The use of readability principles in the ERIC guidelines for abstracting suggests that abstracts should be more readable than source documents. However, information theory suggests that the compression of information would result in less readable abstracts. Abstracts and their source documents were selected randomly from each of four clearinghouses also selected randomly. Sample size was determined by a calculation of power. The Flesch RE formula was used to calculate readability scores. Results of an analysis of variance indicated that abstracts were less readable than source documents but their mean grade levels of 11.7 to 12.7 were within the expected reading ability level of users of the ERIC system. (Author)