tively capture this distinction. Tests that are widely used today require students to answer questions that often assess only superficial aspects of comprehension. If deeper understanding of text is the goal of instruction, those tests will be inadequate to inform decisions about instructional effectiveness. Indeed, Box 2.5 suggests that inferencing ability and recall are different aspects of comprehension, and measuring recall only can be highly misleading. New assessments must therefore be a high priority.

Since comprehension is not a unitary process, it is necessary to assess separately the different components of comprehension. Just how many independent components exist is a matter of some disagreement (Pearson and Hamm, 2002). Rigorous research to push further on a working answer to that question is under way and should be continued and extended. Hannon and Daneman (2001), for example, designed a comprehensive test that measures four different components of comprehension: a reader’s ability to recall a text, the ability to make inferences based on explicitly stated facts, the ability to access general word knowledge, and the ability to make inferences that require integration of prior knowledge with text information. These four components proved to be good predictors of performance on a variety of comprehension tasks. Kintsch et al. (2001) pursued a similar goal when they assessed separately how well people could reproduce a text and how well they could answer simple problem-solving (inference) tasks for which information in that text is required.

Importantly, promising initial developments must be followed through to ensure both their validity and their practicality. Evaluation of internal construct validity examines how well the assessment explains comprehension performance in com-

BOX 2.5
Measuring Recall Alone Does Not Measure Comprehension

In a study by Mannes and Kintsch (1987), students read one of two versions of a chapter: one was well organized and explicit; the other was slightly disorganized and left some things unsaid. When asked to recall the chapter, the well-organized version produced 25 percent more recall. However, when understanding was tested by inference questions, the less explicit version was better by 75 percent. Making readers draw their own inferences when studying had its benefits, but if measured by a test that merely required them to reproduce the text, the reverse would appear true.

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement