An experiment with random assignment examined the effectiveness of a strategy to learn unfamiliar English vocabulary words during text reading. Lower socioeconomic status, language minority fifth graders (M = 10 years, 7 months; n = 62) silently read eight passages each focused on an unknown multi-syllabic word that was underlined, embedded in a meaningful context, defined, depicted, and repeated three times. Students were grouped by word reading ability, matched into pairs, and randomly assigned to one of two conditions. In the strategy condition, students orally pronounced the underlined words during silent reading. In the control condition, students penciled a check if they had seen the underlined words before but did not say the words aloud. Results of ANOVAs showed that the oral strategy enhanced vocabulary learning (ps less than 0.01), with poorer readers showing bigger effect sizes than better readers in remembering pronunciation-meaning associations and spellings of the words. In a second experiment, 32 fifth graders from the same school described the strategies they use when encountering unfamiliar words in context. Better readers reported more word-level strategies whereas poorer readers reported more text-based strategies. Our explanation is that application of the word-level strategy of decoding new words aloud strengthened connections between spellings, pronunciations, and meanings in memory compared to silent reading of new words, particularly among poor readers who were less skilled and less likely to use this strategy unless instructed to do so.