Respondents are often asked to complete difficult tasks in Web surveys, where the technology may facilitate the task. One example is prescription medications, with thousands of possible responses and complex names that are likely to produce misspellings and coding errors when entered into an open text field. We report on two experiments in which we evaluate alternative approaches to collecting such information, specifically using as drop box (select list) or a JavaScript lookup with type-ahead support.  We use a database of over 6,800 names (generic and brand names) of prescription medications available in the U.S., and test the alternative approaches on two surveys of opt-in panel members.  We compare the accuracy of the responses (in terms of the match to the database) and the effort (in terms of time taken) required.