PURPOSE: To evaluate the quality of abstracts of original research articles. DESIGN: Blind, criterion-based survey. SAMPLE: Systematic sample of 33 abstracts of original research articles published in CMAJ in 1989. MEASUREMENT: The quality of abstracts was measured against a checklist of evaluation criteria, which were divided into eight categories. A score for each abstract was obtained by dividing the number of criteria present by the number applicable. The overall mean score was also determined. RESULTS: The overall mean score of abstract quality was 0.63 (standard deviation 0.13) out of 1. Of the abstracts reporting study design 56% did not include specific technical descriptors. About 52% did not explicitly describe the study variables. In describing subject selection 79% failed to use specific technical terms. Of the abstracts reporting results 66% did not provide appropriate supporting data. Of those that gave conclusions 86% did not address study limitations and 93% made no recommendations for future study. CONCLUSION: Most of the abstracts provided some information pertaining to each evaluation criterion but did not provide detail sufficient to enhance the reader's understanding of the article. On the basis of the study sample the abstracts need improvement in description of research design, reporting of subject selection and results, and statements of limitations and recommendations. The small sample from one journal and the absence of comparison between the contents of the abstracts and the contents of the articles were limitations. Future studies should address these issues and compare the quality of traditional and structured abstracts.