The most appropriate statistical model for the small-scale (n<100) studies common in language testing research is the Rasch one-parameter logistic model. The Rasch model provides a wide range of options for conducting research, refining existing examinations, and developing tailored (computerized adaptive) language tests. Three investigations using this analysis in testing for English as a second language (ESL) illustrate its usefulness. One looked for possible item bias in an ESL placement test battery, a second investigated the identification of inappropriate answers suggesting cheating, and a third involved the development of a computerized adaptive grammar and reading test. Several charts and a three-page list of references conclude the report. (MSE)