An Analysis of the Confusion among English Consonants Heard in the Presence of Random Noise

Abstract

The English consonants p, t, k, f, θ (thin), s, ∫ (she), b, d, g, v, ð (they), z, ȝ (beige), m and n were pronounced before the vowel a (father) by female talkers with General American pronunciation. Listeners heard the syllables in the presence of a masking noise and were forced to guess at every syllable. The signal-to-noise ratio, measured between the over-all level of the noise in a band from 20 to 7000 cps and the level of the vowel as read on a VU meter, ranged from −18 to +12 db in steps of 6 db. Five talkers were used with four listeners for each talker. The most obvious aspect of these consonants in noise is the presence or absence of voicing; voiced and voiceless consonants were seldom confused until S/N was about −12 db or worse. The aspect of nasality also seems to disappear at about −12 db. The next most obvious aspect is the difference between stops and continuants; confusions on this aspect did not occur until S/N was about −6 db or worse. These three aspects define five classes of consonants: voiceless non-nasal stops, p, t, k; voiceless non-nasal continuants, f, θ, s, ∫; voiced non-nasal stops, b, d, g; voiced non-nasal continuants, v, ð, z, ȝ; and voiced nasal continuants, m, n. Confusions within these five classes occurred when S/N was Odb or worse. At favorable signal levels the most frequently confused pair is f and θ, closely followed by v and ð, then p and k.