Resolution, spectral weighting, and integration of information across tonotopically remote cochlear regions: hearing-sensitivity, sensation level, and training effects

Abstract

This paper examines how listeners resolve, weight, and integrate redundant sensory information from tonotopically remote cochlear regions, one in the mid (m) the other in the high (h) frequencies. Subjects listened to two-component complexes and attempted to resolve and integrate simultaneous differences in frequency (DFm) at ≈1000 Hz and in duration (DTh) at ≈3500 Hz. Discrimination performance was studied as a function of hearing sensitivity, sensation level (SL) and training. Normal hearing listeners tend to resolve both DFm and DTh and integrate the information if the component SLs are equal, but resolve only the louder component difference if the SLs are very unequal; i.e., spectral weighting is biased and integration is limited or nil. Much the same pattern obtains when the low SL in H results from high-frequency impaired sensitivity (IS). Once established, equating the component SLs (e.g., by high-frequency amplification) is often ineffective to rectify the weighting bias; doing so requires extensive discrimination training with the low SL component. With IS, an anomalous trend obtains when the SL is much lower for (m) than for (h): the resolution of DFm is equal to or better than that of DTh, and integration is quite efficient, possibly reflecting tonotopic map reorganization.