The use of cognitively oriented models in research in reading comprehension requires the examination of the task demand characteristics of the research. Those cognitive processes involved in the performance which are due to the experimental procedures must be separated from those which operate in normal circumstances. "Control programs"--which direct individuals' attempts to perform a task--are created either as a result of trial-and-error experience or on the basis of instruction. Instructions are one obvious source of different demand characteristics, but control programs are also affected by such factors as ambiguity as to the knowledge and procedures needed for a task or by the discovery that efficient control strategies have not been created. Such factors may lead individuals to change or modify their strategies for performing the task. Every task must have a response component, and, when this response is not one made under normal conditions, the cognitive processes necessary to make the response may affect performance. Because of these and other demand characteristics, the tasks employed in research in reading comprehension must be examined closely before the results are extrapolated to the classroom. (AA)