In the quest to unveil the nature of the orthographic code, a useful strategy is to examine the transposed-letter effect (e.g., JUGDE is more confusable with its base word, JUDGE, than the replacement-letter nonword JUPTE). A leading explanation of this phenomenon, which is in line with models of visual attention, is that there is perceptual uncertainty at assigning letters ("objects") to positions. This mechanism would be at work not only with skilled readers but also with preliterate children. An alternative explanation is that the transposed-letter effect emerges at an orthographic level of processing as a direct consequence of literacy training. To test these accounts, we conducted a same-different matching experiment with preliterate 4-year-old children using same versus different trials (created by letter transposition or replacement). Results showed a significantly larger number of false positives (i.e., "same" responses) to transposed-letter strings than to 1/2 replacement-letter strings. Therefore, the present data favor the view that the visual processing of location information is inherently noisy and rule out an interpretation of confusability in letter position coding as emerging from literacy training.