Electrode ranking of "place pitch" and speech recognition in electrical hearing.

Abstract

The ability to distinguish electrical stimulation of different electrodes on the basis of "pitch or sharpness" was evaluated with an electrode ranking procedure in 14 individual users of the Nucleus cochlear implant. Prior to the electrode ranking test, absolute thresholds and maximum comfortable loudness levels were measured, and loudness balancing was accomplished across all usable electrodes. Performance on the electrode ranking task was defined in terms of d' per mm of distance between comparison electrodes. Large individual differences were found among cochlear-implant users. In subjects with good to excellent place-pitch sensitivity, the electrode ranking task was limited by a ceiling effect; however, in those with poor to moderate sensitivity d'/mm was relatively constant with spatial separation between electrodes. Place pitch was typically ordered from apical to basal electrodes, i.e., basal electrodes were judged to be higher in pitch than more apical electrodes. However, instances of reversals in place-pitch ordering were seen on some electrodes in some subjects. Instances were also seen of better electrode ranking in the apical half of the electrode array than in the basal half, and vice-versa. Analyses of the electrode ranking functions in terms of d' per stimulus indicated that, in some subjects, perfect performance was reached with as little as 0.75 mm between comparison electrodes, the minimum possible. In other subjects, perfect performance was not reached until the spatial separation between comparison electrodes was over 13 mm, more than three quarters of the entire length of the electrode array. Ten of the subjects also participated in a closed-set recognition task of intervocalic consonants. Although the maximum transmitted information for place of consonant articulation (which is based primarily on spectral speech cues) was only 34%, correlations between place-pitch sensitivity and transmitted speech information were as high as 0.71. This was surprising considering the excellent place-pitch sensitivity exhibited by some of the subjects, and may reflect limitations of the Nucleus speech coding strategy for representing spectrally coded speech information. The two prelingual subjects performed notably poorer on the speech task than the postlingual subjects, even though one of the prelingual subjects demonstrated very good place-pitch sensitivity.