There is a long-standing debate about how to characterize dialect use among speakers of African American English (AAE). This is a critical question for educational research, as students who speak AAE are at greater risk for difficulty in learning to read. We consider different methods of characterizing AAE dialect use from existing assessments and examine which methods best predict literacy achievement. Kindergarten and first grade students who speak AAE received two assessments of dialect use and two assessments of decoding at the beginning and end of the school year. Item-level analyses of the dialect-use assessments were used to compute measures of dialect usage: (1) an overall feature rate measure, (2) a subscore analysis based on items that pattern together; (3) an alternative assessment where children repeat and translate sentences; and (4) “repertoire” measures based on a categorical distinction of whether a child used a particular feature of Mainstream American English (MAE). There were significant decreases for all measures of AAE from fall to spring longitudinally and from kindergarten to first grade. Changes in the frequency-based measure, particularly its phonological sub-component, significantly predicted reading scores in the spring. Fall scores on multiple measures for grammatical differences also predicted reading measures in the spring.