There is a striking parallel between the hierarchical organization of behavior and the hierarchical organization of the cerebral cortex (Botvinick 2008). It is thus tempting to assign consciousness, at least historically often considered to be one of our highest functions (Jackendoff 1987; Markowitsch 1995), to the prefrontal cortex (PFC), which is positioned at the top of the cortical hierarchy. While the idea that consciousness can be localized to a single brain area has now been discredited, many current theories of consciousness still consider the PFC a key player in the emergence of conscious perception (Dehaene & Changeux 2011; Lau & Rosenthal 2011). And indeed, a multitude of neuroimaging studies has shown differential activation for perceived vs. unperceived stimuli in various parts of the PFC (Dehaene et al. 2001; Lau & Passingham 2006; Sahraie et al. 1997; Schwiedrzik et al. 2014). A very prominent theoretical proposal on the neural correlates of consciousness, the Global Neuronal Workspace (GNW) model by Stanislas Dehaene and colleagues, posits that the PFC (in conjunction with parietal cortex) serves to distribute information that is processed in unconscious modules to the entire brain, and that it is this broadcasting of information that gives rise to conscious experience (Dehaene & Changeux 2011). The PFC may be particularly well equipped to do so, for example because it hosts an abundance of neurons with long-distance connections, so called “von Economo” neurons, which seem ideally suited to both receive and deliver information from all areas of the brain to all areas of the brain (Dehaene & Changeux 2011). A prediction that can be directly derived from this account and that has been eloquently put forward by John-Dylan Haynes is that the PFC should at least temporarily represent the information that we consciously perceive, i.e., it should directly encode the contents of consciousness (Haynes 2009; this collection). To test this idea, Haynes and his coworkers have used a neuroimaging technique that allows for exquisite access to perceptual content, namely multivariate pattern analysis (MVPA) of functional magnetic resonance imaging (fMRI) signals. In this technique, powerful machine learning algorithms are used to analyze spatially-distributed patterns of brain activity, and a brain region is said to represent the content of interest if its activity patterns allow the reliable classification—in the case of consciousness—of which stimulus the subject perceived on a given trial. This contrasts with previous fMRI studies not using MVPA: because these studies do not directly address content, activity in the PFC (and other regions) that differentiates perceived from not perceived trials could in principle reflect other aspects of conscious experience, for example the allocation of attentional resources or working memory. The stunning result of Haynes’ investigations is that while MVPA shows that perceptual content can be decoded from higher sensory areas, PFC activity does not yield decoding accuracies higher than chance level. So what’s up with PFC?