Eight fifth grade students participated in a study that investigated the effects of test context on test performance and subjects' perceptions of tests in varying contexts. In the first three contexts, reading of words in isolation was tested using materials drawn from three different informal reading inventories equivalent in difficulty level. The words were printed on cards and shown to each child. In context one, there were no prompts; in context two, the child was able to listen to his or her reading of the words and choose the ones thought correct; in context three, the child read another set of words, and was prompted for those not identified. In the remaining three contexts, oral reading of connected discourse was tested using materials compiled from informal reading inventory passages. Context four used comprehension question prompts; context five allowed subjects to scan the passages before reading; and context six used another set of passages and involved discussion of the passage topics before reading. Results showed that for most of the subjects, freedom to ask questions and to receive prompts increased the range of their performance markedly. In addition, six of the subjects thought they had done well in context one, while none though he or she had done well in contexts two or three. Three thought they had performed well in context four, and all subjects perceived themselves to have performed poorly in contexts five and six. The results provide support for dynamic assessment contexts that invite subjects to use language in more natural ways than traditional test situations allow. (A copy of a test protocol is appended.) (FL)