Gopnik, Alison and Andrew N. Meltzoff (1997) Words, Thoughts and Theories. Cambridge: MIT Press. 268 pages. $14 (paperback)
Reviewd by Laura Wagner, University of Massachusetts.
Gopnik and Meltzoff (G&M) have written a book that is full of ideas and information. They are interested in what our representations of the world look like and they are pushing the view they call the Theory Theory: that our representations are organized around theories. More than that, they argue that children's representations of the world are also organized around theories and that developmental changes children undergo are in fact changes in their theories. Even more than that, they argue that the theory revisions that children make are of precisely the same type as scientific theory revisions. That is, the difference between a 9 month old and an 18 month old is essentially the same as that between a Newtonian physicist and an Einsteinian physicist: they possess qualitatively different theories of how the world works. The use of ''words'' in the title refers to two claims that G&M make: one, that children's early vocabulary reflects technical notions in their theory du jour and two, that linguistic elements can act as part of the evidence children use in theory building.
The book is divided into three sections, in addition to the introduction and conclusion. The first section addresses the theoretical commitments of the Theory Theory and relates it other theories of cognitive organization and development. The second section examines three domains of knowledge for evidence of children's changing theories. The third section has a neo-Whorfian flavor and addresses the question of how different languages could effect children's theory-building.
Part 1: The Theory Theory
In this section, G&M face head on what is certainly the most difficult of their claims to swallow, namely that children's early theories about the world are qualitatively the same as scientific theories and subject to revision in the same ways. While acknowledging that there are many differences between the ways scientists and infants operate, they nevertheless maintain that they both construct the same sorts of theories in the same ways. The theories of both are abstract and general, and therefore provide a means for making predictions in a variety of domains. Both groups are concerned with the match of the theory to the world and will maintain a theory only as long as the evidence supports it. When evidence mounts against a theory, the theory holder goes into a transitional state, actively seeking out new sorts of evidence, conducting experiments and developing new technical vocabulary as they settle into a new and better theory. One of G&M's strong claims is that the standard adult theory (the so-called ''folk-physics'' and ''folk-psychology'', e.g.) is in principle open to revision; it was developed in response to inadequate child theories of the world, and, as with all scientific theories, it too will be revised, if the need arises.
The second chapter in this section compares the Theory Theory to two other prominent cognitive models, modularity and empirical generalizations (e.g., scripts, connectionism). G&M do not present the Theory Theory as a substitute for either of these models completely but as taking over some roles from each. Thus, with respect to modularity, the Theory Theory accounts well for many central cognitive processes and some of the less perceptually oriented modular processes; with respect to empirical generalizations, the Theory Theory is better equipped to deal with things like causal linkages and can provide a deeper level of explanation in general, but such generalizations may still form an important basis for the creation of theories.
Part 2: Evidence for the Theory Theory
Part 2 consists of three chapters which each take on a different knowledge domain and show how the Theory Theory can explain children's development in these areas. Chapter 4 deals with children's theories of appearances and addresses children's development with respect to object permanence. Chapter 5 takes on children's theories of action and causality with animate and inanimate things. Chapter 5 addresses children's developing theories of natural kinds. Each chapter discusses what the adult theory in each of these domains looks like and then marshals experimental evidence that demonstrates what the infant's initial theory looks like and what subsequent revisions to their theory the children make. In each domain, they argue that children make crucial theory changes around the ages of 9 months and again around the age of 18 months. G&M report the results from many experiments (a large number of them conducted by G and/or M themselves) which amply illustrate that children's competence in these domains changes (and improves) with age. This experimental evidence is at times augmented with anecdotal accounts of how infants affectively interact with different tasks at different ages. G&M are looking for scientific behavior from these children and it is therefore important not only that children can pass more tasks as they get older (and their theories get better) but that they view the tasks differently as they get older. They point to examples of emotional consternation that infants show around the time that their theories are putatively changing. Thus, a 15 month old will fail to find an object that has been secretly moved to a new hiding place but will be unperturbed by her failure; a 21 month old will find the object in the new place but will be unimpressed by her own success. An 18 month old, however, is both deeply disturbed when she fails (and by the structure of the task as a whole) but is joyful with success. G&M argue that such affective differences reflect something like the blase-ness of normal science compared to the eureka moments that mark a theory in change.
With respect to linguistics, one of the claims from these chapters is that early utterances may correspond to technical vocabulary of newly formed theories. G&M's arguments are particularly intriguing with respect to performative utterances, such as ''gone'' and ''uhoh'' which are quite frequent in the speech of 18 month olds but have been largely ignored in language acquisition studies. They claim that, for example, the emergence of the word ''gone'' in child speech is closely linked with children's ability to solve an invisible displacement task (when an object is secretly moved to a new hiding place). ''Gone'' indicates an object that is out of sight for any reason, and it is in effect a technical term, reflecting a concept in the infant's new theory.
Part 3: Language and Thought
In the previous chapters, the kinds of evidence that G&M suggested were driving theory formation and revision arose primarily from children's interactions with objects, actions and people in the world. In this chapter, G&M note that language itself is part of the child's world and different languages could in principle provide different kinds of evidence. Since the children's theories are not innately given but are presumed to be constructed on the basis of the evidence they get from the world, different kinds of linguistic evidence could lead to different theories, or at least to different rates of theory formation. They find support for the effects of language in theory building in a series of longitudinal, crosslinguistic studies looking at children acquiring Korean and English. They do not cover the experiments in detail, but the gist of them appears to be that Korean is more verb-centric while English is more noun-centric and this linguistic difference translates into Korean children developing their theories of action more quickly than their English counterparts, who for their turn, develop their theories of natural kinds more quickly than the Korean children.
Despite the fact that the Theory Theory informs an active psychological research program, this book is primarily a book of philosophy. Thus, the results from a large number of experimental investigations are discussed but no one experiment is discussed in sufficient detail to be critically evaluated. Any reader not already well acquainted with the cognitive developmental literature may find themselves somewhat confused (and perhaps somewhat unconvinced) about how G&M are able to draw the conclusions that they do from the experiments reported. This problem might have been remedied to some extent by detailed explanations of one or two experiments that could act as representative examples and the use of diagrams depicting the sorts of tasks used to test infants. A few more diagrams and/or summary tables would have been helpful in general. The Theory Theory is dynamic in form and G&M trace two or three theory changes within three different knowledge domains. The critical point about theory change comes through clearly, but many of the details get lost in the mass of information; a summary table describing the different theories (and perhaps the evidence that leads to theory revision) would have been a useful reference.
Although G&M devote an entire chapter to persuading the reader that infant theory development and scientific theory development are the same, ultimately this argument remains unconvincing and the insistence on this point throughout the book has the effect of lessening the force of their claims, rather than strengthening them. There are many reasons to be skeptical of the link between the two sets of theory builders but I think the most compelling reason to reject their argument is that it seems to do a grave injustice to the entrenchment of our folk-scientific reasoning. In a nutshell, our adult folk-science is a stable level of knowledge and may not, even in principle, be subject to revision. G&M try to argue against this powerful intuition in two ways. First, they point to the fact that science has moved beyond our folk theories, so those theories can't be the last word in what is true. While I agree that Einsteinian physics is probably more true (i.e., a better model of the world) than my own folk-physics, I'm still stuck with my inadequate theory. G&M appeal to the division of knowledge-labor to claim that the theory really has changed at a societal level, but that only highlights the fact that for the common person, our folk-science is unrevisable. G&M's second argument that our folk-science is open to revision is that we can imagine a world that would require a different theory, and they frequently invoke the TV show Star Trek as illustrating such a world. This fantasy argument leaves me entirely cold, partially because I have no strong intuitions (as G&M seem to) about, for example, the status of objects in a Star Trek transporter beam, and partially because I believe that they are placing far to heavy a burden on a piece of popular fiction. It should be noted, however, that rejecting the most extreme of G&M's claims does not fundamentally undermine their arguments that knowledge is organized around theories and that developmental change reflects changes in children's theories. Even if children's theories are not of the same type as scientific theories, analyzing our knowledge in terms of some kind of theory and children's development as some form of theory change is still a very intriguing and compelling perspective to take.
In sum, G&M have raised a number of extremely interesting questions about the nature of knowledge representation, the nature of theories, the nature of cognitive development and the relationship of language to cognitive development. They offer the Theory Theory as a way to address these questions and, although one might not believe that the Theory Theory actually has the answers, it is certainly an important view in the debate.
ABOUT THE REVIEWER:
ABOUT THE REVIEWER Reviewed by Laura Wagner, Department of Psychology, University of Massachusetts - Amherst. My research centers on children's acquisition of tense and aspect.