This paper, a discussion of the use of written tests to assess second language proficiency and achievement, considers what constitutes a valid writing test task and addresses three questions: (1) To what extent is performance influenced by prior knowledge about the topic? (2) Does it make a difference how the writing task is specified on the test paper? and (3) Do different types of tasks produce significant differences in the performance of learners in a writing test? Three task types varying in the amount and nature of content material provided as part of task specification are distinguished: independent tasks, guided tasks, and experience tasks. A test developed at the English Language Institute of Victoria University (New Zealand) and administered to foreign students before they begin study consists of three tasks, one guided, one guided/experience, and one clearly an experience task. It is proposed that while classification by task type focuses attention on one important dimension of writing test tasks--namely, the extent to which they provide content support for test-takers--other factors in task validity must be considered, including the amount of guidance given concerning the quantity and form of text expected, the clarity of task definition, the choice of general versus field-specific topic, and assessment focus on product rather than the process of writing. Several approaches to increasing writing test validity are suggested. (MSE)