It is long-acknowledged that the consonantal voicing effect—whereby vowel duration is greater preceding voiced vs voiceless consonants (e.g., rib/rip)—is larger in English as compared with other languages, and that the effect’s magnitude generally decreases in multisyllabic forms (e.g., rabbit/rapid). The current study examines consonantal voicing effects in the multisyllabic environment, crucially contrasting non-coronal with coronal cases (e.g., rider/writer). In American English, the latter are subject to a flapping process that surface-neutralizes the voicing distinction. Hence, while both phonological and phonetic sources for a vowel-duration difference are available in non-coronals, flapping eliminates the phonetic source in coronals. We present an analysis of critical vowel durations in elicited productions (target words uttered in a carrier phrase), and confirm the pattern expected if the post-vocalic consonant’s place of articulation matters: the consonantal voicing effect was entirely reliable for non-coronals, but not for coronals. More detailed analyses set aside tokens where flapping failed to apply, and found that the consonantal voicing effect might be altogether absent with coronal place. We speculate that, here, the voicing distinction may have been neutralized in phonological representation, whether that distinction is a matter of orthography (doodle/duty) or is also supported by morphological alternation (rider/writer).