Abstract
Writing learnable instructional text is an art. How can we capture the art and transform it into a science? This paper describes a twinned passages method for capturing text art and demonstrates the method in experiments with 10 United States Army instructional texts. Twinned passages come in pairs: an original and a revised version, with equivalent content. The art has been applied to the revised version to improve its quality. The first step is to verify that the revised version is really better than the original on some meaningful measure; here retention tests were administered after a 24-hr delay. Results showed the revised versions were not forgotten over 24 hrs. For the originals, however, 24% of what had been remembered on an immediate test was forgotten over the 24-hr delay. The next step is to test various hypotheses for the causes of the improvement. Tested here were retrieval, storage, signaling, and readability hypotheses. Tests of other hypotheses are proposed. Several populations of twinned texts are identified that can be used for further research.