Using multivariate pattern analysis techniques, an increasingly growing number of studies have shown that representations can be reliably extracted (‘decoded’) from neuroimaging data (e.g. fMRI, MEG). A current topic of debate is whether these decoded representations are used by the brain in behavior. For example, Williams, Dang, & Kanwisher, (2007) found that visual object category information could be decoded from both retinotopic cortex and lateral occipital cortex, but in the latter, information was only decodable on the trials on which the subject responded correctly, suggesting that only information from that area is directly read-out in behavior. A recent approach to test whether a link exists between representational spaces and behavior is the "representational distance hypothesis" (Carlson, Ritchie, Kriegeskorte, Durvasula, & Ma, 2014), which states that if the brain uses a representational space in behavior, then the distance from a decision boundary (as estimated by classifiers used for decoding) can be used to predict reaction time. In this talk, I will discuss this approach in more detail and I will present our findings on testing the predictions that follow from the representational distance hypothesis.