Comodulation masking release in the inferior colliculus by combined signal enhancement and masker reduction

The detection of comodulation, i.e., coherent level fluctuations in different frequency regions, is an important feature of speech recognition. In this study, we demonstrate how the representation of a signal in comodulated masking conditions changes along the auditory pathway by using a stimulus paradigm from the cochlea nucleus for the first time in the inferior colliculus. This happens on a timescale that makes corticocollicular feedback a likely candidate as the source.

Abstract

Auditory signals that contain coherent level fluctuations of a masker in different frequency regions enhance the detectability of an embedded sinusoidal target signal, an effect commonly known as comodulation masking release (CMR). Neural correlates have been proposed at different stages of the auditory system. While later stages seem to suppress the response to the masker, earlier stages are more likely to enhance their response to the signal when the masker is comodulated. Using a flanking band masking paradigm, the present study investigates how CMR is represented at the level of the inferior colliculus of the Mongolian gerbil. The responses to a target signal at various sound pressure levels in three different masking conditions were compared. In one condition the masker was a 10-Hz amplitude modulated sinusoid centered at the signal frequency while in the other two conditions six off-frequency carriers (flanking bands) were added. For 64 of a total of 94 units, the addition of comodulated flanking bands to the on-frequency masker did not change the response to the target signal. The remaining 30 units showed a change that enhanced target detectability if coherent flanking bands were added, indicative of CMR. The current data demonstrate that the response characteristics of these neurons represent an intermediate stage between the representation in the cochlear nucleus and the auditory cortex by increasing the response during the signal intervals and decreasing the response for the following masker portions.

NEW & NOTEWORTHY The detection of comodulation, i.e., coherent level fluctuations in different frequency regions, is an important feature of speech recognition. In this study, we demonstrate how the representation of a signal in comodulated masking conditions changes along the auditory pathway by using a stimulus paradigm from the cochlea nucleus for the first time in the inferior colliculus. This happens on a timescale that makes corticocollicular feedback a likely candidate as the source.

many natural sounds contain coherent envelope fluctuations in different frequency regions (Nelken et al. 1999). The ability of the auditory system to use these stimulus characteristics as a cue is illustrated, among others, in experiments on comodulation masking release (CMR) (Hall et al. 1984; for a review, see Verhey et al. 2003). CMR experiments investigate how the detection of a masked narrow-band target signal, usually sinusoid, is affected by masker comodulation, i.e., coherent masker envelope fluctuations across frequency. CMR describes the effect that the detection threshold of the target is often lower when the masker is comodulated than in a masking condition where the masker is not comodulated. The exact definition of CMR varies across studies and depends on the masking paradigm that was used as the baseline condition. CMR was primarily investigated in humans but was also observed in other species, e.g., Mus musculus (Klink et al. 2010), European starlings (Sturnus vulgaris) (Klump and Langemann 1995), and bottlenose dolphins (Tursiops truncatus) (Branstetter and Finneran 2008). In humans, CMR was not only shown for artificial signals as sinusoids embedded in noise but also for band-limited speech signals (Grose and Hall 1991). Although the hypothesized underlying mechanism (across-frequency processing) may have a small impact on speech reception threshold (Festen 1993), it seems to be important for the grouping of the speech itself (Carrell and Opie 1992).

Neuronal correlates of CMR were reported at different stages of the auditory system. At the level of the cochlear nucleus (CN), some neurons showed a reduced response to the masker when it was comodulated, which facilitated the detection of an embedded target signal. The reduced masker response was presumably a consequence of wideband inhibition (Neuert et al. 2004; Pressnitzer et al. 2001). Data recorded from higher levels of the auditory system indicate that, in some neurons, the presence of the target signal suppresses the response to the masker. This effect is commonly referred to as envelope locking suppression (ELS, Las et al. 2005; Nelken et al. 1999). Las et al. (2005) showed that ELS does not occur at the level of the IC. Taken together, the data indicate that the neural correlate of CMR changes along the auditory pathway, but comparing the results is difficult because two different CMR paradigms were used.

The CN studies used the flanking-band type of experiment (Dau et al. 2009; Grose and Hall 1989; Moore et al. 1990; Verhey et al. 2013). This type of experiment uses narrowband masker components and usually compares three different masking conditions. The comodulated (CM) condition contains an on-frequency masker which is centered at the frequency of the target signal and flanking bands (FBs) at frequencies remote from the target frequency with the same level fluctuations (envelope) as the on-frequency masker. A common baseline or reference (RF) condition is the masking condition where the masker only consists of the on-frequency masker (Hall et al. 1990). To investigate the importance of the coherent envelope, signal detectability in the CM condition is often also compared to a masking condition where the masker has the same spectral content as the CM masker but where the FBs and the on-frequency masker do not share the same envelope. When the FBs have the same envelope this condition is referred to as the codeviant (CD) condition. Some studies use two definitions of CMR: one is the difference in threshold between the RF and CM condition, and the other is the difference between the thresholds in the CD and the CM condition (e.g., Schooneveldt and Moore 1989). Since RF thresholds are usually lower than CD thresholds the first measure (RF-CM) may be considered as the more conservative estimate of CMR. Following this line of argument, the difference between the CM and the RF masking condition was used in this study as a measure of CMR.

Envelope locking suppression was investigated with the band-widening type of CMR experiment (Hall et al. 1984; for a review, see Verhey et al. 2003). In this paradigm, signal detectability is measured in the presence of a single band-pass noise masker for various bandwidths. CMR is the difference in masking of an unmodulated and a (co-)modulated masker. To use different paradigms for the study of CMR at different stages of the auditory system may be problematic since psychoacoustical modeling studies indicate different mechanisms underlying the masking release in the band-widening and flanking-band types of CMR experiment (Dau et al. 2013; Piechowiak et al. 2007; Verhey et al. 1999). Thus it is unclear if the different neural response characteristics reflect a change in coding strategy along the auditory pathway or are simply due to differences in the masking paradigm.

The present study measures physiological correlates of CMR at the level of the inferior colliculus (IC) using essentially the same flanking-band paradigm as used in the CN studies on CMR. This allows the investigating of how the neural representations change from the CN to the IC. For quantitative analysis we employed an evaluation method that is sensitive to both mechanisms, i.e., an enhanced signal representation due to wideband inhibition and envelope locking suppression-like effect due to the presence of the masker.

MATERIALS AND METHODS

Experiments were performed on 27 adult male ketamine/xylazine anesthetized Mongolian gerbils (Meriones unguiculatus) with an age of 3–16 mo and a body weight of 80–120 g. All experiments were conducted in accordance with the international National Institute of Health Guidelines for Animals in Research and with ethical standards for the care and use of animals in research defined by the German Law for the protection of experimental animals. All experiments were approved by an ethics committee of the State of Saxony-Anhalt, Germany.

Surgical procedure.

Anesthesia was a mixture of 45% ketamine (50 mg/ml; Ratiopharm), 5% xylazine (Rompun, 2%; BayerVital), and 50% isotonic sodium chloride solution (154 mmol/l; Braun). Initially, a dose of 0.3 ml was given intraperitoneally. To maintain a constant state during the surgery and later during the course of the experiment, it was applied subcutaneously at a rate of 0.1 ml/h, or if necessary. The state of the anesthesia was ensured by monitoring the hindlimb withdrawal reflex and respiratory rate (Zandieh et al. 2003). Body temperature was kept at 37°C using a controlled heating blanket. A craniotomy was performed 2 mm caudal and 2 mm lateral of lambda using the major blood sinus as a landmark (Brückner and Rübsamen 1995; Cant and Benson 2005); its size was about 4 mm. A gold-plated pin (Amphenol) with good electric contact to the dura was implanted in the contralateral parietal bone and served as voltage reference. For stereotactic fixation an aluminum bar was attached to the frontal bones with dental cement (Paladur; Heraeus Kulzer). Recordings were made with tungsten electrodes (2–4 MΩ, FHC) using a Plexon Multichannel Acquisition Processor or Omniplex System (Plexon). The electrode was lowered dorsoventrally until a multi-unit response to wideband noise or tone pips were observed. After this procedure, single units were identified via their signal-to-noise ratio (SNR) and postsorted using the software Offline Sorter (Plexon). To ensure that the position within the IC was the central nucleus, the tonotopic gradient in the multi-unit response was measured.

Stimulus generation and setup.

Acoustical stimuli were digitally synthesized, pseudo-randomized and controlled using Matlab (MathWorks), and amplified. Depending on the setup used on the experimental day the amplifier was either an AMP 75 wideband power amplifier (Thomas Wulf, Frankfurt, Germany) or an Alesis RA 150 amplifier (inMusic Brands) using a Canton Plus XS.2 speaker. The speaker was calibrated to ensure a flat response in the frequency between 0.25 and 16 kHz. The sound was presented in an acoustically and electrically shielded chamber. The distance between the speaker and the animal was either 1 m (setup with the AMP 75) or 70 cm (setup with the RA 150). The maximum presented sound level was 90-dB SPL at the position of the animals head, measured with a B&K microphone (4939).

Recording paradigm.

Once a single unit with good SNR was found, it was characterized on the basis of its frequency response area (FRA) using pure tones with a frequency in the range from 0.25 to 16 kHz. The tone duration was 100 ms including 5-ms raised-cosine-squared ramps at on- and offset. The signals were separated by silent intervals of 500 ms. The signal frequencies were equally spaced on a logarithmic scale in one-fourth octave steps. The signal intensities were varied from a highly suprathreshold level to below threshold in steps of 10 dB. Each combination of level and frequency was repeated 10 times.

After the FRA was measured, the unit was stimulated using the same flanking-band experiment that was used in Neuert et al. (2004) to investigate neural correlates of CMR at the level of the CN.

The target signal was a sequence of three 50-ms tone pips that were separated by 50-ms silence intervals. Each tone pip was gated on and off using 25-ms cosine-squared ramps. The signal frequency F_c was within the excitatory region of the unit and was close to the best frequency of the unit. Three different masking conditions were used: reference (RF), comodulated (CM), and codeviant (CD). All maskers were 500-ms long and had an on-frequency component (OFC) in common. This component is described by the following equation:

$$M_{\text{OFC}}=[1+\cos \left(2\pi \cdot 10t+\pi \right)]\hspace{0.17em}\sin \left(2\pi \cdot F_{\text{c}}\cdot t\right)$$ 1

with t the time. Thus the OFC was a sinusoidally amplitude-modulated sinusoidal carrier with a carrier frequency of F_c (from Eq. 1), a modulation frequency of 10 Hz, and a modulation depth of 100%. For this modulation depth and frequency it was shown that a sufficiently large fraction of units (more than 70%) in the IC should respond to sinusoidal amplitude modulations (Joris et al. 2004; Krishna and Semple 2000).

The three tone pips were positioned into the three last minima of the masker, i.e., the onsets of the tone pips were 175, 275, and 375 ms after masker onset. The signal level relative to the level of the OFC were 0, −10, −20, −30 dB, and −∞ (i.e., no signal was present). The latter level will be referred to as masker-only (MO) condition in the following.

For the CM condition, the masker consisted of this OFC and six further components at spectrally remote frequencies, referred to as flanking bands (FBs). The frequencies were chosen to be just outside the excitatory region of the unit according to the FRA and are likely to be inhibitory (Egorova and Ehret 2008). The FBs were amplitude-modulated pure tones (with carrier frequencies FB_i). The frequency and starting phase of the modulation were the same as those of the on-frequency masker. The comodulated masker is described by the following equation

$$M_{\text{CM}}\left(t\right)=M_{\text{OFC}}\left(t\right)+{\displaystyle \sum _{i=1}^6[1+\cos \left(2\pi \cdot 10t+\pi \right)]\times \sin \left(2\pi \cdot {\text{FB}}_{\mathit{\text{i}}}\cdot t\right)}$$ 2

In the present study, the frequencies of the FBs were spaced at least 100 Hz apart from each other. All masker components had the same level.

For the CD condition, the FBs had a starting phase of the modulation of 0 instead of π,

$$M_{\mathrm{C}\mathrm{D}}\left(t\right)=M_{\text{OFC}}\left(t\right)+{\displaystyle \sum _{i=1}^6[1+\cos \left(2\pi \cdot 10t\right)]\hspace{0.17em}\sin \left(2\pi \cdot {\text{FB}}_{\mathit{\text{i}}}\cdot t\right)}$$ 3

For the third masking condition RF, the masker consisted of the OFC only, i.e., the level of the OFC was the same for all conditions, and the overall level of the RF masker was lower than that of the two other conditions. The spectrograms of the different maskers are shown schematically in Fig. 1.

Fig. 1.

Schematic plot of 3D spectrograms of the masker (black) and target signal (gray) for the three masking conditions: reference (RF, left), comodulated (CM, middle), and codeviant (CD, right). The signal-to-noise ratio was set to 0 dB, which can be seen as the amplitude of the target signal is the same as that of the RF masker. In the middle and on the right only 2 flanking bands (FBs) are shown. The actual number of flanking bands that were used in this study was 6.

All stimuli were presented at least 25 times. For units with a stable response, 50 repeats were recorded. For a subset of units (11 units measured in 2 of 27 animals), the stimuli were repeated 20 times and the signal was played at three signal levels relative to the level of the on-frequency masker component (0, −10, −20 dB, and the MO condition). These relative signal levels will be referred to as signal-to-masker ratios in the following. For these units, the closest FB was placed 0.2, 0.5, 1, 2, 3, and, if the recording stability made it possible, 4 kHz above the signal frequency. The other FBs were placed in 0.1-kHz steps above the closest one. The silent interval between the stimuli was at least 500 ms. The FBs were only placed above the target signal to ensure that, for all units, the same set of spectral distances could be used, irrespective of the unit's characteristic frequency. Psychoacoustical CMR measurements on the effect of frequency of the FB on the magnitude of CMR indicate that, for some of the center frequencies of the OFC, CMR tend to be smaller for FBs spectrally above the OFC than for FB positioned below the OFC (Schooneveldt and Moore 1987). Thus it is possible that for some characteristic frequencies of the units larger correlates of CMR would have been found if FBs spectrally below the OFC had been used.

Data analysis.

Before being analyzed, the units were grouped into two subgroups; those that responded with more than half the spikes in the first 100 ms in the combined three MO and hence unable to show CMR were exempted in some of the later group comparisons where appropriate. This was done to avoid statistical underpowering while comparing, by comparing two responses with very few spikes against each other.

For the evaluation of each unit a delay time was determined that was optimal for the detection of the target signal. The target signal was presented in the minima of the masker, i.e., during the time intervals from 175 to 225, 275 to 325, and 375 to 425 ms after stimulus onset, to be called “signal interval” from here on. Conversely, the intervals from 225 to 275 ms, from 325 to 375 ms, and from 425 to 475 ms will be referred to as the “masker interval.” The optimum delay time minimizes the response in the MO conditions during the signal interval. The time was allowed to vary in a range from −10 to +30 ms. Negative delays were sometimes necessary because the amplitude modulation in the masker could cause units to respond early (see the masker scheme in Fig. 2 with the black line denoting the envelope of the RF masker). Units with a low detection threshold could respond to the on-frequency masker during the signal interval and hence a negative delay would be necessary. In the following analysis, the time scale was corrected for this delay.

Fig. 2.

The envelope of the RF masker only (solid line) and the target signal (interrupted line) at 0 dB signal-to-noise ratio. The line at the bottom denotes the signal (interrupted) and masker (solid) interval, which can be shifted by −10 to +30 ms to ensure a minimum of spikes in the MO condition.

The study used the difference between spikes in the signal interval and spikes in the masker interval as the criterion to differentiate between CMR and non-CMR representing units. For every signal level, it was tested whether 1) this difference was largest in the CM condition, and 2) there was an increase in the given signal response compared with the MO in the CM condition. A one-way ANOVA (MatLabR2012b) was performed to examine the statistical significance of the effect. The dependent variable was the masking condition (RF, CM, and CD). If the ANOVA showed a significant effect of condition, post hoc Scheffé's (multcompare; MatLabR2006a) tests were applied to test for significant differences between the masking conditions. If these tests showed significant differences in the aforementioned spike count between CM and RF (P < 0.05) for at least two signal levels, the unit was defined as a CMR unit. All units showing CMR according to this definition would also be classified as CMR units when the CM and CD were compared using the same method and parameters.

To facilitate the comparison to previous studies the data were also analyzed with two further methods. First, we employed the analysis approach presented by Neuert et al. (2004) where the difference in spikes in the signal interval were calculated between the response to the signal embedded in the masker and the response in the MO condition.

In addition and in contrast to the two previous methods, a d-prime analysis adapted from psychophysical studies and detailed in Pressnitzer et al. (2001) was used to analyze the data.

First, the PSTHs were constructed with a 25-ms bin width. Then, for each signal-to-masker ratio (except for the MO condition) the d-prime was calculated as follows

$${\mathrm{d\prime }}^{}=\sqrt{{\displaystyle \sum _i\frac{{\left({\overline{S}}_i-{\overline{NS}}_i\right)}^2}{0.5\left[\sigma {\left(S_i\right)}^2+\sigma {\left(N{S}_i\right)}^2\right]}}}$$ 4

with the bin number i, the number of spikes in that bin in the MO condition NS, and the number of spikes S in that bin when the signal was present at the investigated signal-to-noise ratio. The unit was defined as a CMR unit according to Pressnitzer et al. (2001) if for two consecutive signal-to-masker ratios d′(CM) > d′(RF) > d′(CD).

The data were also analyzed with respect to the suppression of the on-frequency masker when comodulated FBs were added (i.e., CM vs. RF) and there was a possibility of predicting CMR behavior by means of the response to the first cycles (see Fig. 3) for the different masking conditions. To this end the similarity index (SI, Las et al. 2005) was calculated that compared two PSTHs (bin width 10 ms) of equal duration in the MO condition. Each PSTH was smoothed using a moving average technique that had an averaging window of 50 ms. The two smoothed responses of equal duration were plotted against each other and linearly interpolated (polyfit, Matlab R2006a). The slope of this linear interpolation was the SI (Las et al. 2005). An SI of one would indicate the same responses to different stimuli. Slopes of above or below 1 indicate a stronger or weaker response, respectively, for the condition plotted on the ordinate. To compare units that showed CMR against non-CMR units an n-way ANOVA was performed (MatLab R 2006a) on the SIs.

Fig. 3.

Schematic representation for the calculation of the similarity index (SI). A: analysis for response in the time interval 0–100 ms compared with the response to the time interval 100–200 ms. B: comparison of one masker (e.g., RF) against another. High values indicate that the second signal elicited a stronger response whereas low values indicate a stronger response to the first signal.

Analysis according to the FRA and firing pattern.

The units were classified according to the shape of their FRA to provide a quantitative overview of the units of this study. The classification was done according to Hernández et al. (2005), but some groups were merged as a detailed categorization was not the main objective of this study. Units were divided into eight different categories. “V-monotonic” units had a V-shaped FRA and increased their firing rate at higher sound levels, whereas “V-non-monotonic” units responded with lower firing rates to higher sound levels. “Multipeaked” units had several distinct regions in the FRA where they responded. Purely “inhibitory” units had a high spontaneous rate which was reduced as a response to the pure tone stimulation. In contrast to Hernández et al. (2005) the present study did not differentiate between “high” and “low tilted” units. Two other categories in Hernández et al. (2005), “narrow” and “closed” units, were combined in the present study because recording stability did not always allow a complete characterization, i.e., a unit with a “narrow” FRA could have revealed a “closed” FRA if higher stimulus levels could have been tested. U-shaped units responded over a wide range of frequencies that was only slightly narrowed for low sound levels. Units not falling into one of these categories were classified as “mosaic.” For the final steps of data analysis, “mosaic,” “tilt,” “inhibitory,” and “multipeaked” units were unified in the category “other.”

Additionally, the units were also classified according to their temporal firing pattern at the best frequency (BF) and best level as in Le Beau et al. (1996). The BF and best level are defined as the frequency and SPL that elicited the maximum number of spikes during the 100-ms interval of the pure tone stimulation. Units were classified as 1) “onset” units when they only spiked at the beginning of the stimulus, 2) “pauser” units when they showed an onset peak followed by a brief period without spikes before they started to gradually fire again, 3) “on-sustained” when they continued to fire at a lower rate after a pronounced onset response, 4) “sustained” units if they fired in response to the signal without pronounced peaks, and 5) “chopper” units when they had several clear peaks in their response. Unlike Le Beau et al. (1996), no distinction was made between “chopper” and “onset-chopper” units because in the present study their firing patterns were too similar for a clear dissociation.

RESULTS

Comparison with previous studies.

For the study, 94 units were investigated of which 82 were suitable for further investigation according to the definition in materials and methods. In these 82 units 30 showed CMR according to the definition of the difference between the signal and the masker response, while only 6 showed CMR according to the definition by Neuert et al. (2004) and 18 according to the analysis by Pressnitzer et al. (2001). In the following, first the response of typical single units will be presented and then the response of all units is analyzed.

Single unit responses.

Figure 4 shows the response of a unit with a “V-shaped” FRA and a “sustained”-like response pattern (Fig. 4A). The FRA shows the temporal firing patterns in response to a pure tone for different combinations of frequency and sound pressure level, as indicated by the respective coordinates on the abscissa and ordinate. The duration of each of the depicted firing patterns was 100 ms.

Fig. 4.

Results for a unit showing a prominent increase in spikes for the CM masker. A: frequency response area (FRA); for each frequency-SPL combination the FRA shows the PSTH for a 100-ms pure tone with this frequency and sound pressure level. The white dot indicates frequency and level of the target signal at 0 dB signal-to-noise ratio and of the RF masker. The black dots show the level and frequency of the flanking bands (FBs) for the CM and CD maskers. B: spike rate in the signal interval (indicated by the bar at the top in C) and in the masker interval for the different masker conditions. C: PSTHs (bin width 5 ms) for the 3 masking conditions (columns) and signal-to-masker ratios (rows). The bottom panels show the response to the masker only (MO).

The signal frequency for this CMR experiment (and thus the center frequency of the on-frequency masker) was 1.7 kHz. The center frequencies of the FBs were 0.5, 0.6, 0.7 and 3.3, 3.4, and 3.5 kHz. The SPL for each masker band (on-frequency masker and FBs) was 60-dB SPL. In Fig. 4A, the frequency-level combination of the on-frequency masker is indicated by an open circle, those of the FBs by filled circles.

The unit responded to all three maskers alone (MO conditions in Fig. 4C). For the RF masker, the responses to the first (0–100 ms) and second (100–200 ms) modulation cycles is slightly larger than that to the remaining cycles of the masker. The average response to the CM masker is reduced compared with that of the RF masker. The response to each cycle of the CD masker is shorter and the maximum firing rate higher than that to the cycle of the RF masker.

When the signal was added, the minimum in the response to the RF masker was filled by the response to the signal while the response to the masker was slightly reduced. At the highest signal level, the response was sustained, i.e., did no longer follow the masker modulation. In the CM condition, the signal also filled the temporal gaps in response to the masker and reduced the masker response. In contrast to the RF condition, signal response was considerably higher than that to the masker at the two highest signal-to-noise ratios. In the CD condition, the signal only changed the firing pattern at the highest signal-to-noise ratio. This unit showed CMR according to our definition and also according to that of Neuert et al. (2004).

Figure 5 shows a unit that was classified as an “onset” unit with a “monotonic” V-shaped FRA (Fig. 5A). The on-frequency masker had a center frequency of 1 kHz and a SPL of 50 dB (open circle in the FRA). The FBs were spaced in 0.1-kHz intervals from 3.3 to 3.8 kHz and also had a SPL of 50 dB (filled circles in Fig. 5A). The FBs were positioned spectrally above the signal due to the lack of nonexcitatory frequencies below the signal frequency.

Fig. 5.

Results for a unit that first ceased to respond to the CM masker, then to the RF masker, and last to the CD masker with an increasing signal-to-masker ratio. For the 0-dB signal-to-masker ratio it responded to the signal in the CM condition only (B and C). Same notation as in Fig. 4.

The MO responses (bottom row of Fig. 5C) to the different maskers were similar. For all maskers, the response in the interval from 0 to 100 ms differed from the response in the interval from 100 to 200 ms. The difference was smallest for the RF masker: the maximum firing rate was slightly higher and longer for the first interval and longer than for the second interval. For the CM masker, the difference was more pronounced. Whereas there was a strong response to the first cycle of the masker modulation, it was less than half during the second one. A similar behavior was found for the CD masker, but the onset response was less pronounced than for the CM masker. The temporal extension of the firing pattern decreased for the CD masker in the second interval and thus reduced the spike count. The SIs between the first and second interval of the different maskers were 0.5 for the RF and 0.4 for the CM and CD maskers.

Figure 5, B and C, show that the response to the CM masker was largely suppressed when the signal was added to the masker at the lowest signal-to-masker ratio of −30 dB. This is also observed for the RF condition at a signal-to-masker ratio of −20 dB and for the CD condition at −20- to −10-dB signal-to-masker ratios. For the CM condition, there was a weak response to the signal at the 0-dB signal-to-masker ratio that was not present in the RF and CD condition. This is also observed in the analysis of the spike rates in the masker and signal intervals (Fig. 5B). Note that the signal was always well above threshold of the unit (see FRA in panel A) when it started to suppress the masker response. This unit was classified as a CMR unit according to the criterion of the present study but was not a CMR unit according to the evaluation method by Neuert et al. (2004). This was because the change in response of the unit due to the addition of the target signal to the CM masker was only for one signal-to-noise ratio larger than in the other two masking conditions.

Figure 6 shows a unit that was not classified as a CMR unit. The FRA is shown in Fig. 6A, and the CMR response for selected signal-to-masker ratios in the CMR experiment is shown in Fig. 6B. The comparison of the response to the CM and CD masker with that to the RF masker indicates that the FBs suppress the unit's response entirely, regardless of their modulation phase. The unit responded to the signal in the RF masker at the 0-dB signal-to-masker level, while there was no change in the firing patterns of the other two maskers. The FRA was classified as “tilt.” There were only a few frequencies for which it responded reliably for a range of different sound pressure levels.

Fig. 6.

Unit with a “tilt” FRA (A) that did not show CMR. B shows the response to the MO (bottom panels) and −10 and 0 signal-to-masker level. As the frequency response of the unit changed, if the SPL was varied it was an unlikely candidate for CMR in the first place.

Figure 7 shows another unit which did not fulfill the criteria of a CMR unit. The unit had a “narrow” FRA and a “sustained” firing pattern. Its high spontaneous activity was suppressed at an SPL of 60 dB for frequencies ranging from 0.6 to 2.8 kHz and around 4.8 kHz. The on-frequency masker had a center frequency of 4 kHz and a 60-dB SPL (open circle in Fig. 7A). The FBs were positioned spectrally below at 1.5, 1.75, and 2.0 kHz and spectrally above at 4.75, 5, and 5.25 kHz; their SPL was 60 dB (filled circles Fig. 7A). The unit followed the amplitude modulation for all maskers (bottom row of Fig. 7B). The responses to the CM masker and the RF masker were similar, i.e., the addition of comodulated FBs did not reduce the response to the on-frequency masker, even though pure tones played at this level suppressed the spontaneous activity (see Fig. 7A). The unit responded well to the signal in the CM condition, but this response was not significantly different from that in the RF condition. Especially, the contrast between the masker and the signal response was not enhanced by the addition of comodulated FBs. The response in the signal interval was suppressed in the CD condition, especially at a signal-to-masker ratio of −10 dB.

Fig. 7.

A unit that showed less suppression of the RF masker in the CM masker than expected from the FRA and that did not show CMR. The FRA (A) was classified as “narrow/close.” B shows the CMR response at the MO and the −10- and 0-dB signal-to-masker levels.

Population analysis.

Figure 8 shows the individual results for the difference between the signal and the masker responses in the three different masking conditions. On the left the results for the RF condition are shown, in the middle the results for the CM condition, and the results for the CD condition are shown on the right. Solid lines denote units that were classified as CMR units, and dotted lines as non-CMR units. In the CM condition there is an trend for larger responses as the signal-to-masker ratio is increased. Figure 9 shows the difference between the spikes in the signal and the spikes in the masker interval with respect to this difference for the masker alone (MO) signal-to-masker ratio as a function of target signal level. Figure 9A shows the average for all 94 units, with the filled upward triangle denoting the CM condition, the circle the RF condition, and the open downward pointing triangle the CD condition. For this sample, the result is the same for the RF and the CM condition in all but the 0-dB signal-to-masker ratio. At this highest signal-to-masker ratio the response in the CM masker is larger than in the other two conditions. The response in CD conditions is lower for all signal-to-masker ratios. Figure 9B shows the results for the 64 non-CMR units. Here, the results are approximately the same for all masking conditions at the −30- and 0-dB signal-to-masker ratios. For the intermediate ratios, signal detectability in the RF condition is better than in the CM and CD conditions. Figure 9C shows the results for the 30 CMR units. Here, the response in the CM masker has the highest values for all except the −30-dB signal-to-masker ratio. At this signal-to-masker ratio the result is similar for the CM and the RF condition.

Fig. 8.

Individual differences between the spikes in the signal interval and the spikes in the masker interval. Solid lines refer to CMR units and dotted lines to non-CMR units.

Fig. 9.

Population average for the differences between the spikes in the signal interval and the spikes in the masker interval (see also Fig. 8) with respect to the MO signal-to-masker ratio. A denotes the results for all units (94), B for the 64 non-CMR units, and C for the 30 CMR units.

Influence of the analysis method.

The analysis presented so far compares, for a given signal-to-masker ratio, the response to the signal with that to the masker, i.e., the responses to the stimulus at different points in time. This is different from the analyses used in Neuert et al. (2004) and Pressnitzer et al. (2001) where the response to a signal embedded in the masker is always compared with that when the masker is presented without the signal (MO condition). Thus already due to this fundamental difference, it is conceivable that the number of units classified as CMR units were different when analyzed with a different method. The analysis used in Neuert et al. (2004) classifies the units on the basis of the difference in response in the signal interval for a signal added to a masker and the masker alone. The population results for the analysis of a spike rate increase in the signal interval with increasing SNR (as performed by Neuert el al. 2004) is shown in Fig. 10 (compare the corresponding panels in Fig. 9). It should be noted that of the 30 units classified as CMR units based on the current method, only 6 would be classified as CMR by the Neuert et al. (2004) method. The analysis of the present study results in a steeper increase in spike rate with level for the CM and RF condition (Fig. 9C) than with the one used in Neuert et al. (2004) (Fig. 10B). Fig. 11 shows the results of the d-prime analysis according to Pressnitzer et al. (2001) for the example units shown in Figs. 4–7. While the results for the different evaluation methods are the same for the units shown in Figs. 4 (compare Fig. 11A–CMR) and 6 (compare Fig. 11C–no CMR), the other two units show opposite results. As the masker response for the unit shown in Fig. 5 is stronger in the RF and the CD masking conditions, the change in response is larger and hence the d-prime is too. Thus this unit is not a CMR unit according to the d-prime analysis (Fig. 11B, although smaller step sizes might have resulted in CMR for this analysis). For the unit shown in Fig. 7 the change in d-prime was slightly larger for two signal-to-noise ratios for the CM masking condition (Fig. 11D) and hence it is classified as CMR unit according to the d-prime analysis. In total, according to the d-prime method 18 units were CMR units. The average of the d-prime evaluation for the CMR units as defined by the spike-difference method is shown in Fig. 11E and for all units in Fig. 11F.

Fig. 10.

Mean of all responses to the target signal with respect to the MO signal-to-masker ratio for the for all 94 units (A) and the 30 CMR units (B). Similar to Fig. 9 but looking at the target signal only instead of the difference between the target and the masker response. Symbols as in Fig. 9.

Fig. 11.

The d-prime analysis as described in Pressnitzer et al. (2001) for the example units shown previously. The results in A and C give the same result as the analysis applied here, but for B and D the results are different. E and F show the results for the average d-prime analysis as done for the 30 CMR units as defined in Fig. 9C (E) and for all units (F).

Sensitivity to comodulation and unit type.

Figure 12 summarizes the responses of all recorded units. Since units in the inferior colliculus form a continuum rather than discrete classes with respect to FRA shapes (see Palmer et al. 2013), the table in Fig. 12 is meant to only provide a rough estimate of the relation of response characteristics and ability to show a correlate of CMR. For each combination of type of FRA and type of PSTH, three numbers are shown: 1) number of units showing CMR according to the criterion of the present study, 2) number of units that did not respond with more than half the spikes in the first 100 ms of the MO conditions (see materials and methods), and 3) the total number of units. The failure to respond to the entire signal was distributed almost evenly across all response-area and firing-pattern types. The two most common temporal firing types were “chopper” and “sustained,” with the “chopper” types being least likely to show CMR-like behavior with <20%. “Sustained,” “pauser,” and “onset” units were equally likely to show CMR like behavior (50%). “Onset” units had the highest chance of not responding to the amplitude modulation, however. “On-sustained” units had a chance of about 30% to show CMR like behavior and thus their chance was close to the overall average. Units with a “narrow” or “closed” FRA are most likely to show CMR but only about 10% of all units had such a FRA. More common were V-shaped FRA. About 40% of all units with a monotonic or nonmonotonic “V shaped” FRA showed CMR. Thirty-two of the 94 units did not fall into the categories and were thus classified as having “other” FRA. Of these units, 6 were “inhibitory,” 9 were “multipeaked,” and 8 were “tilt”; the remaining ones had a “mosaic”-like FRA. Seventeen percent of them showed CMR. Units with an “other” type of FRA were more than one-third of the entire measured population, whereas “narrow/closed” and “U-shaped” units were the smallest fractions. None of the units with a U-shaped FRA showed CMR. With the exception of the U-shaped units, we found that units with a plateau in the rate level function were more likely to show CMR-like behavior than those without it.

Fig. 12.

Table showing the number of CMR units (left), the number of units that were able to follow the AM signal (middle), and the total number of units (right). Abbreviations of the different firing types: Su, sustained; OS, on-sustained; On, onset, Ch, chopper, Pa, pauser. Abbreviations of the different FRA types: VM, V-shaped monotonic; VNM, V-shaped nonmonotonic; N/C, narrow or closed; U, U-shaped; Oth, other (including inhibitory, tilt, multipeaked, mosaic).

Can the similarity index predict the units' ability to show CMR?

To answer this question the similarity index (SI) was calculated for the responses to the masker only (MO) in two ways. In the first analysis the SI between different masking conditions was calculated. This was either done for the whole duration of the stimulus (0–500 ms) or for the four last cycles of the masker envelope (100–500 ms), i.e., excluding the onset response of the unit. The second analysis focused on changes in the firing pattern from the first cycle (0–100 ms) to the second cycle (100–200 ms) of the masker envelope within a masking condition. The SI was only calculated for the 70 units that showed a response to the amplitude modulation of the on-frequency masker. If the SI is sensitive to the ability to show a CMR-like behavior, then SI of the 26 CMR units should be significantly different from the SI for the remaining 44 units not showing CMR.

Figure 13 shows the SI between the RF and the CM (Fig. 13A) and RF and CD (Fig. 13B) maskers. Each panel shows four columns. The first two indicate the results for the whole duration of the stimulus, the other two when the analysis was restricted to the last 400 ms of the stimulus. The SI between the CM and the RF masker was always smaller than 1, indicating that for all units the response in the CM condition was weaker than that to the RF condition. For CMR units, the mean SI was 0.44 for the comparison of RF and CM condition when the whole duration of the stimulus was analyzed. It was considerably lower than that for the non-CMR units. This may indicate that the response to the on-frequency masker was suppressed by the presence of the FBs in the CM condition. However, the effect was not significant. Interestingly, similar SIs for the RF and the CM masker responses were obtained when the analysis was restricted to the last 400 ms of the stimulus which would not be expected on the basis previously mentioned across-frequency suppression caused by the FBs. When SI was calculated for the comparison of RF and CD condition (Fig. 13B) non-CMR units had a smaller SI than CMR units. The difference was significant when the interval 100–500 ms was analyzed.

Fig. 13.

SI between different signals (RF and CM in A, and RF and CD in B) for different time intervals (0–500 ms and 100–500 ms). The upward pointing triangle denotes the results for unit from Fig. 4, and the downward pointing triangle those from Fig. 5. The triangle pointing to the left denotes the results for the unit from Fig. 7. Due to a lack of spikes the results for the unit shown in Fig. 6 are not displayed in A. *P < 0.05.

The triangles show the values of the units presented in this study in the following way: the unit shown in Fig. 4 corresponds to the upward pointing triangle, Fig. 5 to the downward pointing one, Fig. 6 the triangle pointing to the right, and Fig. 7 the triangle pointing to the left. While the CMR units show approximately similar responses between the RF and the CD masker for the time interval from 100 to 400 ms, the CMR units shown in Figs. 4 and 5 have approximately similar responses between the RF and the CD masker for the time interval from 100 to 400 ms, but so does the unit shown in 7. The unit shown in Fig. 6 has a much weaker response to the CD than to the RF masker and would be correctly indicated as unlikely to show CMR.

Figure 14 shows the SI between the first (0–100 ms) and the second (100–200 ms) cycles of the on-frequency-masker envelope. For the RF and the CM masker, the units seem to be less responsive to the second cycle of masker envelope than to the first, regardless of whether or not the unit showed a CMR-like behavior. This was different for the CD masker, in which units that showed CMR tended to respond with more spikes to the second cycle (SI > 1). In contrast, non-CMR units had a SI of 0.5, i.e., similar to the SI for the other two masking condition. The difference for the CD masker for the CMR units and non-CMR units was statistically significant. As a statistical test, an n-way ANOVA (anovan in MatLab) was used with criterion P < 0.05; the dependent variable was the effect.

Fig. 14.

SI between different intervals of identical signals (RF, CM, or CD masker). Symbols for the different units as in Fig. 13. For the unit in Fig. 6, calculating the SI was not possible in the CM condition as there were too few spikes. *P < 0.05.

Unusually, the units shown in Figs. 4 and 5 responded weaker to the second cycle of the RF modulation in the CD masker condition, which would indicate that they do not show CMR. On the other hand, the increase seen in the unit shown in Fig. 7 would classify it as a likely candidate for CMR.

The comparison of the CMR units and the non-CMR units reveals that 1) the most informative measure of a unit's CMR response is the similarity between the initial and the following 100 ms of the response to the CD masker: if the similarity is high then the unit is likely to show CMR. 2) A high similarity between the responses to the RF and the CD masker also is a good indicator, whereas the suppression of the on-frequency masker in the CM condition is not a reliable criterion to differentiate between units showing CMR and those that do not show CMR but follow the envelope of the masker.

The influence of the carrier and of the distance of the closest FB.

The average frequency of the target signal was 1.41 kHz (minimum 350 Hz, maximum 4 kHz) for all CMR units whereas that of the remaining units was 1.9 kHz (minimum 250 Hz, maximum 6.7 kHz). This difference was not significant. Further, the influence of the closest FB divided by the target signal was investigated. The mean for the CMR units was about 1, meaning that the FBs were one octave higher or lower than the target signal (minimum 0.3, maximum 3.8). For non-CMR units, this distance was 1.3 (minimum 0.2, maximum 4.7). Again, the difference was not significant. However, as Schooneveldt and Moore (1987) showed in their psychophysical study, there is an influence of the distance of the nearest FB on CMR. Thus, in a modified paradigm, 11 units were measured for various spectral distances of the FBs (see materials and methods). From these 11 units, only two showed CMR-like behavior for any of the FB positions, i.e., a smaller portion than found in the overall population. This may be partly due to differences in the experimental paradigm between the 11 units and the other units. For the 11 units, 20 repetitions and 4 instead of 5 signal-to-masker ratios were used and all FBs were placed at higher frequencies. Only FBs spectrally above the signal frequency were used since 1) the distance was varied up to 4 kHz and 2) using lower frequencies for lower distance would have complicated the experimental conditions.

To test the influence of the distance of the closest FB, 11 units were measured for various spectral distances of the FBs (see materials and methods). Figure 15 shows the response of one of the two CMR units. The signal frequency was 0.5 kHz and nearest FB was positioned at 0.2, 0.5, 1, 2, 3, or 4 kHz above the signal frequency (see Fig. 15A). This corresponds to a FB frequency relative to the signal frequency of 0.4, 1, 2, 4, 6, and 8. The unit showed a CMR-like behavior if the closest FBs were positioned 0.2 or 1 kHz above the signal frequency. In Fig. 15B, the increase in the response difference between the signal and the masker response is shown. Black triangles denote the CM, circles the RF, and white triangles the CD masking condition. For the 0.2-kHz FB distance, the difference was best for all signal-to-masker ratios in the CM condition. Interestingly, the response in the CD condition was better than that in the RF condition. For the 0.5-kHz FB distance, the response in the CM condition was not highest at 0 dB signal-to-masker ratio, and, although it was higher than in the other masking conditions for the other two ratios, this effect was not significant and at this distance the unit did not show CMR-like behavior as defined here. At 1-kHz FB distance, the response in the CM condition clearly was higher than that in the RF and the CD conditions and the CMR-like behavior was most prominent. At 2 kHz, the response in the CM condition still was the best but failed to reach significance and at higher distances, the CMR-behavior vanished. Figure 15C shows the SI from Fig. 13 in the lower two plots. While the SI between the RF and the CM masker confirm the trend observed for CMR units, namely lower values for CMR-like behavior, the situation is different for the SI between the RF and the CD masker. The number is close to the mean if CMR-like behavior occurs, but higher when the unit failed to show CMR. The top plot of Fig. 15C shows the significant result of Fig. 14. The trend seen in Fig. 14 is confirmed for 1-kHz FB distance, but not for 0.2-kHz distance. The second unit only showed a CMR-like behavior when the closest FB was placed 0.2 kHz above the carrier frequency (data not shown). Overall, while the unit shown in Fig. 15 was in good agreement with the data by Schooneveldt and Moore (1987), the other two populations (94 normally measured units and 11 with a varying FB paradigm) did not display these results. The reason for this could be the individually fitted CMR paradigm for each unit.

Fig. 15.

Unit for which the spectral position of the FBs was varied. A: the FRA of the unit with a white dot at the stimulation frequency and black dots for the starting points of the closest FB. B: the result for the individual FB distances (as in Fig. 9). C: the similarity indexes. Top, the CD 0–100- vs. 100–200-ms intervals; middle, the similarity between the RF and the CM condition (solid: 0–500 ms; dotted: 100–500 ms), and bottom, the SI between the RF and the CD condition, with same time frames as for the middle panel.

DISCUSSION

The present study showed that a neural correlate of CMR can be found at the level of the IC as an increase of the signal response relative to the masker response when the masker was comodulated. In this CM condition, the comparison with the masker response is very important to detect units that show CMR-like behavior. Using the method by Neuert et al. (2004) focusing only on the signal intervals many units would not have been detected indicating that the correlate of CMR is different between the CN and the IC. An analysis of the temporal response pattern showed that CMR units have a significantly higher similarity between the first and second 100-ms portions of the response to the CD masker. In addition to using an appropriate method for detecting CMR, it was also shown that the spacing of the flanking bands is a contributing factor for CMR. This finding is in line with psychophysical studies (Hall et al. 1990; Piechowiak et al. 2007; Schooneveldt and Moore 1987). This would be supported by the lack of CMR-like responses for “U-shaped” FRA types, where the FBs had to be placed spectrally too distant to elicit CMR-like behavior, due to their broad frequency responses.

As the modified analysis is a generalization of the one used in Neuert et al. (2004) it is likely that it detects more units. According to the analysis used in Neuert et al. (2004) one would conclude that less units are encoding masker comodulation in the IC than in the CN where 26 of their 46 units showed a correlate of CMR. However, this difference is not so large when the new analysis was used. The analysis of the present study is also sensitive to a change in response to the masker due to the presence of the target signal, i.e., envelope locking suppression (Las et al. 2005). Most of the units of the present study showed both effects, a response reduction to the masker and an enhanced signal response in the comodulated condition. Interestingly the d-prime analysis of Pressnitzer et al. (2001) is also sensitive to a change of the masker since it analyzes signal and masker intervals and uses squared differences which are not sensitive to the direction of the change. However, as the results for the unit in Figs. 5 and 11B show, the psychophysically motivated approach has disadvantages when investigating units with a weak response to the CM masker that is suppressed at higher signal-to-noise ratios. The response of the unit shown in Fig. 5 could be interpreted in a way that the responsiveness to the masker was reduced first for the CM, then for the RF and at the highest signal-to-masker ratio for the CD masker. When the signal is played at the same level as the masker, the unit “ignored” the CD masker more than the CM masker, but it responded to the signal in the CM masking condition. An optimal method analyzing neural data would be expected to find signal detection thresholds across the population of cells that are similar to psychophysical studies if the neural activity is causal to task performance. For similar (although not identical) CMR paradigms, both Eddins and Wright (1994) (using 4 instead of 6 FBs) and Verhey et al. (2013) (using 4 FBs and 20-Hz-amplitude modulation) found differences of approximately 10 dB between the CM and the RF masking condition and 10 dB between the RF and the CD masking condition. The population analysis in Fig. 9 shows that the response at 0 dB SNR to the signal in the RF masking condition is equal to that in the CM masking condition at −4 dB SNR. Using the same approach, the CD masking condition shows a −10-dB difference from the RF masking condition. Figure 11F shows that these differences can also be found in the d-prime analysis. The analysis of the signal response alone, although not sufficient to detect many units, seems to be a bit closer, showing a −8-dB difference between the CM and the RF and a −12-dB difference between the RF and CD masking conditions (Fig. 10).

The comparison of the results using the analysis of Neuert et al. (2004) with those using the analysis of the present study indicates that, at the level of IC, the majority of the units respond in agreement with the hypothesis that the correlate of CMR at higher stages is the response reduction to the envelope of the masker, referred to as envelope locking suppression (ELS). ELS originally refers to the reduction of the masker response when a signal is presented at a level at which the signal itself would not elicit a response. It has been observed at the level of the MGB and the AC (Las et al. 2005). Although a different CMR paradigm was used that might even invoke a different mechanism, the results presented here are in good agreement with the observations made by Las et al. (2005). As for the unit shown in Fig. 5 the response to the pure tone alone would start at 10-dB SPL, and the signal at −30-dB signal-to-noise ratio corresponded to 20-dB SPL; this reduction at level above the detection threshold could be a preprocessing step.

Thus the presuming neuronal mechanism underlying CMR along the auditory pathway would start off with an increase in the signal response and end with a reduction of the masker response (although a good signal response could still be important, see Fig. 10). The followup question then is what the underlying mechanism at the level of the IC could be. At the level of the CN, it was argued that CMR might originate from improved signal detection because the CM masker was largely suppressed by a wideband inhibitor. If the wideband inhibitor has only a limited frequency range around the BF, it would also be in agreement with the loss of CMR for distant flanking bands (see Fig. 15B). According to the wideband inhibitor idea, the presence of the signals hardly changed the response of the unit to the CM masker, but this mechanism can be ruled out as the single source due to the comparisons of the introduced evaluation and that one as done by Neuert et al. (2004). Our data would support more recent studies that emphasize the importance of inhibition in the IC as reviewed by Pollak et al. (2011) for units with non-V-shaped response areas over claims in that especially closed units resemble the response from the DCN (Davis 2002). However, as about one-third of the units showing CMR-like behavior have a V-shaped FRA, inhibition could play a crucial role in those as well. Generally, apart from the previously mentioned U-shaped units a constant or broadening frequency response with increasing sound pressure level seems beneficial for showing CMR-like behavior. Given that the DCN projections to the IC are only excitatory (Malmierca et al. 2005; Semple and Aitkin 1980), these studies indicate that the inhibition of the masker signal has a different origin. The transformation of the stimuli response from the CN mainly showing signal facilitation for CM condition to the IC which displays both signal facilitation as well as masker suppression can be accounted for by assuming a parallel feed-forward network. In this model, a CMR unit receives excitatory input from the CN and a delayed stimulus-driven inhibitory input. This network accounts for the reduction in the spike rate to the masker if it includes an appropriate temporal delay. These temporal requirements would make this form of CMR mechanism in the CN unlikely. As the auditory cortex has been shown to modify response properties in the IC (Bajo and King 2013), another possibility to modify or create CMR in the IC could be the corticollicular pathway. The corticocollicular pathway originates from glutamatergic neurons in deep layers of the AC and targets both excitatory as well as local inhibitory interneurons in the IC (Nakamoto et al. 2013a, 2013b). Thus suppression of the masker response in combination with an increased signal response are both possible by corticocollicular feedback. (for the suggested cortical feedback pathway, see Fig. 16).

Fig. 16.

A possible diagram from a cell in the CN to a cell in the IC that would explain the suppressive effects and the temporal course in the SI comparison for the CD masker. The CN neuron would receive input as suggested by Pressnitzer et al. (2001) and Neuert et al. (2004), meaning that the signal representation would be enhanced in the CM masking condition. This enhanced signal representation would be propagated along the auditory pathway to the AC. If the AC would suppress the IC following an excitation, this could build up to a reduction of the masker response.

The importance of the interplay between spectral and temporal dimension is also reflected in the SI results shown in Fig. 14. Simple onset effects could explain the weaker response in the second cycle of the RF masker. For CMR units, these effects tend to be overruled by another mechanism that increases the response to the on-frequency masker, if a full cycle of the flanking bands was played as would be the case in the CD masker. This means that the response to the signal is stronger if the unit was first stimulated with the 100-ms portion of the CD flanking bands.

Overall, we conclude that the evaluation method introduced in this study is most beneficial for detecting CMR on a single-unit basis and that the effects shown here give insight about the connectivity of the IC in the auditory pathway and the transformation of signal representation from lower to higher stages. Theoretically, the suggested method would even be capable of detecting ELS, although not according to the strict definition, as ELS requires knowledge of the pure tone response, which is a prerequisite for the parameters of the paradigm used in this study, but not necessary for the evaluation itself.

GRANTS

This work was supported by the Deutsche Forschungsgemeinschaft (SFB TRR31).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

J.-P.D. and M.J. performed experiments; J.-P.D. and J.V. analyzed data; J.-P.D., M.J., and J.V. interpreted results of experiments; J.-P.D. prepared figures; J.-P.D. drafted manuscript; J.-P.D., M.J., and J.V. edited and revised manuscript; J.-P.D., M.J., F.W.O., and J.V. approved final version of manuscript.