The purpose of this study is to contribute to efforts to improve methods for gathering and analyzing data from classroom observations in early literacy. The methodological approach addresses current problems of reliability and validity of classroom observations by taking into account differences in teachers' uses of instructional actions (e.g., modeling) in specific skill areas (e.g., fluency, reading comprehension). The findings from observations of second- and third-grade teachers' literacy instruction showed that teachers' instructional actions differed by literacy skill area and were more consistent within than across skill areas. Furthermore, teachers' uses of instructional actions in a given skill area were more strongly associated with students' gains in achievement in that skill area than were teachers' uses of actions across all skill areas. The approach offers significant improvements in methods to identify features of effective literacy instruction. (Contains 7 tables, 1 figure, and 1 note.)