Perceptual Study of English /l/ Allophones
Mark P. Haggard · The Journal of the Acoustical Society of America · 1967
Once the broad set of perceptual cues distinguishing utterances is isolated, speech synthesis can be further used to answer detailed questions about the process of perception. Such questions are basic to application of knowledge about speech to synthesis-by-rule. Improvement of synthesis rules may result from adjusting the rules to maximize quality of long utterances, but more restricted, detailed perceptual investigations are also necessary. The /l/ phoneme fits into no convincing matrix of feature oppositions and is difficult to synthesize; so five questions were examined for six /l/ allophones—initial and final in the contexts /i,a,u/. (1) How do the optimal patterns vary as a function of context? (2) What is the importance of spectral discontinuities? (3) Do the /l/ steady state and transition have interdependent perceptual effects? (4) Is the same priority of cues shown in quality judgments at high intelligibility levels and in effects on recognition of an ambiguous stimulus? (5) What modification do analysis data and preliminary synthesis experiments suggest for extension of answers to the above questions to other contexts? Results and some descriptive generalizations are given.