Does language change thought? This classical question has recently received renewed attention, as new lines of evidence have been offered, either supporting or arguing against the idea that speaking a particular language-- or having a language at all-- affects our non-linguistic representation. The domain of space has provided particularly fertile territory for this debate. In this talk, I will present a new hypothesis about the way in which language interacts with visual-spatial representations, arguing that this occurs on a momentary basis, with no repercussions for permanent changes in our non-linguistic spatial representation. These momentary interactions increase over development, resulting in obligatory linguistic encoding in many tasks, but no real change in non-linguistic representations. This hypothesis not only accounts for new data I will discuss; it also accounts for much of the existing data on both sides of the aisle, in domains as different as spatial cognition and the representation of color. Having a language undoubtedly changes human cognition-- but it does not change our non-linguistic representations.