Safe Takeoffs—Soft Landings

The richness of life’s possibilities creates a paradox. It poses fundamental limitations on the ability of an organism to reason and act intelligently, as information in its environment is often ambiguous and consistent with an infinite number of interpretations. At the same time, interpretation of this information is required for intelligent action. To further complicate the situation, many problems that are solvable in principle are intractable in practice. As an example, the number of winning sequences of moves in a chess game is finite. Yet determining these sequences from all possible sequences is impractical for the chess player, as there are more possible sequences than atoms in the universe. Cognitive scientists are certainly familiar with underdetermination and computational complexity in association with work on language learnability theory (e.g., Gold, 1967; Wexler & Cullicover, 1980), but complexity problems extend to virtually every area of cognition from perception to concept learning to decision making. If organisms cannot, and therefore do not, examine all possibilities in some cognitive task, they must. be “prepared” or biased to learn some things rather than others, to draw some plausible inferences rather than others, and in general to favor some possibilities at the expense of others. As evidenced by the present set of exemplary papers by Brown, Gelman, Keil, Markman, Newport, and Spelke, psychologists are increasingly drawing on analyses of constraints or “guiding principles” as a framework for their research. All of the present authors are interested in development, and it is no accident that developmental psychologists have been leaders in recognizing the importance of constraints. Without guiding principles of some form it simply is unclear how the learning process would ever get off the ground. The purpose of this comment is to underline some of the important points and issues growing out of developmental work on constraints as reflected in the contributed papers. We first discuss a couple of distinctions and definitions associated with research on constraints, then shift attention to the per-