In a foundational study published in 1955, Miller and Nicely measured the perceptual confusions among 19 consonants followed by the vowel [a] (ba, da, ga, etc.). The stimuli were subjected to different kinds of linear distortions, i.e., additive noise and variations in bandwidth. They published their results in terms of confusion matrices. In a followup study, Shepard (in a 1972 book chapter, but see a more accessible publication, Shepard, 1980) analyzed the Miller & Nicely data in a Multidimensional Scaling framework, aiming at revealing the underlying perceptual mechanism that caused those confusions. Unfortunately, interpretation of these studies through the prism of oscillation-based models fail to shade any new insights into our understanding of how sub- word units are decoded. This is so because the duration of the stimuli are too short to allow entrainment, resulting in an oscillatory array in idle mode and a system reduced to the conventional model. Currently, at MPIEA, we are repeating the Miller and Nicely paradigm but with the listeners hear the CVs in a “theta regime”; we are testing whether the placement of the CV inside a theta cycle will improve performance. 

back