Recognition of famous voices forwards and backwards

Abstract

To investigate familiar voice recognition, samples of famous male voices were tape-recorded and edited on a PDP-11/34 computer. Three listening tasks were prepared. Subjects indicated whether they recognized voices from (1) 2-s samples; (2) different and reordered 2-s samples presented along with six choices; and (3) 4-s samples presented backwards (rerandomized and refoiled). Ninety-six subjects were divided into four groups, three by age and one given the backwards presentation first. For task 1, the mean recognition rate was 17%. In task 2, subjects correctly identified 69.5% of the voices they knew (ascertained by questionnaires). Recognition for known voices presented backwards was 12.8% less than for voices presented forwards. In the group given the voices backwards first, a similar difference (12.5%) was observed. A two-way ANOVA comparing the four groups on tasks 2 and 3 revealed main effects of group and task, but no group-by-task interaction; thus differences in performance between forwards and backwards presentations were unchanged across the four subject groups. These results indicate that voice recognition can be achieved given only limited information which includes rate, pitch, and pitch range. [Work supported by NIH Grant No. 4-443944-31117.]