Rationale
A major aim of experiment 2 was to determine whether the FFR could provide evidence for, and an objective measure of, modulation specificity in neurons in the upper brainstem. Psychophysical evidence for selectivity in the modulation domain initially came from experiments where modulation detection thresholds for a target modulation were measured in the presence of simultaneous maskers with various modulation rates. Typically, maskers were most effective when the modulation rate of the masker was similar to that of the target (see, e.g., Bacon and Grantham
1989; Houtgast
1989). These modulation-masking data have inspired models of envelope processing based on banks of filters selective for modulation frequency for each peripheral auditory channel (Dau et al.
1997). Psychophysical evidence for modulation-rate selectivity in the modulation domain has also come from experiments investigating long-term adaptation (Richards et al.
1997; Wojtczak and Viemeister
2003) and forward masking (Wojtczak and Viemeister
2005; Moore et al.
2009) in the modulation domain. In addition, physiological experiments have revealed neurons in the IC (Langner and Schreiner
1988; Joris et al.
2004) and in the ventral nucleus of the LL (Batra
2006; Zhang and Kelly
2006; Recio-Spinoso and Joris
2014) that are tuned to different AM frequencies. As mentioned above, the FFR is thought to reflect phase-locked activity at the level of the IC and/or the LL. The best AM frequencies (the AM frequencies that maximize either the synchrony of firing or the firing rates) of neurons tuned to AM depend on species and the recording site. For example, in the IC of the unanesthetized rabbit, the best AM frequencies in terms of synchronized rate (the product of synchronization coefficient and average firing rate) ranged from 11 Hz (mean minus SD) to 193 Hz (mean plus SD) (Batra et al.
1989). In contrast, in the ventral nucleus of the LL of anesthetized rats, best AM frequencies in terms of firing rate ranged from 10 to 300 Hz or from 10 to 400 Hz, depending on the type of neuron, and ranged from 10 to 400 Hz in terms of vector strength (a measure of synchronization) (Zhang and Kelly
2006). In the ventral nucleus of the LL of anesthetized cats, the best AM frequencies in terms of firing rate ranged even up to 900 Hz, with strong synchronization to the envelope waveform over the entire range of responsiveness (Recio-Spinoso and Joris
2014). Furthermore, an fMRI experiment with macaques has provided evidence that modulation rate is represented at the level of the IC along a dimension that is orthogonal to that for CF (Baumann et al.
2011). That study used broadband noise with modulation rates ranging from 0.5 to 512 Hz, thus including rates high enough to evoke a musical pitch.
Here, we investigated tuning to modulation rate in the FFR using adaptation. The experiment employed complex tones with unresolved harmonics filtered into the same frequency region, allowing the effects of modulation rate to be evaluated without a corresponding and potentially confounding co-variation in center frequency. The rationale in this and the following experiments is based on the assumption that an adaptor will reduce the response to a target more when it adapts neurons responding most strongly to the target than when it adapts neurons that respond less strongly to the target (Näätänen et al.
1988). The experimental question was whether the FFR evoked by a target sound with a given modulation rate would be reduced more by an adaptor that had the same modulation rate than by adaptors with different modulation rates. If such specificity for modulation rate of the adaptor were observed in the FFR of the target, this would provide evidence that the FFR originates mainly from neurons that respond selectively to the modulation rate of the stimulus and would provide physiological evidence for the existence of filters tuned to modulation rates in humans. In addition, experiment 2 helped to control for the effects of CF on the amount of peristimulus adaptation as, in contrast to experiment 1, the frequency region of the various complex tones (used as adaptors) was kept roughly constant.
Methods
The FFR evoked by a 100-ms, 75-dB SPL complex-tone target with 213 envelope peaks per second (defined as envelope rate) was measured. The target was always preceded by a 200-ms adaptor and followed the adaptor without any silent gap. The different adaptor conditions are specified below. All tone durations included 5-ms raised-cosine rise/fall times. All complex tones were composed of alternating-phase harmonics (even harmonics added in cosine and odd harmonics added in sine phase). They were filtered into the frequency range between 3.9 and 5.4 kHz, which was high enough to ensure that the harmonics were not resolved and interacted strongly in the peripheral auditory system. Under these conditions, the number of envelope peaks per second is twice the F0. This leads to a pitch corresponding to twice the F0 (Shackleton and Carlyon
1994), and major spectral peaks in the FFR addition waveform at the envelope rate (corresponding to the pitch of the complex) and integer multiples thereof (Krishnan and Plack
2011). Filtering of the complex tones into a high-frequency region and alternating the starting phases of harmonics allowed us to employ complex tones with a pitch in the musical range, and within the range used by Baumann et al. (
2011), without the presence of resolved harmonics.
When the adaptor and target had the same envelope rate, the target component phases were “preserved” relative to those for the adaptor, i.e., the phase of each component in the target was the same as if the adaptor had continued; the adaptor and target comprised the first 200 and the last 100 ms of a 300-ms complex, respectively, with onset and offset ramps applied independently on each. The starting phases of components in the target were identical in all conditions. When the adaptor envelope rate differed from that of the target, a temporal offset was applied to the adaptor so that its last major envelope peak occurred at the same time as for the 213-Hz rate adaptor.
Stimulus generation and presentation and analysis of the recorded signal were done in the same way as in experiment 1, with the following exceptions: stimuli were played with a repetition rate of 1.81/s. The same stimulus condition was played in blocks of 1500 (valid) trials; two blocks were run for each condition in randomized order across subjects, giving 3000 valid trials per condition. The FFR was recorded with a sampling rate of 10 kHz. The FFR was analyzed and compared across five 50-ms time ranges: (1) from 15 to 65 ms after adaptor onset (A-start; the value of the starting time was increased from 12 ms, as used in experiment 1, to 15 ms in experiment 2A because of the later response onset of the FFR for the adaptor with the lowest envelope rate (see below) for which the first pulse in the physical stimulus appears later than for higher rate stimuli); (2) from 65 to 115 ms after adaptor onset (A-mid); (3) from 150 to 200 ms after adaptor onset (A-end); (4) from 12 to 62 ms after target onset (T-start); (5) from 50 to 100 ms after target onset (T-end). The FFR for the 50-ms window before adaptor onset served as baseline.
For spectral analysis, for each subject, the 50-ms waveform was zero-padded symmetrically to make up a 1-s signal, and the magnitude spectrum was calculated via a discrete Fourier transform. The magnitude spectrum is specified in decibels re 0.01 μV. The FFR strength was defined as the highest magnitude present in the spectrum within a 20-Hz range centered at the envelope rate of the signal, i.e., at 213 Hz for the target and at the envelope rate of the adaptor when the adaptor was a complex tone, or the audio frequency of the adaptor when it was a pure tone.
Experiment 2A: Adaptor at 75 dB SPL
An RM two-way ANOVA (with factors time, A-start vs A-end, and four adaptor types) was calculated on the FFR for all adaptors. The main effect of adaptor type was highly significant [F (3,27) = 30.41, p < 0.001], while the effect of time was not (p = 0.16), but there was a significant interaction between type of adaptor and time [F (3,27) = 5.27, p = 0.011]. The interaction was driven by the 90-Hz envelope rate adaptor for which the FFR increased rather than decreased over time. A second ANOVA, without the 90-Hz envelope rate adaptor, showed that the FFR strength was significantly larger during period A-start than during period A-end [F (1,9) = 5.84, p = 0.039]. The main effect of adaptor type was highly significant [F (2,18) = 33.97, p < 0.001], and there was no significant interaction between type of adaptor and time (p = 0.33, after the exclusion of the 90-Hz adaptor). Note that excluding the 504-Hz rather than the 90-Hz adaptor from the ANOVA did not eliminate the significant interaction between type of adaptor and time [F (2,18) = 6.24, p = 0.009], with no main effect being significant, showing that the pattern over time observed for the 90-Hz adaptor differed from that observed for the other adaptors. Individual t tests, calculated for each adaptor type separately, showed that the FFR strength was significantly larger during period A-start than during period A-end for condition AC213 [difference = 1.0 dB, t (9) = 3.63, p = 0.022], but missed significance for condition AC504 and AP213 [difference = 2.1 dB, p = 0.296 and difference = 0.9 dB, p = 0.967, respectively] and was non-significantly smaller during period A-start than during period A-end for condition AC90 [difference = −1.2, t (9) = −2.43, p = 0.152]. For condition AC213, the FFR strength was significantly larger during period A-start than during period T-start [difference = 1.2 dB, t (9) = 2.95, p = 0.016], showing that the presence of the short offset and onset ramps in this otherwise “continuous” complex did not lead to complete recovery from adaptation for the target.
The observed increase in FFR over the time course of the 90-Hz envelope rate adaptor (which led to a significant interaction between type of adaptor and time) was unexpected, and we do not have a definite explanation for it. One possible explanation is that, for the 90-Hz adaptor, two generators contributed to the FFR; if the two generators were out of phase, then adaptation of one of them could lead to an increase in the FFR. Another possible explanation is connected with the small number of envelope periods within the analysis time window for the 90-Hz rate adaptor. Reducing the duration of the analysis time window (from 50 to 21.13 ms) for the 213-Hz rate adaptors (conditions AC213 and AP213), so as to contain the same number of cycles as for the 90-Hz rate in the 50-ms windows, did not generally result in an increase of the FFR from the first to the fourth time windows. However, it did introduce non-monotonic FFR strengths over the first four time windows for both 213-Hz rate adaptors, indicating that increased measurement variability might contribute to the observed increase in the FFR for the 90-Hz adaptor.
Any effects of the adaptor on the target FFR were expected to be strongest at the beginning of the target. For this reason, in this and the following experiments, the statistical analyses were restricted to T-start, but the figures include T-end for completeness. To assess whether there was any evidence for envelope-rate-selective adaptation in the FFR for the target, an RM one-way ANOVA was calculated on the FFR strength for the target during period T-start (Fig.
2, fourth cluster of bars from the left) for all conditions with a complex tone adaptor. The effect of adaptor envelope rate was not significant (
p = 0.12). Thus, there is no evidence that the complex tone adaptor with an envelope rate equal to that of the target led to more adaptation of the target FFR than the complex tone adaptors with envelope rates different from that of the target.
To compare the FFR strength at time T-start across all adaptor conditions, i.e., the complex tone adaptors and the pure tone adaptor which occupied a different frequency region, an RM one-way ANOVA was calculated on the FFR during period T-start for all adaptors. This showed that the FFR strength differed significantly across adaptor conditions [F (3,27) = 9.27, p < 0.001]. To assess whether there was any evidence for carrier-frequency-selective adaptation, the FFR strength during period T-start was compared between the pure tone and the complex tone adaptor conditions. Individual t tests showed that the FFR strength during period T-start was significantly larger after the pure tone adaptor (AP213) than after any of the complex tone adaptors [p = 0.003, p = 0.009, and p = 0.006 for the comparison of AP213 with AC213, AC90 and AC504, respectively]. Thus, there was a significant effect of frequency region (or/and the absence or presence of modulation), with more adaptation when the adaptor was in the same frequency region as the target (or/and was modulated) than when it was below the target. After the pure tone adaptor, the target FFR was slightly greater than at the beginning of the AC213 adaptor [difference = 1 dB, t (9) = 2.34, p = 0.044]. This could be the consequence of accumulated adaptation across trials, and will be discussed below.
The lack of adaptation of the target FFR in condition AP213 also indicates that the FFR to the target is not mainly driven by neurons with CFs corresponding to the QDP produced by the target complex tone. This suggests that, for our stimuli at least, the FFR to the envelope of a complex tone can be measured without a major confounding effect of the QDP and without the need for a noise masker.
Experiment 2B: Adaptor at 80 dB SPL
Stimulus generation and presentation, and analysis of the recorded signal, were done in the same way as for experiment 2A, with the following exceptions: (i) Four blocks of 1500 trials were run for each condition, with conditions being alternated from one block to the next. (ii) The first and second time ranges for FFR analysis were from 12 to 62 ms after adaptor onset (A-start) and from 62 to 112 ms after adaptor onset (A-mid).
As in experiment 2A, adaptation over time is visible in the FFR in response to both adaptors. An RM two-way ANOVA (with factors time, A-start vs A-end, and adaptor type) was calculated on the FFR for both adaptors. The main effects of adaptor type and time were both highly significant [F (1,10) = 129.12, p < 0.001 and F (1,10) = 37.07, p < 0.001 for adaptor type and time, respectively]. There was also a significant interaction [F (1,10) = 16.39, p = 0.002], showing that there was significantly more adaptation (in dB) for the 504-Hz than for the 213-Hz adaptor. This is consistent with the results of experiment 1 that showed more adaptation for higher than for lower frequencies and F0s. Individual t tests, calculated for each adaptor type separately, showed that the FFR strength was significantly larger during period A-start than during period A-end for condition AC213 [difference =1.1 dB, t (10) = 4.69, p = 0.002] and condition AC504 [difference = 3.7 dB, t (10) = 5.40, p = 0.001].
To assess whether there was any evidence for envelope-rate-selective adaptation in the FFR for the target, a paired t test compared the FFR strength for the target during period T-start between the two adaptor conditions. The FFR strength in condition AC213 was not significantly smaller than for condition AC504 [difference = 0.1 dB, p = 0.85; two-tailed]. Thus, there is no evidence that the complex-tone adaptor with an envelope rate equal to that of the target reduced the FFR strength to the target more than the complex-tone adaptor with an envelope rate different from that of the target.
Discussion
Like experiment 1, experiment 2B showed that FFR adaptation, measured in dB, was greater during presentation of a 504-Hz rate stimulus than during a lower-rate stimulus. Measured in this way, the results show an effect of modulation rate for stimuli filtered in the same frequency region and with similar excitation patterns. Thus, it seems that the effects of stimulus frequency and F0 on peristimulus adaptation are not exclusively driven by CF. Rather, F0 or envelope rate as such can affect the degree of adaptation in the peristimulus FFR. Such differences in amounts of adaptation might affect comparisons between FFR measures for stimuli with different F0s or envelope rates and duration.
We did not find any evidence that the envelope-related FFR in response to a complex-tone target is adapted (reduced) more by a complex-tone adaptor that has the same envelope rate as the target than by one with a different envelope rate. Hence, we found no evidence for tuning in the modulation domain in the FFR. Factors that would make it difficult to observe selectivity of adaptation with regard to envelope rate are: First, not all neurons in the IC and LL, and not all neurons that contribute to the FFR, have responses that are selective for envelope rate (Langner and Schreiner
1988; Joris et al.
2004). The responses of non-selective neurons would dilute the measured effects of any differences in adaptation produced by the three complex tone adaptors. Second, because the FFR requires averaging over several thousand trials, there is the possibility of accumulated adaptation effects across trials from the target itself. Specifically, stimuli were played with a repetition rate of 1.81/s, resulting in an inter-target interval (end of target to start of next target) of 452.5 ms. Possibly, envelope-rate-sensitive neurons firing in response to the target had not yet fully recovered from adaptation due to the previous target. This too would dilute the measured effect of any differential adaptation produced by the three complex-tone adaptors. Accumulated adaptation effects across trials could also occur from the adaptors, which were separated by an inter-adaptor interval of 352.5 ms. The observation of experiment 2A, that the FFR for AC213 at A-start was slightly smaller than that at T-start in condition AP213 suggests that recovery from adaptation produced by the adaptors was not complete at the beginning of a trial. However, this is unlikely to have reduced possible differences in adaptation between conditions, as the different adaptor conditions were tested in a blocked design (with breaks between blocks).
Interestingly, there is some recent evidence, from single-unit recordings in the IC of rabbits, that only a subset of units exhibit suppression in the discharge rate to a signal modulation when this is preceded by a masker modulation of the same rate (Wojtczak et al.
2011). Wojtczak et al. (
2011) recorded from units that were sensitive to sinusoidal amplitude modulation (AM), and the units were stimulated at their CF with their best modulation frequency. The carrier was not interrupted between masker and signal modulation. Some units’ responses were even enhanced following the masker modulation. Overall, Wojtczak et al. (
2011) reported only slightly lower firing rates when the signal modulation was preceded by a masker modulation of identical modulation rate than when it was preceded by an unmodulated carrier. If this observation of weak adaptation of neural responses to modulation per se holds in humans as well as rabbits, it might account for the present finding of equal target adaptation for all complex-tone adaptor rates in the FFR.
The results of experiment 2A showed that the envelope-related FFR in response to a complex-tone target showed a smaller reduction following a pure-tone adaptor with a frequency corresponding to the envelope rate of the target than following a complex-tone adaptor that was filtered into the same frequency region as the target complex. This provides strong evidence that the FFR response to the complex tone was not primarily driven by a propagated QDP and shows that an adaptor that is several octaves lower than the center frequency of the target did not produce any measurable adaptation. Carrier-frequency-specific adaptation in the FFR was investigated in experiment 3.