Considering the wide variety of statistical software packages available for data analysis, how important is it for today's practicing statistician to understand the details of programming in low level languages like C or FORTRAN? In Numerical Methods of Statistics, (John F. Monahan, ISBN 0-521-79168-5, Cabridge University Press) this question is answered by a balanced presentation of topics ranging from computer arithmetic to Markov chain Monte Carlo methods. The author explains that higher level statistical software packages have usually been rigorously tested and produce accurate computations, but when "pushed to their limits" for research purposes, time and memory use becomes inefficient, even with the increasing availability of computing resources. At that point, the statistician who needs to implement new or modified methods requires knowledge of numerical techniques to develop new software or to optimize the needed building blocks of existing software. In this text, algorithms are presented as pseudocode, along with examples of FORTRAN code, and a disk with programs and demonstrations is included inside the back cover. In summary, this textbook seems ideally suited for its intended purpose of providing graduate students with the knowledge necessary to write programs I to implement new methods in statistics. To use the examples and demonstration software requires prerequisite knowledge and the ability to compile and execute FORTRAN program code. Object-oriented programming is mentioned but generally ignored in this textbook. While the emphasis is on low-level general-purpose computer languages, some additional consideration of higher- level interpreted languages, considering their more widespread use, would have been a useful addition. Otherwise, this textbook would be a good choice for teaching statistical computing.