We may view the development of logical or formal thought as a truly revolutionary accomplishment. It is an accomplishment that not only separates the educated adult from the child; but, the accomplishment that separates the culture of modern man from the culture of primitive man. This two-fold perspective should always be kept in mind, for each developmental transition involves at once the ontogenic dimensions of the individual's growth and the "phylogenetic" dimension of social transmission of information across generations (Piaget, 1970; Piaget and Inhelder, 1969). From a phylogenetic perspective, formal thought may be viewed as the distinguishing characteristic of modern man: the characteristic that provides an unlimited range of possibilities with a concomitant source of unlimited potential danger. And as for the developmental aspect, as Hunt (1969) has noted, the formal operations of adolescence are such that "thought directs perception in investigation...toward reforming the world that exists." A careful investigation of what constitutes formal thought and of the factors which influence its rise and subsequent course is a topic of prime importance. This paper will draw attention to some of the factors that seem to be important in the development of inference abilities in the child.
At one time it was believed that small children were incapable of making certain types of logical inference that were easy for adults. It was also believed that small children were quite "egocentric" in their approach to social relations. Current opinion is that, while there certainly are differences, the gap between the thinking of children and adults is much less than we thought twenty years ago.
The reason for this change of view is that much attention has been given in recent years to the design of experimental situations with the goal of determining the earliest ages at which children show signs of using forms of deductive inference. This research was sparked by various claims made by Piaget and colleagues to the effect that children up to the age of 6 or 7 years were not capable of making adult-like inferences on certain kinds of logical tasks (Inhelder & Piaget, 1958). Subsequent research has made use of ingenious techniques for circumventing those aspects of experimental situations that investigators have felt to be especially problematic for the child.
As a result of this rather vigorous research effort we know something today about the kinds of situational variables that are important determinants of the young child's performance on problem solving tasks. We still lack, however, an adequate heuristic model and experimental method that will help us organize the findings of this research and guide future explorations. We have shown that children at younger and younger ages are able to engage in forms of inferential reasoning, but this research has left open the issue of tracing and explaining its development (Gelman, 1969). One of the problems seems to be technical - that as our subjects get younger, experimental techniques that will display the child's inferential abilities are more difficult to design. We have rapidly come up against an "experimental wall". An important thing shown by this research is that we can never be sure we've reached the bottom age for a particular ability. We have, however, made some progress in clarifying a number of possible situational and developmental constraints on inferential ability.
Behind this technical difficulty, however, there lies a conceptual problem. Within Piaget's model development always proceeds along the same unilineal track and children differ only in the speed at which their formal logical operations developed, not in what develops. For this reason investigators have been led to narrowly focus research questions on the issue of the earliest age at which forms of logical inference appear. We have been slow in coming to question Piaget's idea that development is best characterized as increasingly complex systems of formal operations.
A larger problem then is to be found in our lack of attention to re-considering the nature of reasoning and the developmental conditions that give rise to it in light of these research findings. We are still too tied to Piaget's penchant for mathematical formalizations and also, perhaps, overly influenced by structural linguistic's emphasis on formal rules. Under the influence of Piaget's cognitivism, development has become defined as the development of formal logico-mathematical models, far removed from actual processes of social/environment/person interchange, the analysis of which should be central to any perspective which seeks consistency with notions of developmental and evolutionary adaptation. We forget, as academically oriented adults, that logic and language as formal systems are conceptions that require a great amount of education in order to acquire, and we are on dangerous ground when we assume that it is such systems themselves that develop. The use of formal logic in problem solving and an appreciation of language as a formal system are skills taught and highly valued by western civilization and are a novelty in human evolution; there is nothing "natural" about them. They do not arise "spontaneously" and they are not acquired without a great deal of guidance and some difficulty, if we consider the amount of resistance many of our children display towards education. Perhaps we might go so far as to say that these skills are thoroughly alien to the human mind (Donaldson, 1978). At the very least it is necessary to emphasize that our rarified conceptions of thought processes may not reflect the reality of how the mind of the child (or adults, for that matter) operates. The role of formal logic in human behavior has not been clearly articulated, and it undoubtedly plays a lesser role at the unconscious level than we commonly suppose (Dreyfus, 1979).
An important consequence of using our adult-developed notions of formal thinking is that the development of logical skills is treated as an individually located and isolated "competence". The stated goal of all cognitive research is to isolate and identify this competence as a set of skills contained within the child. Consequently, all task/situational characteristics are treated as confounding factors to be controlled in achieving an understanding of this competence. For this reason their significance has not been truly explored.
Elements of a Heuristic Model
An interesting heuristic model for the discussion to follow has been taken from an article by B. Webb in Scientific American. It has lessons for the field of artificial intelligence, which attempts to model the behavior of organisms.
The author's goal was to construct a robot that would model the behavior of a female cricket in moving toward the sound made by a possible mate. This is a quite robust behavior in the female cricket.
The author points out the difficulties with modeling even the most apparently simple, automatic behaviors of organisms. The interaction of sensors and actuators with the environment is always complicated and requires complicated sets of algorithms to achieve the desired result, if it can be achieved at all (see also Dreyfus (1992)for a discussion of this problem). The approach usually taken is to try to handle complex environmental circumstances with equally sophisticated algorithms. In other words, the environment is considered as something to be overcome by a central processing unit which calculates the most appropriate response to an infinite number of possible situations.Her goal is stated as follows:
"...to design the robot so that its interaction with the environment is exploited rather than resisted. For example, instead of attempting to force the robot to travel a straight- line course, it could be programmed to follow contours of the terrain that lead to its destination - circumventing rather than conquering hills in its path. Through this type of approach, what seems like complex behavior in a robot can come from a surprisingly uncomplicated control algorithm."
The approach to the problem consisted of three aspects:
The amount of "information processing" required by the robot was minimal since it's morphology only allowed it to respond to just any stimulus. The morphology was the first phase in information processing. It could not respond to the wrong cricket song.
"The evidence from the robot suggests that situational factors, rather than additional neural processing mechanisms, may explain the effects of chirp structure on the female's movement toward a potential mate."
"Because I programmed the robot, I knew it was not capable of distinguishing or deciding between the sounds. Yet again it appears that it is the interaction of the robot's uncomplicated mechanisms with particular sound fields that produces this interesting - and useful - behavior."
"More generally, it shows that a rather competent and complex performance can come from a simple control mechanism, provided it interacts in the right way with its environment."
This research suggests that artificial intelligence experts, and cognitive psychologists, may be attempting to place too much central control in the "brain". Instead, the boundaries between organism and environment are blurred, and control is more evenly distributed between the two. In this case the investigator has been able to specify the relevant characteristics of a particular environmental stimulus and relevant characteristics of the organism's morphology. The "code" necessary for programming the microchip we may consider represents the unknown factors of the cricket's neurophysiology, the "black box." It is the stuff we fill in when part of the system is unobservable, and in this case it takes the form it does because of the requirements of constructing a functioning machine in contrast to the organism. Most importantly, by looking at the code itself I am sure that we could not understand how the machine does its thing. The behavior of the machine is not to be found in any simple location - not in the machine, simply because we constructed it, nor in the nature of the environmental stimulus.
An adequate model must therefore focus on the integrative, interactive relation between the child and the context. This implies adopting a view that behavior is an output of multiple factors, these factors being mutually dependent as effective causes. That is, both organismic and contextual factors are given equal theoretical status. Thus, there is no intelligible meaning to "ability" as an individually localized and isolated characteristic, and no single situation can give an adequate picture of the child's developmental state. An appropriate analysis of competence must examine the child's behavior across a variety of situations in which the contextual constraints are systematically varied and considered as a part of the possibility for performance itself.
As an alternative to identifying development with formal operations, a naturalistic perspective may be adopted that postulates the continuity of lower (less complex) and the higher (more complex) activities and forms. The problem to be investigated is the transformation of organic behavior to intellectual behavior marked by logical properties. Inferential activities of a covert, "mental" sort could not exist without prior overt existential activities - arranging the constraints of situations in such a manner that a solution is evident. To understand the development of intellectual behavior, it will be necessary to consider it within the broader development of the child and consider the conditions which must be satisfied in order for particular types of inference to be made. (Donaldson, 1971). Such preconditions may involve for example the development of motor restraint and the capacity for sustained activity guided by a plan or strategy, the ability to use language directives in guiding behavior, a sense of when one does not have sufficient information for the resolution of a problem, the internalization of speech, the capacity for sustained activity, a sense of language as a formal system that may be explored as such, and so on.
A naturalistic perspective also implies a functional orientation. Inferential activities are here assumed to function in the control of exploration or inquiry, such as to lead to warrantable assertions. This means that to be understood, inferential activities must be viewed in relation to the context in which they represent a solution to a particular set of conditions. By "warrantable assertions" is meant that inferences are subject to social constraints, that is, they have a social significance. Adequate inferences must be warrantable as such in a social context. Not just any inferential judgment will do, as will be demonstrated below.
Following these regulative principles, inferential processes may be viewed as arising in actual momentary interpersonal processes and manipulation of materials. The possibility of inference is the outcome of, and therefore expresses, a relationship between ourselves and the instrumentalities used to achieve an understanding of what particular situations are all about. The relationship between task factors and the use of logical inference is more than simply confounded - logical inference is not possible and cannot be understood without analyzing situational factors as constraints on performance. Logic is not adequately characterized as something carried around by the child, to be applied here and there as conditions permit, but as locally produced each time and in every situation as a result of the child's particular coupling achieved with the task and social constraints. The guiding notion of this paper is that the ability to make any logical inference depends on the situatedness of activity. It depends on achieving an appropriate relationship with the materials at hand as well as with other people with whom the task is shared. To paraphrase John Dewey, the characteristic human experience is one of doing acts, dividing up, cutting up, placing, arranging, marking off, extending, piecing together, joining, assembling and mixing - in general, selecting and adjusting things as means for reaching consequences. An understanding of inferential behavior will require the study of how control over activity emerges during development, involving the changing character of information and strategies for its creation and use.
The notions of control and information are inextricably linked. Information changes in form and function as new developmental structures arise, that is, the characteristic functioning of the child is what it is because of the special way in which situational features enter into its activities. Information has a developmental history. In our testing situations, how the child accesses information from the spatial arrangement and perceptual prominence of physical materials, as well as from the activities and speech of the experimenter become questions of central concern. The general issue is how children come to control their behavior (or how it comes to be controlled) in increasingly complex ways, and how information sources for control become available at what developmental age. Our attention must turn to how the child is able to use situational constraints for the resolution of a task, and how the child is able to handle situations in which the situational constraints are at odds with the formal task.
In this paper the question of information availability will be discussed as an issue of attentional salience of situational features. How the child views particular aspects of experimental situations as relevant or irrelevant to the formal task is crucial to achieving an appropriate inference. Situational salience depends on the child's developmental state as well as on circumstantial constraints such as the spatial arrangement of materials and the ordering activities of the adult. From this point of view, a major task for the child and the experimenter is to achieve a shared way of partitioning the situation into relevant and irrelevant features such that a mutually understandable resolution is realized. Thus logic development is to be linked to the more general issue of behavioral coordination and the reading of intentions.
Some of the research that has been done in the areas of conservation and perspective-taking will first be reviewed with an eye toward noting the factors used by experimenters to elicit correct performance by children on these tasks. These manipulations may be interpreted as the provision of contextual support for inferential activity and may be classed as either physical or social in character. It will be suggested that one of the primary tasks of the child in the experimental situation (of both participants, really) is that of coordinating intentions, of reaching a mutual understanding of what is going on. This coordination of intentions is required for making appropriate inferences, and hence the child's ability to read the other's state may be considered a prior skill. Directions for future research will be discussed throughout.
Manipulating Task Characteristics: A Paradigmatic Case.
With the publicity of Piaget's work in the United States an impetus was provided toward a cognitive analysis of the child's behavior. His perspective was that the child's understanding of the physical and social worlds, and consequently the child's ability to move freely in both of those worlds, depended on the development of logico-mathematical operations that were applied to both domains. His research seemed to show that the young child's thinking abilities were quite different from those of adults, and many of his experiments have become paradigmatic in this field.
Criticism was rapid, however. Many investigators felt that the subjects in Piaget's experiments were being asked to perform in situations that were far removed from their everyday life. An argument was developed to the effect that children were more capable than their performance in Piaget's experimental situations indicated, and that this could be demonstrated by simplifying the experimental tasks in various ways while maintaining their formal demands.
As a case in point, we may consider Helen Borke's (1975) modification of Piaget's mountain experiment. This study contains many of the types of manipulated elements that have been found to be successful in other areas of research on cognitive development.
In the now classic experiment designed to demonstrate childhood egocentrism, Piaget and Inhelder (1956) showed a table-top scene of three mountains to children, each mountain being of a different size and marked with different objects, such as a cross or a house. A doll was positioned at the various sides of the table and the children were asked to indicate what the doll would see from its particular position. To indicate their response the children were to choose one of a series of pictures or small replicas representing views of the scene from different sides. Children between 4 and 6 years of age were most likely to choose a scene that replicated their own view of the scene, and this was interpreted as an egocentric response, meaning that the children were not able to distinguish the doll's perspective from their own. It is not until ages 9 or 10 that children can accurately choose the correct perspective under these conditions.
Borke found that simplifying the stimulus and response aspects of the mountain task made problem solution easier for young children. She used simpler, more familiar objects and had children rotate the table instead of choosing a picture or replica as Piaget and Inhelder did. She designed four displays. The first was a single fire-engine used for a pre-test familiarization with the task. Display 2 consisted of a lake with a sailboat, a house, and miniature animals, the third was a replication of Piaget's mountains, and display 4 contained eight different groupings of people or animals in natural settings. In addition, she provided a story-like context by having a familiar television character, Grover from the program Sesame-Street, drive around the display. The children had to indicate Grover's view by turning a model display so they could see the same thing as Grover. Furthermore the children practiced with help on the fire-engine display until it was clear that the task was understood. Grover parked at each side of the various displays but sequence was randomly varied. Children between the ages of 3 and 4 did considerably better than the children in Piaget's experiment, even on the mountain replica. Forty two percent of 3 year olds and 67% of 4 year olds were successful on the mountain task; 80% of the combined group of 3 and 4 year olds were correct on display 2; 79% of three year olds and 93% of 4 year olds correct on display 4. Of the total number of errors on all three displays, almost one third or 31% were "egocentric" and slightly over two thirds or 69% were random. Children thus found the familiar object displays easier, with no apparent effect for quantitative complexity of the arrays. The form of response also made a difference - fewer children made incorrect responses on the mountain display than found by Piaget. Varying task complexity both in terms of stimulus and response demands thus affected the child's ability to perform as required on the task.
This experiment controlled for complexity of the stimulus arrays and the comparison with Piaget's findings made it possible to determine the effect of response demands. Two of the other changes made to the traditional task were not examined for their effects. First, the fictional context that was provided for the child. The Sesame Street character was well-known to the children and his imaginary activity of driving around the scene stopping to look out the window makes human sense. Second, the specific effect of the practice trials with the simplest display (the fire engine), involving the experimenter's attempts to achieve a mutual understanding of what was required, was also not examined.
The experiment thus varied a number of both physical and social aspects of the experimental situation. It is considered to be a paradigmatic case because a great number of studies have made similar "age appropriate" changes in the original Piagetian experimental paradigms with successful results. Investigators generally proceed intuitively when designing age appropriate tasks, and an analysis of what is meant by age appropriateness has been slow in coming about. The changes center around aspects that make it easier for the child to understand what is required - at times this involves using familiar or simpler material, while other studies change the problem to a formally simpler one (today it is generally agreed that Piaget's mountain task, as well as Borke's variation, both ask the child to indicate how something is seen to another, which is considerably more difficult than indicating what is seen by another person). In this way, we have lowered the ages at which children appear to have various cognitive skills, to the point where the central issue no longer seems to be that of logical development but of how the child's performance changes with these manipulations. We can now clearly state that there is more going on in the experimental situation than a simple testing for a logical skill. There is more to a child's behavior here than the ability to use logic, and the use of logic itself depends on more than being in some sense developmentally capable of using it. The child reacts to, or makes use of, features of the task situation that a purely formal analysis would deem unimportant or irrelevant.
With the apparent age at which a child can use logical skills declining, we must now face the broader problem of how the child makes use of the various features of each situation and for what ends. "Age appropriate" must be turned into the investigation of concrete social and structural cues that a child is able to manipulate at various stages of development.
The interesting effects on performance due to manipulations of task factors has been known for some time. Kohler's (1925) early work on insight learning with chimps showed that the physical placement of materials was very important in encouraging chimps to make the relations necessary for the solution of problems.
An early study of deductive reasoning in children carried out by the Kendler's (1967) from a Hullian perspective appeared to show that children could not reason deductively. In this experiment a rather "cold" machine was used of the style often used in experimental learning studies. Children were trained to perform two separate tasks, first to get a marble by pressing a button, and using a marble handed to them to get a toy by putting the marble in a hole. After learning these two tasks, the children were asked to get the toy by themselves, meaning they had to combine the two behaviors they had learned previously. It was found that they had great difficulty in first getting the marble by pressing the button and then putting the marble in the hole to get the toy. They could not reason deductively and put the two acts together in order to achieve the appropriate solution.
Interestingly, Cole (1971) found that African adults could not do the task either. However, when he changed the task to one that was more familiar and meaningful (using matchboxes with keys) his subjects had few problems. Formally the task was the same, only the materials were different. Hewson (1974) also changed Kendler's machine into something less forbidding and found a tremendous jump in children's success rate. He replaced the button with drawers the child could open and shut, and took away the mystery of the task by showing the child how a marble could open a door. The success rates more than doubled for 4 year olds and tripled for 5 year olds.
These studies show clearly the effects of using familiar task materials - the same formal problems may be solved with the familiar set of materials but not with others. Why this should be is a crucial question.
It may certainly be the case that the forbidding apparatus' often used by investigators creates uncertainty and consequent tension in subjects. More developmentally, familiar materials already have a certain certitude about them. In theory, a "familiar" object is surely one that has a functional significance for the person. It is an object that has been explored, not perhaps for its "objective" properties as Piaget proposed, but for its possible relationship to the completion of other environmental structures and/or its function in structuring social interactions. In this process of fitting or familiarizing objects we might suggest that objects achieve particular characteristics for the individual. Certain properties become salient, i.e. relevant to particular activities, uses, or relationships. It is reasonable to assume that a person's tendency to "see" particular properties and relationships will carry over to new situations to be used in the resolution of new problematic relations. And rather than being a completely individualized activity, to "familiarize" oneself with an object is to at the same time socialize it, to see it and use it in socially shared ways in which "purpose" is integrated with "properties". It then has a place, it is fit into the world. It is also reasonable to assume that the functional significance of objects changes developmentally, and that social interaction in crucial to this process. On his own the child cannot comprehend the social purpose of objects nor master their appropriate use, for these qualities of objects are not directly perceivable. The familiarization of a "cup" involves using it in a certain way, shaping the hand to fit it, and thus bears on the issue of cultural transmission (Zaporozhets & Elkonin, 1971). By means of personal and social exploration of an object it achieves a particular "salience" and relevance in relation to activities. "Familiarity" thus involves adopting at least partially shared ways of using the objects.
This would suggest that the social/psychological sense of a situation may be more important for the young child that the formal sense of the task. We as academically oriented adults link experimental situations according to their formal characteristics and assume that this is the central issue, forgetting that this is but one of many possible ways to see relationships between tasks.
An interesting hypothesis has emerged in the past few years that directly bears on this issue. An increasing number of investigators are suggesting that an important factor in experimental situations is how the child sees certain factors as relevant or irrelevant to the task at hand (Donaldson, 1982; Gelman, 1978; Botvin & Murray, 1975; Bryant, 1974;). These investigators have come to the conclusion that the child does not always differentiate between relevant and irrelevant factors in the same way as the adult investigator when participating in experiments. This has come out of some interesting research on conservation.
Relevance and Irrelevance
In Piaget's conservation tasks the child is required to judge whether or not two quantities (weight, volume, mass, number) are equal before and after a transformation of form. First the child is to agree with the experimenter that the two quantities are equal, the child then witnesses the transformation, and the experimenter asks the child if any change in quantity has occurred. Piaget's findings suggested that the child under 6-7 years of age is incapable of conserving equality of quantity across the transformation, largely because the children are dominated and led astray by the appearance of the quantities. The perceptual characteristics take precedent until the time when the child acquires the notion of reversibility, i.e. "You've only poured it into a taller glass and you could pour it back again so it's still the same amount".
But consider the following explanation applied to a conservation of number of problem:
"You've only added two and you could take them away again, so it's still the same number".
This is an example of logical reversibility that is parallel to the length transformation, but clearly it doesn't work, as pointed out by several authors (Wallach, 1969; Schultz et. al., 1979; Donaldson, 1982). The difference between the two problems is that in one case the action which is reversed is irrelevant to the question being considered, while in the other case it is not. The act of adding is relevant, the act of moving apart is irrelevant. What is fundamental here is the understanding of relevance. The ability to logically reverse the situation is important too, but it is not sufficient to account for the child's performance. The use of logic to reach appropriate conclusions is linked to the determination of relevant from irrelevant aspects of the situation as a necessary precondition.
Bryant(1974) made a similar proposal; given that children in certain situations seem to be capable of conserving number, the problem with conservation experiments is not that of the child being incapable of quantitative invariance but of the child using one of two conflicting rules for answering the questions. His work focussed on number invariance. The paradigm for this type of study is to present two equal rows of counters to the child, assure that the child will say they are numerically equal, spread out one of the rows so it is longer but still contains the same number of counters and then again ask the child which has more. Children between 3 and 5 will generally say the longer row has more. Piaget interpreted this to mean that the child confuses length with numerosity and for that reason cannot conserve number.
Bryant suggested that the child has two rules for answering the investigators question. The first rule is a one-to-one correspondence rule, which draws the child's attention to whether each counter in one row is opposite a counter in another row. The second rule is a length rule, which says that the longer of two rows is more numerous. The child has both rules available and may use one or the other. In this light the traditional conservation task contains an ambiguity. With the first array, in which the rows are equal, the child may apply either rule and reach the conclusion that they are the same. When the length of the rows is changed in the second part of the experiment, the child can no longer apply the one-one correspondence rule and so applies the length rule. Thus the child's responses are consistently in favor of the length rule. Moreover, a condition in which children could not use either rule (a "chance" condition) was created and indeed the children did not respond in any consistent manner. A series of studies designed to control for this conflict (by changing the perceptual structure of the task problems) showed that children could indeed conserve number under appropriate conditions.
Bryant concludes that the child's problem lies in not having adequate information for distinguishing relevant from irrelevant cues. In the original conservation task, the child has no systematic way of resolving the conflict of which rule to use and which not. Thus, the structure of the task materials make it easy or difficult for the child to access the information leading to an appropriate response.
It must be pointed out that this is a highly constrained situation in which the child is encouraged to focus on a task array with a limited set of features. It would be inappropriate to conclude that such rules would be dominant in other task situations, or in other words that the child is limited to them.
Gelman's(1969) work on attentional training is consistent with this hypothesis. In this study a series of displays were presented to the child, consisting of three rows of counters, two of which varied in length but not number, and a third varying in number and length. The children were asked which was the odd one out, and given feedback in the form of indicating an incorrect or correct answer. In this way the children were taught to attend to the number and not the length cue. Children who had this training were more successful at conserving than children who did not have this training. They were learning, in Bryant's terms, what rule was relevant to the problem, in other words, not to respond to the irrelevant cue of length when judging number.
There are a number of other studies that may be interpreted in the same way. Miller & Heller(1976) and Miller & West(1976) made efforts to increase the salience of one-one correspondence and achieved better conservation, on small sets of numbers. A similar result was found with length conservation in a study by Silverman, Vanderhorst & Eull(1976). Botvin & Murray(1975) have concluded from their own work that the first step in learning conservation is to first recognize the role of relevant (addition-subtraction) and irrelevant transformations (displacement), and this conclusion is shared by Gelman(1978).
This work suggests that what appeared to Piaget as a lack of an ability may be more constructively viewed as a lack of information regarding what is relevant to the adult in the situation. Furthermore, we must be careful about what information we presume is lacking, because children generally seem to presume they know what the correct answer is. At bottom we are dealing with the child partitioning the features of the situation, i.e. creating information, in a way that is inconsistent with the way the adult experimenter does.
It appears then that children may be responding to factors in the experimental situation of which the adult is not aware. Most experiments designed to test the child's logical skills may be expected to involve a number of other skills in reading situations - in fact we might suggest that designing a task to test purely for logical abilities would be extremely difficult, or at the very least that the conservation task is not the best one for such purposes (Donaldson, 1982).
Salience and Understanding
It appears then that the child's understanding of the task depends on what is salient for them and that the child uses cues from the structure of the task itself to interpret the experimenter's purpose and question. Going further with this idea, both Gelman(1978) and Donaldson(1974) have advanced the hypothesis that failure on a task very often results from the child's misunderstanding of what the experimenter is up to. Gelman presents a number of studies in various fields in which the authors conclude that such a misunderstanding occurred, after having investigated the causes for failure by changing task requirements (Paris & Lindauer, 1976; Girgus et. al., 1975; Estes, 1976; Bullock & Gelman, 1977).
Donaldson(1978) reports a couple of interesting studies on this theme, apparently showing how the salience of various task features is influenced by the arrangement of the task materials and that this salience influences the child's interpretation of the experimenter's question. In one study, children are required to tell a "talking" toy panda if it's statements are correct or incorrect. They were shown four garages, connected together in a row, and some toy cars in the garages. Sometimes there were three cars, in which case one garage was empty, and sometimes five, in which case one car was outside. The panda's statements included:
All the cars are in the garages.
All the garages have cars in them.
When there were a total of three cars, they were all placed inside the garages so the first statement was true, and the second one false. When there were five cars, the truth value of the statements was reversed.
Some of the children, however, judged both statements to be false when there were only three cars, and judged both statements to be true when there were five cars. After ruling out the possibility that the children did not understand the meaning of "all" (tested in another situation by asking them if "all the garage doors were shut"), Donaldson reaches the conclusion that the children were interpreting "all the cars" to mean "all the cars which ought to be there". She notes a similar meaning to a common household question, "Have you put all the knives and forks on the table?", which doesn't mean all the silverware that exists, but the silverware that belongs on the table. At any rate, it seems as if the children were actually judging the statement "All the garages are full", as suggested by children's comments on why the panda was wrong. She concludes, "Watching the children and listening to them, one had the powerful impression that the empty garage was somehow salient for them, and that they interpreted everything they heard in ways affected by this salience" (p. 67).
The second study she reports is particularly interesting because it is more similar to a conservation experiment, in that the child gives one answer to a question, a transformation that is irrelevant to the meaning of the words is made, and the child then gives a different answer. In this study, two shelves were used, one above the other. Each shelf had two rows of cars, one with five and the other four. Children were asked which row had more cars in it, and this didn't present any difficulties for them. Then a row of garages was placed over each row of cars. The row of four cars was enclosed by a row of four garages, and the row of five cars was covered by a row of six garages, thus leaving one garage empty. The question of which row had more cars in it was repeated, and nearly a third of the children changed their judgments, saying that the shelf of four cars had more than the shelf of five. Again, for some children the empty garage had a salience that influenced their interpretation of the question. We may note the similarities to Bryant's findings in that in the number invariance tasks one factor, length, is salient to children. On the standard number conservation task, the children appear to be confusing number with length. Piaget interpreted this as a lack of differentiation between number and length. In Donaldson's task however, both the number and length of the rows are unaltered and the children still change their answer.
One might argue that the act of repeating the question may in and of itself have led children to change their response (Rose and Blank, 1974). It is not likely however that the children change their response simply because they think the experimenter wants them to. Repeating the question after the transformation leads them to look for new information regarding what is relevant in the situation and a number of them look to the empty garage. Their attention to the empty garage, in other words, may be overdetermined and not simply due to either the question "in itself" (which may function as an instigation to search for further information) or the addition of the garages. Given that children seemed attracted in the previous study to the emptiness of a garage it may be that here they are too. In any case, we may be justified in concluding that children arrive at an interpretation of the question based on the placement of the task materials.
All of these studies considering relevance or salience, and particularly Bryant's and Donaldson's hypotheses, are interesting because they suggest that the most crucial issue for problem solving in many experimental situations may not be the use of logic in and of itself but an activity of attentional partitioning of the situation such that certain features or relationships stand out. The experimental issue thus changes from a cognitive deficit hypothesis of what logical forms the child is capable or not capable of using, to one of how the world comes to be salient for participants. The problem is that the adult and child do not share the same set of relevancies, and consequently the same expectations or understanding as to what is happening. It is increasingly clear that the use of logic in a "correct" way (or at all?) depends on achieving shared ways of organizing experience, or ways of determining the relevant from the irrelevant. A crucial question for the study of intelligence may be, then, how do various features of the world become salient? One might suggest that the issue of saliency is basic to understanding the child's correct use of logic in experimental situations. Hence we now have the basis for the suggestion made at the beginning, that the experimental situation presents a problem of coordinating behavior and understanding, both for the investigator designing and presenting the task and for the child who's performance is the focus.
What factors determine what features in any given situation the child will attend to? The child's expectations apparently depend in part on features of the situation that are not immediately related to the adult's intentions as expressed through moves or language. When the task features are changed, the child's interpretation may also change. Many studies have shown how children can be taught to attend to the features that are relevant from the experimenter's point of view. Bryant(1974), for example, explores some of the conditions in which salience changes as a result of physical changes in the task situation. Whether or not we may speak of a consistent set of differential weightings given to features of the task at particular ages is not yet clear, and may be a good question for further research. But given the inherent uniqueness of every situation it is not likely that an account in terms of learned rules will be very useful as an explanatory mechanism. No set of rules could possibly foresee all eventualities. We may be able to speak of general tendencies, but the factors determining the salience of features in any given situation probably depend on an entire range of attentional and other ordering resources available to the child for making sense of what's happening in conjunction with the features of the situation itself. It may be more fruitful to view the "rules" the child uses as being created then and there, in the situation in relation to the particular material and social constraints at hand. We might grant that under the highly constrained circumstances of Bryant's task that the child's attention seems attracted to both length and number - the choices as well as which one is dominant would seem to depend on various features of the physical and social setting.
An important difference between adults and children that becomes apparent in experimental situations is that children often go "beyond" what is given. This makes sense when we note that the usual experimental situation is one in which the child lacks information and is blocked from the usual means for attaining it (asking for help, manipulating the materials, looking to see). He or she is required to suspend for the time being whatever else is known about similar situations and how to find a solution. Often, little by way of an explanation is presented to the child regarding the nature of the task or what is required. Therefore it is not surprising that children will make assumptions not directly related to the "task at hand" when they have insufficient resources for garnering information from what is given.
The differential salience of task features can thus lead to different understandings of the task by adult and child. But the child has access to other information in the situation that hasn't been directly addressed - the moves and speech of the adult. Indeed, the fact that the child's non-conserving response in the second part of the traditional conservation task is at least in part due to the fact that the experimenter asks the same question twice is an example of situational social constraints (Rose & Blank, 1974). If it is true that social factors in the experimental situation are important determinants of the child's performance, we must next look at the child's ability to read the intentions of the adult from the adult's verbal and behavioral behaviors.
III. Social Factors in Cognitive Performance: Access to the Social Context and the Achievement of Understanding.
We've seen in the previous section that in the child's negotiations with the experimental task that the salience of features and relationships is basic for arriving at a shared reading of the situation and for using logic appropriately. It is being assumed that the child is an active organizer looking for cues that will be relevant for achieving an understanding of what's what and will at times arrive at their own expectations based on these characteristics more or less alone. We must not assume, however, that the properties to which the child attends are wholly objective. It is by means of the links made by adults via their focal attention that the world takes on salience for the child and this is the means by which the child learns to organize his activities in relation to that of others and the task at hand. This occurs situationally throughout development. The child's ability to visually (an manually) organize critical features of the task situation is a function of the degree to which the experimenter has made visible his-her own visual organization of the situation through the visual linking of the child and critical features (Wertsch, 1980) This involves bodily posturing on all levels - from gross postural orientation to gestures such as hands, eye movements and speech.
In recent years we have come to have a greater appreciation of the child's sensitivity to and abilities in the social sphere. Next we look at some of the research in cognitive development with a view toward understanding the importance of the social element in the experimental situation.
Piaget's original hypothesis of childhood egocentrism held that children between 3 and 5 years old should hardly be aware that others have a perspective that is different than their own, and certainly not be able to coordinate viewpoints with another. Today we know that this isn't the case, that children are quite sensitive to the purposes of the investigator in experimental situations and that they construct readings of the investigator's intentions, with or without being able to verbalize these. Our experimental situations often ask children to perform tasks that are too complicated for them. Behaviorally, their sensitivity to the intentions of others is evident in their ability to adjust their behavior in accordance with the desires of adults and, in fact, in their ability at recruiting adults to adapt to their own.
For present purposes we may identify two types of studies that are particularly relevant to this issue. In the first, the task is presented in terms of motives and intentions on the part of the experimenter or fictional characters that are familiar to the child. In the second, the more general issue of framing the task is studied. Attention to both of these factors points towards new directions to take in the study of the resources used by children to achieve logical conclusions.
A perspective-taking task created by Hughes(1975) demonstrates that the use of familiar motives and intentions aids the child in responding in a way the adult experimenter considers appropriate. In this study, a "wall" in the form of a cross was constructed on a table-top and two small dolls were used, one representing a policeman and the other a boy. The policeman was placed at the end of one of the arms of the cross, so it appeared as if he had visual access to two of the four possible areas demarcated by the wall. Initially, the researcher placed the boy doll in each of the areas in turn and asked the child if the policeman could see him. Afterwards, the policeman was placed on the opposite side, and the child was asked to place the doll in a position where the policeman could not see him. If the child made any mistakes the investigator corrected him. Then the actual testing began by placing two policemen at adjacent corners and asking the child to hide the doll in the unique area that wa visually inaccessible to both policemen. The youngest children in the experiment, with an average age of 3.9 years, achieved a success rate of 88% on this task.
This type of research has led many investigators to believe that the child is not as egocentric as was once thought, at least in Piaget's sense.
We must go further though and ask the following question: if the study by Hughes showed that children as young as 3.9 years could solve the task, why was Piaget's mountain task so difficult? First it should be noted that the Hughes task only required that the child determine what the policemen saw, not how they saw the scene. This would surely make the task easier than the mountain task, in which the child had to indicate the actual visual perspective of the other. However, it cannot explain why children in the mountain experiment tend to choose their own point of view. If we accept the idea that the responses of children at age 6 do not demonstrate a basic egocentrism, then how do we explain this preference? If the problem is simply one of lacking information, then more of the children's "errors" should have been random, rather than "egocentric" (in the mountain experiment there were at least four possible choices). In the previous description of Donaldson's garage study it was suggested that children were interpreting the task according to the salience of an empty garage. An alternative hypothesis for the results of the mountain experiment then is that the children do not understand what they are supposed to do from the experimenter's point of view, and furthermore that they arrive at their own understanding of the situation that is different from the experimenter's but which leads at least a large group of them to choose their own visual perspective (Donaldson, 1978). What this interpretation might be is not clear in this case, unless it is that the child assumes the experimenter is asking for his or her own visual point of view.
One new element generally added by the investigators who have modified the mountain task (e.g. Borke, 1975; Hughes, 1975) is that of presenting the task in terms the child could understand - that is, in terms of familiar human motives and intentions. As part of her successful modification of Piaget's mountain experiment, we saw that Borke simplified the stimulus array and response requirements, which was the focus of her study. But she also provided an understandable story context for the task by using a familiar Sesame Street character - Grover - and set him about a familiar activity - driving his car around the table and stopping to look at the scenery. Not only that, but she began each child with a set of practice trials to make sure the child understood what was expected in the situation. It must be assumed that all the factors taken together - a simplified array, simplified response, fictional context, and training - contributed to the increase in perspective-taking responses, though the respective weight of these factors was not separated.
Similarly, the Hughes situation makes sense to children. They may be able to generalize from their experience of hiding from someone, they know what it means to be naughty (the motive to escape and the motive to pursue and capture). The motives and intentions of the other are entirely comprehensible. This is not the same as a literal understanding of what another person sees from a given standpoint, but with what he is feeling or planning to do.
This same sensitivity may be found in other studies that have been successful in the area of testing for egocentrism. Flavell's group has used a format very familiar to the child - that of showing something to another person (Masangkay et.al., 1974; Lempers et.al., 1977). Mothers and their children conduct this sort of business very often. It therefore appears that the mountain task is abstract in an important psychological sense - it is abstracted from common human purposes and feelings. Also, no attempt is usually made in the standard task to make the intentions of the experimenter clear to the child. This suggests that we ought not take the literal perspective-taking of the highly cognitivized sort as a basic skill. It is, rather, a later developmental accomplishment, one that develops in a context of more fundamental skills of reading intentions.
Training studies may be considered to function in a similar way. Providing the child with training is a way of coordinating intentions between child and experimenter - the experimenter makes clear the type of factors that are important in a situation and then tests the child on a formally similar task to see if the training will transfer. In general, some success has been achieved on various Piagetian tasks by providing this type of context (Gelman & Baillargeon, 1983). Some studies provide the child with conflicting perceptual and quantitative cues (Beilin, 1977), others provide reversibility training (Wallach & Sprott, 1964). There is another research literature on observational learning and modeling (Smiley & Brown, 1979; Murray, 1972; 1981; Botvin & Murray, 1975; Silverman & Geiringer, 1973) that has also been successful at times in inducing appropriate classification or conservation.
Given that the child is sensitive to the purposes, motives or intentions of the experimenter, what sources of social information can the child use? Here we must include as possibilities all the actions or "moves" of the investigator in manipulating materials, gesturing to the child, and verbal communications. Some relevant research has been done on the influence of social factors in recent years, focussing on the child's understanding of how actions by the investigator or a third party are related to the focal task (McGarrigle & Donaldson, 1974; Light et.al. 1979; Hargreaves, et.al. 1982). These studies have used the conservation paradigm to explore the influence of "accidental" or "incidental" changes to the test array on children's conservation response. After asking the first question regarding the equality of the display, an event occurs to change the perceptual characteristics of the display that is not apparently related to the purpose of the experimenter. In the McGarrigle and Donaldson study, a "naughty" teddy bear comes in and "accidentally" messes up the display. Children between 4 and 6 years had less problems with this conservation task than in the usual paradigm, with about 70% conserving. In an attempt to increase the children's success rate, Light et.al. introduced an "incidental" transformation by having the experimenter change one of two beakers filled with an equal number of pasta shells ostensibly because one of the beakers was chipped and then asking the child to indicate whether the two quantities were still equal. Conservation responses increased even more than in the accidental situation. Hargreaves et.al. replicated this finding with a "mischievous" monkey (reasoning that this makes more sense to children than a naughty teddy) that was manipulated by a second experimenter to avoid the child linking the original experimenter with the change. Another study was carried out with a second experimenter who comes in and begins to remove the testing displays as if he thought everything was finished. Both of these "incidental" manipulations led to better performance on the second conservation question (88% of 5-6 year olds conserved in the second condition).
Hargreaves and colleagues conclude at the end of their study: "Social and interpersonal aspects of the test situation are fundamental to our understanding of the child's response, and should not be regarded merely as surface phenomena (p. 234)".
These studies indicate a sensitivity by the child to the purpose of the experimenter, as displayed in the ways the task framework is partitioned from the broader ongoing flow of events. Given this fact, then one must assume that the results of other manipulations or "social moves", both purposeful and incidental from the experimenter's point of view, may serve to change the child's understanding of what is going on. Rose & Blank's(1974) one-question conservation test is another case in point. The authors of this study hypothesized that the very fact that the experimenter asks the same question both before and after a transformation of the array suggests to the child that the transformation is relevant to the second question and so the child gives a non-conserving response. They designed a one-question test and increased the number of conserving responses over the standard situation. Thus we may conclude that children are sensitive to the "moves" of the experimenter and will change their interpretation of the question light of these actions.
These studies contain a wealth of information on the situated use of logic, although the focus of the investigators is generally on logical schemata and not on the situational factors as components of the very use of logic itself. Therefore crucial information about the child's behavior in these situations is often not reported. A more detailed process analysis of what occurs in these situations is needed in order to determine what resources the child is able to make use of in achieving a coordinated understanding of the task. Instead of viewing development as a matter of changing logical schemata, attention should focus on how children attentionally organize and otherwise interact with the physical situation, their attention to gestural and vocal moves of the other, and how children are able to use these resources for accessing information needed to function adaptively in a situation. In short, more focus is needed on the actual ecological and social conditions which both make possible and support logical inference.
One important resource used by participants in the experiment and in all social interaction is speech, so this deserves special consideration.
In the foregoing it has been alluded that the child's understanding of the situation often seems independent of the speech used by the experimenter. In Bryant's work, for example, the child interprets the number conservation question of "more, less or the same" in terms of salient features and is therefore answering a different question than that understood by the experimenter. In Donaldson's garage study the child's attention appears to be drawn to the "fullness" of the garages and so the question is interpreted accordingly. In other words, the child's interpretation of the situation depends on more than just the adult's speech. It depends on other skills for making sense of situations involving direct and immediate human interaction. Speech is often interpreted in accordance with other information available to the child in the situation.
Considering observational data, it appears that children interpret situations before they can adequately use language. John McNamara(1972) has suggested that the child's ability to learn language in fact depends on a prior ability to understand human situations and that language is fit to the contours of the situation, thus providing a mechanism for language learning. The coordination of behavior begins before the child can effectively use language or understand it for its own sake. An understanding of language thus depends heavily on the presence of other cues. Consider this example given by Donaldson(1978):
An English woman is in the company of an Arab woman and her two children, a boy of seven and a little girl of thirteen months who is just beginning to walk but is afraid to take more than a few steps without help. The English woman speaks no Arabic, the Arab woman and her son speak no English.
The little girl walks to the English woman and back to her mother. Then she turns as if to start off in the direction of the English woman once again. But the latter now smiles, points to the boy and says: `Walk to your brother this time'. At once the boy, understanding the situation though he understands not a word of the language, holds out his arms. The baby smiles, changes direction and walks to her brother. Like the older child, she appears to have understood the situation perfectly (p. 37).
The speech of adults generally fits the patterns of interaction, and its meaning is often highly predictable in the human context of its occurrence for young children.
In the experimental situation, one in which we impose a pre-determined task on the child, a shared comprehension and ability to use language is crucial. A number of investigators have suggested that children do not perform well on reasoning tasks because they do not share with adults an understanding of the meaning of the terms used in the questions. From the present point of view, it would do us well to investigate the influence of circumstantial constraints on the child's interpretation of the adult's speech. It would seem that the child's performance on the formal task imposed by the adult depends a great deal on the force of immediate circumstantial constraints such as perceptual prominence or spatial arrangement, as well as the social moves of the adult.
The question of how language or speech is bound up with the development of reasoning abilities is a crucial one that requires more investigation. How able, for example, is the young child to engage in strict inference from verbal premises presented to him or her? Evidence suggests that by 3 years old children are able to act logically in accordance with simple verbal statements (Donaldson, 1978). Another question concerns how the child is able to reason in situations in which the circumstantial constraints are at odds with the formal task demands as presented verbally by the adult. Children often have difficulty operating only in terms of the latter. There is a difference between adults and children in the weight given to linguistic form. For young children the meaning of the words does not seem to carry enough weight to over-ride the meaning of the situation as constructed from other cues. They have difficulty paying attention to language in its own right. As language comes to be employed more and more as a resource, it increasingly comes to dominate attention. Initially, language is heavily context-bound. We might say that the child first learns to read intentions by non-linguistic cues and fits what is said into the context of these cues, later coming to rely more and more on the words per se.
Language has a central role in the development of behavioral control and more specifically in the manner in which situations are partitioned into regions of salience and non-salience. Language compels an individual to take the standpoint of other individuals and to see and inquire from a standpoint that is not strictly personal but is common to them as participants in a joint undertaking. It may be "about" physical objects, but it has a prior reference with others for whom it constitutes a communication, or the making of something common. Hence, to that extent its reference becomes general and "objective". Therefore it is the means by which both individual and interpersonal control are simultaneously achieved. By means of it situations acquire new properties for the individual. These properties are shared with other members of the culture and provide the means by which the coordination of behavior is made possible. Language makes possible deliberate recollection and expectation and allows for the selection of new combinations of outcomes without the final commitment inherent in other kinds of action. This ability of deliberate exploration (inferential activity) is, however, constrained by the social context which in part structures the way attention moves.
The learning of language is thus crucial for furthering a process we can assume has already begun - the entering of the child into a world of feature salience and relevance that is shared by others. Further progress may be assumed to occur when the child begins to develop a sense of language as something that may be manipulated "in itself" as a formal system.
We must conclude therefore that the study of speech development viewed as a means for personal and interpersonal control is central to the study of reasoning skills.
This discussion of the problem of relevance or salience and the emergence of salience in interactive situations has led us to consider that the development of inferential skills may best be viewed in the context of behavioral coordination and the reading of the behavioral intentions of the others. All experimental situations may be so viewed as involving first and foremost the basic problem of coordinating intentions. A number of processes are involved in order that the experimenter and child reach this mutual understanding. It appears that we can make new information available by creating perceptual distinctions, changing the language of the instructions, change the manner in which the child participates, simplify the response requirements, or changing the salience of the various features of the situations. In fact, we may say that the first task facing the child in these situations is not the resolution of a logical cognitive problem but that of reaching an understanding of what's happening. In this situated task, questions of expectations and understanding are of prime importance and this requires that the child has resources for seeing the situation in the same manner as the adult. It is within the context of behavioral coordination that behavior comes to assume logical properties. The more literal and advanced skill develops within the context of, and functions within, the former.
For Piaget, the young child is faced with the primary goal of understanding the physical world. From the present point of view, the child's fundamental adaptive goal is to coordinate its behavior with that of others. This makes sense from an evolutionary-developmental point of view - the first adaptive task of the infant is to establish and maintain a link with the mothering context. This is the essential ecological niche, the infant's link to the future. Later, adaptation to the adult world is the important developmental task; of preparing oneself for functioning in the adult world, contributing to the welfare of the group in many ways, a task for which the continued development of abstract thinking is required (Mason, 1970). From Piaget's perspective, the salience of situations is an outcome of the child's exploration of an objective reality, and it by virtue of this that adults share the same world. An alternative is to see this shared salience as a social-developmental outcome of coordinating behavior.
Before an inferential conclusion can be reached, a general reading of the relevant features of the situation must occur. We cannot in the end make much distinction between a deductive inference in a formal sense and questions of salience, of attentional control, of interactive information and ecological structure. This reliance on the physical layout of the task, the activities of the experimenter and child, is what is meant by saying that the use or appearance of logic depends on situated activity, or as Dewey said, the proper ordering of materials towards the end of achieving certain consequences.
We must focus our future research efforts on the central issue of how children come to control their behavior in increasingly complex ways, which is the same as asking how their behavior comes under the control of external contextual features. The study of inferential abilities must be more closely tied to the study of the child's general development. The growth of motor restraint, for example, is likely to be very important in the development of the ability to accept an imposed task. How able is the child to sustain behavior that he or she has not initiated? Under what conditions do children recognize that they do not have enough information to act appropriately and how do they seek out that information? When does language come to dominate over circumstance as a guide to behavior, particularly in situations of discordant information? The development of inferential skills depends on issues such as these.