A r c h i v e d  I n f o r m a t i o n

Toward a New Science of Instruction: Programmatic Investigations in Cognitive Science and Education--August 1993

Reading, Comprehension, and Self-Explanation

Evidence that good learners can draw more inferences from their reading than the straightforward, text-based inferences described by Perfetti arose in the course of a larger investigation into the nature of background knowledge and knowledge construction in two science subjects, physics and biology. A considerable body of research had already documented how difficult it was for instruction to remove students' common, deeply held yet erroneous preconceptions about physical science phenomena; but less had been written or understood about initial misconceptions in biology. In order to test whether and how biology misconceptions might differ from those in physics, and whether they might also be easier to remove, Michelene Chi conducted empirical studies of eighth-grade students learning from a text on the human circulatory system. Questions motivating the studies included: What is the nature of the knowledge structures students build as they learn? What causes students to misunderstand key concepts in a text on a given topic? Do misunderstandings arise out of students erroneous background knowledge? Can instruction help students to reshape their existing knowledge structures?

Chi chose to study learning about the circulatory system for several reasons. Not only had the topic been identified as one of the five most important ones to be learned in biology, but, from an instructional perspective, it was a complicated topic to understand and one in which students were likely to have varied and unpredictable misconceptions. The topic also offered an opportunity to examine closely the causal reasoning involved in learning about the complex relations among circulatory system components and their functions. Finally, such a causally integrated system permitted researchers to locate points at which misunderstandings occur and to learn more about the nature of these misunderstandings.

Researchers examined eight textbooks that covered the circulatory system and created a composite set of text materials for students to read. Text analyses had revealed that all the texts contained numerous inadequacies, and even the most highly rated texts failed to explain the ways in which the behavior of local mechanisms of the circulatory system--that is, sets of interacting components such as cells, cell density, and membranes--served the functioning of the circulatory system--as a whole. Another major textual fault was the failure to fully specify the functions of individual components of the circulatory system, a gap that made it very difficult to grasp the causal structure of the system.

According to Chi, when texts are seriously inadequate, "there are two alternative routes one can take to improve instruction. The most obvious one is to modify the text, but this seems overly difficult since there are so many texts out there." Furthermore, even though research data strongly indicate that incorrect intuitions about biology can be removed if texts directly contradict them and supply accurate information instead, "the problem is, you don't know which intuitions a given student will have. They're not predictable." Chi decided to pursue the second, less obvious alternative for improving instruction when she began to see that some students who were reading the inadequate circulatory system texts were--contrary to expectations that Perfetti's work might create--correctly inferring the functioning of system components even when the text did not supply that information either directly or indirectly.

Instead of attempting to analyze and revise texts, therefore, Chi began to investigate whether these successful students were applying cognitive skills that compensated for problematic texts. "We started to ask, 'How can students who hold incorrect mental models of the circulatory system or some of its parts be taught to construct accurate mental models from particular, inadequate texts?'" As the work progressed, and researchers took detailed, sentence-by-sentence protocols from students reading the troublesome biology text, evidence emerged that the readers who were able to induce the function of circulatory system components without any textual explanation were generating accurate self-explanations as they read. Researchers gave all the students the task of monitoring their understanding after each sentence they read, but, says Chi, "Monitoring in general is just an assessment of understanding. It is not the same as generating a piece of knowledge. Self-explanation generates pieces of knowledge."

As an example, Chi cites two self-explanations that one student generated after reading the sentence: These substances (including vitamins, minerals, amino acids and glucose) are absorbed from the digestive system and transported to the cells. The first fact that the student was able to induce, though the text did not directly state it, was, "Okay, so that's the point of what hepatic portal circulation is. To pick up these nutrients . . . ." The second self-explanation was prompted by the researcher's question, "Okay, why would you say that?" The student responded, "Well, because it says that it's absorbed from the digestive system--um, vitamins, minerals, and amino acids and glucose --and so that's why it's important to eat a balanced diet, or else your cells won t get the right vitamins, minerals, and amino acids and glucose." Both self-explanations--regarding the purpose of hepatic portal circulation and the importance of diet to the health of cells--compensated for gaps in the text's explanations.

Overall, Chi found that this process of self-explanation in successful students involved the ability to draw numerous low-level inferences that, cumulatively, led to the higher-level induction of circulatory-system components--in other words, the generation of that new piece of knowledge. Chi's most recent work identifies four kinds of low-level inferences that are necessary to self-explanations. These inferences link textual content with (1) common-sense knowledge, (2) episodic personal experience, (3) prior information from reading preceding sentences in the text, and (4) background knowledge in the topic. Chi stresses that these are "minute inferences" and that self-explanation is likewise "small." "It does not have to be complete at any level. Instead it builds up a network of understanding and allows knowledge to integrate."

Chi assumes that nearly all readers are able to make text-based, or syntactical, inferences of the kind described by Perfetti. But she believes that, in order to understand a broad and complex topic like the circulatory system they must also make these other, incremental kinds of inferences, drawing them from many sources. The eighth-graders in her studies were able to do so in response to "high-level prompts" from the researchers, questions that elicited the right kinds of inferences by "making the students work harder and prodding them to study the text material in a deeper way." Chi believes students can be taught to give their own high-level prompts and thereby to generate knowledge through this process of drawing several kinds of inferences. It is a possibility she is currently investigating. If self-explanation does prove teachable, it would readily cross the boundaries of subject matters and would therefore be a higher-order general skill of the kind Charles Perfetti has characterized as so difficult to find. Its implications for instruction would therefore be very great.
-###-

[What Does It Mean to Learn From Texts?] [Table of Contents] [Reading As Reasoning]