Article Figures & Data
Figures
- Figure 1.
Acoustic stimuli. A–C. Pressure waveforms for one second of speech, Poisson click train, and standard periodic click train, respectively. Vertical scale is plotted in arbitrary units (AU), but is consistent across plots. D–F, Spectrograms of a smaller excerpt of the above stimuli, with darker colors corresponding to higher power. G–I, PSD plots of the above stimuli, calculated from 30 s of data using Welch’s method with a segment length of 5.67 ms, segment overlap of 50%, and Hann window. Note that although the speech recordings were gently high-pass filtered at 1000 Hz, there remains plenty of power in the 125- to 1000-Hz range (G) and pitch information is clearly preserved (D; vertical striations between 200 and 300 ms correspond to individual glottal pulses).
- Figure 2.
Comparison of ABR to standard periodic click trains and Poisson click trains. A, The average ABR waveform evoked by the standard, periodic click train at 44.1 clicks/s (black) and the pseudorandom Poisson click train (gray; 44.1 clicks/s overall rate). Areas show ±1 SEM. Both responses are high-pass filtered at 150 Hz. The spike at −1 ms is a stimulus artifact, and occurs before 0 ms to compensate for the 1 ms tube delay of the earphones. B, The histogram of per-subject correlation coefficients between the standard and Poisson click-evoked ABRs. Solid/dotted black lines show median/quartiles. C, Comparison of the Poisson click-evoked ABR with 150-Hz high-pass filtering (gray) and without (i.e., broadband; blue). The latter is used as the benchmark response for the remainder of the study.
- Figure 3.
Comparison of click-evoked responses (blue) with speech-derived responses (red). A, The average waveform across subjects (areas show ±1 SEM). B, The histogram of correlation coefficients between the click-evoked and speech-derived stimuli for each subject. Solid/dotted black lines show median/quartiles. C, Individual subject responses, sorted by descending correlation coefficient. The correlation is shown in the upper right corner.
- Figure 4.
Correlation of speech-derived and click-evoked Wave V latencies (A) and amplitudes (B) across subjects. Because the click-evoked Wave V is known to be subcortical, the strong correlations across subjects point to brainstem neural generators for the speech-derived response as well. Points have been jittered slightly to prevent visual overlap. Regression lines are shown with the 95% confidence interval shaded.
- Figure 5.
Comparison of female-narrated responses (green) with male-narrated responses (purple). A, The average waveforms across subjects (areas show ±1 SEM). B, The histogram of correlation coefficients between the female-evoked and male-evoked stimuli for each subject. Solid/dotted black lines show median/quartiles. The speech-derived ABRs from the male and female narrators show strong similarities, but are not identical, indicating some talker dependence.
- Figure 6.
SNR as a function of data accumulation. SNR was calculated for each subject using the mean of all the data up to each recording epoch by computing the variance in the ABR time interval 0–20 ms, and in the prestimulus noise interval −125 to −10 ms.
- Figure 7.
Changes to the range of lags and filtering parameters allow early, middle, and late responses to be analyzed from the same recording. A, The average speech-derived ABR with Canonical Waves V and VI labeled. B, The middle latency response with its canonical waves labeled (low-pass frequency: 200 Hz). C, The late auditory evoked potential (low-pass frequency: 20 Hz). Due to adaptation, the amplitudes for the later waves are much smaller than typically seen in the event-related potential literature. These peaks are also not given canonical labels as in A, B because their latencies do not directly correspond to the standard N1 and P2 peaks. Shaded areas show ±1 SEM.
In this issue
