SCHEME is a list-oriented programming language derived from LISP. It is a powerful, extensible language featuring first-class procedures, dynamic memory collection, and garbage collection.

Watson’s goal is to educate the reader in the power of SCHEME programming by means of artificial intelligence examples. This book is divided into nine chapters. The chapters cover an introduction to SCHEME, design for reuse, portability, design, an implementation of a neural network library, complex data structures, and AI programs for the games of chess and go.

As an introduction to SCHEME, Watson’s text is good: it is succinct and covers all of the main features of the language intelligently and accurately. Programming skills are undoubtedly required, so the book’s audience is undergraduates or interested professionals. No exercises are supplied.

Watson’s compact style does not, however, carry through to the remainder of the book. SCHEME and artificial intelligence are introduced through examples: a neural network library, a chess program, a go program, and others. Since the underlying principles are fairly complex, one would expect rigorous design documentation. However, Watson chooses to present the entire program text for each of the AI examples, with only a few pages of explanation. In some cases, more pages are given over to program output than to explanations. For example, the chess program consists of 4 pages of design, 15 pages of source code, and 12 pages of program output. This hardly promotes software reuse.

Watson’s book comes with two floppy disks that contain Windows 95 and NT environments for developing SCHEME programs, featuring the MIT compiler. The complete text of each program is available on each disk. I would rather have seen a book containing a good deal more on the principles of encoding AI algorithms using SCHEME, with pointers to the supplied code, rather than the reams of listings supplied.

Another problem with this book is in the presentation of listings. The code is annotated using SCHEME comments: a better approach would have used literate programming techniques to document the programs as text. Norman Ramsey’s excellent noweb is my recommendation.

Watson tries to do too much with this book--that is, to educate readers about both AI and SCHEME. The content is too slight to achieve this. The book is an excellent introduction to the features of SCHEME, but it would be difficult to use for learning about artificial intelligence techniques.