This experiment used the masked priming lexical decision task to address previous contradictory evidence about the relative strength of priming for (i) transparent pairs (e.g., worker WORK ) which are morphologically and semantically related; (ii) opaque pairs (e.g., corner CORN ) which appear to be morphological relatives but are not semantically related; and (iii) form pairs (e.g., turnip TURN ) that are only orthographically related. The average data for 92 university students showed stronger priming effects for transparent than opaque or form pairs, due to a constant ‘headstart’ for related relative to unrelated pairs across the RT distribution. However, these average effects were significantly modulated by individual differences in independent measures of spelling and vocabulary. A ‘semantic profile’, defined by relatively higher vocabulary than spelling, was associated with robust priming for transparent pairs, particularly for slower responses, but little priming for opaque or form pairs. In contrast, individuals with an ‘orthographic profile’ of relatively higher spelling than vocabulary showed sustained priming for opaque pairs that was at least as strong as for transparently related pairs. This evidence of systematic individual differences amongst skilled readers has important implications for theories of lexical representation and processing.