The present study investigated how response mode (oral vs. manual) modulates the Stroop effect using a picture variant of the Stroop task in which participants named orally, or identified with a manual keypress, line drawings of animals (e.g., camel). Consistent with previous color-response Stroop studies, relative to the nonlinguistic neutral distractor (a row of "#" symbols), incongruent distractors (e.g., GIRAFFE) interfered with responding to pictures, and that interference was reduced for the manual, compared with the oral, response. Additionally, pseudoword distractors with no phonological overlap with the picture name (e.g., NUST--camel) interfered with the oral, but not the manual, response. The novel finding is that relative to this pseudoword distractor, the oral response was facilitated when the distractor shared the onset segment with the picture name, regardless of orthographic overlap (e.g., CUST--camel = KUST--camel < NUST--camel); in contrast, for the manual response, there was no difference between the three pseudoword distractor conditions. These results are explained in terms of phonological encoding, a speech production process involved in computing a phonetic plan for generating an oral, but not a manual, response.