Discussion

The existence of considerable variability in the routes by which Nigerian children change their minds about dream-events is not surprising. As far as we know no study has ever shown a single invariant sequence of changes in dream-event understandings. Kohlberg (1966, 1969) drops between 30% and 40% of "aberrant" subjects from his analysis of transitional invariance, and Laurendeau and Pinard (1962:103-104, 114) explicitly mention the considerable variation observed in the transitional period between "realism" and "subjectivism." This period includes "all the possible steps between the total reification of the dream, attenuated by unskillful attempts at interiorization, on the one hand, and the almost complete subjectivation of the phenomenon together with a residual expression of realism on the other," for example, the form of understanding in which the dream is understood to be "like a story, like a little play" happening inside the child's head, potentially visible to anyone who could open the head to have a look (is this an internal perception reminiscent of our Hausa subjects or a shared fantasy?). What is surprising is how little theoretical interest has been taken in this undeniable variability. We believe any satisfactory account of the evolution of dream-event understandings must render this variability intelligible not invisible.

Throughout this essay we have been skeptical of the view that

changes in children's understandings of dream-events must follow a single and logically determined order. The principle of the "doctrine of invariant sequence" ("if there is logical priority there must be temporal precedence") and Kohlberg's analysis of the "inner logic of the concept of reality" both seem questionable to us. Documentation of transitional variability in dream-event understandings among the Hausa makes us that much more doubtful. But, so does the initial "realism" of children from diverse cultures.

How are we to interpret the agreement between six-year-old Hausa children and four-year-old American children that the events in their dreams are external perceptions (real, capable of public perception, and externally located)? This seems to us remarkable; the agreement is not made intelligible by suggesting, as do cognitive-developmentalists, that these children lack relevant distinctions that can only appear in a logically determined, nested order. For if six-year-old Hausa children did not distinguish real/unreal, public/private, external/internal, one would expect free variation in their responses; they should fail to comprehend the distinction and thus randomly produce responses. Cognitive-developmental theory fails to explain how it is that lacking, for example, the distinction between real and unreal, young children should understand dream experiences as characteristically any one pole of an absent distinction, and why, given that they do respond so characteristically, it is that one pole and not the other that is preferred. Either young children understand their dream experiences and everyday waking perceptions to be neither real nor unreal (which tells us very little about how they do understand them but is at least consistent with the notion of "lacking a distinction") or, as we think more likely, they distinguish by some criteria the real and unreal but have no good reasons to view dream-events as anything but real.

We believe these "good reasons" for applying one form of understanding and not another, these criteria for evaluating the differential applicability of concepts to experience have received too little attention in the study of cognitive development. Toulmin's philosophical investigations of "rationality" (1971b , 1972) are germane in this regard. He argues that an account of understanding has less to do with the creation of logical systems, the avoidance of inferential errors, and the formally coherent interdefinition of con-

cepts, and more to do with "the manner and circumstances in which a person is prepared to change or modify his ideas," "the criteria of reference by which choices are made between rival ways of understanding." He believes that concepts and forms of understanding most typically co-exist in conceptual populations or aggregates (i.e. they are very loosely integrated and more often independent) from which choices among potentially relevant members are made with reference to very ungeneralized (local) intellectual problems, contexts, and criteria of relevance.

With regard to dream events, it seems to us the choice among rival forms of understanding is subject to diverse constraints. All knowledge available to the organism, whether preadapted in the evolutionary history of the species or the cultural history of the group, or postadapted in the life history of the individual, begins with external perception. There may well be a preparation of homo sapiens to understand experiences as external perceptions until such an understanding is shown to be deficient. The informational conditions under which such an evaluation of "deficiency" is likely to be made include the instability over authoritative observers of the effect associated with a phenomenon, that is, its lack of reliability and the consequential difficulty in finding consensual validation for one's judgments about experience (see Jones and Nisbett 1972, and Kelley 1967). Certain other kinds of understandings may be differentially selected against or amplified by one's culture. Schooler and Caudill (1964) for example, in a comparative study of symptoms among schizophrenics in Japanese and American mental hospitals, have commented upon the significantly higher incidence of "hallucinations" among those committed as insane in America, a fact they hypothesize may be related to American cultural concerns about "clear-sightedness" and the "accurate perception of reality." Japanese may be able to have "hallucinations" without being quite so readily committed as "mad."

We are raising as an issue the question of the explanatory adequacy of a particular form of understanding in the face of certain kinds of evidence (some of which is and some of which is not culturally variable), certain criteria for what counts as relevant evidence and certain standards for what counts as an adequate explanation (again some of which are and some of which are not culturally variable). For example, certain kinds of evidence are made more

intelligible by certain kinds of understandings. In the face of knowledge that dream-events are not perceived by those who should have been in a position to perceive them, certain understandings of the dream-events (e.g. they are mirages) are simply inadequate.

Our position with regard to children's changing understandings of dream-events is as follows:

Children come to know what it is they mean by any particular form of understanding, and come to reject their initial understanding of dream-events piecemeal and in relation (among other things) to evidence that their way of construing dream-events is inadequate. The specific orderings of changed understandings that do occur reflect the children's differential access to evidence disconfirming their working hypotheses. Some kinds of disconfirming evidence may ultimately be presented to children in all cultures, but at the very least there is no reason to believe all relevant disconfirming evidence is universally presented to children in the same order either within or across cultures. Cross-cultural or intracultural variability in the order in which disconfirming evidence is presented to children may have no influence on the understanding of dreams ultimately arrived at, but such variability in the order of evidence presentation will be related to the routes by which children achieve this understanding.

In summary, the understanding that dream-events are fantasies may be the most adequate understanding "spontaneously" available to the child in the face of certain universal facts about waking experience, that is, untutored in the entailments of adult dream concepts and the subtleties of their application to everyday experience.^[3] But the child's sequence of understandings in arriving at this relatively more adequate of childhood understandings depends heavily on the order in which these overwhelming facts of everyday experience (e.g. the sister who did not drown in the well) are encountered.

Our position places primary emphasis on the adequacy of childhood understandings given access to certain kinds of evidence. As far as we know, no study has ever shown a single invariant se-

[3] There is no reason to view ultimate childhood understandings of dream-events as adequate in any absolute sense as anyone must realize who ponders how mental phenomena (which by ultimate childhood definition are inherently private and lack extention in space) can be located anywhere (whether internally or externally).

quence of changes in dream-event understandings, but if such an invariant sequence had been reliably documented we would have looked for an invariance in the order in which evidence disconfirming of one aspect or another of children's initial dream-event understandings had become available to them, and not for a logically necessary overall direction to conceptual change.

At this time we are not in a position to carry our general observations any further. In the context of our critique, however, the existence of alternative ways of changing one's mind about dream-events among Hausa children, and, as discussed above, its possible relationship to aspects of these children's culturally influenced experience (i.e., Hausa sleeping arrangements) should, at the very least, (1) disenchant us from too heavy a reliance on a logical analysis in an account of children's changing understandings of concepts, (2) emphasize the importance of careful studies of these changing understandings in relation to relevant evidence encountered by a child from his cultural as well as noncultural experience, and (3) encourage studies of cross-cultural variations (or lack of variation) in the criteria on the basis of which children change their minds.

William James put nicely one the themes of this essay and we will let him have the last word (as quoted by Durkheim 1960:409).

"the enormously rapid multiplication of theories in these latter days has well-nigh upset the notion of any one of them being a more literally objective kind of thing than another. There are so many geometries, so many logics, so many physical and chemical hypotheses, so many classifications, each one of them good for so much and yet not good for everything, that the notion that even the truest formula may be a human device and not a literal transcript has dawned upon us."