Distortion products and their influence on representation of pitch-relevant information in the human brainstem for unresolved harmonic complex tones

Abstract

Pitch experiments aimed at evaluating temporal pitch mechanism(s) often utilize complex sounds with only unresolved harmonic components, and a low-pass noise masker to eliminate the potential contribution of audible distortion products to the pitch percept. Herein we examine how: (i) masker induced reduction of neural distortion products (difference tone: DT; and cubic difference tone: CDT) alters the representation of pitch relevant information in the brainstem; and (ii) the pitch salience is altered when distortion products are reduced and/or eliminated. Scalp recorded brainstem frequency following responses (FFR) were recorded in normal hearing individuals using a complex tone with only unresolved harmonics presented in quiet, and in the presence of a low-pass masker at SNRs of +15, +5, and -5 dB. Difference limen for F0 discrimination (F0 DL) was obtained in quiet and in the presence of low-pass noise. Magnitude of DT components (with the exception of components at F0 and 2F0), and the CDT components decreased with increasing masker level. Neural pitch strength decreased with increasing masker level for both the envelope-related (FFR(ENV)) and spectral-related (FFR(SPEC)) phase-locked activity. Finally, F0 DLs increased with decreasing SNRs suggesting poorer F0 discrimination with reduction of the distortion products. Collectively, these findings support the notion that both DT and CDT, as reflected in the FFR(ENV) and FFR(SPEC), respectively, influence both the brainstem representation of pitch relevant information and the pitch salience of the complex sounds.

Figure 1

Spectra of the tonal complex and low-pass noise masker to illustrate the experimental paradigm. Tonal complex contains six equal amplitude harmonics (12^th–17^th harmonic of a 90-Hz F0) falling in the peripherally unresolved region. Low-pass noise (500 order FIR; Fpass=885 Hz, Fstop=965 Hz, Astop=60 dB) spectra show the three levels used in the study. The spectrum of the low-pass noise shown represents an average of 1000 presentations of the noise stimuli.

Figure 2

FFR_ENV waveforms (left panels) and their spectra (right panels) from a representative participant are plotted for the quiet and three noise conditions. The multiple DT components are also identified. Note that these spectra were obtained from the electrical signals.

Figure 3

Mean magnitude of F0, 2F0 (left axis), and 3F0-8F0 DT distortion components (right axis) for the quiet and three noise conditions. The asterisk indicates statistically significant differences in magnitude across experimental conditions. Note the difference in amplitude scale for each response type. Error bars correspond to 1 SEM.

Figure 4

FFR_SPEC waveforms (left panels) and their spectra (right panels) from a representative participant are plotted for the quiet and three noise conditions. Vertical dotted lines in the spectral plots demarcate the frequency regions containing the CDT distortion components and the FFR response to stimulus temporal fine structure (FFR_STIM). The low-pass noise region is also identified.

Figure 5

Mean magnitude of FFR_SPEC response components for the stimulus temporal fine structure (FFR_STIM), and CDT distortion product components for the quiet and noise conditions. The asterisk indicates statistically significant differences in magnitude across experimental conditions. Error bars correspond to 1 SEM.

Figure 6

Mean autocorrelation (ACF peak amplitude) magnitude for FFR_ENV (circle) and FFR_SPEC (square) data plotted for the quiet and noise conditions. The ACF peak magnitude, for both responses, corresponds to a delay equal to the period of the 90-Hz F0. Error bars correspond to +/− 1 SEM.

Figure 7

Mean F0 DL (triangle and left y-axis), mean magnitude of 3F0-8F0 DT distortion components (circle and right y-axis) and, mean magnitude of CDT distortion components (square and right y-axis) for the quiet and noise conditions. Error bars correspond to 1 SEM.