What is the route that is used for writing non‐words to dictation? Does it have to pass through phonological output? Two possibilities are found in the literature. In one, writing non‐words requires access from the phonological input buffer to the phonological output buffer and from there, via phoneme‐to‐grapheme conversion, to the orthographic output buffer. The second approach maintains that writing non‐words can proceed directly from the phonological input buffer to the orthographic output buffer. In this study, we discriminate between these two options using a cognitive neuropsychological approach. Specifically, we present a multiple case study of 24 individuals with a developmental impairment to the phonological output buffer, who nevertheless show unimpaired non‐word writing. These data lead to the conclusion that the phonological output buffer is not necessary when writing non‐words and that writing non‐words to dictation can proceed directly from the phonological input buffer to the orthographic output buffer. We suggest that the cognitive assumption that non‐word writing proceeds through the phonological output buffer may have resulted from graphic conventions and the depiction of the lexical processing model: in the common depiction of the model drawing a line through the phonological output buffer is visually simpler than a direct line (which would require lines to cross or long bypass lines).