A large-scale study of British English stop consonants investigated the usefulness of perceptually motivated spectral scaling (Bark scale, phones, broad band integration) in the algorithmic discrimination of place of articulation. 2947 tokens of syllable-initial stop consonants were processed to give representations scaled in dB/Hz., phon/Bark with a 1 Bark frequency smearing filter and phon/Bark with a 3 Bark frequency smearing filter. Each representation was then examined in a search for features robustly associated with place of articulation distinctions. Two types of features were sought, dynamic (patterns of spectral change) and static (features of the burst spectrum). No reliable dynamic features were established, but a static feature based on the frequency distribution of the two main spectral peaks in the burst was effective, especially in the phon/Bark/1 Bark frequency-smeared representations (77.4% correct classification). The addition of a second feature based on the amplitude difference between these two main peaks further improved classification to 81.3% for the phon/Bark/1 Bark frequency-smeared representations.