Cognates are known to facilitate second language acquisition and use, as learners tend to assign to a new L2 word the meaning of a similar L1 word. Consequently, for L2 tests that rely largely on lexical items, performance may prove inflated for examinees whose L1 shares many cognates with the language being tested. This article examines the possibility of L1 bias on the Peabody Picture Vocabulary Test (PPVT), a well-established measure of receptive vocabulary knowledge in English. To investigate if performance on the PPVT is affected by cognates, we tested 293 speakers of French and 150 speakers of Polish, since those two languages differ markedly in the number of cognates they share with English. After demonstrating that both groups yield clearly distinct response patterns, descriptive and multivariate statistics confirmed that cognate items enhance test performance: the items with the highest score difference in favour of a language group overwhelmingly consist of cognates for that group only. Mantel-Haenszel and logistic regression show that items that are cognates for one of the two groups are more likely to show differential item functioning than the average items. The results suggest that scores on L2 vocabulary-based tests could be biased by the presence of cognates with the examinee's first language.