Intelligibility of VCV segments excised from connected speech

Confusions between pairs of intervocalic consonants excised from connected speech were investigated under several conditions of speech degradation: digital voice processing, noise, and bandpass limiting. The distribution of the types of errors that were made, e.g., voicing, nasality, place of articulation, differed from those made on citation form syllable-initial consonants. Intervocalic consonants that were taken from word-initial position scored higher with no degradations than those taken from word-medial or word-final position, but the medial and final segments suffered less under mild degradations than did the initial ones.