Language communities differ in their stock of reference frames (coordinate systems to reference locations and directions). English typically uses egocentrically-defined axes ("left-right"). Other languages like Tseltal lack such a system but use geocentrically-defined axes ("north-south"). Several researchers have recently argued that the lexical resources for encoding frames of reference determine the availability or salience of spatial concepts: in support of such language-on-thought effects, a series of experiments has shown that Tseltal speakers do not use a left-right coordinate system to solve spatial tasks (Brown and Levinson, 1993; Levinson 2003). Here I present the results of a collaborative project which re-examined spatial reasoning in Tseltal Mayan speakers using a battery of novel tasks. Contrary to previous studies, our data show that Tseltal speakers are quite capable of solving "egocentric" problems, despite the absence of a left-right coordinate system in their language. Furthermore, participants were often better on the egocentric than the geocentric version of our tasks. Taken together, our results strongly suggest that the linguistic encoding of spatial relations does not determine speakers' performance in nonlinguistic spatial tasks.