There is substantial debate around the nature of letter position coding in reading. Research on a variety of Indo-European languages suggests uncertainty in position coding; for example, readers perceive transposed-letter stimuli (jugde) as similar to their base words (judge). However, these effects are not apparent in all languages. We developed a powerful new method to delineate how specific properties of a writing system shape the representation of letter position. Two groups of 24 adults learned to read novel words printed in artificial scripts. One group learned a dense orthography (i.e. with many anagrams) and one group learned a sparse orthography (i.e. no anagrams). Following four days of training, participants showed a larger transposed-letter effect in the sparse orthography than in the dense orthography. These results challenge existing models of orthographic processing in reading, and support the claim that orthographic representations are shaped by the nature of the writing system. This work has been accepted for publication: Lally, C., Taylor, J.S.H, Lee, C. & Rastle, K. (2019) "Shaping the precision of letter position coding by varying properties of a writing system", to appear in Language, Cognition and Neuroscience.