ISCA Archive Interspeech 2006
ISCA Archive Interspeech 2006

50 years late: repeating miller-nicely 1955

Andrew Lovitt, Jont B. Allen

Portions of the procedure and analysis of the wide-band noise masking experiment in Miller-Nicely’s 1955 JASA paper (MN55) was repeated and a new set of data was collected in 2005. This classic paper is a commonly referenced work in which confusion matrices were collected for a set of consonant-vowels (CVs). From an analysis of the original results, they made conclusions about the robustness of various distinctive features when the CVs are degraded in masking noise. Our repeat experiment shows a number of similarities and differences. The two experiments show significantly different amounts of relative information transmitted for each distinctive feature. In the repeat experiment the voicing feature is less robust whereas the place feature is more robust.