The author reviews Hole and Shaw's book detailing experiments with computer programs designed to find the best one-dimensional ordering of archaeological provenance unit assemblages. The single best order, or seriation, is commonly assumed to be a chronological order. Hole and Shaw discuss their computer program, which calculates several coefficients of similarity, and then go on to a review of seriation methods and theory. Most commonly used methods are based on a matrix of Brainerd-Robinson similarity coefficients. A number of positive aspects of the book are noted, including the clear, jargon-free discussion of statistics and computers. Goodness of fit measures indicate that best orders in frequency seriations are usually obtained with straight permutation searches. Presence-absence seriations are always worse that frequency seriations. The really surprising result of the experiments, not noticed by Hole and Shaw, is that, when the true chronological order is known, the best seriation does not generally coincide with that order. Cowgill points out that this result should make archaeologists skeptical about interpreting seriation results literally. He also suggests that Hole and Shaw and some other recent works had perhaps carried the frequency seriation technique about as far as it could go. (Abstractor's note: A. Gelfand (American Antiquity 36: 263-274) later discovered how to obtain the best one-dimensional order in any set of frequency data, thus making much of Hole and Shaw's discussion obsolete.)