Using the same-different matching task, Massol and Grainger (2022) reported similar effects of low-level perceptual noise for different character strings (letters, digits, or symbols), whereas transposed-character effects were greater for letter strings. This pattern has been interpreted as follows. At the level of location-specific complex feature detectors, representations are subject to a certain amount of positional noise, independently of stimulus type. These then activate location-invariant object identities where object order is encoded. Here, a distinction is drawn between a generic mechanism for encoding order information and a more flexible mechanism that is only used to represent letter order. We investigated the role played by both mechanisms using the same-different task. Critical pairs of stimuli (letter or symbol strings) could differ by transposing two characters (e.g., FBJK-FJBK), substituting one character by another one that could be visually similar (e.g., FJDK-FJBK) or dissimilar (e.g., FJVK-FJBK), or substituting two characters (e.g., FHSK-FJBK). We predicted that effects of visual similarity will operate independently of stimulus type whereas transposed-character effects will be greater for letters.