Recent research investigating embedded stem priming effects with the masked priming paradigm and pseudoword primes (e.g., quickify–quick) has shown that priming effects can be obtained even when the embedded target word is followed by a non-morphological ending (e.g., quickald–quick). Here we examine the specific nature of such priming effects by testing whether they are modulated by morphological family size. We reasoned that if the effects are driven by pre-lexical orthographic processing then they should not be influenced by the family size of the embedded target word. On the contrary, we found that embedded words having several morphologically related family members (e.g., serpentoche–SERPENT [English: snakerel–SNAKE]—serpents, serpentin, serpentine, serpenter, serpentant) generated greater priming than embedded words having only the plural form in the morphological family (e.g., dauphingri–DAUPHIN [English: dolphinald–DOLPHIN]). We therefore conclude that embedded stem priming is at least partly driven by processing at the level of lexical and morpho-semantic representations.