Temporal Suppression of Clicked-Evoked Otoacoustic Emissions and Basilar-Membrane Motion in Gerbils

Abstract

Otoacoustic emissions evoked by a click (CEOAEs) can be reduced if the evoking sound is preceded or followed by another (“suppressor”) click. Studying the temporal suppression of CEOAEs can provide insights into the dynamics of cochlear nonlinearity. However, temporal suppression has never been measured in basilar-membrane (BM) motion. Thus, it remains unclear whether the characteristics of CEOAE temporal suppression are reflected in BM responses. Here we simultaneously measured ear-canal pressure and BM velocity (laser Doppler vibrometry) in response to clicks in gerbil. When the suppressor click preceded the evoking click, CEOAEs were maximally suppressed for interclick intervals (ICIs) equivalent to ~2 periods of the analyzed frequency (9–14 kHz). Maximal temporal suppression at nonzero ICIs has been previously observed in human CEOAEs. BM responses to clicks were maximally reduced when the suppressor click preceded the evoking one by ~1 period of the characteristic frequency (CF ~14 kHz). Thus, the “delayed” characteristics of CEOAE temporal suppression are reflected in BM motion, although on a different time scale. When the suppressor click followed the evoking click, CEOAEs were augmented rather than suppressed, while enhancement was not observed in BM motion at the CF. This result indicates that some aspects of CEOAE temporal suppression are intrinsic to CEOAE generation mechanisms and/or to places that are not reflected in a BM motion at a single-location.

INTRODUCTION

In a pioneering study, Kemp and Chum [1] showed that human click-evoked otoacoustic emissions (CEOAEs) can be diminished in magnitude if another click either precedes or follows the evoking one. Subsequent exploration of this phenomena revealed that maximal suppression of the CEOAE waveform often occurred for nonzero time separation between the two clicks (interclick interval, ICI) [2, 3, 4]. The origin of this delay in CEOAE temporal suppression has been hypothesized to stem from nonlinear interactions between basilar membrane (BM) responses to the evoking and suppressor clicks [6]. Here, we test this hypothesis by simultaneously recording ear-canal pressure and BM velocity in response to acoustic clicks in gerbils.

METHODS

After inducing a deep anesthesia (ketamine with xylazine, intital dose of 80 and 10 mg/kg, respectively), the animals were tracheotomized and the pinna was resected, allowing the tip of the OAE probe to be sealed to the bony rim of the ear canal. The bulla was widely open to visualize the round-window membrane, which was then removed. A few reflective micro-beads (1.14 g/cm³ density) were spread over the BM and the opening then covered with a glass cover slip. The laser-Doppler vibrometer (LDV) was focused on one of the beads to measure its velocity (e.g., [5]). The animals were sacrificed upon termination of data collection. Animal care protocols were approved by the VA Loma Linda Healthcare System.

Stimuli were generated digitally (System III, Tucker-Davis Technologies; 200 kHz sampling rate) and delivered through one of the sound sources of an OAE probe system (ER10X, Etymotic Research). Clicks of ~100-μs duration (Fig. 1A) were shaped in the frequency domain using a recursive exponential window to generate a flat spectrum (3-dB bandwidth 11–28 kHz) at the OAE probe microphone. The microphone output voltage was high-pass filtered, amplified, and corrected for the microphone transfer function. Similarly, the LDV output voltage was high-passed and corrected for the LDV sensitivity.

FIGURE 1.

Ear-canal pressure (panels A, C, E) and BM velocity (panels B, D) responses to click stimuli. The pressure waveform of the click stimulus (75 dB peSPL) at the OAE probe microphone is shown in A, and the corresponding BM and CEOAE responses are shown in B and C using black dashed lines. The first 0.5 ms of the CEOAE response was windowed out to remove the evoking click artifact. The frequency spectra of the response waveforms at varying click levels (from 75 to 35 dB peSPL, thick to thin lines, see numbers in D) are shown in D (BM) and E (CEOAE). The responses to 81-dB peSPL click were scaled down and subtracted from the lower level responses to reveal the part of the response that grows nonlinearly with stimulus level (red in panels B–E) [1].

The “double-click” paradigm was utilized for stimulus presentation [2]. Four different stimuli (buffer size: 2048 samples) were presented within each block in an interleaved manner: [C₀], [C₀C₀], [C₀C_π], and [2C₀], where the subscripts indicate the phase shift of the click (C) stimulus. Unless stated otherwise, click levels were 75 dB peSPL. The two clicks in each double-click pair (i.e., in [C₀C₀] and [C₀C_π]) were separated by a fixed ICI (varied from 0–0.5 ms in steps of 5.1 μs and for longer ICIs in larger steps up to 20 μs). ICI conditions were presented in a random order. Although the first click in each stimulus buffer was presented with a 2-ms delay (re. the beginning of the buffer), the time origin for all analyses was placed at the centroid of the evoking click. Stimulus blocks were repeated 1200 times and the noisiest measurements (~6–10%) were later identified and excluded from the average. The first three stimulus conditions were used to evaluate the suppression of the click responses; the stimulus [2C₀] was included as a control measure (see below).

Suppression was evaluated by comparing the click-alone response (v_C0) for each measurement channel (i.e., OAE probe output and LDV output, v) with the corresponding derived-suppressed (DS) response, defined as

$$v_{\text{DS}}(\pm \text{ICI},t)=\frac{1}{2}\left[v_{{\mathrm{C}}_0{\mathrm{C}}_0}(\text{ICI},t)\pm v_{{\mathrm{C}}_0{\mathrm{C}}_{\pi }}(\text{ICI},t)\right].$$ (1)

The DS waveform represents the response to the evoking click obtained in the presence of the suppressor click. In our definition of v_DS(ICI, t), ICI is measured relative to the centroid of the evoking click. Thus, negative ICIs mean that the suppressor click precedes the evoking click and positive ICIs mean that the suppressor click follows the evoking click. Because the phase of C_π in the second click pair is shifted by 180 degrees, the response to the suppressor click is (ideally) canceled from the derived-response [1].

Temporal suppression was quantified in two ways. First, the difference between the rms amplitudes of the v_C0 and v_DS waveforms was calculated (in dB) for each ICI. Due to temporal overlap between the stimulus waveform and the CEOAE response, the first 0.5 ms of the response was windowed out before calculating the rms suppression (see caption of Fig. 2). Although rms suppression quantifies the overall magnitude and direction of the change due to the suppressor click (i.e., reduction or augmentation of the evoked response), the metric ignores any phase and/or frequency-specific effects. To circumvent these limitations, we calculate the suppression residual waveform as the difference between the v_C0 and vDS response waveforms in the time domain (i.e., [v_SR(ICI, t) = v_C0 (ICI, t)−v_DS(ICI, t)]). The suppression residual waveform captures any change in the response due to the suppressor click (e.g., phase shifts). If the suppressor click has no effect, v_SR(ICI, t) has zero amplitude (or, more realistically, falls into the noise). The residual waveforms were converted to the frequency domain (via fast Fourier transform) and the residual levels at specific frequencies were plotted against ICI to form residual suppression curves. Note that because v_SR(ICI, t) is free from stimulus artifacts due to the evoking click, no supplemental windowing is needed and the first ~0.5 ms of the response can be preserved in the analysis. For each nonzero ICI a control residual waveform was extracted as the difference between v_C0 and ½ v_2C0; this is equivalent to extracting the residual at ICI = 0.

FIGURE 2.

Change in the rms amplitude of the BM click response (A) and CEOAE (B) due to presentation of the suppressor click at varying ICIs (x-axis); raw data: thin lines, gently smoothed: thick). Response waveforms were windowed with 10^th-order recursive exponentials (from 0.5–1.5 ms for CEOAE and from −0.5–2 ms for BM velocity) before calculating the rms amplitudes (N.B.: extending the window duration up to 5 ms has little effect on the results). The insets show the modulation of the rms suppression at fixed ICI caused by varying the phase of the suppressor click (x-axis). At 0 phase shift, the two ICI conditions shown correspond to a local minimum (ICI = −133 μs, solid) or a maximum (−118 μs, dashed) in BM rms suppression. Representative evoking click (v_C0, dashed black) and derived-suppressed (vDS, solid red; see Eq. 1) time-waveforms are shown for BM velocity (C, G) and CEOAE (E, I) for both negative (left) and positive (right) ICI. The bottom panels (D, H and F, J) show corresponding suppression residual waveforms in solid red for the ICI condition listed and in dashed black for zero ICI (control test). The down-pointing triangles mark the centroid of the suppressor click. Noise floors (i.e., response waveform delayed by additional 5 ms) are shown in gray.

RESULTS

Basilar-Membrane and Ear-Canal Responses to a Click Stimuli

In two of the four animals tested, opening the cochlea abolished measurable CEOAEs and data collection was aborted. Although CEOAEs were preserved in the other two animals, only in animal GN111 was the change less than 3 dB near the BM characteristic frequency (CF); the data for this animal are presented here. CEOAE temporal suppression was measured in five other animals with intact cochleae, and the results were similar to the data from GN111.

Figure 1 shows examples of BM and ear-canal responses to click stimuli. The BM velocity response to an acoustic click (B, D; dashed black) is similar to click responses reported previously in gerbils and other species at basal locations [7, 8]. The early and late portions of the BM response grow linearly with click level (B; red; see caption). The early part of the response displays a frequency modulation (glide), with the instantaneous frequency increasing from 9–14 kHz over the time period from 0.1–0.5 ms (the CF is ~14 kHz). The secondary response lobe visible in the BM time waveform near 1.2 ms likely results from intracochlear reflection of traveling wave, as previously reported for chinchillas [9]. In the frequency domain (D), the BM responses demonstrate compressive nonlinearity near the CF (abolished post mortem). At the CF, the phase-gradient delay increases from ~5.9 to 7 periods of the CF with decreasing click level.

Due to temporal overlap between the evoking click and OAE pressures at the microphone, the early part of the CEOAE waveform cannot be directly visualized (Fig. 1C; black dashed). To reduce contamination from the stimulus pressure, the nonlinear CEOAE response was extracted (C, E; red; see caption for details). The CEOAE phase-gradient delay varies little with stimulus level and is roughly ~6 periods in the 9–15 kHz frequency range. When compared at the BM CF, the CEOAE phase-gradient delay was shorter than expected based on round-trip travel times. However, the significant variability inherent in CEOAE phase-gradient delays [10] implies that a larger sample size is needed to determine reliable relationships between mean BM and CEOAE phase-gradient delays.

Temporal Suppression of the Basilar Membrane and CEOAE Responses

Many characteristics of human CEOAE temporal suppression are also present in the gerbil otoacoustic and mechanical responses (Fig. 2). In both the BM and CEOAE responses, maximal suppression occurs at nonzero ICIs (−87 μs for BM velocity; −113 μs for CEOAE; A and B, thick lines). The reduction in the click response due to the preceding suppressor click usually has little effect on the response phase (C and E, for BM velocity and CEOAE measurements, respectively).

The rms suppression curves for both BM and CEOAE measurements display a quasi-periodic pattern of local maxima and minima (A, B, thin lines). Local maxima in the BM rms suppression curve occur at spacings of ~36 μs, an interval corresponding to one-half the period of the CF. The pattern of local suppression maxima and minima can be altered by shifting the relative phases of the evoking and suppressor clicks (inset in A). For example, local maxima in rms suppression (inset, dashed line) become local minima when the suppressor click phase is shifted by 90 degrees but are not affected by a 180 degree shift. These results are consistent with the interpretation that microstructure in the temporal suppression curve arises from interactions between the oscillating BM responses produced by the evoking and suppressor clicks. The periodicity of the CEOAE rms suppression curve was less pronounced than that seen in BM motion, plausibly due to broadband nature of the OAE response. Fourier analysis of the CEOAE suppression microstructure curve (thin line minus thick line in B) reveals two broad peaks corresponding to periodicities of ~38 and ~43 μs. The modulation of CEOAE rms suppression with suppressor-click phase was less consistent (B, inset) than that seen in the single-location BM measurements.

When the suppressor click follows the evoking click (i.e., ICI > 0), the rms suppression of the BM velocity decreases monotonically with increasing ICI (Fig. 2A). The envelope of the BM response to the suppressor click presumably peaks at about ~300 μs after it is presented (Fig. 1B), and this is the time when the suppressor click has the largest effect on the BM response (e.g., Fig. 2G). In contrast, CEOAEs at positive ICIs shows a region of rms augmentation rather than suppression (B, note rms suppression < 0; also see Fig. 2I). Although augmentation might arise artifactually due to imperfect cancellation of the suppressor click from the derived response (Eq. 1), this seems an unlikely explanation of the data because (i) no artifact is evident at the time of the suppressor click (see the down-pointing triangle in Fig. 2); (ii) suppression disappears post mortem (verified for ICI of 200 μs); and (iii) the rms suppression curve has microstructure at positive as well as negative ICIs. The origin of the augmentation remains unclear.

Examples of residual suppression waveforms obtained for nonzero (red) and zero (dashed black) conditions are shown in Fig. 2D,H for BM velocity and in F,J for CEOAEs. For each ICI condition, the BM and CEOAE residual waveforms were transformed to the frequency domain and the level of the suppression residual at a given frequency was plotted versus ICI (Fig. 3, BM and CEOAE in top and bottom panels, respectively; analyses frequency listed on the top). Independent of the analysis frequency (< CF, at CF, and > CF, left to right), the BM residual suppression curves demonstrate a pattern of peaks and valleys with period equal to ~36 μs. At frequencies at and below CF, the BM residual suppression curves reach their maxima at negative ICIs (−67 and −174 μs corresponding to 1 and 1.8 periods of the respective analysis frequencies), but smaller, secondary peaks are often present at positive ICIs. At frequencies above CF, the BM suppression residual reaches its maximum at a positive ICI (240 μs or 3.9 periods) and displays a smaller peak at the negative side (−118 μs or 1.9 periods). In all cases, if the suppression residual is measurable for ICI ⩾ 0.5 ms it represents the effect of the suppressor click on the second “ringing” lobe of the BM click response (Fig. 1B, dashed black). At the CF, the phase of the BM residual suppression curve varies little with ICI (not shown), suggesting that the suppressor click predominantly affects the amplitude of the response (e.g., see Fig. 2D and H—the suppression residual waveforms for nonzero and zero ICI conditions are in phase). Below CF, the phase of the residual suppression curve is constant for ICI < 0 but lags for ICI > 0 (over ~13 cycles between 0 and 0.6 ms ICI). In contrast, above CF, the phase of the residual curve is constant, except for ICIs between −0.2 and 0 ms where it lags over ~13 cycles. Thus, at analysis frequencies either below or above CF, the secondary (i.e., smaller) peak in the residual suppression curve results in part from the suppressor click affecting the phase of the BM-click response at that frequency.

FIGURE 3.

Examples of temporal-suppression curves (red) constructed for single frequency components (columns) of the suppression residual spectra for BM velocity (upper panels) and CEOAEs (bottom panels). Examples of the suppression residual waveforms are shown in Fig. 2D,H,F,J. The residual waveforms were windowed in the time domain before calculating their spectra (see caption to Fig. 2; for the CEOAE residual waveform the window was expanded to start at −0.05 ms). The level of the residual extracted with the control test is shown with the black dashed line. Noise levels are shown in gray. Raw data (thin lines) were gently smoothed (thick) to better visualize the trends.

The CEOAE residual suppression curves qualitatively resemble the BM curves. They display periodic microstructure and are often double-peaked, with maximal suppression occurring at nonzero ICIs (Fig. 3B, D, and F). The fine-structure period in the CEOAE residual suppression curves varies with analysis frequency: the higher the frequency, the shorter the period (varying from 45 to 29 μs as frequency increases from 9.6 to 16.11 kHz). The period of the suppression microstructure tends to be somewhat less than half the period of the analysis frequency, unlike for BM data at the CF. However, as observed for the BM velocity, the level of the residual at a given ICI varies periodically with the phase of the suppressor click—valleys became peaks, and vice versa, when the suppressor click phase is shifted by 90 degrees (data not shown). This suggests that the microstructure of the CEOAE residual suppression curves arises from local interactions of the oscillating BM responses to the evoking and suppressor clicks. The CEOAE suppression curves usually reach their maximum for positive ICI (ranging from 220 to 277 μs or from 2.6 to 3.9 periods of the analysis frequency). If present, the peak on the negative ICI side occurs at shorter ICIs (−195 and −143 μs or 1.9 and 2.1 periods for frequencies of 9.6 and 14.1 kHz, respectively). The phase of the CEOAE residual suppression curve varies little with ICI for ICI < 0 and lags for ICI > 0 (over ~13 cycles from 0–0.6 ms ICI for frequencies of 9.6 and 14.1 kHz; for the 16.3 kHz response, the signal-to-noise ratio is poor, obstructing phase trends). This contrasts with the BM data, where very little phase rotation occurs at CF. The change in suppression residual phase at positive ICIs is related to the augmentation of the CEOAE response in the presence of the suppressor click (Fig. 2B, J).

DISCUSSION

In this study, we explored whether the shift in the maximal temporal suppression of CEOAE from nonzero ICIs can be understood in terms of BM mechanics. Data obtained in a sensitive gerbil cochlea reveal that, as with CEOAE responses, the suppression of BM click responses is greatest for clicks separated in time. When the suppressor click precedes the evoking click (ICI < 0 ms), the suppression of the CEOAE response resembles the BM results at the CF. In both cases, the residual suppression curve peaks at nonzero ICIs, displays periodic microstructure, and has a phase independent of ICI. However, BM residual suppression peaks at shorter ICIs than the CEOAE curve (−67 versus −154 μs or 1 versus 2 periods at the CF; Fig. 3C and D). In addition, the period of the CEOAE residual curve microstructure appears smaller than for the BM at the CF. These results suggest that whereas the CEOAE temporal suppression for ICI < 0 ms originates in BM nonlinearities, the suppression characteristics of single-frequency CEOAE components do not match the single-location BM responses. When the suppressor click follows the evoking click (ICI > 0 ms), augmentation, rather than suppression, is observed in the CEOAE data (Fig. 2B). In contrast, only suppression is observed in the BM responses (Fig. 2A). Although it remains unclear what causes the augmentation of the CEOAE response for these conditions, the augmentation appears unlikely to be artifactual (e.g., based on post mortem controls and measurements in acoustic test cavities).

The occurrence of maximal BM temporal suppression at nonzero ICIs may result from delays in the activation of cochlear amplification [12]. Interestingly, the ICI at peak suppression appears to depend on the strength (or “health”) of cochlear nonlinearities. In another animal, opening the cochlea greatly diminished but did not completely abolish nonlinear responses in BM motion. In this animal, the BM temporal suppression curve at CF peaked at 0 ms ICI. Future experiments might probe the dependence on the strength of the nonlinearities by controlling the degree of sensitivity loss (i.e., by inducing varying amounts of acoustic trauma).

Prior to this report, CEOAE temporal suppression has been measured only in humans. Here we show that gerbil CEOAEs are also maximally reduced when the suppressor click precedes the evoking click. However, in gerbil, the residual suppression curves tend to peak at ICIs corresponding to ~2 periods of the analysis frequency, while in humans, the peak occurs at ICIs corresponding to 4–5 periods (for the 3–8 kHz frequency range, [4]). This result suggests that the human cochlea has more prolonged BM impulse responses and sharper frequency tuning than common laboratory animals [13, 11]. When the suppressor click followed the evoking click, gerbil CEOAEs appear augmented—a result not observed in human data obtained with the same “double-click” paradigm. One possibly relevant difference between human and gerbil CEOAEs is that the former are often dominated by synchronized emissions that can ring for over 180 ms. Synchronized or spontaneous emissions were not detected in any of the gerbils tested to date.