The overarching goal of this chapter is to provide a better understanding of the characteristics of text as they vary across genres and grade levels. We regard this as a first step toward creating reading assessments for which the nature of the passages is more fully understood. Research has shown that text cohesion is an important facet of text difficulty and that the effects of text cohesion depend on the readers’ domain knowledge and reading skill. The current study further investigates the importance of text cohesion and other signatures of text difficulty by examining the characteristics of a corpus of texts that vary in grade level and text genre. The results show that referential cohesion increases across grade levels. However, while lower-grade texts have lower referential cohesion, they have higher verb cohesion. This suggests that the smaller overlap in objects is compensated by larger overlap in actions. The results collectively point to a tradeoff between difficulty at the lexical and cohesion levels. Such a result suggests that it is insufficient to define text difficulty simply on the basis of word frequency, word length, and sentence length. Text difficulty is also a result of cohesion.

There are many issues to consider when developing measures to assess text comprehension. Most traditional reading comprehension assessments comprise brief passages that cover a relatively wide range of topics (eg, Nelson-Denny, Gates MacGinitie tests). This breadth of coverage is assumedly intended to reduce the effects of prior knowledge on performance. Choosing a readability range for the passages and then writing test items to the passage along some construct model (eg, main idea, detail, reasoning, etc.) is considered state-of-the-art in test design. The assumption is that text difficulty is a unidimensional construct, most often reflected by a readability