We use Woodcock Johnson III child assessment data in the New Immigrant Survey to examine language assimilation and test score bias among children of Hispanic immigrants. Our identification strategy exploits the test language randomization (Spanish or English) to quantitatively measure the degree and speed of language assimilation, in addition to the potential costs associated with taking a test in one’s non-dominant language. We find that U.S. born children of Hispanic immigrants are not bilingual as predicted by most language assimilation models but rather are English dominant. English language assimilation occurs at a rapid pace for foreign born children as well; children who arrive in the U.S. at an early age or who have spent more than four years in the U.S. do not benefit from taking the tests in Spanish. Results are robust to a fixed effects specification that controls for household level characteristics constant across siblings.