Empirical Lexile Measures for Words

MetaMetrics® has historically provided measures of text complexity for books, articles, and other texts using The Lexile® Framework for Reading. A text is made up of many words, which raises the question: can a single word have a Lexile® measure? Quantifying word difficulty on the Lexile scale would provide a variety of educational benefits. Example utilities could include identifying challenging words in a text and more precise word selection for both human- and machine-generated assessment items. This research brief describes recent efforts to find empirical measures of the complexity (difficulty) of words.

The research-based solution described in this brief met several requirements selected to increase applicability. First, empirical word measures are connected to the Lexile scale. That is, the word measures should have some meaningful relationship to the text measures and not be arbitrarily or independently scaled. Second, whereas an equivalent text complexity and student ability measure yields an expected 75% comprehension rate within the Rasch model as used within the Lexile Framework (Rasch, 1960), an equivalent text complexity, word difficulty, and student ability measure should also be anchored to a 75% comprehension rate in a modified Rasch model. Finally, when determining the empirical complexity of words, circularity should be avoided such that the word measure is determined independently of a student measure that resulted from that student’s performance on that particular word.

Data from EdSphere® (Hanlon et al, 2015), a personalized online learning platform was used in this analysis. The Lexile empirical word measures are used to generate targeted vocabulary lists in the Lexile “PowerV®” Word Selector (www.Lexile.com/powerv) and have also been used to develop a corpus-based measure for a more expansive set of words (Elmore, 2016).