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

What Does It Mean to Learn From Texts?

Charles Perfetti, who leads this line of inquiry, conducted a decade or so of early work on reading ability, in which he identified the main components as word recognition, immediate memory of what was read, and related abilities to track textual sequences and syntax. He found that readers comprehend text successfully to the degree that these abilities are automatic and do not interfere with concentration. He concluded that "students who don't learn from reading don't have full control over these basic processes."

The fairly low-level skills--not "learning skills" but basic reading skills--are important in text learning but are not sufficient. These skills allow a reader to identify words and figure out the meanings of sentences, but they do not provide the connections to the subject matter. The question became what else is needed to learn from texts. This "what else?" question marked a shift in Perfetti's research from basic abilities to a concern with higher-order processes--"from asking how these basic processes work to asking how they connect with learning and whether that connection is different in different subject matters." Part of the answer to the question was obvious, he says. Beyond good reading skills, students in every subject need good texts, clearly and coherently written. They also need enough background knowledge in the subject to support a connection with what a text contains. If a student already knows what a text is about, Perfetti says, "he can read one sentence after another in a great flow, and you don't see any problem. But you show that same text to someone who doesn't have the background knowledge, and you'll suddenly see how implicit this text is--how much information is left out, or left up to the reader to fill in." Thus, reading is always and inevitably a matter of drawing inferences, a process that is made far easier if the reader has a reservoir of related knowledge from which to fill textual gaps.

Perfetti undertook text-related studies on the effects of background knowledge and reading skill in history, science, and even football. "In one study," he says, "we showed that students who did not have much knowledge about football could understand a story about it pretty well if they had good reading skills--but not if they were poor readers." A conclusion Perfetti drew from this held even across later research in history and science: "We found that if you had both knowledge and skills, your comprehension was very good. But if you had reading skills without background knowledge or if you had background knowledge without reading skills--either way--then comprehension was not so good." Nevertheless, Perfetti points out, it does not necessarily follow that content knowledge and skill are equally important to learning. Both have limits. "If someone gives me a technical article on subatomic physics, I'll have trouble understanding it no matter how skilled I am at reading. Ability can compensate only so far for a lack of knowledge. And vice versa. If you know a lot about a subject but can't read well, a text on that subject will remain to some extent incomprehensible. Texts are always going to compromise between leaving wide gaps that a reader has to fill with background knowledge and supplying every piece of information a reader needs to understand it."

With the limits of both skill and background knowledge in mind, Perfetti turned to studies of the differences in learning from science and history texts. "We chose the domains to be dramatically different," he says. "The physics text, for instance, was written in flat, declarative sentences describing the structure of the atom, the behavior of particles, and so on. If you do a content analysis, you find that this particular science text has a logical, hierarchical kind of structure. The history text, which is about the United States' acquisition of the Panama Canal, takes a narrative form, describing many causal relationships between events."

As in the football study, Perfetti found both knowledge and skill important to readers understanding of these texts. It was clear that basic reading ability, as Perfetti defined it in his earlier work, was a general skill, one that crossed the boundaries of subject matter and predicted good learning in science as well as history. In fact, it may have been more important in science, which is more difficult to read because it lacks the narrative flow of history and introduces more unfamiliar vocabulary and abstract concepts. The next question was whether some higher-order skill, closely related to basic reading ability, might also generalize across subject matters. "One candidate is the ability to make inferences from a text," Perfetti says. "Texts are more implicit than they are explicit. That is, they always demand that you provide some knowledge to fill in the gaps. So maybe drawing inferences, or understanding the implied content of texts, contributes to learning across domains."

After some studies, however, Perfetti is skeptical. Although certain inferences can be made directly from text--for example, the inference that a given pronoun refers to a given antecedent--"making inferences to fill in actual gaps in the text is very dependent upon knowledge. You can't do it if you don t have the background information. So at that level, inference-making is not likely to be a completely generalizable skill. . . . I guess the story continues to be that in order to learn things from texts, students have to have a certain amount of background knowledge."

Perfetti thinks that one reason it is so difficult to identify higher-order learning skills that apply across subject matters may be that texts always remain, to some extent, abstract. "Texts are not a very good way to make concrete the basic concepts that you want to communicate. You want students to be able to see what happens to an equation as certain values increase. That's where an interactive computer laboratory, for example, is really, really important. I would be curious to know whether any kind of reading ability that we measure predicts performance in deep science learning of the kind that hands-on experiments can generate."

What Perfetti is suggesting here obviously relates to the similar problem in mathematics instruction of linking concrete referents to abstract mathematical symbols and calculations. Just as that connection in math can be forged with the help of manipulatives, real-life story problems, and demonstrations, Perfetti believes a similar approach might be useful in science. He speculates that texts might contribute to a deeper scientific understanding if, for example, they described the activities of scientists more often. "I'm not convinced it's as important for the child to read some flat description of the atom as it would be to read a paragraph on the excitement of scientific discovery and how people have actually learned what the atom is like. That's the kind of thing that could be put in a narrative structure. Then the child could go to a lab to watch hypothetical models of the atom--see the electrons spinning, things like that."

Returning to the focus of his research, Perfetti says it has contributed "a clearer picture of how the trading relationship goes between basic ability and basic knowledge." He stresses that "to have identified both of these as important in learning does not mean that texts should resort to independent instruction on basic ability and basic knowledge." Instead, Perfetti recommends building both background knowledge and plenty of reading practice into classroom instruction. "Good instructional practice is aware of this," he says. "I think it's always been part of what good teachers do--assess background knowledge before beginning a lesson. Engage in a bit of dialog about relevant concepts. See what level of existing information emerges. Examine the text. Make sure it's coherent--that one sentence builds carefully on another, that you never have more than one knowledge gap between sentences, that the text is vivid and narrative."

This is a tall order, as Perfetti knows. Few textbooks in widespread use are able to fill it. Hardly any are "vivid," and most fail the coherence test. Another line of work at NRCSL has therefore raised the question whether it is possible for readers to develop learning strategies that compensate for the inadequacies of texts.
-###-

[Thinking Through Language]

[Reading, Comprehension, and Self-Explanation]