This quasi-experimental study was conducted in 30 classrooms in three urban middle schools and involved 495 students in grades 6, 7, and 8. One-third of the students received a monolingual English version of the cognate curriculum; one-third received a cross-linguistic version that made explicit connections across English and Spanish; and one-third served as a control group. Students were administered an English vocabulary assessment (EVOCA), that includes taught cognates, non-taught cognates, and other non-taught words. The Test of Reading Comprehension (TORC) and the Test of Silent Word Reading Fluency (TOSWRF) were also administered as pre-tests of general English proficiency and English literacy, respectively. To estimate treatment effects, we ran a series of multilevel models. After controlling for native language, grade level, and pre-test scores on the EVOCA, TOSWRF, and TORC assessments, students from the two intervention conditions outperformed control students on the EVOCA post-test total score (Cohen’s d = .30). The effect size was larger when only taught cognates were considered (Cohen’s d = .50). However, there were not statistically significant differences between the two conditions (monolingual vs. cross-linguistic) on the EVOCA total score. Moreover, native Spanish speakers scored significantly lower than other students, and treatment did not moderate this effect.