We compared the susceptibility of two emotional intelligence (EI) tests to faking. In a laboratory study using a within-subjects design, participants completed the EQ-i and the MSCEIT in two sessions. In the first session (i.e., the ‘applicant condition’), participants were given a job description and asked to respond to the EI measures as though they were applying for that job. Participants returned 2 weeks later to repeat the tests in a ‘non-applicant’ condition in which they were told to answer as honestly as possible. Mean differences between conditions indicated that the EQ-i was more susceptible to faking than the MSCEIT. Faking indices predicted applicant condition EQ-i scores, after controlling for participants' non-applicant EQ-i scores, whereas the faking indices were unrelated to applicant condition MSCEIT scores, when the non-applicant MSCEIT scores were controlled. Using top-down selection, participants were more likely to be selected based on their applicant condition EQ-i scores than their non-applicant EQ-i scores, but they had an equal likelihood of being selected based on their MSCEIT scores from each condition. Implications for the use of these two EI tests are discussed. Copyright © 2007 John Wiley & Sons, Ltd.