A descriptive study examined reading errors of emergent at-risk readers to determine which cueing system(s) they used initially, and whether or not there were changes over time in cues they used during reading. Running records of 30 first-grade children (participating in Reading Recovery programs at urban or suburban public elementary schools in five counties in California) taken during the first three months of school were analyzed for the types of cues used. Responses were coded as Accessing meaning (M), accessing structure (S), accessing visual aspects of print (V), meaning and structure (MS), structure and visual (SV), meaning and visual (MV), and meaning, structure, and visual (MSV). Results indicated that: (1) MS represented the majority of the cueing sources for substitutions for the first five lessons, and M, S, and MS represented virtually all of the substitutions made in all lessons; (2) the use of M, S, and MS increased as text level and difficulty increased between lessons 1 and 30; (3) the use of V information gradually increased between lessons 1 and 30; and (4) in lessons 10 through 30, V, MV, and MSV combined represented virtually all of the self-corrections made. Conclusions from the data are constrained by the reliability of running records and the small sample size. Findings suggest that, just as other children, at-risk emergent readers draw upon their knowledge of the world and their familiarity with language structure when they approach the initial task of learning how to read. (RS)