Purpose: To evaluate whether the validity of the Receptive Vocabulary scale of the MacArthur Communicative Development Inventory for Infants (MCDI-I; L. Fenson et al., 1991), a parent-report measure of early vocabulary, could be improved for children with autism spectrum disorders (ASD) by removing items that are biased. Method: Logistic regression was used to identify biased items. Items are considered biased if characteristics other than those being measured by the instrument change the probability that a person will get an item correct. Participants in the current study included 272 typically developing infants younger than 18 months of age and 209 toddlers with ASD older than 18 months of age. The age difference between the 2 groups is a result of matching on total size of the receptive vocabulary. Results: Twenty-five items were identified as showing large bias. Conclusion: Deletion of these items from the test should increase the degree to which the authors are measuring the size of the respondent's mental lexicon with the total score from the MCDI-I.