Multivariate pattern analysis (MVPA) has become an increasingly popular approach to fMRI research because these methods offer the attractive possibility of "decoding" the content of brain representations. One weakness of MVPA is that the source of decodable information is not always apparent, as evidenced by the ongoing debate about orientation decoding in human visual cortex. In a recent study (Carlson, 2014), we used an unbiased model of visual cortex to reveal a new source of decodable information that may account for orientation decoding. Clifford and Mannion (2015) take issue with the model's capacity to decode spiral sense. Here, we discuss their findings in the context of the ongoing debate on orientation decoding and further highlight the limitations of using MVPA to infer the content of brain representations.