In this study, we used Reading Recovery as the context to examine the relationship between three types of contingent teaching (temporal, instructional, and domain contingency) and student outcomes in a one-to-one literacy tutoring setting. We first created a National Teacher Effectiveness Index for all Reading Recovery teachers in the country and then used that index to identify two distinct groups of teachers from an existing data repository of 38 Reading Recovery teachers in training: those whose students had either lower average gain scores at the end of the school year (n = 6) or higher average gain scores (n = 4). We coded 1,199 teacher and student moves when they interacted at difficulty while reading a new book with teacher support. Results from hierarchical linear modeling analyses showed no significant differences for instructional contingency (the amount of information provided at difficulty), no main effects, and a nonsignificant interaction. Nor did we find statistically significant differences between the two groups of teachers for temporal contingency (the frequency of help). Dramatic differences existed, however, for domain contingency (what the teachers focused on when providing help). Specifically, teachers of students with higher outcomes had 8.3 greater odds of prompting students to use sources of information that they were neglecting while trying to decode a word. These findings provide needed information about the processes of scaffolding and have important implications not only for reading instruction and intervention but also for the ways in which scaffolding is studied.