Reading development involves several changes in orthographic processing. A key question is, "how does the coding of letters develops in children learning to read?" Masked priming effects of transposition and substitution primes have been taken to index the importance of letter position and identity coding. Somewhat contradicting results for developing readers have led to opposing theories. Here, we present new evidence from a large longitudinal study with over 100 children from grade 2 to 4. We investigate three different issues concerning transposition and substitution priming: (a) comparing priming both against an all-different and an identity baseline, (b) testing priming effects for nonword targets, and (c) taking into account interindividual differences in orthographic knowledge. The analyses of the longitudinal data show, respectively, (a) an increase of priming effects over development in comparison to both baselines, (b) identity priming for nonword targets in the elementary school years, and (c) an additional impact of orthographic knowledge on priming effects for word, but not nonword, targets that is similar to the effect of grade. Taken together, our examination suggests that letter identity is coded relatively strictly, whereas letter position is coded relatively flexibly already early in reading development for words, but not for nonwords. We discuss how this pattern fits with different developmental models of orthographic processing.