Analogical Reasoning in Children

Usha Goswami, Institute of Child Health, University College London.

 

Analogical thinking lies at the core of human creativity. It has been argued that the very act of forming an analogy requires a kind of ‘mental leap’, as it necessitates seeing one thing as if it were another (Holyoak & Thagard, 1995). Famous analogies in science frequently reveal their inventor’s ability to make these mental leaps. However, as well as being an important cognitive mechanism in creative thinking, analogy is the basis of much of our everyday problem solving. Small-scale ‘mental leaps’ are being made all the time by children and adults and form a core part of our everyday mental repertoire. This chapter examines the availability of analogical reasoning to young children. Far from being a sophisticated reasoning strategy characteristic of older children and adults, it will be argued that analogical reasoning is available from infancy onwards. ‘Analogy pervades all our thinking, our everyday speech and our trivial conclusions as well as artistic ways of expression and the highest scientific achievements’ (Polya, 1957).

The Development of Analogical Reasoning

The Relational Similarity Constraint

Even though analogies require mental leaps, they are guided by certain basic constraints. The most important is the need for relational similarity. In many analogies, the objects in the analogy are not similar at all. Their similarity is at a purely relational level. A striking example of this is the analogy that led to Kekule’s (1865) theory about the molecular structure of benzene (see Holyoak & Thagard, 1995). In a dream, Kekule had a visual image of a snake biting its own tail. This gave him the idea that the carbon atoms in benzene could be arranged in a ring. The similarity between the snake and the carbon atoms was at a purely relational level - circular arrangement. The fact that the objects being compared in an analogy should be linked by the same relations is widely accepted to be the hallmark of analogical reasoning, and has been called the ‘relational similarity constraint’ (Goswami, 1992).

Quite frequently, objects in the two situations being compared in an analogy do bear some resemblance to each other - they share ‘surface’ similarity. This similarity of appearance can support the analogical mapping. An example is the invention of Velcro, which followed the observation by Georges de Mestral that burdock burrs stuck to his dog’s fur (see Holyoak & Thagard, 1995). The surface similarity in the appearance of the small hairs coating burdock burrs and the fuzz on Velcro supports the relational similarity of ‘effective sticking mechanism’. These two factors - relational similarity and surface similarity - both affect analogical reasoning (e.g., Gentner, 1989). Gentner’s work shows clearly that surface similarities between objects can support relational mappings and hence affect analogical performance. In order to examine the development of analogical reasoning in children, therefore, we need to examine their understanding of the relational similarity constraint in the absence of surface similarity.

The Item Analogy Task

The standard test for analogical reasoning (used in IQ testing) is the ‘item analogy’ task. In item analogies, two items A and B are presented to the child, a third item C is presented, and the child is required to generate a D term that has the same relation to C as B has to A. Successful generation of a D term requires the use of the relational similarity constraint. For example, if the child is given the items ‘cat is to kitten as dog is to ?’, she is expected to generate the solution term ‘puppy’. The response ‘bone’, which is a strong associate of dog, would be an error. Another example is the analogy ‘Bicycle is to handlebars as ship is to ?’. Here the relation constraining the choice of a D term is ‘steering mechanism’, and so a child who offered the completion term ‘bird’ would not be credited with understanding the relational similarity constraint.

The first developmental psychologist to study analogical reasoning, Piaget, used a pictorial version of the item analogy task, and his data suggested that understanding of the relational similarity constraint did not develop until early adolescence (Piaget, Montangero & Billeter, 1977). Younger children tested by Piaget offered solutions like ‘bird’ to the bicycle/ship analogy, giving reasons like ‘both birds and ships are found on the lake’. Piaget concluded that younger children solved analogies on the basis of associations (see also Sternberg & Nigro, 1980). He argued that children only became able to reason on the basis of relational similarity at around 11-12 years of age. This conclusion was accepted in developmental psychology for many years. Piaget’s conclusions about analogical development fitted neatly into his influential theory of the development of logical reasoning in children. Analogies appeared to be characteristic of the final stage of logical development - the stage of ‘formal operational’ reasoning. Formal operational reasoning required children to operate mentally on the results of simpler operations. A simpler operation was finding relations between objects (‘first-order’ relations). As analogies required children to reason about similarities between the relations between objects (‘second-order’ relations), it appeared to be a typical formal operational skill.

The Role of Relational Familiarity in Analogical Development

Closer inspection of Piaget’s experimental methods, however, suggest that his conclusions about analogical development were too negative (Goswami, 1991). A key methodological problem was that Piaget did not check whether the younger children in his experiments understood the relations on which his analogies were based (for example, the relation ‘steering mechanism’ in the bicycle:handlebars::ship:rudder analogy). This failure to ensure that the first-order relations were familiar means that the younger children’s failure to solve the item analogies in Piaget’s experiments could have arisen from a lack of knowledge of the relations being used. Item analogies based on unfamiliar relations would obviously underestimate analogical ability.

One way to test this possibility is to design analogies based on relations that are known to be highly familiar to younger children from other cognitive developmental research. Simple causal relations such as melting, wetting and cutting are known to be acquired early in development, and to be available for use in picture-based tasks by the ages of 3 and 4 years. Relations between real world objects such as ‘trains go on tracks’ and ‘birds live in nests’ are very familiar to 4- and 5-year-olds. Item analogies such as ‘playdoh is to cut playdoh as apple is to cut apple’ and ‘bird is to nest as dog is to doghouse’ can thus be used to examine whether 3- to 5-year-olds have the ability to reason by analogy.

For this young age group, a picture-based version of the item analogy task is required (Goswami, 1989, Goswami & Brown, 1989, 1990). The analogy task can then be presented as a ‘game’ about matching pictures. In our studies, the children were shown a ‘game board’ with four slots for pictures, the slots being grouped into two pairs for the A:B and C:D parts of the analogy (see Figure 1). As the children watched, the experimenter presented the first three terms of a given analogy (e.g., pictures of a bird [A], a nest [B], and a dog [C]). As the pictures were presented, the child was asked to name each one to ensure that they were familiar. The child was then asked to predict the picture that was needed to finish the pattern. The prediction task was included to see whether the children could generate an analogical solution spontaneously, without seeing the solution pictures. This would be evidence for truly ‘mental’ operations.

----------------------

                                    Figure 1 about here

                                    ----------------------

Following the request for a prediction, the experimenter showed the child a choice of solution terms. For the bird/dog analogy, these were pictures of a doghouse, a cat, another dog, and a bone (see Figure 1). The different choices were designed to test different theories of analogical development. The correct choice, which would indicate analogical ability, was the doghouse. The associative choice was the bone. Selection of the bone would be expected if younger children rely on associative reasoning to solve analogies, as Piaget and Sternberg had claimed. The other choices were a ‘mere appearance match’ choice (the second dog), and a category match (the cat). ‘Mere appearance’ matching is a term coined by Gentner (1989) to refer to the matching of object or ‘surface’ similarities when attempting to solve analogies (such as choosing another dog to match the dog in the C term, or choosing something that looks like a nest to match the B term). Gentner has suggested that younger children might rely quite heavily on object similarity in reaching analogical solutions (Gentner, 1989).

The matching game showed that all children tested (4-, 5- and 9-year-olds) performed at levels significantly above chance in the analogy task, selecting the correct completion term 59%, 66% and 94% of the time respectively. There was no evidence of mere appearance matching. Although many younger children were shy of making predictions prior to seeing the solution choices, those who were more confident showed clear analogical ability on this measure as well. For example, when 4-year-old Lucas was given the analogy bird is to nest as dog is to ?, he first predicted that the correct solution was puppy. He argued, quite logically, "Bird lays eggs in her nest [the nest in the B-term picture contained three eggs] - dog - dogs lay babies, and the babies are - umm - and the name of the babies is puppy!" Lucas had used the relation type of offspring to solve the analogy, and was quite certain that he was correct. He continued "I don't have to look [at the solution pictures] -- the name of the baby is puppy!" Once he looked at the different solution options, however, he decided that the doghouse was the correct response.

The matching game also included a control task to ensure that the correct solution to the analogy was not simply the most attractive pictorial match for the C term picture. In the control task, the children were simply shown the C term picture along with the correct solution term and the distractors, and were asked to choose which picture ‘went best’ with the C term picture. For example, the children were shown the picture of the dog, and were asked to choose the best match from the pictures of the doghouse, bone, second dog and cat. In this unconstrained task, the children were as likely to select the associative match (bone) as the analogy match (doghouse). Additionally, although the children readily agreed that another match could be correct in the control condition (9 year olds: 76%, 4 year olds: 82%), they were not so flexible in the analogy condition, where most of them said that only one answer could be correct (9 year olds: 89%, 4 year olds: 60%). This shows awareness of the relational similarity constraint that governs truly analogical responding. The children understood that the correct completion term for the analogy had to link the C and D terms by the same relation that linked the A and B terms. Notice also that Lucas was using the relational similarity constraint when he generated the solution ‘puppy’ for the bird/dog analogy. This cognitive flexibility displays a full understanding of analogy, and provides evidence of truly mental operations, thereby meeting Piaget’s original criteria for the presence of ‘true’ analogical reasoning.

The Relationship between Relational Knowledge and Analogical Responding

From the picture analogy game, we know that the ability to reason by analogy is present by at least age 4. However, the analogy game may still have underestimated analogical ability. This is because relational knowledge was not measured independently of analogical success. Instead, it was simply assumed that familiar relations had been selected for the analogies, leaving open the possibility that the younger children may have failed in some trials because the relations used in those particular analogies were unfamiliar to them. Alternatively, some children may have failed some analogies because they were actually reasoning about relations that were different from those intended by the experimenter - like Lucas.

The idea that children’s analogical performance depends on their relational knowledge has been called the ‘relational familiarity’ hypothesis’ (Goswami, 1992). In order to establish whether children's use of analogical reasoning is knowledge-based, dependent on relational familiarity rather than analogical ability, relational knowledge as well as analogical ability needs to be assessed. This can be done by changing the control task in the picture matching game. The appropriate control task measures children's knowledge of the relations being used in the analogies that are presented in the item analogy task.

A second set of analogy experiments using the picture matching game were thus carried out to test the relational familiarity hypothesis. This time, item analogies based on physical causal relations like melting, cutting and wetting were used. These relations are known to be available by 3 - 4 years (Bullock, Gelman & Baillargeon, 1982). Children aged from 3 to 6 were given analogies like ‘chocolate is to melted chocolate as snowman is to ?’, and ‘playdoh is to cut playdoh as apple is to ?’. This time the distractors were (1) a different object with the same causal change (e.g., cut bread for the cutting analogy), (2) the same object with a different causal change (e.g., a bruised apple), (3) a mere appearance match (e.g., a small ball), and (4) a semantic associate of the C term (e.g., a banana). Knowledge of the causal relations required to solve the analogies was measured in a control condition. Here the children were shown three pictures of items that had been causally transformed (e.g., cut playdoh, cut bread, cut apple), and were asked to select the causal agent responsible for the transformation from a set of pictures of possible agents (e.g., a knife, water, the sun).

The results showed that both analogical success and causal relational knowledge increased with age. The 3-year-olds solved 52% of the analogies and 52% of the control sequences, the 4-year-olds solved 89% of the analogies and 80% of the control sequences, and the 6-year-olds solved 99% of the analogies and 100% of the control sequences. By far the most frequent error was to select the correct object with the wrong causal change (e.g., the bruised apple, a dirty snowman). This suggested that the children knew that a causal transformation was required, but did not always select the correct one. The next most frequent error was to select the correct causal transformation of the wrong object. There was also a significant conditional relationship between performance in the analogy condition and performance in the control condition, as would be predicted by the relational familiarity hypothesis. This conditional relationship arose because individual children’s performance in the analogy task was linked to their knowledge of the corresponding causal relations. Analogical reasoning in children is thus highly dependent on relational knowledge. This raises the possibility that children younger than 3 years of age may be able to reason by analogy - as long as they have the requisite relational knowledge.

Analogical Reasoning in Infants and Toddlers

It is difficult to use the item analogy format to demonstrate analogical competence in children younger than 3 because of the abstract nature of the task. Researchers working with very young children have thus devised ingenious problem analogies in order to show analogical reasoning at work. In problem analogies, a young child is faced with a problem that they need to solve. Let us call this problem B. The use of an analogy from a previously-experienced problem, problem A, offers a solution. The measure of analogical reasoning is whether the children think of using the solution from problem A in order to solve problem B.

Chen and his colleagues devised a way of giving problem analogies to infants as young as 10 months of age, using a procedure first developed by Brown (1989) for 1 1/2- to 2-year-olds. Brown's procedure depended on seeing whether toddlers could learn how to acquire attractive toys that were out of reach. Different objects (such as a variety of tools, some more effective than others) were provided as a means to a particular end (bringing the desired toy within grasping distance). The analogy was that the means-to-an-end solution that worked for getting one toy in fact worked for all of the problems given, even though the problems themselves appeared on the surface to be rather different. Brown and her colleagues used this paradigm to study analogical reasoning in children aged 17-36 months. Chen, Campbell and Polley (1995) were able to extend it to infants.

In Chen et al.'s procedure, the infants came into the laboratory and were presented with an Ernie doll that was out of their reach. The Ernie doll was also behind a barrier (a box), and had a string attached to him that was lying on a cloth (see Figure 2). In order to bring the doll within reach, the infants needed to learn to perform a series of actions. They had to remove the barrier, to pull on the cloth so that the string attached to the toy came within their grasp, and then to pull on the string itself so that they could reach Ernie. Following success on the first trial, two different toy problem scenarios were presented, each using identical tools (cloths, boxes and strings). However, each problem appeared to be different to the problems that preceded it, as the cloths, boxes and strings were always dissimilar to those encountered before. In addition, in each problem two strings and two cloths were provided, although only one pair could be used to reach the toy.

                                    ----------------------

                                    Figure 2 about here

                                    ----------------------

Chen et al. tested infants aged 10 and 13 months in the Ernie paradigm. They found that although some of the older infants worked out the solution to reaching Ernie on their own, others needed their parents to model the solution to the first toy acquisition problem for them. Once the solution to the first problem had been modelled, however, the 13-month-olds readily transferred an analogous solution to the second and third problems. The younger infants (10 months) needed more salient perceptual support in order for reasoning by analogy to occur. They only showed spontaneous evidence of using analogies when the perceptual similarity between the problems was increased (for example, by using the same goal toy, such as the Ernie doll, in all three problems).

Freeman (1996) devised a series of analogies for 2-year-olds using real objects and models. Her analogies were based on the simple causal relations of stretching, fixing, opening, rolling, breaking and attaching. For example, a child might watch the experimenter stretching a loose rubber band between two plexiglass poles in order to make a 'bridge' that she could roll an orange across ("Look what I'm going to do, I'm going to use this stuff to roll the orange! Stretch it out, put it on - wow, that's how I roll the orange!"). Following an opportunity to roll the orange by themselves, the children were given a transfer problem involving a loose piece of elastic, a toy bird and a model with a tree at one end and a rock at the other. They were asked "can you use this stuff to help the bird fly?". The intended solution was to stretch the elastic from the tree to the rock, and to 'fly' the bird along it. In a third analogy problem, the children were asked to "give the doll a ride" by stretching some ribbon between two towers of different heights that were fixed to a base board. Children in a control condition were simply asked to "help the bird fly" and "give the doll a ride" without first seeing the base analogy of rolling the orange.

Freeman found that whereas only 6% of the children in the control condition thought of the stretching solution to the transfer problem, 28% of 30-month-olds in the analogy condition did so, and this figure rose to 48% following hints to use an analogy ("You know what? To help the bird fly, we have to change this", said while pointing to the elastic). When the same hint was given to the children in the control condition, only 14% thought of the stretching solution. Although these performance levels may appear modest, they are comparable to the spontaneous levels of analogical transfer found in adults. Problem analogy studies conducted with adults typically find spontaneous transfer levels of around 30%, at least in unfamiliar problem scenarios (e.g., Gick & Holyoak, 1980).

The studies reviewed above suggest that analogical reasoning is available at 1 and 2 years of age. Children younger than about 1 year may need the support of perceptual or surface similarity in order to use analogies successfully (see Gentner, 1989). However, the degree of surface similarity required may depend on the extent to which the relations relevant to the analogy have been represented by the infant. Remarkable research evidence from studies of babies suggests that a focus on relations in the absence of surface similarity is strong even in the first year of life.

The Origins of Relational Reasoning

The knowledge that children have about objects and events in their worlds, and the knowledge that they have about the causal and explanatory structures underlying relations between these objects and events, is known to expand at a terrific rate from the very first months of infancy. Infants and young children gain a lot of their knowledge by observing and interacting with the world around them. Equally importantly, they seek to explain their everyday worlds to themselves. The result is that the knowledge base is continually restructured and revised in the light of experience. Analogy may play an important role in this process of acquisition and self-explanation of knowledge (Goswami, 1998). This process of revision and restructuring is aided by the development of abstract knowledge structures (schemas and scripts) for representing the temporal and causal frameworks of events, frameworks that facilitate the storage and organisation of incoming information and presumably also the application of analogy. However, the recognition of relational similarity does not appear to depend on the development of such abstract representations. The ability to recognise relational similarities appears to be present from very early in development. This has been called the ‘relational primacy’ hypothesis (Goswami, 1992, 1996).

Relational Processing in Infants: The Physical World

The roots of relational competence can be found in the mechanisms that infants use to process perceptual information. The early representation of perceptual prototypes, for example, appears to depend on the extraction of correlational information about visual features, suggesting sensitivity to the relations between those features (see Goswami, 1996, 1998). For example, infants who were shown a series of ‘cartoon’ animals attended not only to the distinctive features of these animals (e.g., their necks and legs), but to the correlations between the features (long legs went with short necks, short legs went with long necks, see Younger & Cohen, 1983). New exemplars were then categorised according to their relational similarity to the prototype. Structural similarities among perceptual stimuli can also be demonstrated, and these may well form the basis of relational representations of those stimuli at a conceptual level (e.g., Smith & Heise, 1992). For example, structural similarities in auditory and visual perceptual events can convey relational information. A tone of music may be ascending because of acoustic properties such as timbre. A visual stimulus may have properties that convey ascension, such as a locus of highest density at its uppermost point (e.g., an arrow). Infants who are shown a choice between two visual stimuli, an arrow pointing upwards and an arrow pointing downwards, and who hear an ascending tone of music coming from the mid-line, preferentially look at the ‘up’ arrow (Wagner, Winner, Cicchetti & Gardner, 1981). This is a form of analogy. These inherent perceptual properties of rising tones of music and of arrows pointing upwards could convey the shared relation 'ascending', which is part of our mental representation of these physical events. Infants may first be sensitive to the similarity between the inherent structure of physical events, and then come to represent them as relationally similar (see Goswami, 1998, for further discussion).

Further support for the notion that infants are sensitive to relations from early in development comes from evidence that infants preferentially attend to changes in the visual scene. Change is informative, because change signals the occurrence of events. Events in the visual world are usually described by relations between objects (such as ball collides with teddy, child pushes friend). The ability to detect structural regularities in these relations (to notice relational similarity) would be cognitively powerful in terms of knowledge acquisition, as events in the visual world are frequently causal in nature. The detection of regularities in causal relations like collide, push and supports between different objects (such as ball collides with teddy, toy car collides with toy garage, mummy collides with daddy) would involve a rudimentary form of analogical reasoning.

There is quite a lot of evidence that infants are sensitive to cause-effect relations by about 6 months of age (see Goswami, 1998, for a review). Other types of relations, such as spatial relations (above and below) and quantitative relations (more than and less than) are also detected by this age. One way of measuring infants' ability to process and represent spatial, numerical and causal relations is to introduce violations of typical regularities in the relations between objects, which then result in physically 'impossible' events. For example, an object with no visible means of support can remain stationary in mid-air instead of falling to the ground. The experimental investigation of infants' ability to detect such violations is the main source of evidence for infants’ ability to process relations between events.

Consider the causal relations involved in support. Adults are well aware that if they put a box of cookies down on a table and the box protrudes too far over the table’s edge, the cookies will fall onto the floor, whereas if only a small portion of the bottom surface of the box protrudes they will not. The recognition of how much contact is required to support an object in different situations must require a degree of relational comparison between one situation and the next. Baillargeon, Needham and DeVos (1992) studied 6.5-month-old infants' expectations about when a box would fall off a platform. The infants were shown a box sitting at the left-hand end of a long platform, and then watched as the finger of a gloved hand pushed the box along the platform until part of it was suspended over the right-hand edge. For some infants, the pushing continued until 85% of the bottom surface of the box protruded over the platform, and for others the pushing stopped when 30% of the bottom surface of the box protruded over the platform. In a control condition the same infants watched the box being pushed to the right-hand end of the platform, but the bottom surface of the box remained in full contact with the platform. The infants spent reliably longer looking at the apparatus in the 85% protrusion event than in the full-contact control event. This suggests that they expected the box to fall off the platform (the box was able to remain magically suspended in mid-air via a hidden hand). The infants in the 30% protrusion event looked equally during the protrusion event and the control event. Baillargeon et al. argued that the infants were able to judge how much contact was required between the box and the platform in order for the box to be stable. In fact, Baillargeon has repeatedly found that infants reason on the basis of relational information before they reason on the basis of absolute information when making judgements about the physical properties of objects. Bryant (1974) has reported a similar precedence for relational over absolute information in young children. The ease with which infants and young children process relational information suggests that relational processing is present in the cradle, and that relational comparisons are an important source of information about the physical world. No-one has so far devised a direct experimental test of this idea, however.

Relational Processing in Infants: The Psychological World

As noted above, one powerful aspect of relational information is that it frequently provides information about causality. Causal relations are not only particularly powerful relations for understanding the everyday world of objects and events, they also provide an insight into the psychological world. For example, recent work with infants has suggested that a sensitivity to causal relations may partly underlie the development of a notion of agency. Things that move on their own are agents, and things that move because of other things obey certain cause-effect, or mechanical, laws (see Leslie, 1994). Infants' interest in things that move helps them to sort out the source of different cause-effect relations in the physical world. Again, this seems likely to require a form of relational comparison.

For example, Gergely, Nadasdy, Csibra and Biro (1996) have shown that 12-month-old infants can analyse the spatial behaviour of an agent in terms of its actions towards a goal, and will apply an ‘intentional stance’ to this behaviour when it appears rational, thereby attributing a mental cause for the goal-directed behaviour. The adoption of an ‘intentional stance’ towards agents entails an assumption of rationality - that the agent will adopt the most rational action in a particular situation to achieve his or her goal. To examine whether 12-month-old infants would generate expectations about the particular actions an assumed agent was likely to perform in a new situation to achieve a desired goal, Gergely et al. designed a visual habituation study in which a computer display gave an impression of agency to the behaviour of circles. The infants saw a display in which two circles, a large circle and a small circle, were separated by a tall rectangle (see Figure 3). During the habituation event, each circle in turn expanded and then contracted twice. The small circle then began to move towards the large circle. When it reached the rectangular barrier it retreated, only to set out towards the large circle a second time, this time jumping over the rectangle and making contact with the large circle. Both circles then expanded and contracted twice more. Adult observers of this visual event described it as a mother (large circle) calling to her child (small circle) who ran towards her, only to be prevented by the barrier, which she then jumped over. The two then embraced.

                                                            --------------------

                                                            Figure 3 about here

                                                            --------------------

Following habituation to this event, the infants saw the same two circles making the same sequence of movements, but this time without a barrier being present. However, although the relations between these objects were the same, the absence of a barrier meant that the causal significance of the relations differed - in analogy terms, the ‘relational structure’ was no longer comparable. In a ‘new action’ event, the small circle simply took the shortest straight path to reach the large circle. Gergely et al. predicted that, if the infants were making an intentional causal analysis of the initial display, then they should spend more time looking at the ‘familiar action’ event than the ‘new action’ event, as even though the former event was familiar it was no longer rational (nor analogous). This was exactly what they found. A control group who saw the same habituating events without the rectangle acting as a barrier (it was positioned to the side of the screen) showed equal dishabituation to the old and new action events. This intriguing result shows just how powerful the ability to represent and compare causal relations between objects may be for the young infant. Causal analyses of the everyday world of objects and events supply information about the physical world of inanimate objects and also about the mental world of animate agents.

Relational Processing in Infants: Imitation

Relational processing and relational comparison may also lie at the heart of the early facial imitation of gestures such as tongue protrusion, which has been reported in infants as young as 1 - 3 days (e.g., Meltzoff & Moore, 1983). Meltzoff and Moore (1997) describe the essential puzzle of facial imitation as follows: ‘Infants can see the adult’s face but cannot see their own faces. They can feel their own faces move, but have no access to the feelings of movement in the other. By what mechanism can they connect the felt but unseen movements of the self with the seen but unfelt movements of the other?’ (p. 180). Meltzoff and Moore suggest that the solution to this puzzle depends on the appreciation of the similarity of ‘organ relations’. They suggest that infants might compute the configural relations between organs, such as ‘tongue-to-lips’, and use these to represent both their own behaviour and that of the adult. These perceived organ relations provide the targets which the infants attempt to match. On this account, imitation involves an early form of analogising.

Interestingly, Piaget also suggested a role for analogy in explaining imitation behaviour in infants. He noted the occurrence of ‘motor analogies’ in his own babies, in which the infants imitated certain spatial relations that they had observed in the physical world with their own bodies. For example, they imitated the opening and closing of a matchbox by opening and closing their hands and mouths. Piaget suggested that this behaviour showed that the infants were trying to understand the mechanism of the matchbox through a motor analogy, reproducing a kinesthetic image of opening and closing. Again, the infant is suggested to be representing relations, and analogy is suggested as a mechanism for knowledge acquisition and for explaining events in the everyday world.

The coding of ‘organ relations’ proposed by Meltzoff and Moore (1997) would also allow infants to recognise when they are being imitated by an adult. Such recognition is found by at least 14 months (Meltzoff, 1990). Recognising that someone else is imitating you implies a recognition of the structural equivalence (or relational similarity) between another agent’s behaviour and your own (see also Meltzoff, 1990; Meltzoff & Moore, 1995). Again, this is a form of analogy. Such analogies may play an important role in the growth of psychological understanding. For example, Meltzoff has argued that imitation, broadly construed, serves as a discovery procedure for understanding persons. Although the links between imitation, analogy and the development of an understanding of the psychological world noted here must remain speculative given the current absence of detailed research, it is perhaps not a coincidence that imitation, analogy and the understanding of mental states (‘theory of mind’) are absent in the animal kingdom (with the possible exception of highly ‘language-trained’ chimpanzees (e.g., Thompson, this volume; Heyes, 1996, 1998; Tomasello, 1990; Visalberghi & Fragaszy, 1990)

Analogies in Foundational Domains

Recently, a number of developmental psychologists have argued that the developing knowledge base can be divided into three foundational ‘domains’ or sets of representations sustaining different areas of knowledge. These are the domains of naïve biology, naïve physics, and naïve psychology (e.g., Wellman & Gelman, 1997). Of course, many concepts will be represented in more than one of these foundational frameworks (for example, persons are psychological entities, biological entities and physical entities). Nevertheless, Wellman and Gelman suggest that children will use at least two levels of analysis within any framework, one that captures surface phenomena (mappings based on attributes) and another that penetrates to deeper levels (mappings based on relations). This means that analogies should be at work within foundational domains. We have already seen that evidence from infancy research suggests that analogies play a role in helping the child to understand the physical and psychological worlds. We turn now to an examination of the role of analogies in developing knowledge in these foundational domains in early and later childhood.

Analogy as a Mechanism for Understanding Biological Principles

Evidence that analogy is an important mechanism for understanding biological principles comes from a series of studies by Inagaki and her colleagues (Inagaki & Hatano, 1987; Inagaki & Sugiyama, 1988). They were interested in how often children would base their predictions about biological phenomena on analogies to people: the personification analogy. As human beings are the biological kinds best known to young children, it seems plausible that children should use their biological knowledge about people to understand biological phenomena in other natural kinds. For example, Inagaki and Sugiyama (1988) asked 4-, 5-, 8- and 10-year-olds a range of questions about various properties of 8 target objects, including  "Does x breathe?", "Does x have a heart?", "Does x feel pain if we prick it with a needle?", and "Can x think?". The target objects were people, rabbits, pigeons, fish, grasshoppers, trees, tulips and stones. Prior similarity judgements had established that the target objects differed in their similarity to people in this order, with rabbits being rated as most similar and stones being rated as least similar. The children all showed a decreasing tendency to attribute the physiological properties ("Does x breathe") to the target objects as the perceived similarity to a person decreased. Apart from the 4-year-olds, very few children attributed physiological attributes to stones, tulips and trees, and even 4-year-olds only attributed physiological properties to stones 15% of the time. A similar pattern was found for the mental properties ("Can x think?"). This study supports the idea that preschoolers' understanding of biological phenomena arises from analogies based on their understanding of people.

Analogy as a Mechanism for Understanding Physical Principles

Evidence that analogy is an important mechanism for understanding physical principles comes from a series of studies by Pauen and her colleagues. Pauen has studied children’s understanding of the principles governing the interaction of forces by using a special apparatus called the ‘force table’ (Pauen, 1996). The force table consists of an object that is fixed at the centre of a round platform. Two forces act on this object, both represented by plates of weights. The plates of weights hang from cords attached to the central object at either 45', 75' or 105' to each other. The children's job is to work out the trajectory of the object once it is released from its fixed position. Their predictions concerning this trajectory are scored in terms of whether they consider only a single force (one plate of weights), or whether they integrate both forces in order to determine the appropriate trajectory. The force table problem is presented to the children in the context of a story about a King (central object) who has got tired of skating on a frozen lake (the platform) and who wants to be pulled into his royal bed on the shore (see Figure 4). Children aged 6, 7, 8 and 9 years of age were tested.

----------------------

                                    Figure 4 about here

                                    ----------------------

Pauen found that most of the younger children (80 - 85%) predicted that the king would move in the direction of the stronger force only (the larger plate of weights). An ability to consider the two forces simultaneously was only shown by some of the 9-year-olds (45%). Such integration rule responses were shown by the majority of the adults tested (63%). Pauen speculated that this may have been because the children who received the plates of weights applied a balance scale analogy to the force integration problem. A balance scale analogy gives rise to one-force-only solutions, which are incorrect.

This idea about the balance scale analogy was prompted by the comments of the children themselves, who said that the force table reminded them of a balance scale (presumably because of the plates of weights). This led Pauen to propose that the children were using spontaneous analogies in their reasoning about the physical laws underlying the force table, analogies that were in fact misleading. To investigate this idea further, Pauen and Wilkening (1997) gave 9-year-old children a training session with a balance scale prior to giving them the force table problem. One group of children received training with a traditional balance scale, in which they learned to apply the one-force-only rule, and a second group of children received training with a modified balance scale that had its centre of gravity below the axis of rotation (a 'swing boat' suspension). This modified balance scale provided training in the integration rule, as the swing boat suspension meant that even though the beam rotated towards the stronger force, the degree of deflection depended on the size of both forces.

Following the balance scale training, the children were given the force table task with the plates of weights. A third group of children received only the force table task, and acted as untrained controls. Pauen and Wilkening argued that an effect of the analogical training would be shown if the children who were trained with the traditional balance scale showed a greater tendency to use the one-force-only rule than the control group children, while the children who were trained with the modified balance scale showed a greater tendency to use the integration rule than the control group children. This was exactly the pattern that they found. The children's responses to the force table problem varied systematically with the solution provided by the analogical model. These results suggest that the children were using spontaneous analogies in their reasoning about physics, just as we have seen them do in their reasoning about biology. At the present time, studies of children’s use of analogy in the domain of naïve psychology (theory of mind) do not feature in the research literature. However, given the promising developments noted in infancy research, this is surely only a matter of time.

Analogies in Piagetian Tasks

Halford’s Structure Mapping Theory of Logical Development

Most neo-Piagetian theorists of cognitive development incorporate the concept of relational mapping (e.g., Case, 1985; Halford, 1987, 1993; Pascual-Leone, 1987). However, Halford is the only neo-Piagetian who has formally proposed that analogy plays a central role in the development of logical reasoning, and who has linked analogical processes to performance in traditional Piagetian tasks. In his structure-mapping theory of cognitive development (1987, 1993), Halford proposed that most logical reasoning was analogical. He also proposed that limitations in primary memory (the memory system that holds any information that is currently being processed) constrained the kinds of analogy that children could use at different points in cognitive development. He suggested that this explained the relatively late emergence of the Piagetian ‘concrete operations’. Piaget used the term ‘concrete operations’ to refer to children’s symbolic or representational understanding of the properties of concrete objects and the relations between them (such as transitivity and class inclusion). His argument was that capacity limitations prevented children from representing and mapping transitive or class inclusion relations prior to approximately 5 years of age.

Halford suggested that the critical capacity limitations governing the use of analogy in logical reasoning concerned the number of relations that could be represented in primary memory at any one time (see also Halford, Wilson & Phillips, in press). The number of relations determined the processing load entailed in solving the analogy. A simple causal inference analogy of the kind used by Goswami and Brown (1989, e.g., vase:pieces of vase::egg:broken egg) was thought to entail a relatively low processing load, as it required the child to represent and map a single relation (breaking) from one object (vase) to another (egg). These relational mappings were said to be within the information-processing capacity of children as young as 2 years of age. An analogy that required two relations to be processed jointly was thought to entail a higher processing load (e.g., A above B above C :: Tom happier than Bill happier than John). The information-processing capacity capable of supporting such system mappings was thought to emerge between 4 and 5 years of age. Prior to this, it was hypothesised that even familiar pairs of relations could not form the basis of an analogy, as there was insufficient capacity in primary memory to hold and compare the representations of both relations simultaneously. Analogies based on pairs of relations were thought to be necessary for the successful solution of Piagetian concrete operational tasks. In order to test this theory, evidence that young children can solve analogies based on pairs of relations and evidence that they use analogies in Piagetian tasks such as transitive reasoning and class inclusion is required.

Analogies based on Pairs of Relations

In order to examine whether young children can solve analogies based on pairs of relations, Goswami, Leevers, Pressley and Wheelwright (1998) designed a set of analogies based on pairs of physical causal relations (extending the technique used by Goswami & Brown, 1989, described above). They asked 3-, 4-, 5- and 6-year-old children to make relational mappings based on either single causal relations like cut, paint, and wet, or pairs of causal relations, like cut + wet and mend + paint. Their experiment had four conditions, a single-relation analogy condition (e.g., apple: cut apple:: hair: cut hair), a double-relation analogy condition (e.g., apple: cut, wet apple:: hair: cut, wet hair), a single-relation control condition and a double-relation control condition. The 4 distractors in both analogy conditions were the same (representing two double relational changes and two single relational changes respectively). The child was thus forced to attend to the A:B pairing in each condition in order to select the correct solution term to each analogy. In the control conditions, the children were asked to select the picture of the causal agent or the pair of causal agents responsible for the causal changes shown in the analogies, following Goswami and Brown (1989).

Children's performance in the analogy and the control conditions was then examined as a function of Condition and Age. The pattern of the results was remarkably similar to the pattern found in the causal relations analogies used by Goswami and Brown (1989). There was a close correspondence between analogy performance and performance in the relational knowledge control conditions for both the single relation and the double relation analogies. For the single relation conditions, the 3-year-olds solved 33% of the analogies and 46% of the control sequences, the 4-year-olds solved 51% of the analogies and 63% of the control sequences, the 5-year-olds solved 72% of the analogies and 76% of the control sequences, and the 6-year-olds solved 89% of the analogies and 88% of the control sequences. For the double relation conditions, the 3-year-olds solved 13% of the analogies and 31% of the control sequences, the 4-year-olds solved 50% of the analogies and 50% of the control sequences, the 5-year-olds solved 62% of the analogies and 74% of the control sequences, and the 6-year-olds solved 78% of the analogies and 91% of the control sequences. Analyses demonstrated no interaction between age and number of relations, although the main effect of number of relations almost reached significance, reflecting the fact that children of all ages found the double relation analogies and control sequences more difficult than the single relation analogies and control sequences. Goswami et al. concluded that the ability to solve analogies based on pairs of relations was governed by relational familiarity. As long as familiar relational structures are chosen as a basis for analogy, therefore, young children should be able to use analogies to help them to solve Piagetian reasoning tasks.

Analogies in a Transitive Mapping Task

Halford has suggested that familiar ordered structures may provide useful analogies for transitive reasoning tasks. For example, a transitive sequence such as Tom happier than Bill happier than John may be solved by using an analogy to a familiar series of height relations such as A above B above C. As relational familiarity is known to determine analogical reasoning in young children, we need a familiar instantiation of height relations for the children to use as a basis for analogy. The family provides a familiar example of an ordering structure based on size, as in most families the father is taller than the mother, and the mother is taller than the child or baby. According to Halford’s structure-mapping theory, young children who have mentally represented the relational structure Father > Mother > Baby should be able to use this structure as a basis for analogies in transitive tasks using less familiar relations.

Goswami (1995) examined this hypothesis using Goldilocks and the Three Bears as a familiar example of family size relations (Daddy Bear > Mummy Bear > Baby Bear). Three- and 4-year-old children were asked to use the relational structure represented by the Three Bears as a basis for solving transitive ordering problems involving perceptual dimensions such as temperature, loudness, intensity, and width. The transitive mapping test was presented by asking the children to imagine going to the Three Bears' house, and then to imagine looking at their different belongings. This imagination task constituted a fairly abstract test. For example, the imaginary bowls of the Three Bears' porridge could be either boiling hot, hot, or warm, and the child had to decide which bowl of porridge belonged to which bear. In order to give the correct answer, the child had to map the transitive height ordering of Daddy, Mummy, and Baby Bear to the different porridge temperatures, giving Daddy Bear the boiling hot porridge, Mummy Bear the hot porridge, and Baby Bear the warm porridge (these mappings do not follow the original fairy tale, in which Daddy Bear's porridge was too salty, and Mummy Bear's was too sweet).

The results showed that the percentage of correctly ordered mappings approached ceiling for the 4-year-olds for most of the dimensions used. The lowest levels of performance occurred for width (of beds, 62% correct), and hardness (of chairs, 76% correct), and the highest occurred for temperature (of porridge, 95% correct). Performance with the width dimension (wide bed, medium bed, narrow bed) was possibly affected by worries that a baby could fall out of a narrow bed, as many children allocated the medium bed to Baby Bear. They were then left without a bed for Mummy Bear. The 3-year-olds produced correctly ordered mappings for only some of the dimensions, performance being above chance (17%) for the dimensions of temperature of porridge (31% correct), pitch of voice (31% correct), and height of mirrors (62% correct, but an isomorphic relation). Relational familiarity and real-world knowledge about family size relations seem to have helped the 3-year-olds with these particular dimensions. The children are unlikely to have based their correct mappings on the story, as none of these dimensions was mentioned in the Three Bears book that was read to them as part of the study. Note that this mapping task did not require a transitive inference, however. Evidence for the use of analogies by 3- and 4-year-olds in a ‘Daddy, Mummy, Baby’ paradigm has also been demonstrated in a mapping task requiring the recognition of monotonic size ordering for successful performance (Gentner & Ratterman, 1991; Ratterman & Gentner, 1998). 

Analogies in a Class Inclusion Task

Halford also suggested that families provide potentially  useful analogues for class inclusion tasks (Halford, 1993). His argument was that families were very familiar entities to young children, that they fulfilled the criteria for inclusion of having two clearly defined sub-sets (parents and children), and that they typically had a small set size, making it easy for children to represent the relevant relations in order to make an analogy. In order to see whether the family could act as a basis for successful performance in Piagetian class inclusion tasks, Goswami, Pauen and Wilkening (1996) devised the ‘create-a-family’ paradigm.

The children in Goswami et al.'s study (4- to 5-year-olds) had all failed the traditional Piagetian class inclusion task, which was given as a pretest ("Are there more red flowers or more flowers?"). In the ‘create-a-family’ paradigm, children were shown a toy family, for example a family of toy mice (2 large mice as parents, 3 small mice as children). Their job was to create analogous families (2 parents and 3 children) from an assorted pile of toys (such as toy cars, spinning tops, balls and helicopters). After the children had correctly created 4 analogous families, they were given 4 new class inclusion problems involving toy frogs, sheep, building blocks and balloons. As ‘family’ is also a collection term, and as collection terms are known to influence class inclusion reasoning, the class inclusion problems were posed using collection terms (‘group’, ‘herd’, ‘pile’, ‘bunch’). A control group of children received the same class inclusion problems using collection terms, but did not receive the ‘create-a-family’ analogy training session.

Goswami et al. found that more children in the 'create-a-family' analogy condition than in the control condition solved at least 3 of the 4 class inclusion problems involving frogs, sheep, building blocks and balloons. This effect was particularly striking at age 4, in which no improvement at all was found in the control group with the collection term wording, but criterion was reached by 50% of the experimental group. It should be remembered that all of the children had previously failed the standard Piagetian class inclusion task. Goswami et al. argued that this change was due to the use of analogies based on a representation of family structure. Goswami et al.’s data suggests that Halford’s structure-mapping theory of how analogies might contribute to the development of logical reasoning is both powerful and plausible. However, further research is required, particularly regarding the developmental appropriateness of the notion of capacity limitation as an upper limit on children’s use of analogies (see Gentner & Ratterman, in press; Goswami, in press).

Conclusion

It is not unusual in developmental psychology for researchers to demonstrate that apparent changes in children’s cognition are in reality changes in the knowledge that children have available as a basis for exercising a particular skill. Analogical reasoning appears to be no exception. If measures of analogy are based on unfamiliar relations, then these measures seriously underestimate children’s analogical skills. Hence early research concluded that analogical reasoning was absent until early adolescence because it depended on experimental tasks that used analogical relations that were unfamiliar to younger children. Later research has demonstrated that analogical reasoning is used by children as young as 1, 2 and 3 years of age. It has also been shown that forms of relational reasoning which probably involve relational comparisons are present in young infants.

Nevertheless, it is important to note that the nature of a cognitive skill measured at time 1 may differ completely from the nature of a cognitive skill measured at time 2. For example, Strauss (in press) has argued that the kind of analogies made by infants may be completely different from the kind of analogies made by older children. He suggests that the kind of analogies made by young infants are perceptual in nature, whereas those made by young children use conceptual knowledge. Such questions about the continuity of cognitive skills are important ones for answering the question of ‘what develops’ in analogical reasoning.

My argument here has been that the early age at which analogies appear suggest that they provide a powerful logical tool for explaining and learning about the world. Analogies also contribute to both the acquisition and the re-structuring of knowledge, and play an important role in conceptual change. As children's knowledge about the world becomes richer, the structure of their knowledge becomes deeper, and more complex relationships are represented, enabling deeper or more complex analogies. This means that, as children learn more about the world, the type of analogies that they make will change. Another important developmental question is whether these changes are driven solely by changes in the knowledge base, or whether information processing factors, such as the number of relations that can be represented in primary memory at any one time, determine these changes.

Finally, it may be worth noting that the developmental role of analogy in cognition is not limited to childhood. The role of analogy in the history of science can also be explained in a knowledge-based fashion. Scientific breakthroughs often depend on the right analogy (Gentner & Jeziorski, 1993; Gordon, 1979), but the scientists who make the breakthroughs seldom have extra information that is unavailable to their colleagues. Instead, the analogy occurs to them and not to their fellow scientists because of the way that their conceptual understanding of their field is structured, and the richness of their representations. This in turn may be correlated with their intelligence. If intelligence is important, then its importance may explain why classical analogy performance is a good correlate of I.Q. It has been reported that more efficient processing of stimuli as a neonate (using a habituation paradigm) is related to performance on a test of analogical reasoning at age 12 (e.g., bread is to food as water is to beverage, Sigman, Cohen, Beckwith, Asarnow & Parmelee, 1991). There is clearly still much research to be done before we can claim to understand the role of analogical reasoning in cognitive development in all its complexity.


References

Baillargeon, R., Needham, A., & De Vos, J. (1992). The development of young infants' intuitions about support. Early Development & Parenting, 1, 69-78.

Brown, A.L. (1989). Analogical learning and transfer: What develops? In S. Vosniadou & A. Ortony (Eds.) Similarity and Analogical Reasoning, (pp. 369-412). Cambridge: Cambridge University Press.

Bryant, P.E. (1974). Perception and Understanding in Young Children. London: Methuen.

Bullock, M., Gelman, R., & Baillargeon, R. (1982). The development of causal reasoning. In W.J. Friedman (Ed.), The Developmental Psychology of Time, pp. 209-254. New York: Academic Press.

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

Chen, Z., Sanchez, R.P., & Campbell, T. (1997). From beyond to within their grasp: Analogical problem solving in 10- and 13-month-olds. Developmental Psychology, 33, 790-801.

Freeman, K.E. (1996). Analogical reasoning in 2-year-olds: A comparison of formal and problem-solving paradigms. Unpublished Ph.D. thesis, University of Minnesota, 1996.

Gentner, D. (1983). Structure-mapping: A theoretical framework for analogy. Cognitive Science, 7, 155-170.

Gentner, D. (1989). The Mechanisms of Analogical Learning. In S. Vosniadou & A. Ortony (Eds.) Similarity and Analogical Reasoning, (pp. 199-241). London: Cambridge University Press.

Gentner, D., & Jeziorski, M. (1993). The shift from metaphor to analogy in western science. In A. Ortony (Ed.), Metaphor and thought, 2nd edition (pp. 447-480). Cambridge, England: Cambridge University Press.

Gentner, D., & Ratterman, M. J. (1991). Language and the career of similarity. In S. A. Gelman & J. P. Byrnes (Eds.), Perspectives on thought and language: Interrelations in development  (pp. 225-277). London: Cambridge University Press.

Gentner, D., & Ratterman, M.J. (in press). Deep thinking in children: The case for knowledge change in analogical development. Behavioural & Brain Sciences.

Gergely, G., Nadasdy, Z., Csibra, G., & Biro, S. (1995). Taking the intentional stance at 12 months of age. Cognition, 56, 165-193.

Gick, M.L., & Holyoak, K.J. (1980). Analogical problem solving. Cognitive Psychology, 12, 306-355.

Gick, M.L., & Holyoak, K.J. (1983). Schema induction and analogical transfer. Cognitive Psychology, 15, 1-38.

Gordon, W. J. J. (1979). Some source material in discovery-by-analogy. The Journal of Creative Behaviour, 8, 239-257.

Goswami, U. (1991). Analogical reasoning: What develops? A review of research and theory. Child Development, 62, 1-22.

Goswami, U. (1992). Analogical Reasoning in Children. Hillsdale, NJ: Lawrence Erlbaum Associates.

Goswami, U. (1995). Transitive Relational Mappings in 3- and 4-year-olds: The Analogy of Goldilocks and the Three Bears. Child Development, 66, 877-892.

Goswami, U. (1996). Analogical Reasoning and Cognitive Development. Advances in Child Development and Behaviour, 26, pp. 91-138. San Diego, California: Academic Press.

Goswami, U. (1998). Cognition in Children. Hove: Psychology Press.

Goswami, U. (in press). Is relational complexity a useful metric for cognitive development? Behavioural & Brain Sciences.

Goswami, U., & Brown, A.L. (1989). Melting chocolate and melting snowmen: Analogical reasoning and causal relations. Cognition, 35, 69-95.

Goswami, U. & Brown, A.L. (1990). Higher-order structure and relational reasoning: Contrasting analogical and thematic relations. Cognition, 36, 207-226.

Goswami, U., Pauen, S., & Wilkening, F. (1996). The effects of a 'family' analogy on class inclusion reasoning in young children. Unpublished manuscript, Institute of Child Health, University College London.

Goswami, U., Leevers, H., Pressley, S., & Wheelwright, S. (1998). Causal reasoning about pairs of relations and analogical reasoning in young children. British Journal of Developmental Psychology, 16, 553-569.

Halford, G.S. (1987). A structure-mapping approach to cognitive development. International Journal of Psychology, 22, 609-642.

Halford, G.S. (1993). Children's Understanding: The Development of Mental Models. Hillsdale, NJ: Erlbaum.

Halford, G.S., Wilson, W.H., & Phillips, S. (in press). Processing capacity defined by relational complexity: Implications for comparative, developmental and cognitive psychology. Behavioural & Brain Sciences.

Heyes, C.M. (1996). Genuine imitation? In C.M. Heyes & B.G. Galef (Eds.), Social learning in animals: The roots of culture, pp. 371-389. San Diego: Academic.

Heyes, C.M. (1998). Theory of mind in non-human primates. Behavioural & Brain Sciences, 21, 101-134.

Holyoak, K.J., & Thagard, P. (1995). Mental Leaps. Cambridge, MA: MIT Press.

Inagaki, K., & Hatano, G. (1987). Young children's spontaneous personification as analogy. Child Development, 58, 1013-1020.

Inagaki, K., & Sugiyama, K. (1988). Attributing human characteristics: Developmental changes in over- and under-attribution. Cognitive Development, 3, 55-70.

Leslie, A.M. (1994). ToMM, ToBY and Agency: Core architecture and domain specificity. In L.A. Hirschfeld & S.A. Gelman (Eds.), Mapping the Mind, pp. 119-148. New York: Cambridge.

Meltzoff, A. (1990). Foundations for developing a concept of self: the role of imitation in relating self to other and the value of social mirroring, social modeling, and self-practice in infancy. In D. Cicchetti & M. Beeghly (Eds.), The Self in Transition: Infancy to Childhood, pp. 139-164. Chicago: University of Chicago Press.

Meltzoff, A.N. (1995). Understanding the intentions of others: Re-enactment of intended acts by 18-month-old children. Developmental Psychology, 31, 838-850.

Meltzoff, A.N., & Moore, M.K., (1983). Newborn infants imitate adult facial gestures. Child Development, 54, 702-709.

Meltzoff, A.N., & Moore, M.K., (1997). Explaining facial imitation: A theoretical model. Early Development & Parenting, 6, 179-192.

Pascual-Leone, J. (1987). Organismic processes for neo-Piagetian theories: A dialectical causal account of cognitive development. International Journal of Psychology, 22, 531-570.

Pauen, S. (1996). Children's reasoning about the interaction of forces. Child Development, 67, 2728-2742.

Pauen, S., & Wilkening, F. (1996). Children's analogical reasoning about natural phenomena. Journal of Experimental Child Psychology, 67, 90-114.

Piaget, J., Montangero, J. & Billeter, J. (1977). Les correlats. In J. Piaget (ed.) L'Abstraction Reflechissante. Paris: Presses Universitaires de France.

Polya, G. (1957). How to Solve It. Princeton: University Press.

Ratterman, M.J., & Gentner, D. (1998). The effect of language on similarity: The use of relational labels improves young children’s performance in a mapping task. In K. Holyoak, D. Gentner & B. Kokinov (Eds)., Advances in Analogy Research: Integration of theory and data from the cognitive, computational and neural sciences, pp. 274-282. Bulgaria: New Bulgarian University.

Sigman, M., Cohen, S.E., Beckwith, L., Asarnow, R., & Parmelee, A.H. (1991). Continuity in cognitive abilities from infancy to 12 years of age. Cognitive Development, 6, 47-57.

Smith, L.B., & Heise, D. (1992). Perceptual similarity and conceptual structure. In B. Burns (Ed.), Advances in Psychology: Percepts, concepts and categories, pp. 233-272. Amsterdam: Elsevier.

Sternberg, R.J., & Nigro, G. (1980). Developmental patterns in the solution of verbal analogies. Child Development, 51, 27-38.

Strauss, S. (in press). Review of Goswami, U.: Analogical reasoning in Children. Cognition & Pragmatics.

Tomasello, M. (1990). Cultural transmission in the tool use and communicatory signalling of chimpanzees? In S. Parker & K. Gibson (Eds.), Language and intelligence in monkeys and apes: Comparative developmental perspectives (pp. 274-311). Cambridge: Cambridge University Press.

Visalberghi, E., & Fragaszy, D. (1990). Do monkeys ape? In S. Parker & K. Gibson (Eds.), Language and intelligence in monkeys and apes: Comparative developmental perspectives (pp. 247-273). Cambridge: Cambridge University Press.

Wagner, S., Winner, E., Cicchetti, D., & Gardner, H. (1981). Metaphorical mapping in human infants. Child Development, 52, 728-731.

Wellman, H.M., & Gelman, S.A. (1997). Knowledge acquisition in foundational domains. In D. Kuhn & R.S. Siegler (Eds), Handbook of Child Psychology, Volume 2: Cognition, Perception and Language, pp. 523-573 (4th Edition).

Younger, B.A., & Cohen, L.B. (1983). Infant perception of correlations among attributes. Child Development, 54, 858-867.


Figure Captions.

1. The gameboard (top row), analogy terms (middle row) and correct answer and distractors (bottom row) for the analogy bird:nest::dog:doghouse (Goswami & Brown, 1990).

2. Depiction of the problem scenarios used to study analogical reasoning in infants by Chen, Campbell & Polley, 1995.

3. Schematic depiction of the habituation events shown to the infants in the rational approach group by Gergely, Nadasdy, Csibra & Biro (1995), depicting (a) the expansion and contraction events and the first approach, and (b) the retreat, jump and eventual contact events.

4. Schematic depiction of the force table used by Pauen (1996).