand on its own. The fundamental flaw in it is that it misleadingly treats early concepts as isolated capacities that spring up with no apparent developmental context to explain their origin. By ignoring the gradual and sequential way in which even the earliest concepts are formed, this approach makes early abilities seem to lie outside the realm of standard interpretations.

Inaccurate Accounts of Piagetian Predictions

In the most blatant cases, investigators reject Piagetian interpretations on the basis of findings that are, in fact, fully consistent with Piaget’s own observations and predictions, which they misrepresent. In this regard, consider a case study of the development of spatial knowledge in a very young blind girl (Landau, Spelke, & Gleitman, 1984). In a series of studies, the investigators skillfully elicited evidence for a complex system of spatial knowledge in a 2-year-old girl who had been blind since shortly after birth. The girl was familiarized with the pathways between several landmark objects in a room and then required to find novel pathways to various objects. Under a variety of circumstances she was consistently able to choose appropriate novel pathways to the objects, demonstrating systematic skills for locating objects in space. Because of the child’s early blindness, the authors concluded correctly that she could not have gained her spatial knowledge from visual perception, thus ruling out one empiricist hypothesis about spatial knowledge.

However, by comparing their findings with an incomplete account of Piaget’s predictions, the authors falsely concluded that the results disconfirmed Piagetian theory. In fact, the findings are completely consistent with Piaget’s predictions for children of this age. Piaget’s well-known and widely substantiated position is that before 2 years of age children construct a system of sensorimotor spatial knowledge by which they can navigate their world, retrace their routes, and surmount obstacles by inventing novel routes to goals (Fischer & Hogan, 1989; Piaget, 1936/1952, 1937/1954; Uzgiris & Hunt, 1975). In agreement with the findings on the blind girl, Piaget argued that this system of spatial knowledge is the result not of visual perception but of the child’s coordination of her actions into a system of sensorimotor knowledge. Blindness would indeed change important aspects of the sensorimotor knowledge by eliminating visual perceptual cues, but the child would still build spatial knowledge based on her actions and her other senses (Fraiberg, 1974). According to Piaget, the system of sensorimotor knowledge contains remarkable abilities like the blind child’s, including practical problem-solving abilities that are so systematic that they appear logical. These abilities continue to serve people into adulthood, as when one locates a coffee cup on the table without looking up from the newspaper.

This system of knowledge is also quite limited, in that its apparent logic cannot be applied beyond the world of practical actions. Only years later, when children have developed concrete operational skills, are they able to think systematically about spatial relations, as opposed to only acting systematically on them (Piaget & Inhelder, 1948/1967). During early preoperational development, children’s emerging symbolic capacities allow them only limited sets of comparisons, leading to geometrical reasoning that is similar to topological geometries. Later, in middle childhood, they systematize their operational skills and then can think logically about space, using metric or Euclidean geometric principles. Piaget based these analyses on an entirely different set of tasks from those for assessing sensorimotor knowledge, tasks designed to elicit operational thinking, not practical skills.

The findings with the blind 2-year-old actually confirm Piaget’s predictions of a systematic practical spatial knowledge, but the authors neglected to mention this aspect of Piagetian theory in their interpretation. Instead, they compared their findings of systematic practical navigational activities at age 2 with Piaget’s prediction that systematic spatial thinking emerges around age 8 to 9. On this basis, the authors concluded that Piagetian theory was not supported and that the blind girl possessed a precocious Euclidean geometric knowledge. Furthermore, believing that they had ruled out Piagetian theory, and citing Descartes’ view of the origin of knowledge, the authors strongly implied that this seemingly precocious Euclidean geometry must be explained by innate determination.

Confounding Piagetian and Cartesian Frameworks: The Relations Between Perception and Conception

Besides such obvious misinterpretations, there are more subtle ones that also set up an erroneous opposition between neo-nativist findings and Piagetian theory and research. One issue where such misinterpretations abound is the relation between perceptual and conceptual development. The sophisticated Piagetian view of the relation between perception and conception is often confounded with the Cartesian view that informs both nativist and empiricist theories of knowledge. Before examining a case of such confounding, it will be helpful to review the theoretical positions involved.

The Cartesian and Piagetian models of the relation between perception and conception are compared in Fig. 7.1. In the Cartesian view, there is a dualism of mind and action. The perceptual system is considered the sole source of the information about the world that informs the mind, and action is assigned mainly the role of carrying out the bidding of the rational mind (Bernstein, 1985). In the nativist version of the Cartesian view, our concepts are ready-made and only need a minimum environmental input to trigger them (for instance, in the preceding interpretation of the blind girl’s behavior). In the empiricist version, concepts are not ready-made and must be impressed on the mind by an atomistic accretion of perceptual experiences.

The Piagetian approach, on the other hand, does not accept a dualism of thought and action. Feedback from actions on the world is considered to be the main source of knowledge about the world. In the Piagetian view, this feedback comes through two related, partly independent systems: the perceptual system and the sensorimotor system. Both of these are considered systems of action, but the nature of the activities they support differs because of genetic constraints. The perceptual system supports activity within the perceptual field and provides information mainly about the form that the world takes. The sensorimotor system supports actions directly transforming the world and provides information about what one can do with the world, including possible relations among objects in the world.

FIG. 7.1. Acomparison of the Cartesian and the Piagetian constructivist models of the relations between perceptual and conceptual development.

In Piaget’s (1961/1969) theory, the two tracks interact and influence one another over the course of development. Perceptual information helps inform sensorimotor activities, and the information about what it is possible to do in the world helps to extend and correct the information from the perceptual system. Furthermore, both the perceptual and sensorimotor systems are considered to be organized from the beginning. With development, the organization of each system changes. The perceptual system begins with highly automatized perceptual structures like Gestalt field effects that are tightly genetically constrained and dominated by external cues. However, these perceptual structures undergo development influenced by the constructions of the sensorimotor system; they become more mobile and therefore more able to control and organize perceptual stimuli (Piaget, 1961/1969, pp. 189–198).

The development of the sensorimotor action system also begins with an organization, based on various subsystems of actions such as grasping, reaching, and looking. As infants coordinate these sensorimotor subsystems, they control more information about what can be done in the world and therefore about the possible relatipns among people and things.

Thus, Piaget’s theory of the relations between perception and conception differs sharply from the empiricist and Gestalt positions, both of which share the Cartesian framework. Unlike the empiricists, Piaget believed that knowledge, both sensorimotor and perceptual, has organization from the beginning. Unlike the Gestaltists, Piaget believed that the initial organization changes, being reconstructed with development. Unlike both of these positions, Piaget believed that conceptual knowledge is constructed not from the perceptual system alone but from the information about relationships among people and objects obtained initially through the sensorimotor system (which of course makes use of information from the perceptual system, as indicated by the “sensori” in its name).

Research that allegedly contradicts Piaget’s position but really contradicts the Cartesian position turns out to support these broad conclusions. Consider Elizabeth Spelke’s (1988) interpretation of research on infant object knowledge. Spelke placed Piaget among the empiricists, because the idea that sensorimotor systems are initially not fully coordinated seemed to suggest the empiricist position that knowledge is built up atomistically from fragments of sensations. On this basis, she argued that recent evidence showing early organization of infant perception contradicts the empiricist position and therefore Piaget’s views. Then, Spelke argued for a rationalist-nativist position in which the 4-month-old infant is innately endowed with the ability to rationally carve up the perceptual field into concepts like that of object permanence.

The evidence at the base of these claims is not at all inconsistent with Piaget’s actual position. The evidence involves the ability of infants at an early age to see a moving rod as whole even when its center is visually blocked or occluded. This conclusion is based on a habituation paradigm, in which infants viewed a display until they lost interest in it and then were presented with a slightly changed display. If they looked for a longer time at the changed display, they were said to notice the change. For the present, we accept this interpretation of the data, but we come back to consider the habituation paradigm later.

When a rod was moved back and forth behind a block that occluded the rod’s center from an infant’s vision, the infant treated the rod as if it were continuous behind the block, as shown by the infant’s subsequent dishabituation to a broken rod (Kellman & Spelke, 1983). When the object was stationary, dishabituation to broken and continuous objects did not differ, suggesting that infants did not see the stationary object as necessarily continuous. Spelke (1988) correctly concluded that this evidence is incompatible with the empiricist argument that object knowledge is built up from bits of sensory information. Infants at 4 months of age “do not appear to perceive the object as a collection of visible fragments” (Spelke, 1988, p. 202) because they treat it as continuous behind its occluder.

Spelke also presented evidence ruling out other interpretations. For instance, infants did not distinguish between moving dissimilar-misaligned surfaces and moving similar-aligned surfaces. Spelke (1988) concluded that, unlike adults, who can analyze the properties of static objects, young infants’ “apprehension of an object depends only on the object’s motion” (p. 205).

These findings in fact support Piaget’s arguments about the relations between perception and conception in development. First, as we saw earlier, the coordination of sensorimotor systems does not imply atomism but the integration of organized systems of knowledge. For example, contrasting earlier and later development in one of his children, Piaget (1936/1952) wrote that, early on,

… the infant, by manipulating things, constructed a series of simple schemes … such as “shaking,” “rubbing,” etc. These schemes, while still not at all coordinated with each other, nevertheless each comprise an organization of movements and perceptions and, consequently, a beginning of putting objects into relations with each other. (p. 263, our translation, emphasis added)

The integration of knowledge systems does not imply a prior unorganized or fragmentary state. Prior to Maxwell’s equations, for example, concepts about electricity and magnetism were not fragmentary and atomistic. They were simply not integrated into a more encompassing system.

Secondly, for Piaget, perception is structured from the beginning, but the structure changes with development. Here too, the evidence presented by Spelke as a disconfirmation of Piagetian theory is consistent with it. The evidence shows that infants’ perception of objects is indeed structured from an early age. At the same time, the structuring is not complete at this age, because it remains narrowly dependent on the stimulus array, functioning only when the object is in motion. Furthermore, developmental change is in the direction predicted by Piaget’s theory of sensorimotor influence on perception: Older infants are less dependent on the properties of the stimulus array, because they do not require motion to perceive an object as a whole.

The fact that such findings from neo-nativist research programs are consistent with Piaget’s theory does not in the least detract from the value of the findings. It is fundamentally important to determine precisely what infants are able to do at early ages in order to build a more articulated framework for understanding development. Piaget’s ideas may be helpful, and they may be supported by the findings, but much more specific models of developing perceptions and actions are needed.

Ever Younger Ages: The Argument from Precocity

The argument from precocity provides a more serious nativist challenge to conceptions of cognitive development, because it seems to stand on its own, not resting on arguments against some particular cognitive theory. According to this sometimes implicit argument, new findings place certain infant abilities, like number or object concepts, so far below the standard age norms that they simply must be innately determined. As the argument goes, if behaviors related to a concept like object permanence can be demonstrated several months earlier than usual, then no conventional developmental theory can explain them. Innate determination seems to be required by the precocity of the behavior.

Valuable though recent findings about the early capabilities of infants are, a problem arises when they lead researchers to such misleading conclusions. These arguments ignore two well-established principles of developmental research: (a) the principle of developmental variability, especially in age of acquisition, and (b) the principle of gradual, sequential formation of new skills and abilities.

Principle of Developmental Variability. To begin with developmental variability, one of the most widely accepted facts in the study of development is the large variability in developmental patterns (Bidell & Fischer, 1989; Flavell, 1982; Piaget, 1941). Skills do not appear at a single age in all children and all places. They vary in time of appearance through a surprisingly wide range of ages, as a function of factors like children’s learning history and cultural background and as a function of differing assessment conditions. For any particular assessment condition in any one study, the age of acquisition represents only a point within the possible range of ages for that type of task—usually neither the top nor the bottom of the range.

Two especially important aspects of assessment conditions are degree of environmental support and degree of task complexity. Environmental support can involve training, prompting, modeling, or even scaffolding, in which an adult assists a child in a task and thus reduces the cognitive load for the child. Variations in support produce dramatic variations in age of performing a task. In studies in our laboratory, for example, acting out a story about social reciprocity (“I was mean to you because you were mean to me”) varies in age of emergence from 4 to 9 years, depending on the degree of environmental support (Fischer & Bullock, in press).

Task complexity varies with factors like the size and salience of stimulus arrays, the number and organization of actions, and the kind of response required of a child to “pass.” Variations in task complexity produce powerful variations in age of acquisition for a given task (Case, 1985; Halford, 1989). Researchers have produced a large number of task simplifications, resulting in earlier ages of acquisition than Piaget found with his tasks (Fischer, Hand, Watson, Van Parys, & Tucker, 1984; Gelman, 1978), and they have also produced many task complexifications showing later ages of acquisition (Fischer, Hand, & Russell, 1984; Flavell, 1982).

When a change in assessment conditions produces a major change in age of acquisition, a serious question arises: Is the new assessment condition measuring the same task, skill, or concept as the original one? In the Piagetian number conservation task, for instance, children must determine by logical reasoning that the number of objects in an array remains constant even when the configuration of the array is changed. The age of acquisition for giving the correct answer can be lowered by decreasing the number of objects used in the task (Gelman & Gallistel, 1978). At the lower limit for number of items—two or three—children can solve the task by a simple counting procedure, without having to employ logical reasoning at all (Silverman & Briga, 1981). Although children “pass” the simplified task at the surprisingly early age of 3 years, it is not legitimate to treat the behavior as showing the same skill or concept as in the Piagetian version passed at 5 to 7 years. In this segment of the research literature, the difference between the two performances has been clearly recognized by important neo-nativist researchers (Gelman, 1972; Gelman & Gallistel, 1978).

For these reasons, comparing the age of acquisition from studies using very different assessment conditions can be misleading. It can be hazardous to proclaim the presence of a concept on the basis of a performance that seems precocious by comparison with standard findings when the assessment conditions have been altered drastically from standard procedures. Fortunately, the second developmental principle suggests a way of dealing with these interpretation problems.

Principle of Gradual Formation of Skills. The second important developmental principle that is ignored in the argument from precocity is the principle of gradual, sequential formation of concepts and skills. Most developmental researchers agree that skills and concepts do not simply spring up fully formed, but are built up over an extended time period through a sequence of increasingly complex forms. The concept of number, for instance, is not a property just of a single stage, emerging suddenly somewhere around, say, 7 years of age. Instead, the concept of number develops gradually, through a sequence of increasingly complex numerical skills and concepts, beginning in infancy and early childhood and continuing through adolescence and even adulthood.

Yet the argument from precocity depends on an image of cognitive capacities that appear suddenly, in isolation from known sequences of cognitive skills in the same or related domains. Instead of attempting to contextualize apparently precocious behaviors by relating them to a sequence of increasingly complex skills, this argument takes the presence of such behaviors as prima facia evidence that a sophisticated concept, normally appearing at much later ages, is “really” present in early infancy.

What could it mean to say that a concept like number or object permanence is present at an age when in most relevant tasks the baby shows no inkling of it? It is misleading to first locate a precocious behavior through ingenious tasks and procedures and then treat it as if it were indicative of a stable capacity functioning independently of the highly supportive environment in which it was observed. The very ingenuity needed to detect such abilities suggests that they represent early steps in a long developmental trajectory eventually leading to an independently functioning skill.

The argument for precocity seems to stem from what might be called adultocentrism. If an adult looking at an infant’s behavior sees it as implying a concept of number or a concept of object, the inference is made that the infant must be using the concept. The jump from an adult interpretation to inference of a skill or concept in an infant is gigantic. Behaviors can look alike in terms of adult categories without actually being alike in the way the child produces them. It is no easy matter to differentiate behaviors that are truly homologous, reflecting the same underlying skill or concept, from those that are merely analogous, reflecting different skills that produce a superficial similarity (Tomasello & Farrar, 1984).

To explicate the principles of developmental variability and of gradual, sequential development, we present some concrete examples in the next section. Following that, we consider the methodological implications of these principles, considering how to take account of them in the design of research and re-examining a well-known study from this methodological perspective.

Interpreting Apparently Precocious Behaviors: Visually Guided Reaching

The early development of visually guided reaching highlights the problems with the argument from precocity that we have been discussing. When she was just one day short of 2 months old, Johanna Fischer (daughter of one of the authors) showed a remarkable “capacity”: She repeatedly reached out and touched her father’s mouth so that he could kiss her hand. Thus, she demonstrated effective visually guided reaching at an age long before other studies have found it (von Hofsten, 1984; Piaget, 1936/1952), even long before the ages when Spelke found object perception. Here is an excerpt from the diary description of the event:

We were in an airplane, with Johanna in her mother’s arms in the seat next to me. She was moving her right arm around in the awkward manner of a 2-month-old. Her hand happened to graze against my mouth, and I kissed it. A few seconds later, she moved her hand again in the general direction of my face, and I kissed her hand again. Over the next 5 minutes, she repeated this movement at least 10 times, and each time I kissed her hand. The movements were global, only coming in the general vicinity of my face, not specifically touching my face or mouth; but I always made sure that my mouth was situated so that I could kiss her hand.

This behavior was shocking, despite its crudeness. The baby was clearly showing a behavior that is not “supposed” to develop until 3–5 months later. A difference of several months is huge when a baby is only 2 months old.

Later observations provide a clue as to the reasons for this early behavior:

Five minutes later, she was in a different position in her mother’s arms. When her arm came toward me, I kissed her hand. She did not repeat the movement toward me. Several other times when she was in different positions, she again did not move her hand systematically toward my face. It appears that the anticipation/operant movement could be repeated only in a limited posture.

In the posture where she repeated the movement, her arm was constrained from shoulder to elbow, so that she could only move her forearm and hand. Also, her hand was already pointed at her father’s face and close to it. She needed only to make a coarse swipe with her hand to reach his face and receive the kiss. Other research indicates that 2-month-olds are capable of such swipes (Bruner & Koslowski, 1972; von Hofsten, 1984).

Was Johanna capable of visually guided reaching, or wasn’t she? Clearly, that is not a satisfactory phrasing of the question. Under very limited circumstances, she was capable of it, but under others she was not. Only at 7½ months of age could she reach out to touch her father’s face from a wide variety of postures and positions. And even then there were many situations where she did not produce such a behavior.

To check out the generality of this skill, his father held a familiar small toy rattle directly over the beads. Seth looked at it and reached for it. But instead of skilled reaching, he made an awkward swipe, flailing and groping with his hand and missing the object. Similarly in his crib, Seth reached skillfully for familiar objects, such as a rolling cylinder on his activity box. Again, this skill did not generalize. When a toy was held out to him in the same location, he swiped gropingly and awkwardly at it instead of effecting a skillful visually guided reach. Over eight repeated trials, he continued to make an awkward, groping swipe.

Seth did not have a capacity for visually guided reaching in the normal sense of “capacity.” Instead, he had limited skills for using vision to reach skillfully in highly familiar situations. Not until 7 or 8 months of age could he reach skillfully in many novel situations.

It is hard to resist the temptation to attribute broad capacities to the child based on the evidence of a few surprising behaviors. In both of these cases of precocity, the behavior was unquestionably an instance of visually guided reaching—a behavior that is sometimes thought to be even more difficult than visual perception of objects. Consequently, the temptation was strong to make a straightforward generalization: The infant had the “capacity” for visually guided reaching. But, as the other observations make evident, this “capacity” is not a simple unitary competence that emerges at one point in development.

The behaviors of Johanna and Seth are relatively complex—a pattern of movement guided by visual information. When a behavior is less complex, as with habituation, it is even less obviously a direct reflection of a hypothesized capacity, and the problems of inferring the capacity become even more difficult. In studies of habituation, a capacity is inferred when an infant dishabituates to one type of stimulus and not to another. From such behaviors it is a great leap to infer that the infant possesses a concept of number or of object (Moore, Benenson, Reznick, Peters, & Kagan, 1987; Spelke, 1988, this volume). The complex capacity is inferred from a very simple behavior, such as looking at one object for a few seconds longer than a second object. It is no easy matter to ensure that the looking actually arises from the stimulus characteristic that interests the researcher. Many other interpretations are possible, including artifacts of familiarity and regression to the mean (Bertenthal, Haith, & Campos, 1983). Consequently, even more supporting evidence is needed to justify the interpretation that this simple behavior reflects a complex capacity.

METHODOLOGICALGUIDELINES FOR RESEARCH: PLACING A BEHAVIOR IN A DEVELOPMENTAL FRAMEWORK

The principles of developmental variability and gradualness are not only relevant to the interpretation of research but also to its design. In order to draw sound conclusions about precocious behaviors, developmental researchers must design their research to take into account both variability in age of acquisition and gradualness of development. The failure to address these issues can lead researchers to reject findings or interpretations when rejection is uncalled for and to jump to unwarranted conclusions about the innate nature of concepts. Problems of interpretation become much more straightforward when research is designed to place findings within a developmental framework.

Dealing with Variability and Gradualness in Development

The variability in age of acquisition as a function of assessment conditions and the gradualness of development make it imperative that assessment conditions be varied or controlled. Unless assessment conditions are systematically varied or are equated with those in comparison studies, comparative ages of acquisition can be greatly misleading (Fischer & Canfield, 1986). When variations in conditions are neglected, it becomes difficult to specify any meaningful age of acquisition, and developmental relations with other skills are obscured. Nevertheless, researchers searching for precocious abilities have at times treated age as something that varies in only one direction—down—and ignored the wide variability in age as a function of condition.

This problem is compounded in cases where, as discussed earlier, manipulations of the assessment condition alter the complexity of the task so as to make questionable its comparability with the comparison task from prior research. If there are no controls or variations in assessment conditions built into the research design, there is no way of telling whether one is measuring the age of acquisition for a target concept, like object permanence, or for some less complex precursor concept, like the understanding that objects exist in the first place.

The issue of relative task complexity is closely associated with the second developmental principle, the gradualness of skill acquisition. Ignoring the influence of variations in task complexity implicitly ignores the gradual development of abilities through a long sequence of increasingly complex skills. Developmental research across a wide range of domains and at every age level supports the conclusion that new skills do not spring up full grown but emerge from previously existing skills and in turn provide the basis for further development.

The variability and gradual nature of cognitive development makes it essential that research be carefully designed to define newly discovered behaviors in terms of three patterns: (a) the developmental sequence of acquisitions that lead up to and follow it in the domain, (b) developmental synchronies or correspondences, the acquisitions in other domains that are typically contemporaneous with it, and (c) clusters of behaviors that “move” together with the target behavior as a result of manipulations that produce performance shifts (Fischer & Bullock, 1981; Fischer & Farrar, 1987).

Developmental Sequences

When researchers can show that a behavior such as a particular type of visually guided reaching forms a step in a developmental sequence with other behaviors, the sequence aids them in interpreting the behavior. Placing the behavior in the sequence circumvents fruitless arguments about whether a particular behavior “really” demonstrates a particular capacity. Each successive step in the sequence is a further realization of the capacity, which emerges in the gradual manner typical of epigenesis.

Infants show a regular developmental sequence of visually guided prereaching in the situation of sitting upright facing a ball (Bruner & Koslowski, 1972; von Hofsten, 1984, 1989). The work of Piaget (1937/1954) and many others on object permanence documents a sequence for more advanced visually guided reaching up to 2 years of age (Diamond, this volume; Uzgiris & Hunt, 1975). Fischer and Hogan (1989) presented a detailed framework for describing and integrating these sequences into an epigenetic portrait for visually guided reaching. The precocious visually guided reaching that was described earlier gains meaning through its place in these sequences.

This framework moves beyond arguments about whether a behavior shows or does not show a capacity. The behavior is interpreted as fitting a point in an epigenetic landscape for what eventually becomes a rich and powerful skill of visually guided reaching. Following is a sketch of the broad sequence for visually guided reaching for middle-class American and European infants.

At 1 month of age, infants facing a small ball in front of them readily look at it so long as their posture does not prevent looking. Likewise, they grasp a ball placed in their hand (Thelen & Fogel, 1989).

By approximately 2 months, when Johanna showed her visually guided reaching in the airplane, infants sometimes show a simple reaching behavior: They look at the ball and produce a coarse reach, in which they move their arm toward it with their fist closed (von Hofsten 1984, 1989). Johanna’s arm was positioned so that she could only make a limited set of movements, and all that was necessary was something like coarse reaching toward the ball. Her behavior thus seems to fit at about this point in the sequence for visually guided reaching. She groped with her hand toward her father’s face, with no need for precise localization nor opening her hand to grasp. In a highly constrained situation, she could thus produce a limited type of visually guided reaching, coordinating looking at a face with reaching for it. But the behavior occurred only with strong contextual constraints on her action. The contextual constraints needed for the action are as important as the biological constraints that helped her to develop this early form of visually guided reaching.

By 3 months, infants extend their arm toward the ball and open their hand as they do so, thus coming closer to true visually guided reaching. But even then they often miss the ball.

More general, effective visually guided reaching does not develop until 7 or 8 months of age, when infants show many visually guided reaching skills: They skillfully use looking at an object in many different positions to guide how they reach for it. They even begin to search for objects hidden under cloths or behind screens (Diamond, this volume; Piaget, 1937/1954).

Seth at 5 months apparently demonstrated early evidence of such skill when he reached adeptly for familiar objects and flailed awkwardly at unfamiliar objects. In a manner that is typical of the skill generalization that occurs across developmental domains (Fischer & Farrar, 1987), he initially mastered visually guided reaching in a few limited, familiar situations. Over the next few months, he then generalized and elaborated those skills until by 7 to 8 months he had a broad capacity for visually guided reaching.

Of course, even this advanced capacity is not the end of the developmental sequence. By 12 to 13 months, babies become much more facile with coordinating vision and prehension. They can skillfully pick up a ball and move it around, using what they see to guide what they do and anticipating many of the consequences of moving the ball a particular way. They even carry out little experiments in action relating how they move the ball with what they see so as to determine how to accomplish special goals, such as dropping the ball through a small hole in a box. This skill continues to grow for years to come.

As these examples demonstrate, placing individual behaviors in the framework of a developmental sequence illuminates their significance. Each behavior becomes a step in a pathway toward a broad capacity. No particular behavior is treated as the one true demonstration of the capacity. Arguments among researchers about which behavior indeed shows the capacity and when the capacity first develops are virtually eliminated.

Developmental Synchronies

Besides the vertical framework of a developmental sequence, the interpretation of a behavior is also aided by relating it to a horizontal framework—the other behaviors that the infant demonstrates at about the same age. Although children show only a few tight synchronies in development at any given age, there are general correspondences between behaviors that develop on average at about the same time (Case, 1985; Fischer & Farrar, 1987). Developmental correspondences in the epigenetic landscape help the investigator to interpret individual behaviors.

One simple way to investigate synchronies is by seeing how changes in context affect a behavior. In the cases of both Johanna and Seth, the significance of their early visually guided reaching was immediately clarified by observations of how their reaching at that age varied as a function of changes in the situation. Johanna could reach out to touch her father’s face only when her arm movements were drastically constrained. In other situations, she did not successfully reach. Seth could reach to adeptly grasp things only when they were familiar objects in familiar situations. Changing to a novel object or situation immediately disrupted the skilled reaching.

More generally, researchers can use the set of skills that develop at about the same period to illuminate the significance of individual skills. For example, we compared Johanna’s and Seth’s visually guided reaching to von Hofsten’s (1984) findings about visually guided reaching in a different situation.

One of the most powerful examples of the importance of this strategy comes from the literature on object permanence. At approximately 8 months, infants search for a toy hidden under a screen. At the same period, when they are presented with two screens, they make the A-not-B error: They search under the screen where they previously found the object (screen A) instead of under the screen (B) where they just saw the adult hide it (Piaget, 1937/1954; Diamond, this volume). The occurrence of this error helps to limit the interpretation of babies’ skill at reaching for a hidden object: They have learned to search for objects that they see hidden, but they do not generally understand the importance of where the object actually disappeared. Diamond (this volume) uses other behaviors, including detour reaching, in a similar way to illuminate the baby’s capacity at this age.

Finding Developmental Clusters: The Role of Experience

In an epigenetic process, many factors in both the child and the environment contribute to the form of development. Even the sequence of behaviors and the synchronies among behaviors change with variations in important organismic and environmental factors (Fischer & Farrar, 1987; Fogel & Thelen, 1987). When these variations occur, some behaviors move together, remaining tied in a sequence or developing in close synchrony even though the age of emergence has changed. Such developmental clusters provide important evidence about the effects of specific experiences on development.

One of the best examples relates closely to visually guided reaching. The onset of crawling at about 8 months of age seems to induce development of a cluster of spatial skills, including visually guided search, according to Campos (in press; Campos & Bertenthal, 1987). In normally developing infants, the appearance of upright crawling appears to induce advances in skills of searching for hidden objects and appreciating the danger of heights, among others. A cluster of these behaviors seems to develop a few weeks after infants can crawl on their hands and knees.

Various studies converge on the conclusion that it is the experience of crawling itself that induces the new behaviors. For example, upright crawling—on the hands and knees—seems to be necessary to bring about the change. Cruder forms of crawling, such as dragging oneself along on the belly, do not seem to have the same effect. Likewise, handicaps that prevent crawling, such as spina bifida and orthopedic problems, delay the development of these spatial skills in infants who are otherwise cognitively normal. The handicapped infants do not demonstrate the relevant spatial skills until they have developed crawling, even when the delay involves several months.

The development of a cluster of spatial behaviors after the emergence of crawling suggests that the behaviors are all induced by the experiences that arise from crawling. The covariation of the behaviors with crawling thus helps to constrain inferences about how the behaviors develop. This kind of activity-based experiential induction is a central component in epigenesis and is a crucial part of a number of models of how innate factors affect development (Changeux, 1983/1985; Gottlieb, 1983; Marler, this volume).

Constraining Generalizations about a Behavior

When a behavior is taken out of the epigenetic context of sequence, synchrony, and clustering, there is no basis on which to constrain generalizations about the capacity it represents. Without these constraints, it is easy to overgeneralize and create the impression of remarkably precocious abilities where they do not in fact exist. To take an extreme example, a species of sunflower found in the Rocky Mountains (Hymenoxys grandiflora) always faces east. It would be absurd to conclude that the sunflower possesses a concept of east or a cognitive capacity to determine directionality. What makes this an absurd proposition is the constraint on generalization derived from knowledge of the kinds of other “abilities” exhibited by these plants. In order to avoid such misleading conclusions about cognitive development, researchers need to design studies so that the results can be contextualized in terms of sequence, synchrony, and clustering of acquisitions.

Consideration of concepts or skills in isolation, outside the developmental context, heightens problems of interpretation and makes faulty generalizations likely. The behavior of constantly adjusting to face east, viewed out of developmental context, seems to indicate a concept of directionality and to lead to the conclusion that the Rocky Mountain sunflower possesses that concept. Similarly, when researchers consider a behavior, such as dishabituation to a perceptual stimulus, without regard for its place in the developmental context for its domain, they jump to mistaken conclusions, such as that it demonstrates object permanence. The best way to constrain such inferences is to systematically design research to reveal the developmental context of behaviors.

RE-ASSESSING APPARENTLY PRECOCIOUS OBJECT-PERMANENCE BEHAVIOR

With these methodological considerations in mind, let us examine some recent research by Baillargeon and her colleagues (Baillargeon, 1987; Baillargeon, Spelke, & Wassermann, 1985). These findings have been taken to indicate the precocious presence of object permanence and have been given strongly innatist interpretations. Using innovative procedures, the investigators elicited behaviors that seemed to indicate the presence of object permanence 4–6 months sooner than found by Piaget (1937/1954).

In the study by Baillargeon (1987) infants were habituated to the sight of a small door rotating horizontally 180° from a flat position lying toward them, then to a flat position lying away from them, and then back again. Next, they witnessed two scenes involving a block behind the door. In one scene, called the possible event, the door swung up, stopped at the block, and swung back to a flat position, revealing the block again. The second scene, called the impossible event, was the same except that the door appeared to swing right through the space occupied by the block, as if the block had disappeared. In reality, the block sank down in a way that the infant could not see, thus allowing the door to pass. The door then swung back up and into its initial flat position, again revealing the block. Infants as young as 3½–4½ months dishabituated to the latter, “impossible” situation.

Baillargeon argued that the infants’ dishabituation was indicative of a concept of object permanence, because the infants apparently expected the object to continue to exist when out of sight behind the door, and to be substantial enough to resist the door. She argued further that this finding disconfirmed Piagetian theory, because the behavior appeared several months earlier than the standard Piagetian indicator of object permanence, search for hidden objects. Furthermore, Baillargeon argued that because the object concept seems to come before the search for objects, infant’s object knowledge must come from some source other than the sensorimotor coordinations posited by Piaget. Having rejected Piagetian theory on the basis of this apparent age discrepancy, Baillargeon proposed two innatist hypotheses: Innate mechanisms either directly determine the object concept, or they determine a highly specific learning process that determines the object concept at a very early age.

This research illustrates the methodological issues we raised about innatist interpretations. The studies are not designed to take account of developmental variability, nor are they placed in the context of developmental sequence and synchrony. The research questions in these studies are posed in a yes-or-no fashion that presupposes the meaning of the behaviors under study. Under these conditions, the results appear to reject Piagetian interpretations and to strongly support an argument from precocity, with the implication that the object concept is innate. When the findings are considered in light of the methodological issues of developmental context, however, it becomes clear that they neither reject Piagetian theory nor call for theories of innate concepts.

When is Object Permanence Present?

Regarding the issue of age variability, these studies made no attempt to vary systematically the aspects of the assessment conditions that might affect age of acquisition. This omission makes it hard to interpret the finding that 3½–4½–month-old infants dishabituate to the disappearing-object stimulus. Even if we temporarily grant that the dishabituation behavior heralds the type of object permanence assessed in the Piagetian Stage 4 task (we will see shortly that this is unlikely), the assessment tasks were designed to drive the age of acquisition to the lower limit of the age range for object knowledge.

The Piagetian age norm of 8–9 months is based on the complex Stage 4 double-search task, which involves hiding an object not once but twice and is designed specifically to probe for a mature and independent understanding of the object concept. When Piaget (1937/1954) used simpler tasks, children showed object-related behavior at much earlier ages—at 3–4 months or even earlier.

Without systematic variation in assessment conditions to reveal the range of variation for the new habituation task, there is no way of judging whether or not the range of ages for this task matches that for the Piagetian task. It is methodologically unsound to compare a task sampling the lower age limit of object knowledge with Piaget’s double-search task, which samples higher in the range. We propose a model that takes account of the range, analyzing it in terms of a developmental sequence of types of object permanence.

What Kind of Object Permanence is Present?

When the issue of task complexity is examined, it becomes clear that Baillargeon’s habituation task is not comparable in complexity to Piaget’s double-search task. One reasonable possibility is that the dishabituation behavior involves an understanding that is one or two stages below the kind of object permanence tested by the double-search task.

Both Piaget’s (1936/1952, 1937/1954) original theory of sensorimotor development and Fischer and Hogan’s (1989) skill-theory analysis of infant cognitive development offer similar alternative hypotheses to Baillargeon’s: We show how within these frameworks, the dishabituation behavior represents an early stage in the development of object permanence.

Piaget (1936/1952, 1937/1954) described six stages of sensorimotor development, during which children gradually construct increasingly complex understandings about themselves and the world, including object knowledge. With the exception of the first two stages, in which objects are not very differentiated from the infant’s own actions, each stage involves some knowledge about objects. The nature of the object knowledge changes with each stage.

The most frequently cited forms of object permanence in Piaget’s scale come at Stages 4 and 5, but Stages 3 and 6 also involve acquisitions of important information about the permanence of objects. After all, just knowing of the existence of objects is an important component of knowledge about their permanence, but it is not the whole story. Understanding that objects are permanent no matter what you do with them is a hard-won piece of knowledge that comes late.

In Piaget’s theory, the first clear understanding of the independent existence of objects comes at Stage 3. At this stage, the infant of about 4 or 5 months recognizes that the object constitutes a whole and actively searches for hidden objects, as long as there is some part of the object still showing. This is a big step in the development of object permanence. In this step the infant has gone from treating objects as part of his or her actions to recognizing their independent existence and searching for them in a way that is consistent with that level of object knowledge. Piaget’s Stage 3 infant already has a sophisticated knowledge about objects, even if there is much more to learn.

At Stage 4 of Piaget’s sequence, the infant of about 8–9 months makes further progress in his or her understanding of object permanence. The infant at this stage understands that a completely hidden object can be retrieved. What the infant does not yet understand is many of the complexities of retrieval. In a classic task, the object is hidden in one place, where he or she finds it, and then with the infant watching, it is hidden in a different place. The infant will usually look for the object in the first hiding place, not the second (the A-not-B phenomenon, discussed by Diamond, this volume).

Piaget’s criterion for Stage 5 object knowledge requires a sophisticated and systematic type of search. Infants of about 12 months can understand where the object is through multiple hidings, so long as they see the movements of the object. According to Piaget, they now understand the object not only in itself but in its relation to other objects.

Even this highly sophisticated understanding of objects is not the last step in the development of object permanence. At the sixth stage, infants of about 18 months acquire the ability to represent object locations even when some of the movements of the object are hidden so that they do not observe the final hiding of the object. Infants look for the object systematically in one place and then another until they find it.

Which of these several kinds of object permanence does Baillargeon’s dishabituation behavior relate to? Baillargeon’s (1987) discussion does not specify the Piagetian stage but refers to the behaviors and ages of Stages 4 and 5. In rejecting the Piagetian theory she wrote that in Piaget’s view, “it is not until infants reach about 9 months of age that they begin to view objects as permanent” (Baillargeon, 1987, p. 662). Yet Piaget believed that infants begin to view objects as permanent several months earlier, at Stage 3. It seems likely that the dishabituation behavior she observed actually involved Stage 3 of object knowledge.

At least two hypotheses about the dishabituation behavior are consistent with early stages of object knowledge. First, infants who understand that objects exist independently will exhibit expectancies based on that knowledge. Indeed, Haith, Hazan, and Goodman (1988) have shown that 3½–month-old infants quickly learn to expect to see an object where they have seen one before. When an object is occluded, an infant at Stage 3 could legitimately be disappointed at the disappearance, without necessarily knowing that the object continues to exist behind the occluder. Furthermore, an infant who was interested in the object before it was occluded could hope for the appearance of at least some interesting object in the place where the object had been seen.

In Baillargeon’s possible event, the object was occluded by the door and then appeared again. When the object was occluded (disappeared), it was replaced with another object—the door—which was familiar to the infant because of the habituation trials. If, as would be consistent with Stage 3, the infant did not know that the first object continued to exist (because no part of it showed), he or she would simply find the door a boring substitute and stop looking. However, in the case of Baillargeon’s impossible event the object disappeared, and the door swung down flat leaving an empty field, with no object to replace the missing object of interest. In this case, it would be consistent with a Stage 3 understanding, based strictly on the knowledge that an object (door or block) did exist there, to expect some object to appear. This would account for the longer period of looking (dishabituation) at the “impossible” condition.

Such behavior, though consistent with Stage 3 knowledge that objects exist independently, does not presuppose a knowledge that they continue to exist when hidden. To understand that an object continues to exist behind an occluder requires not only knowledge of object existence but also knowledge of the relations between objects and occluders, which develops at Stages 4 and 5.

The second hypothesis is that the dishabituation behavior indicates a transition from Stage 3 to Stage 4, the stage in which infants search for even completely hidden objects but only in the place where they saw them vanish. From this perspective, the infants who looked longer at the impossible situation would be demonstrating the beginnings of this search pattern. In the possible situation, the familiar door blocks visual access to the object of interest, and, because the infant is located too far away to manually reach the door and finds no possibility of visual search, he or she may simply give up. However, in the impossible situation, the object is missing for a time, and a visual search for it where it disappeared would be both possible for the infant and consistent with the hypothesis that the infant is beginning to show Stage 4 behavior. This transition hypothesis seems to be further supported by the fact that less than half of Baillargeon’s 3½–month-olds showed the dishabituation behavior, suggesting that it may reflect an emerging ability.

Either of these hypotheses (and possibly others) could explain the dishabituation behavior in ways that are consistent with Piagetian theory, with no need to introduce a dualistic separation between the infant’s object knowledge and search activity. Fischer and Hogan’s (1989) neo-Piagetian skill theory supports similar hypotheses. This theory describes a developmental sequence of levels of action-based cognitive control structures or “skills.” This sequence includes greater explication of developments in the first months of infancy than provided by Piaget. At each new level, infants comprehend and control more extensive information about themselves and the world, including objects. An especially important level emerges at 3½–4 months, when infants reorganize their previously reflex-like control structures into more stable skills controlling sensorimotor actions. This new capacity allows them to understand the results of simple actions on objects and, therefore, the independent existence of objects. This theory too, then, is consistent with the hypothesis that the dishabituation to the impossible situation was due to a simple expectancy about independently existing objects.

From these considerations it should be clear that Baillargeon’s findings do not allow clear determination of which kind of object knowledge is being assessed. If the dishabituation task requires Piaget’s Stage 3 of object permanence knowledge, it can hardly be used to reject Piagetian predictions. A crucial part of testing hypotheses about early object knowledge is to systematically vary task complexity to evaluate the sequence of acquisitions for the domain of each type of task (for relevant data, see Spelke, this volume).

Which Object Permanence is THE Object Permanence?

In developmental research, an ability cannot be legitimately defined in an all-or-nothing fashion, with one task marking the presence or absence of that ability. An ability like object permanence is not a fixed capacity that is either present or absent. Some form of object permanence is present throughout most of infancy. To ask whether object permanence is present at 3½ months of age is to ask the wrong question. Because there are different kinds of object permanence, such a question only leads to fruitless arguments over which kind is the real object permanence, which task is the right assessment, or what is the true age of acquisition.

Because of the changing nature of skills with development, the meaning of any behavior, including dishabituation to a missing object, has to be determined by examining its place both vertically, in the developmental sequence for its domain, and horizontally, in the cross-section of contemporaneous abilities in other domains. These vertical and horizontal directions have to be determined partly on an empirical basis, although a theory of skill complexity and generalization is certainly helpful as well.

Our two hypotheses considered the dishabituation behavior relative to a developmental sequence. It also must be considered in relation to synchronous developments in other domains. For instance, Spelke’s (1988, this volume) research has shown development of the ability to perceive the continuity of a moving rod at an age that is approximately synchronous with Baillargeon’s dishabituation to the missing object. Visually guided reaching also seems to show approximately synchronous developments, as illustrated in the description of the development of reaching. Taken together, all these behaviors seem to support the notion that infants around age 3–4 months begin to understand objects as independently existing wholes.

The context of developmental sequence and synchrony provides powerful scientific tools for placing precocious behaviors within an epigenetic landscape. These tools provide a set of reference points from which one can begin to ask difficult questions about the relations among the genome, the activity of the organism, and the environment. Instead of asking when infants “really” have object permanence, a more fruitful set of questions includes: What is the developmental sequence of object knowledge from earliest infancy through early childhood? How is the development of this sequence related to developments in other domains? How is it constrained by the nature of perceptual and sensorimotor processes, which are partly regulated by the genome, and by environmental inputs? How are such constraints evident at various points in developmental sequences?

A MODEL FOR RESEARCH ON DEVELOPMENTAL CONSTRAINTS: THE CASE OF CLASSIFICATION BY COLOR AND SHAPE

Research on biological constraints needs to move beyond the overly simple interpretation of early appearing behaviors as innate capacities to more complex questions about the mechanisms by which biological constraints influence the course of development. The developmental framework we have been describing can provide powerful methods for testing developmental constraints, methods that can go beyond analyzing merely one point in development to investigating systematic developmental patterns over longer periods. By assessing developmental sequences and synchronies in detail, it is possible to examine the operation of a general constraint or bias over a broad developmental period. A study can be designed to test for a constraint at every step in a developmental sequence, with several tasks matched for cognitive complexity at each step to assess whether there is a bias (Pipp, Fischer, & Jennings, 1987). A constraint that shows up repeatedly throughout a developmental sequence is clearly more general and powerful than one that appears only for a single task, although generality and power do not necessarily imply a simply innate basis (Fischer & Bullock, 1981).

The classification of objects is a promising candidate for examining the operation of some perceptual constraints (Gelman & Baillargeon, 1983). For example, shape has often been found to be more salient to preschool children than other attributes, such as color (Bornstein, 1985; Landau, Smith, & Jones, 1988; Rollins & Castle, 1973), although the relation can be reversed by using bright colors and obscure shapes (Odom, Astor, & Cunningham, 1975). One possible hypothesis is that children show a general bias toward noticing shape more than color.

Pervasive Constraint for Shape over Color

To study the development of early classification skills, Fischer and Roberts (1989) devised a series of tasks where children sorted blocks by shape and/or color. The hypothesis that development is constrained by a bias toward shape over color was tested by administering separate shape and color versions of each task. The blocks used for sorting were 6 mm thick and comprised of three colors (red, yellow, or blue) and three shapes (triangle, circle, or square). In most tasks, the colors were highly saturated and represented the prototype for that color. The shapes were also prototypical—circles, squares, and equilateral triangles.

As shown in Table 7.1, there were four main tasks, with a color and shape version of each. The first task tested children’s sorting in terms of a single category when two uniform categories were present. The categories were uniform in that all blocks within a category were identical. For example, in the color task for Step 1, a child was presented with a pile of circles that were identical except that some were red and some yellow. Within the color categories, all the reds were identical and all the yellows were identical. The child had to sort out the yellow ones. In the several versions of this task, the categories were yellow and red for color, and circle and triangle for shape. To pass the task for shape or color, children needed to sort at least one of the two categories correctly.

TABLE 7.1
A Developmental Sequence of Classification Tasks

Step	Classification Task	Task
1	Single Category among Two Uniform Categories	Blocks vary on two salient categories, and each category includes only uniform blocks (e.g., all circles are exactly the same shape, color, and size). Child can sort them into one category (e.g., red).
2	Multiple Uniform Categories	Blocks vary along only one dimension (shape or color), and each category includes only uniform blocks. Child sorts all three categories correctly.
3	Multiple Variable Categories	Blocks vary along one dimension and each category is variable, including multiple types of blocks (e.g., different shades of red). Child sorts all three categories correctly.
4	One Dimension of Uniform Categories with Interfering Dimension	Blocks vary along two dimensions (color and shape), and each category is otherwise uniform. Child sorts correctly along each single dimension separately.

The second task required sorting three uniform categories, either three shapes (triangle, square, circle) or three colors (red, yellow, blue). The third task was similar to the second one, except that the blocks now varied within each category, with different shades of a color or different versions of a shape.

In the fourth task, the children again had to sort blocks into three shapes or three colors, but the composition of each category changed in a new way. All blocks of a given color or shape were once again uniform within that category, but they varied along the other dimension. That is, the children had to sort blocks by shape when the color categories were also present or by color when the shape categories were also present. They had to sort along one dimension when the blocks presented the option of using the interfering second dimension.

A total of 70 middle-class children were tested on the tasks, as well as some more complex classification tasks that did not provide clear tests of the hypothesized constraint. Five boys and five girls were tested for each year between 1 year 3 months and 7 years 3 months. For the purposes of this chapter, we report only the 34 subjects whose performance allowed differentiation of the four tasks shown in Table 7.1 (that is, those who passed some of the tasks shown in Table 7.1 and failed all the more complex tasks).

TABLE 7.2
Scalogram Profiles for Classification Tasks

Note: + indicates correct performance, – incorrect performance. Age is in months.

TABLE 7.3
Distribution of Profiles for Shape and Color Tasks Ordered by Highest Step Passed

Note: + indicates correct performance, – incorrect performance. S and C stand for color and shape versions, respectively, of the task for a step.

The children were assessed individually, sitting on the floor of a carpeted playroom with the experimenter. Each task required the child to sort blocks into open boxes (one box for the first task and three for the other tasks). The boxes were designed so that the blocks could be easily placed into them and seen by the child. The experimenter demonstrated each task to the child and then described the arrangement of blocks to highlight how he had sorted them. For example, when the blocks were sorted into three boxes by color, he said, “All the red ones go here in this box, all the blue ones go here, and all the yellow ones go here.” Then he removed the blocks from the boxes, put them in a scrambled pile in front of the child, and said, “Put the blocks in the boxes so they go together like the way I put them in.” If the child sorted the blocks incorrectly, he or she was given a second opportunity to sort the blocks.

The tasks were administered in a scrambled order so as to minimize the production of artifactual scaling from practice or fatigue effects. The order of the color and shape versions of each task was counterbalanced across children.

The children showed a straightforward developmental sequence for the four tasks. Every child fit the scalogram profiles predicted for the sequence, as shown in Table 7.2 for the combination of shape and color tasks. Most 1-year-olds could sort single categories (Step 1). At 2 and 3 years, children correctly sorted first multiple uniform categories and later multiple variable categories (Steps 2 and 3). By 4 years, most children could sort along a dimension of categories when an interfering dimension was present (Step 4).

For testing the constraint hypothesis—that children can more readily sort prototypical shapes than prototypical colors—the four tasks were broken down into shape and color versions. The results strongly supported the shape-over-color hypothesis. For all four tasks, every child who showed a difference between shape and color versions passed shape and failed color, as shown in Table 7.3.

Also, the shape version of the interfering dimension task (Step 4) proved to fall at approximately the same point in the developmental scale as the shape version of the task from the previous step (Step 3). The color version, on the other hand, developed later. This finding makes sense in terms of the bias toward shape. Dealing with shape in the presence of color is equivalent to dealing with shape alone; color does not interfere. But dealing with color in the presence of shape is difficult, because the bias toward shape means that the child must constantly guard against sorting by shape.

In addition, all children except the 1-year-olds were given a fifth task to test how they would sort spontaneously. Before they did the dimension task (Step 4), they were presented with the same blocks as in that task but without modeling. They were instructed to put the ones together that “go together.” To demonstrate a preference, children had to sort seven out of nine blocks correctly on one dimension. By this criterion, 9 of the 24 children preferred shape, and 15 showed no systematic preference.

In summary, the results supported the shape-over-color constraint for all five tasks. It existed not only at one point in the developmental sequence but for all four steps, spanning a period from approximately 1 to 4 years of age. Children showed a strong constraint in favor of shape.

Developmental Scope of a Constraint

This method for detecting constraints provides a valuable tool for studying hypotheses about constraints or biases in development. By testing for the constraint at each step in a developmental sequence, it allows assessment of the developmental scope of the constraint.

Some constraints may operate across a wide developmental span, as the shape-over-color constraint seems to operate from 1 to 4 years. Others may function for one part of development and not for another, as suggested by the results of a study of the acquisition of knowledge about self and mother. Pipp, Fischer, and Jennings (1987) argued that biological constraints will produce two opposite biases for conceptions of features and agency of self and mother. The architecture of the body leads children to naturally understand their mothers’ features before their own, because their eyes and ears point outward toward people and objects in the world. But for actions, they tend naturally to understand their own actions before their mother’s actions, because they have more direct control over their own actions. Consequently, for tasks focusing on features of the person, a bias will operate in favor of the mother over the self. But for tasks focusing on agency (the control of action), a bias will operate in the opposite direction—self over mother.

To test this hypothesis, they constructed developmental scales for assessing infants’ understanding of features and agency of self and mother between 6 months and 3½ years. The features assessed included physical features, such as location of a sticker on various parts of the body, and social features, such as name and gender. Agency tasks included control of actions, social categories, and interactions in pretend play. Tasks at parallel steps for self and mother in the scales were as similar as possible in content and were matched for skill complexity.

The constraint hypothesis was supported, but only for a portion of the age span tested. Between 6 months and 2 years, babies understood their mothers’ features before their own, and they understood their own control of actions before their mothers’, as predicted. After 2 years of age, however, the bias disappeared for features and became much more complex for agency. Thus, the developmental scope of the bias seemed to be limited primarily to the first 2 years. The authors interpreted this change as arising from the emergence at 2 years of the capacity to represent people independently of the child’s own actions.

Age of Development of the Capacity to Classify

The method of analyzing developmental sequences in detail thus allows the detection of constraints at multiple steps in development. But it also places behaviors firmly in a developmental context and constrains explanations in an epigenetic direction. The use of multiple tasks in the sequence greatly lessens the temptation to make broad generalizations about capacity based on a single task. In developmental scales, the complexity and content of a task clearly affect whether a child will show a given “capacity.” For the classification sequence, it would be possible to argue from the data in Table 7.2 that children understand classification at a very early age (Gelman & Baillargeon, 1983): At 1 year of age, babies sort correctly for single categories of shape or color (Step 1), and at 2 years of age they sort correctly for a dimension of shape or color (Step 2).

As appealing as this argument is, it succumbs to the same problems as arguments about when infants “really” have the capacity for visually guided reaching. Is it legitimate to say, for example, that a child has classification capacity for a color dimension when he or she can sort into single uniform categories of color (Step 2) but cannot sort when there are simple variations in the shades of red, blue, and yellow (Step 3)? Even when a child can sort correctly for variations within a color category (Step 3), does he or she have the capacity for a color dimension when he or she cannot sort in the presence of variations in shape (Step 4)?

Such arguments are no more fruitful for classification than for visually guided reaching. A more fruitful approach is to describe the epigenetic landscape for classification by color and shape—including how the child’s skills gradually build from very simple sorting tasks to complex ones and how there is a bias in sorting for shape over color.

EVALUATING THE EFFECT OF ENVIRONMENTAL CONTEXT: DEVELOPMENTAL RANGE

Earlier, we emphasized the dramatic variability in an individual’s skills in relation to the degree of immediate contextual support provided by the environment, both physical and social. To adequately evaluate developmental hypotheses about the meaning of early behaviors and the role of biological constraints, the variability due to environmental context must be considered along with measures of sequence and synchrony.

Vygotsky was one of the first to provide a specific theory about the nature of this variability with his concept of the “zone of proximal development” (Rogoff, 1982; Vygotsky, 1978; Wertsch, 1985). Vygotsky (1978) defined the concept as “the distance between the actual developmental level as determined by individual problem solving and the level of potential development as determined through problem solving under adult guidance or in collaboration with more capable peers” (p. 86). By defining developmental level in terms of a zone, susceptible to social influence, Vygotsky challenged the view of cognitive capacity as a static competence isolated within an individual. He suggested instead that children’s cognitive skills are plastic and influenced upwardly or downwardly by the relative degree of environment support available.

In recent research based on skill theory, the concept has been further specified in the distinction between functional, optimal, and scaffolded cognitive levels (Fischer & Elmendorf, 1986; Lamborn & Fischer, 1988). Like Vygotskian theory, skill theory attributes cognitive abilities not to a person but to a person-in-a-context. A person’s skill or ability exists (and is assessed) not as a fixed capacity but over a range of levels similar to the zone of proximal development. Table 7.4 shows such a developmental range. The left column represents a sequence of known steps in a given task domain. The right column shows the range of performances of which an individual child may be capable, given differing environmental support conditions.

Within this developmental range, the optimal level for a given domain is defined as the highest level of behavior under conditions designed to evoke a person’s best performance—in a familiar context, with the opportunity to practice the skill, and with contextual support for high-level performance. Among the most effective forms of contextual support are modeling a high-level skill or prompting its key components (Rotenberg, 1988; Fischer & Bullock, in press).

The lower end of the developmental range for a given domain is defined by the spontaneous functional level, which is related to what Vygotsky referred to as independent problem solving. Here, there is no contextual support for high-level performance—no prompting or modeling. Children are left to do the task on their own, and typically their best performance is far below that under optimal conditions, even when they are highly motivated to perform well and have the opportunity to practice their response. Without contextual support they cannot demonstrate the skill in most domains.

TABLE 7.4
Developmental Range of Levels for a Given Individual Under Varying Environmental Support Conditions

Finally, although the optimal level is generally the highest performance level that a child can consistently produce on his or her own, an even higher scaffolded level can sometimes be achieved. In scaffolding, an adult or older child actually performs part of the task for the child, thus allowing him or her to participate in the task at a level beyond what the child can achieve without such intervention (Bruner, 1982; Wood, 1980).

The developmental range offers a way of characterizing variation in a skill due to context and constraining the meaning of apparently precocious behaviors. Evaluation of a person’s skills under functional, optimal, and scaffolded conditions for a given domain provides a clearer picture of what he or she can do than a single assessment that ignores contextual variation. As in the examples of visually guided reaching and classification, a behavior that appears at an early age also varies with the context in which it is assessed. Effects of variation due to contextual support—and many other aspects of context—should be included in neo-nativist assessments of seemingly precocious behaviors.

CONCLUSION: THE CONSTRUCTION OF COGNITIVE SKILLS

When existing theoretical predictions are closely examined, and when specific behaviors are considered in relation to appropriate methodological guidelines for interpreting developmental hypotheses, there is no compelling evidence in support of the view that knowledge is specified innately. At the same time, current research on infant cognition does provide evidence of rich sets of perceptual and behavioral skills early in life. This information promises to help specify the ways in which biological constraints on early perceptual and action systems influence the child’s development of knowledge.

In order to make meaningful generalizations about early behaviors, researchers need to constrain the scope of their inferences by locating individual behaviors within an epigenetic map. Just as the meaning of a street on a map is determined by the whole length of the street and its connections with other streets, the nature of a particular cognitive skill is specified by information about where it came from, what it is leading to, and what concurrent skills are connected with it.

Throughout this chapter, we have used the terms capacity and skill differentially. Arguments over the age of acquisition of a capacity are based on the notion that the child can be characterized as either having or not having some fixed capacity regardless of the context. The concept of skill recasts this notion in terms of gradually emerging abilities that are context-dependent. Skills are cognitive structures constructed in and for specific contexts. A bottle has a certain fixed capacity for holding water, but children do not have any such fixed capacities for visually guided reaching, classification, object permanence, or any other task. Instead, they have skills, which emerge first in a limited form in limited circumstances and are gradually generalized to more powerful forms applicable to broader circumstances (Fischer & Farrar, 1987; Fogel & Thelen, 1987).

An approach that explains development as the control and construction of skills avoids the many problems that arise from concepts of capacity (Fischer, 1980; Fischer & Bullock, 1984, in press). With a skill approach, it is obvious that any single behavior can be interpreted only within the framework of the epigenetic landscape for behaviors in that domain. A behavior may be discovered in isolation, but interpretation of its significance requires relating it to similar behaviors that develop before and after it as well as other ones that develop at about the same time. The manner in which sets of behaviors hang together and come apart provides central clues to the meaning of each behavior individually. Only by placing behaviors within such an epigenetic framework—by considering them as skills rather than capacities—can researchers build a full understanding of cognitive development.

ACKNOWLEDGMENTS

The work reported in this chapter was supported by grants from the Carnegie Corporation and Harvard University. The authors thank Daniel Bullock, Claes von Hofsten, Ralph Roberts, Jr., and Louise Silvern for their contributions to the arguments presented.

REFERENCES

Baillargeon, R. (1987). Object permanence in 3½– and 4½–month-old infants. Developmental Psychology, 23, 655–654.

Baillargeon, R., Spelke, E. S., & Wasserman, S. (1985). Object permanence in five-month-old infants. Cognition, 20, 191–208.

Bernstein, R. J. (1985). Beyond objectivism and relativism: Science, hermeneutics and praxis. Philadelphia: University of Pennsylvania Press.

Bertenthal, B. I., Haith, M. M., & Campos, J. J. (1983). The partial-lag design: A method for controlling spontaneous regression in the infant-control habituation paradigm. Infant Behavior and Development, 6, 331–338.

Bidell, T. R., & Fischer, K. W. (1989). Commentary (Durability and variability in cognitive development). Human Development, 32, 363–368.

Bornstein, M. H. (1985). Colour-name versus shape-name learning in young children. Journal of Child Language, 12, 387–393.

Bruner, J. S. (1982). The organization of action and the nature of adult-infant transaction. In M. Cranach & R. Harre (Eds.), The analysis of action. New York: Cambridge University Press.

Bruner, J. S., & Koslowski, B. (1972). Visually preadapted constituents of manipulatory action. Perception, 1, 3–14.

Campos, J. J. (in press). Locomotor experience: Domain specificity in its consequences for cognitive development. In R. Wozniak & K. W. Fischer (Eds.), Specific environments: Thinking in contexts. Hillsdale, NJ: Lawrence Erlbaum Associates.

Campos, J. J., & Bertenthal, B. I. (1987). Locomotion and psychological development in infancy. In F. Morrison, K. Lord, & D. Keating (Eds.), Advances in applied developmental psychology (Vol. 2, pp. 11–42). New York: Academic Press.

Case, R. (1985). Intellectual development: Birth to adulthood. New York: Academic Press.

Changeux, J.-P. (1985). Neuronal man: The biology of mind (L. Garey, Trans.). New York: Oxford University Press. (Original work published 1983)

Fischer, K. W. (1980). A theory of cognitive development: The control and construction of hierarchies of skills. Psychological Review, 87, 477–531.

Fischer, K. W., & Bullock, D. (1981). Patterns of data: Sequence, synchrony, and constraint in cognitive development. In K. W. Fischer (Ed.), Cognitive development. New Directions for Child Development: Iss. 12 (pp. 69–78). San Francisco: Jossey-Bass.

Fischer, K. W., & Bullock, D. (1984). Cognitive development in school-age children: Conclusions and new directions. In W. A. Collins (Ed.), The years from six to twelve: Cognitive development during middle childhood (pp. 70–146). Washington, DC: National Academy Press.

Fischer, K. W., & Bullock, D. (in press). The failure of competence: How context contributes directly to skill. In R. Wozniak & K. W. Fischer (Eds.), Specific environments: Thinking in contexts. Hillsdale, NJ: Lawrence Erlbaum Associates.

Fischer, K. W., & Canfield, R. L. (1986). The ambiguity of stage and structure in behavior: Person and environment in the development of psychological structures. In I. Levin (Ed.), Stage and structure: Reopening the debate (pp. 246–267). New York: Plenum.

Fischer, K. W., & Elmendorf, D. (1986). Becoming a different person: Transformations in personality and social behavior. In M. Perlmutter (Ed.), Minnesota symposium on child psychology (Vol. 18, pp. 137–178). Hillsdale, NJ: Lawrence Erlbaum Associates.

Fischer, K. W., & Farrar, M. J. (1987). Generalizations about generalization: How a theory of skill development explains both generality and specificity. International Journal of Psychology, 22, 643–677.

Fischer, K. W., Hand, H. H., & Russell, S. L. (1984). The development of abstractions in adolescence and adulthood. In M. Commons, F. A. Richards, & C. Armon (Eds.), Beyond formal operations (pp. 43–73). New York: Praeger.

Fischer, K. W., Hand, H. H., Watson, M. W., Van Parys, M., & Tucker, J. (1984). Putting the child into socialization: The development of social categories in preschool children. In L. Katz (Ed.), Current topics in early childhood education (Vol. 5, pp. 27–72). Norwood, NJ: Ablex.

Fischer, K. W., & Hogan, A. (1989). The big picture for infant development: Levels and sources of variation. In J. Lockman & N. Hazen (Eds.), Action in social context: Perspectives on early development (pp. 275–305). New York: Plenum.

Fischer, K. W., & Roberts, R. J., Jr. (1989). The development of classification skills in the preschool years: Developmental level and errors. Cognitive Development Laboratory Report, Harvard University, Cambridge, MA.

Fischer, K. W., & Silvern, L. (1985). Stages and individual differences in cognitive development. Annual Review of Psychology, 36, 613–648.

Flavell, J. (1982). On cognitive development. Child Development, 53, 1–10.

Fogel, A., & Thelen, E. (1987). Development of early expressive and communicative action: Reinterpreting the evidence from a dynamic systems perspective. Developmental Psychology, 23, 747–761.

Fraiberg, S. (1974). Blind infants and their mothers: An examination of the sign system. In M. Lewis & L. A. Rosenblum (Eds.), The effect of the infant on its caregiver (pp. 215–232). New York: Wiley.

Gelman, R. (1972). Logical capacity of very young children: Number invariance rules. Child Development, 43, 75–90.

Gelman, R. (1978). Cognitive development. Annual Review of Psychology, 29, 297–332.

Gelman, R., & Baillargeon, R. (1983). A review of some Piagetian concepts. In P. H. Mussen (Ed.), Handbook of child Psychology: Vol. 3. Cognitive development (pp. 167–230). New York: Wiley.

Gelman, R., & Gallistel, C. R. (1978). The child’s understanding of number. Cambridge, MA: Harvard University Press.

Gottlieb, G. (1983). The psychobiological approach to developmental issues. In P. H. Mussen (Ed.), Handbook of child psychology: Vol. 2. Infancy and developmental psychobiology (pp. 1–26). New York: Wiley.

Haith, M. M., Hazan, C., & Goodman, G. S. (1988). Expectation and anticipation of dynamic visual events by 3.5-month-old babies. Child Development, 59, 467–479.

Halford, G. S. (1989). Reflections on 25 years of Piagetian cognitive developmental psychology, 1963–1988. Human Development, 32, 325–357.

Hofsten, C. von (1984). Developmental changes in the organization of prereaching movements. Developmental Psychology, 20, 378–388.

Hofsten, C. von (1989). Transition mechanisms in sensorimotor development. In A. de Ribaupierre (Ed.), Transition mechanisms in child development (pp. 233–258). New York: Cambridge University Press.

Kellman, P., & Spelke, E. S. (1983). Perception of partly occluded objects in infancy. Cognitive Psychology, 15, 483–524.

Lamborn, S. D., & Fischer, K. W. (1988). Optimal and functional levels in cognitive development: The individual’s developmental range. Newsletter of the International Society for the Study of Behavioral Development, No. 2 (Serial No. 14), 1–4.

Landau, B., Smith, L. B., & Jones, S. S. (1988). The importance of shape in early lexical learning. Cognitive Development, 3, 285–298.

Landau, B. Spelke, E. S., & Gleitman, H. (1984). Spatial knowledge in a young blind child. Cognition, 16, 225–260.

Lewontin, R. C., Rose, S., & Kamin, L. J. (1984). Not in our genes: Biology ideology and human nature. New York: Pantheon.

Moore, D., Benenson, J., Reznick, J. S., Peters, M., & Kagan, J. (1987). The effect of numerical information on infants’ looking behavior: Contradictory evidence. Developmental Psychology, 23, 665–670.

Odom, R. D., Astor, E. C., & Cunningham, J. G. (1975). Effects of perceptual salience on the matrix performance of four- and six-year-old children. Child Development, 46, 758–762.

Piaget, J. (1941). Le mecanisme du developpement mental et les lois du groupement des operations. Archives de Psychologie, Geneve, 28, 215–285.

Piaget, J. (1952). The origins of intelligence in children (M. Cook, Trans.). New York: International Universities Press. (Original work published 1936)

Piaget, J. (1954). The construction of reality in the child (M. Cook, Trans.). New York: Basic Books. (Original work published 1937)

Piaget, J. (1969). The mechanisms of perception (G. N. Seagrim, Trans.). New York: Basic Books. (Original work published 1961)

Piaget, J., & Inhelder, B. (1967). The child’s conception of space (F. J. Langdon & J. L. Lunzer, Trans.). New York: Norton. (Original work published 1948)

Pipp, S. L., Fischer, K. W., & Jennings, S. L. (1987). The acquisition of self and mother knowledge in infancy. Developmental Psychology, 22, 86–96.

Rogoff, B. (1982). Integrating context and cognitive development. In M. E. Lamb & A. L. Brown (Eds.), Advances in developmental psychology (Vol. 2). Hillsdale NJ: Lawrence Erlbaum Associates.

Rollins, H., & Castle, K. (1973). Dimensional preference, pretraining, and attention in children’s concept identification. Child Development, 44, 363–366.

Rotenberg, E. J. (1988). The effects of development, self-instruction, and environmental structure on understanding social interactions. Unpublished doctoral dissertation, University of Denver.

Silverman, I. W., & Briga, J. (1981). By what process do young children solve small number conservation problems? Journal of Experimental Child Psychology, 32, 115–126.

Spelke, E. S. (1988). Where perceiving ends and thinking begins: The apprehension of objects in infancy. In A. Yonas (Ed.), The Minnesota symposia on child psychology: Vol. 20: Perceptual development in infants (pp. 191–234). Hillsdale, NJ: Lawrence Erlbaum Associates.

Thelen, E., & Fogel, A. (1989). Toward an action-based theory of infant development. In J. Lockman & N. Hazan (Eds.), Action in social context: Perspectives on early development (pp. 23–63). New York: Plenum.

Tomasello, M., & Farrar, M. J. (1984). Cognitive bases of lexical development: Object permanence and relational words. Journal of Child Language, 11, 477–493.

Uzgiris, I. C., & Hunt, J. Mc. V. (1975). Assessment in infancy: Ordinal scales of psychological development. Urbana, IL: University of Illinois Press.

Vygotsky, L. S. (1978). Mind in society: The development of higher psychological processes. Cambridge, MA: Harvard University Press.

Wertsch, J. V. (1985). Vygotsky and the social formation of mind. Cambridge, MA: Harvard University Press.

Wood, D. J. (1980). Teaching the young child: Some relationships between social interaction, language, and thought. In D. R. Olson (Ed.), The social foundations of language and thought (pp. 280–296). New York: Norton.