This dissertation examines the theoretical construct of a ‘morpheme’ as a unit that is independent of semantics and phonology. Specifically, the dissertation examines multi-morphemic words that are semantically opaque, i.e., words whose meaning is not derivable from the meaning of the separate morphemes. While some models of lexical access posit abstract morphological representations (i.e., Full-Decomposition models), other models posit whole-word storage and processing either for all forms or for certain forms depending on factors like semantic transparency. Semantically opaque words are crucial for testing predictions of these different models of lexical access, as they allow us to investigate whether morphological processing occurs in the absence of semantic relatedness. Building on the extensive prior literature from both theoretical linguistics and experimental psychology, this dissertation probes questions about morphological relatedness using an auditory primed continuous lexical decision task. In particular, the dissertation examines the processing of the following types of words: (i) Dutch prefixed verbs, which differ in meaning relatedness between the stem and the complex verb from fully transparent (e.g., aanbieden ‘offer’) to fully opaque (e.g., verbieden ‘forbid’, with the stem bieden ‘offer’) (Chapters 3 and 6); (ii) English suffixed words like treatment and their relation to pseudo-suffixed words like pigment (Chapter 4); and (iii) compound words, which may be transparent (e.g., bedroom) or opaque (e.g., strawberry, with an opaque modifier, or staircase, with an opaque head) (Chapter 5). The results in Chapter 3 show equal and robust priming effects for transparent and opaque prefixed verbs. A significant difference is found between the priming effects for suffixed words and pseudo-derived words in Chapter 4. Chapter 5 shows priming effects for both constituents in a compound, regardless of semantic transparency. Finally, the results in Chapter 6 show associative priming effects for the meaning of the stem in semantically opaque complex words. Together, the results offer a window into the issue of how (apparent) multi-morphemic words are processed and represented in the mental lexicon during auditory word recognition. In line with a Full-Decomposition view, the results suggest that morphemes form the basic units of lexical processing, and provide evidence that morphological relatedness does not require shared meaning. Moreover, the results provide evidence that the meanings of morphemes are accessed also in opaque forms.