Scripts, plans, goals, and understanding: an inquiry into human knowledge structures By Roger C. Schank and Robert P. Abelson (review)

Abstract

BOOK NOTICES 779 purposes, because her 'test is not commercially available' but has to be constructed by the clinician from her description (p. 119). Theory is foreign to this book. No controversial issues are raised ; no past controversies that have been resolved by observing the linguistic behavior of children are mentioned. This tone is set by Sylvia Richardson's foreword, where she observes that the 'heated controversy among several disciplines' regarding ' remedial programs for children with language and reading disabilities ... is [not] helpful to the advancement of knowledge' (p. iii). Nonetheless, the linguist who wants to learn about recognized clinical disabilities and who wants a guide to available clinical diagnostic tests will profit from reading this book. [Thomas E. Armbruster, University of California, Irvine.] Scripts, plans, goals, and understanding: an inquiry into human knowledge structures. By Roger C. Schank and Robert P. Abelson. Hillsdale, NJ: Lawrence Erlbaum Associates, 1977. Pp. 248. Schank (a specialist in artificial intelligence studies) and Abelson (a social psychologist) here describe the conceptual framework concerning human thought and language which has allowed them to organize and integrate their various separate computer simulations of human cognitive processes, and to make each more sophisticated and realistic than it would otherwise have been. The problem with their book is that it is about the conceptual framework rather than about the actual simulations. They treat the conceptual framework as if it described human thought— although, in fact, when taken in isolation from the programs (as it is here), it becomes only another abstract metaphor for how certain conceptual processes might work. The authors' discussion of human productions (such as story tellings, descriptions of events, and questions) contains dangerous oversimplifications of word meanings, contextual effects, communicative intents, syntactic structure, and conceptual development. More attention to related fields would have alerted the authors to many of these problems and caused them to project claims from their computer programs onto their child subject (Schank's daughter Hana) with considerably more care than was in fact the case. Labels of operations from their programs, such as MTRANS, PTRANS, D-PROX, and LOC, after being only minimally defined, are used without paraphrase or further description throughout the book in discussions of computer and human activities. This practice exacerbates rather than eases the reader's problem of trying to understand and apply the work. It is important for S & A's own work that they have such a conceptual framework; however, the framework is not of much general interest until it is spelled out in sufficient detail so that others can use it for their own research, or until it shows itself able to generate a wide range of useful problems and solutions. A more useful book would have been one that told us enough about S & A's actual computer simulations (which seem impressive) so that we could each independently assess the implications of the simulation work for our own cognitive and linguistic concerns. The results of S & A's work imply that their simulations embody real insights into how we understand verbal accounts of experienced events (such as restaurant visits or trips), how we construct questions and answers about these events, and how we reason about their behavioral constituents. The insights appear compatible with a variety of theoretical approaches. At least one of S & A's particular findings—that it is much easier to write a program that 'understands' a story (and that deals effectively with the syntax and semantics of the story) when the program has knowledge of the pragmatic contexts in which the story is told, and in which the story takes place— should be of interest to syntactic and semantic theorists, as well as more directly to those involved in machine translation work. It is regrettable that this and other such findings are not discussed in more detail. [David B. Kronenfeld, University of California, Riverside .] ...