In recent years, the Education for All movement has focused more intensely on the quality of education, rather than simply provision. Many recent and current education quality interventions focus on literacy, which is the core skill required for further academic success. Despite this focus on the quality of literacy instruction in developing countries, little rigorous research has been conducted on critical issues of assessment. This analysis, which uses data from the Primary Math and Reading Initiative (PRIMR) in Kenya, aims to begin filling this gap by addressing a key assessment issue--should literacy assessments in Kenya be administered orally or silently? The authors compared second-grade students' scores on oral and silent reading tasks of the Early Grade Reading Assessment (EGRA) in Kiswahili and English, and found no statistically significant differences in either language. They did, however, find oral reading rates to be more strongly related to reading comprehension scores. Oral assessment has another benefit for programme evaluators--it allows for the collection of data on student errors, and therefore the calculation of words read correctly per minute, as opposed to simply words read per minute. The authors therefore recommend that, in Kenya and in similar contexts, student reading fluency be assessed via oral rather than silent assessment.