Spelling ability is not static; rather, as children age, learning how to encode morphophonologically complex words in conventional ways is motivated by the increasingly complex demands imposed by academic experiences with morphologically complex words. Success requires ongoing integration of phonological (P), orthographic (O) and morphological (M) knowledge. However, current research on the development and assessment of spelling has not sufficiently accounted for the way word features and participant characteristics interact with students' POM knowledge in the spelling of derived words. This study used a linear mixed effects regression approach to provide new insights about how both word characteristics and students' linguistic knowledge affected the application of POM from grades 3-7 in the spelling of derived forms. Spelling data (WIAT-II) were taken from a larger longitudinal study focused on reading development (Garcia et. al., 2010). Eleven words from the WIAT-II with derivational morphology (which included one inflected form with a derived homophone possibility) were analyzed first with the Phonological Orthographic Morphological Analysis of Spelling (POMAS; an unconstrained scoring system) in order to identify linguistic feature errors within misspellings. Next, misspellings were quantified with the POMplexity metric to evaluate the individual and combined influences of phonology (P), orthography (O), and morphology (M) to derivational misspellings over time. A linear mixed effects regression approach evaluated the impact of item-level characteristics (derivational frequency and shifts), participant characteristics (rime, spelling choice and morphological awareness task scores), and time (grade level) on POMplexity scores. Results indicated item-level characteristics, participant characteristics and time significantly predicted variation in P, O, M, and total POMplexity scores. Frequency had a significant impact on scores, with high frequency words resulting in lower POMplexity scores than low frequency words and these effects were most obvious in grades 3 and 4. Slope differences between words suggested that low frequency misspellings resolve more rapidly than high frequency words. Derivational shift was shown to have a significant interaction with time for O, M and Total scores, but not P scores. In all cases, the slopes for derived words with no shift improved more quickly than shift categories. Finally, performance on measures on the measures of linguistic skill correlated to improved scores for the related POMplexity code. These results strongly suggest that the developmental course of learning to spell derivations is not a linear accumulation of POM knowledge, but instead is a recursive process with both general and word-specific knowledge affecting how an individual student produces a derivational spelling at any given point in time. Contributions of word characteristics, such as frequency and number/type of derivational shift, suggest that morphemic features challenge encoding; that is, increased complexity taxes the system's ability to represent both sound and meaning orthographically. Educational and clinical implications will be described.