Cross-linguistic lexical similarity in the form of cognates and loanwords has been shown to positively impact second language learning and use, as well as performance on tests of lexical knowledge, even when the learners' languages differ in script. The present study utilizes Japanese cognate frequency, as an indication of cognate knowledge, to predict accuracy on the "English Vocabulary Size Test." Japanese participants (n = 113) completed either the monolingual or bilingual version of the English test. Generalized linear mixed models revealed a significant effect of cognate frequency on accuracy in both tests. In the bilingual version, this cognate effect was shown to be slightly greater for lower level participants. These findings contribute further evidence that cognates in different-script languages may significantly impact accuracy in language tests. Moreover, a comparison of the proportion of cognates in the test with the proportion in the languages in general revealed that the "Vocabulary Size Test" may over-estimate Japanese learners' knowledge of English vocabulary. The primary implication of the findings is that the proportion of cognates must be controlled in tests of lexical knowledge that are aimed at specific populations and that cognate frequency is useful to determine the likely extent of the cognate effect.