Coh-Metrix: Providing multilevel analyses of text characteristics

Graesser, Arthur C., McNamara, Danielle S., & Kulikowich, Jonna M. (2011). Coh-Metrix: Providing multilevel analyses of text characteristics. Educational Researcher, 40(5), 223-234.

Here is cutting-edge research on what we used to call readability or text difficulty. Researchers like the authors of this article are now thinking more in terms of text characteristics. The new system described here, called Coh-Metrix (Coh stands for cohesion) makes traditional measures of text difficulty look like anachronisms. Although the old readability formulas like the Fry Graph and the Flesch-Kincaid measure broke ground in their day, and though they still allow us to make some inferences that help us match readers with texts, they only looked at a few text characteristics like sentence length and word length, and sometimes, as with what used to be considered the more “sophisticated” formulas like the Harris-Jacobson formula, measures of word frequency were used. These formulas were helpful in that they could reveal characteristics that tended to make texts more difficult, but anyone who has ever worked with readability formulas knows, they don’t tell the whole story about texts, and are not an exact science. Compared to those formulas, Coh-Metrix, which looked at 53 measures that then were statistically distilled into eight dimensions, is a much more sophisticated and sensitive tool. In this article, the authors specifically discuss four of the eight dimensions: Narrativity, Syntactic Simplicity, Word Concreteness, and Causal Cohesion. Even though there are portions of the article that are highly technical, the descriptions of the text dimensions and their logical relationships to text difficulty are clear, accessible, and will make sense to most educators.

One thing that stands out here is how complex texts are, and how many things can determine whether a text is easy, difficult, or “just right” for a given reader. Coh-Metrix provides a bridge between the quantitative traditional measures and the qualitative, descriptive rating systems that have been used to augment them. What fascinates me about this line of research is the ways that factors one might consider entirely qualitative are being represented by multiple quantitative measures. The old readability formulas did some of that, but now with advances in technology, texts can be minutely examined on many factors at once. The research described here applied the Coh-Metrix tool to 37,520 texts, which the authors say represent the range of texts a typical high school senior would have encountered in his or her entire K-12 career! We can do so much more now than could be done in the days when Ed Fry developed his readability graph.
Something that really resonated with me here was the discussion of how a tool like Coh-Metrix could be used to provide a balanced “diet” of texts for young readers, tailored specifically to their reading needs. According to the authors, young people need some mostly “just right” texts, but they also need some easy texts to build fluency, efficacy, and confidence, and some difficult texts to provide an opportunity to “stretch.” That view fits with my own about how we should be using measures of text difficulty to help learners find texts. It is refreshing to hear this view, in contrast to some of the more rigid “leveling” approaches I’ve witnessed lately. Overall, this was an important development and a key article for any literacy educator to read, and I plan to look further into Coh-Metrix.