Perceptual learning accounts of orthographic coding predict that transposed-letter (TL) priming effects should be smaller when the prime and target stimuli are not presented in their canonical (left-to-right horizontal in English) orientation (Dehaene, Cohen, Sigman, & Vinckier, 2005; Grainger & Holcomb, 2009). In contrast, abstract letter unit accounts would propose that TL priming effects should be essentially unaffected by presenting stimuli in most unfamiliar text orientations (Witzel, Qiao, & Forster, 2011). In the present experiments, we examined masked TL priming effects with primes and targets presented in 3 different text orientations (e.g., 0 degrees, as well as 90 degrees and 180 degrees rotations). Results revealed that the magnitude of the TL priming effect with native English readers was equivalent for stimuli presented in these three orientations, providing support for abstract letter unit accounts of orthographic coding.