An alternate approach to constructing distortion product otoacoustic emission (DPOAE) suppression tuning curves

Abstract

DPOAE suppression tuning curves (STCs) typically are constructed using suppression criteria of 3 or 6 dB. This paper describes a technique for generating DPOAE STCs using criteria ranging from 3 dB to complete suppression. The criterion effect was examined at various primary levels (f₂ = 4 kHz) in normal-hearing ears. As expected, Q_ERB and tip-to-tail differences decreased as probe level (L₂) increased. Q_ERB values were not systematically dependent on criterion. For low L₂’s, tip-to-tail differences decreased as criterion increased. Given similarities in methodology, DPOAE STCs constructed from completely suppressed responses might be most appropriate for comparison to psychophysical tuning curves.

I. INTRODUCTION

Distortion product otoacoustic emissions (DPOAEs) have been measured using a suppression paradigm. Suppression tuning curves (STCs) constructed from these measures may be compared to tuning curves obtained from other methods, such as psychophysical masking experiments in humans. Because the slope of functions relating amount of suppression to suppressor level varies with suppressor frequency, the shape of the DPOAE STC constructed from these functions might depend on the suppression criterion used to construct it. DPOAE STCs frequently are constructed using suppression criteria of either 3 or 6 dB (e.g., Abdala, 2001; Abdala and Fitzgerald, 2003; Abdala et al., 1996; Brown and Kemp, 1984; Gorga et al., 2002; Gorga et al., 2003; Martin et al., 1998b, 2003). These criteria, however, represent only partial reduction of the DPOAE response and, as such, are not directly analogous to tuning curves generated from psychophysical masking experiments in which the probe is rendered inaudible. In this paper, we propose a method for generating DPOAE STCs, based on the fully suppressed condition, that may be more similar to the conditions from which psychophysical tuning curves are constructed. We also examine the influence of suppression criterion on the shape of the DPOAE STC as described by Q_ERB and the tip-to-tail difference.

II. METHODS

A. Subjects

Data from 22 ears of 20 subjects with normal hearing were included in the analyses presented here. Normal hearing was defined as thresholds ≤ 15 dB HL (re: ANSI, 1996) at 4 kHz. The mean threshold at 4 kHz was 5.7 dB HL.

B. Stimuli

Stimulus generation (and data collection) was accomplished using custom designed software (EMAV, Neely and Liu, 1993) that controlled a sound card (CardDeluxe, Digital Audio Labs) housed in a PC. The stimuli were delivered via loudspeakers housed in an ER-10C probe-microphone system. The probe frequency (f₂) was fixed at 4 kHz with a primary frequency ratio (f₂/f₁) of 1.22. The level of f₂ (L₂) varied in 10-dB steps from 20 to 70 dB SPL. The level of f₁ (L₁) was set according to the equation L₁ = 0.4L₂ + 39 dB (Kummer et al., 1998). Seventeen suppressor frequencies (f₃) were used, varying from 1 octave below to 1/2 octave above f₂. The level of the suppressor (L₃) was varied in 5-dB steps from 5 to either 85 or 90 dB SPL, depending on f₃.

C. Data collection

For each L₂, DPOAE levels were measured first in response to the primaries presented alone (the control condition). Next, the primaries were presented in the presence of a fixed suppressor frequency (f₃) that varied in level. This process was repeated until the f₃-level sequence had been repeated for each of 17 f₃’s and each of 6 L₂’s.

D. Analysis

Following data collection, DPOAE levels were converted to decrements (amount of suppression) by subtracting the level measured during each suppression condition from the level measured during the control condition. Mean decrement-versus-L₃ functions for 3 f₃’s (one below, one approximately equal to, and one above probe frequency, f₂) are shown as solid lines in the upper panel of Fig. 1. In this figure, f₂ = 4 kHz and L₂ = 40 dB SPL. Because we were interested in constructing DPOAE STCs for a range of suppression criteria, including the condition where the response is fully suppressed, the maximum amount of decrement was calculated for each L₂. The signal-to-noise ratio (SNR) in the control condition (28 dB at this L₂) defines the maximum measurable amount of suppression and was used to define the point at which the response was suppressed to the noise floor. In the top panel of Fig. 1, the upper horizontal line is equivalent to the control-condition SNR when L₂ = 40 dB SPL. The lower horizontal line in the same panel indicates the point at which the response had been decremented (suppressed) by 3 dB.

Fig. 1.

Illustration of the approach to data analysis. In both panels, the probe frequency (f₂) = 4kHz and the probe level (L₂) = 40 dB SPL. Upper panel: Mean decrement-versus-L₃ functions for 3 f₃’s are shown as solid lines. The upper horizontal line represents the maximum amount of decrement and was determined by calculating the SNR in the control condition (28 dB in this example). This horizontal line, therefore, represents the point at which the response is fully suppressed into the noise floor. The lower horizontal line represents the point at which the response has been suppressed by 3 dB. The dashed lines represent the linear fits to the transformed data (as described in the text). The linear fits were solved for a range of decrements. The 3-dB decrement points are shown in the figure as open symbols for three different suppressor frequencies. The corresponding fully suppressed points are shown as closed symbols. Lower panel: Mean DPOAE STCs plotted as suppressor level (L₃) versus suppressor frequency (f₃) for two different suppression criteria, 3 dB (lower curve) and complete suppression (upper curve). The symbols plotted on the STCs correspond to the symbols displayed in the upper panel. The points on the upper curve above the horizontal, dashed line represent extrapolated values of L₃ (see text for more details).

Before generating DPOAE STCs, the decrement-versus-L₃ functions were transformed according to the following equation:

$$\mathrm{D}=10\log ({10}^{\mathrm{decr}∕10}-1).$$ (1)

When D = 0, the decrement is 3 dB. This equation served to linearize the functions and allowed points with high SNRs (i.e., decrements between 1 and 3 dB) to be included in the generation of a linear model for the data. The transformed data were then fit with a linear equation with the constraint that only decrements ≥ 1 dB were included in the fit and that there must be at least 3 L₃’s producing decrements greater than 1 dB. Additionally, only functions including at least one condition for which at least 3 dB of suppression occurred were transformed with a linear equation. The r² values resulting from these fits ranged from 0.8 to >0.99 with 86% greater than 0.9. Those r² values falling below 0.9 were equally divided between low- and high-frequency suppressors. The linear equations were subsequently solved for decrements ranging from 3 dB to the fully suppressed condition (i.e., the point when the decrement equals the SNR in the control condition). DPOAE STCs were generated using these different suppression criteria.

These approaches to data analysis are also illustrated in Fig. 1. In the upper panel, the linear fits to the transformed data are shown as dashed lines. The open symbols represent 3-dB decrements (calculated from these linear fits) for the three suppressor frequencies shown in the figure. The closed symbols represent the corresponding complete-suppression condition. In the lower panel, these points are re-plotted as a function of frequency. The lines shown as tuning curves represent spline fits to these data, in combination with the data from the other 14 suppressor frequencies. The STCs that were generated using decrement criteria of both 3 dB and complete suppression are shown. It is important to note that, for some conditions (especially when f₃ > f₂), it was not always possible to fully suppress the DPOAE response, even when L₃ = 85 or 90 dB SPL. However, using the approach described above, it was possible to extrapolate to the L₃ that would be expected to fully suppress the response. In the lower panel, points above the horizontal, dashed line represent extrapolated values of L₃. The greatest difference in the STCs constructed from the 3-dB and fully suppressed criteria was observed when f₃ > f₂, less when f₃ ≈ f₂, and least when f₃ < f₂.

III. RESULTS

A. Suppression tuning curves

Mean DPOAE STCs are shown in Fig. 2, with data for a different L₂ shown in each panel. DPOAE STCs are shown for suppression criteria ranging from a decrement of 3 dB to the fully suppressed condition (in 6-dB increments). In each panel, the lowest curve represents the DPOAE STC constructed from the 3-dB criterion and the highest curve represents the fully suppressed condition. In each panel, the circles shown on the upper- and lower-most curves indicate f₃’s where the decrement-versus-L₃ function met the linear-fit inclusion criteria. Even for the highest L₂ (70 dB SPL), where it was more difficult to suppress the response, linear-fit inclusion criteria were met for 10 suppressor frequencies. At lower L₂’s, the inclusion criteria were met for more suppressor frequencies. Data falling above the dashed, horizontal lines represent extrapolated values of L₃, as described above. As the criterion progressed from minimum (3-dB decrement) to maximum suppression, there was a systematic upward shift in the STC. At low L₂’s, there tended to be a greater shift at the tip than on the low-frequency tail of the tuning curve. This effect was less evident at higher L₂’s. Additionally, the low- and high-frequency slopes of the STC’s changed as the suppression criterion was varied. As the suppression criterion increased, the slope of the high-frequency flank became steeper while the slope of the low-frequency flank became shallower.

Fig. 2.

Mean DPOAE STCs. Each panel represents data for a different L₂. Within each panel, STCs are shown for suppression criteria ranging from 3 dB (open symbols) to complete suppression (closed symbols) in 6-dB increments. As in Fig.1, points falling above the horizontal, dashed lines do not represent measured suppressor levels, but rather are values extrapolated from the linear fits to the transformed decrement-versus-L₃ functions.

B. Q_ERB and Tip-to-Tail Difference

The upper panel of Fig. 3 plots the Q_ERB values for the mean DPOAE STCs as a function of the suppression criterion used to construct the DPOAE STC. The lower panel plots the tip-to-tail difference for the same conditions, where the tip was defined as the lowest L₃ on each curve and the tail was defined as the L₃ when f₃ = 2 kHz (an octave below f₂). Within both panels, the parameter is L₂. Q_ERB is defined as the best frequency (BF) divided by the equivalent rectangular bandwidth (ERB). For any filter, the corresponding ERB is the bandwidth of the rectangular filter with the same BF response that passes the same total power. Q_ERB values for low probe levels were larger (indicating sharper tuning) than Q_ERB values for higher probe levels. The criterion used to construct the DPOAE STC had little systematic influence on Q_ERB. Across all criteria, the tip-to-tail difference decreased as L₂ increased, going from a maximum of about 45-48 dB for L₂ = 20 dB SPL down to approximately 16 dB for L₂ = 70 dB SPL. For L₂ = 20-50 dB SPL, the tip-to-tail difference decreased as the DPOAE response was more fully suppressed. At higher L₂’s, the suppression criterion did not influence the tip-to-tail difference.

Fig. 3.

Q_ERB (upper panel) and tip-to-tail difference (lower panel) as a function of suppression criterion (dB). In both panels the parameter is L₂.

IV. DISCUSSION

With the exception of the tip-to-tail differences at L₂ ≤ 50 dB SPL, the criterion used to construct the DPOAE STC had little influence on measures of cochlear tuning derived from the DPOAE STC (e.g., Q_ERB). Given the variation in the slopes of the decrement-versus-L₃ functions across suppressor frequency (see Fig. 1), it was expected that the shape of the DPOAE STCs constructed using different suppression criteria would change. Indeed, there are several reports in the literature that estimates of tuning became sharper as the criterion for constructing the DPOAE STC was varied from minimal suppression to greater amounts of suppression (e.g. Kummer et al., 1995; Martin et al., 1998a; Taschenberger and Manley, 1998). In the present data, the slope of the high-frequency side of the DPOAE STC increased as the response was more fully suppressed; however, this effect had little influence on estimates of tuning (Q_ERB) around the tip of the STC.

Tip-to-tail differences on DPOAE STCs previously have been viewed as measures related to “cochlear-amplifier gain” (Mills, 1998; Pienkowski and Kunov, 2001), and have been shown to decrease as probe level increases (Gorga et al., 2002, 2003). The decrease in gain, based on these measures, is a consequence of differences in growth of response to on- and low-frequency suppressors. The observations in the present study, in which tip-to-tail differences decreased as the response is more fully suppressed, is a consequence of the same underlying processes. That is, growth of response to suppressors below probe frequency (f₂) was more rapid, compared to suppressors close to f₂. As the criterion increases, this frequency-dependent difference in response growth results in a decrease in the difference between the L₃ required to produce the criterion amount of suppression for on- and low-frequency suppressors.

In our view, the most important result was the demonstration of an approach that enabled us to construct DPOAE STCs from any suppression criteria, including the fully suppressed condition. Because decrement-versus-L₃ functions were well described by linear fits to transformed data, it was possible to extrapolate to suppressor levels that were above the levels that could be produced by the hardware. Assuming this extrapolation was appropriate, this enabled us to extend the dynamic range of our measurements. Furthermore, by using the entire decrement-vs.-L₃ function to develop a linear model, reliable estimates of L₃ were possible for all conditions, including those for which the response was close to the noise floor (i.e., those conditions approaching the fully suppressed condition).

While there was little systematic influence of criterion on estimates of Q_ERB from DPOAE STCs, criterion did influence estimates related to cochlear-amplifier gain, particularly at the low to moderate L₂’s that are similar to the levels used to construct psychophysical tuning curves. Furthermore, constructing DPOAE STCs based on the fully suppressed criterion may be methodologically more similar to conditions under which psychophysical tuning curves are constructed. In psychophysical tuning-curve experiments, the probe frequency and level are fixed and the level of the masker is varied until the probe is “at threshold” or just audible. Contrast this condition with the DPOAE condition, in which the response is only partially suppressed. The need to present probe and suppressor simultaneously complicates comparisons to other measurements that describe only the “excitatory” pattern of response. However, the completely suppressed response may be a more appropriate criterion for constructing DPOAE STCs when comparisons will be made with tuning curves constructed from threshold measurements such as psychophysical masking experiments.