According to the edge-aligned embedded word activation theory and the principle of full decomposition (Grainger & Beyersmann, 2017), (pseudo)morphemes are extracted from both edges of the whole word based on orthographic information. For example, car is extracted from the compound carport, where it functions as a morpheme, and from the pseudo-compound carpet, where it does not, and is suppressed by the pseudo-compound. Embedded words should receive activation when the whole word can be decomposed into constituent (pseudo)morphemes, and accessing one embedded (pseudo)morpheme should facilitate access to another one. Using English compound words, Taikh et al. (2024) found that replacing boundary letters slowed down processing of the whole compound word, while replacing constituent-internal letters did not, which is consistent with the principle of full decomposition. Whereas replacing boundary letters interferes with access to both constituents, replacing constituent-internal letters allows access to the unaltered constituent. In the present study, we first replicate the findings of Taikh et al. using English compounds and then examine the principle of full decomposition in English pseudo-compounds. We find that replacing pseudo-boundary and pseudo-constituent letters slows down the processing of a pseudo-compound to a similar extent, suggesting that accessing an unaltered pseudo-constituent does not facilitate processing the altered pseudo-constituent. Our findings are consistent with the idea that embedded pseudo-constituents are suppressed by the pseudo-compound and thus do not facilitate its access. Additionally, embedded words may not receive facilitation if the recognition of the entire word requires that they be suppressed strongly enough.