Learning from educational texts: is reduction in global text cohesion a desirable difficulty as structure building skill increases?

One possibility for improving learning from text is to introduce “desirable difficulties” into the text. To do so, we manipulated the global cohesion of a textbook chapter and examined the possibility that individual differences in structure-building skill would interact with level of global cohesion. The interaction (not accounted for by working memory capacity) was significant: Those with lower structure-building skill performed poorly on educationally authentic comprehension questions after reading a low-cohesion (difficult) text, relative to a high-cohesion text. For more-skilled structure builders, comprehension was similar across text-cohesion levels; however, the low-cohesion text did not produce better comprehension for more-skilled structure builders. More-skilled structure builders appear adept at building relations across presented information, thereby accurately capturing interconnections among subtopics and ideas at a global level, even for texts that are less cohesive. Supporting this notion, the representations generated by learners became more closely aligned with experts' structures as structure-building skill increased.

Educational relevance and implications

To facilitate learning from texts, a standard recommendation in education is to construct a cohesive text that helps readers understand the interconnections among the concepts and ideas presented in the text (rather than as a list of facts). A provocative alternative suggested by basic theory and research is that in some cases a less cohesive version of the text can improve learning by prompting readers to more actively engage with the text. The present study decreased the global cohesion of a textbook chapter by removing the text signals (topic headers, bolded terms, and “take-home” text boxes) intended to assist the reader in understanding. We found that the presence/absence of these text signals did not affect the learning performance of readers who were relatively skilled in constructing a coherent understanding of things they read. By contrast, the absence of these text signals impaired the learning performance of readers who were less skilled at constructing a coherent understanding. The strong implication is that for students who struggle to form a coherent understanding of their reading assignments, it is especially important for textbooks to provide explicit signals that facilitate understanding.