A study investigated the usefulness of instructional materials on the writing of scientific articles by comparing the descriptions of abstracts offered in the textbook with a sample of abstracts drawn from four fields of medicine (clinical medicine, surgery, epidemiology, basic sciences). The comparison was confined to abstracts of results-focused papers, and papers were divided evenly between British and North American journals. Analysis focused on discourse features in three areas: purpose; methods and results; and conclusion. Results indicate close similarities between the textbook and the abstracts, but also show the textbook to be overly simplistic and rigid. The relationship between information structure and linguistic elements was found to be more complex than was implied in the book. Specifications for sections introducing purpose ignore some frequent lexical signals; methods tend to be amalgamated into adjacent sections; and it is not always necessary to conclude with a generalization. While the analysis detected no systematic differences between abstracts in British and North American journals, some evidence was found that conventions vary according to field or type of research design. Contains 16 references. (Author/MSE)