The last two decades have been marked by excitement for measuring implicit attitudes and implicit biases, as well as optimism that new technologies have made this possible.  Despite considerable attention, this movement is marked by weak measures.  Current implicit measures do not have the psychometric properties needed to meet the standards required for psychological assessment or necessary for reliable criterion prediction. Some of the creativity that defines this approach has also introduced measures with unusual properties that constrain their applications and limit interpretations.  We illustrate these problems by summarizing our research using the Implicit Association Test (IAT) as a case study to reveal the challenges these measures face. We consider such issues as reliability, validity, model misspecification, sources of both random and systematic method variance, as well as unusual and arbitrary properties of the IAT’s metric and scoring algorithm. We then review and critique four new interpretations of the IAT that have been advanced to defend the measure and its properties. We conclude that the IAT is not a viable measure of individual differences in biases or attitudes. Efforts to prove otherwise have diverted resources and attention, limiting progress in the scientific study of racism and bias.