For 2 years we followed lower-performing English learner (EL) and native English speaking (non-EL) students who participated in an efficacy trial of a supplemental first-grade code-oriented intervention implemented by paraeducators. At the end of grade three, across all students (n = 180 of the original 187 students), treatment effects were maintained on word reading (approximate d = 0.45), spelling (0.36) and reading comprehension (0.24). However, treatment effects tended to be smaller for EL students, and were significantly smaller for spelling in particular. While pretest grade one word reading did not moderate treatment response for either ELs or non-ELs, it was found to strongly predict all three end-of- grade-three outcomes, although to a lesser extent for ELs on reading comprehension. Findings add support to previous research on the benefits of early code-oriented tutoring.