Three experiments are reported investigating the role of letter order in orthographic subset priming (e.g., "grdn"-GARDEN) using both the conventional masked priming technique as well as the sandwich priming technique in a lexical decision task. In all three experiments, subset primes produced priming with the effect being considerably larger when sandwich priming was used. More importantly, there was very little difference in the degree of priming produced by subset primes with transposed (i.e., "gdrn") vs. nontransposed ("grdn") internal letters. The priming effects with transposed letter subset primes contradict Peressotti and Grainger's claim that letter order must be maintained in order to produce subset priming effects (i.e., their "relative position priming constraint"). (Contains 3 footnotes and 6 tables.)