Guidelines for text reduction were developed and evaluated to advance the art of manually preparing informative abstracts. The study was intended to: (1) develop guidelines that result in abstracts which provide maximal support to abstract-users, and (2) develop these guidelines so that they result in reliable, i.e., consistent, abstracts of scientific/technical material. The abstracting procedure, and Abstracting Form and associated instructions, produces reasonably consistent abstracts. An expert judge rated 13 subsections of six technical papers prepared by three different abstracters as 88% consistent, i.e., contained identical information. The abstracts prepared were a substantial reduction of the original text. Considering the six abstracts used in a performance test (judged as containing the most information, but not necessarily the longest) mean percentage reductions obtained were: 47% reduction of words, 28% reduction of figures, and 27% reduction of equations. Level of performance, as measured by accuracy on use-tests, supported by abstracts was equivalent to that supported by original text, regardless of test time restriction. However, total test time required was less using abstracts than with full text. (Author).