Auditory signal detection appears to depend on temporal integration of subthreshold activity in auditory cortex

BR A I N R ES E A RC H 1 3 8 5 ( 2 01 1 ) 2 0 6 –21 6

available at www.sciencedirect.com

www.elsevier.com/locate/brainres

Research Report

Auditory signal detection appears to depend on temporal integration of subthreshold activity in auditory cortex Bernd Lütkenhöner⁎ Section of Experimental Audiology, ENT Clinic, Münster University Hospital, Münster, Germany

A R T I C LE I N FO

AB S T R A C T

Article history:

The threshold of hearing decreases with increasing sound duration up to a limit of a few

Accepted 3 February 2011

hundred milliseconds, whereas other auditory time constants are orders of magnitude

Available online 5 March 2011

shorter. A possible solution to this resolution–integration paradox is that temporal

Keywords:

resolution. But this would require information about subthreshold events in the periphery

Threshold of hearing

to reach higher centers. Here we show that this prerequisite is fulfilled. The auditory evoked

Auditory brainstem response

response to a just perceptible pulse series does basically not depend on whether single

Auditory evoked potential

pulses are below or above behavioral threshold. The failure to find evidence of temporal

Resolution–integration paradox

integration up to response latencies of 30 ms suggests that the integrator is located more

integration occurs more centrally than computations depending on high temporal

centrally than primary auditory cortex. By using noise to its advantage, the auditory system apparently has established a central integration mechanism that is about as efficient as the peripheral one in the visual system. © 2011 Elsevier B.V. All rights reserved.

1.

Introduction

In various sensory channels, a percept can be formed by the aggregation of subthreshold activations over time, which is often denoted as temporal integration. Within limits, this process yields a roughly reciprocal relationship between stimulus duration and threshold intensity. A particularly well-studied case is the human auditory system, where a tenfold increase in stimulus duration reduces the threshold by 8–10 dB (perfect energy integration would lead to an improvement by 10 dB) and thresholds continue to improve up to stimulus durations of 500 ms (Florentine et al., 1988). Auditory temporal integration at threshold has puzzled researchers for decades (for a review, see, e.g., Algom and Babkoff, 1984;

Eddins and Green, 1995), and the subject is still a matter of controversy (see, e.g., the discussion triggered by Heil and Neubauer, 2004). While the idea of temporal integration at a peripheral site, in the synapse between hair cell and afferent auditory nerve fiber, is still being discussed (Heil and Neubauer, 2003; Heil et al., 2008; Neubauer and Heil, 2008), psychoacoustic experiments suggest a central origin of auditory temporal integration (e.g., Viemeister and Wakefield, 1991). The latter hypothesis naturally raises the question as to where and how the integration takes place. Since unequivocal electrophysiological correlates of auditory temporal integration are missing, the nature of the underlying processes can only be speculated about. It is noteworthy in this context that the auditory system has the most extensive subcortical

⁎ Corresponding author at: Section of Experimental Audiology, ENT Clinic, Münster University Hospital, Kardinal-von-Galen-Ring 10, 48129 Münster, Germany. Fax: +49 251 83 56882. E-mail address: [email protected]. Abbreviations: ABR, auditory brainstem response; AEF, auditory evoked field; AEP, auditory evoked potential; SEM, standard error of the mean; SL, sensation level 0006-8993/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.brainres.2011.02.011

207

BR A I N R ES E A RC H 1 3 8 5 ( 2 01 1 ) 2 0 6 –2 16

component of all sensory systems. Thus, any response property described in the auditory cortex could be generated subcortically (Nelken et al., 2004). Single-unit studies devoted to temporal integration in the auditory pathways are scarce, and only limited conclusions can be drawn from them (see the recent review by Recanzone and Sutter, 2008). While providing a less detailed view than singleunit data, auditory evoked potentials (AEP) recorded from the scalp have the advantage of being representative for populations of neurons. Moreover, different levels of the auditory pathways can be studied simultaneously. In the present study, wave V of the auditory brainstem response (ABR) and the early cortical response Pa are considered. A series of eight tone pulses with interpulse-intervals of 16 ms was presented at a rate of 4/s. At the lowest sound level tested, a single pulse was just below behavioral threshold, whereas the pulse series just exceeded threshold. A previous study (Lütkenhöner and Seither-Preisler, 2008) had shown that ABR measurements at such low levels are extremely time-consuming, but feasible. The idea underlying the present study was as follows. If there is no temporal integration occurring more peripheral to the generator of a specific AEP component, the response to the first pulse should not differ from the responses to the subsequent pulses in the series. Such an outcome would consequently suggest a more central origin of temporal integration.

V Pa

V Pa

V Pa

V Pa

V Pa

V Pa

Results

2.1.

Visual inspection of the response waveforms

Fig. 1 shows, for four different stimulus levels, the time course of the AEP elicited by the tone pulse series. The upper curve represents the highest level tested (60 dB SL1). Most conspicuous is a positive wave that occurs about 26 ms (gray vertical lines) after each of the eight pulses. Regarding its latency, the wave can be identified as the middle-latency response component Pa (see e.g., Pratt, 2007). Clearly visible is also an early positive wave, which occurs about 6 ms after each pulse (dotted vertical lines; response peaks additionally marked by filled triangles). This is wave V of the ABR (see e.g., Burkard and Don, 2007). With the exception of the response to the first pulse, wave V interferes with the rising slope of the Pa response to the preceding pulse. The other three curves represent the three lowest sound levels. The response amplitudes are much smaller now (note the different amplitude scale), and the response latencies are significantly prolonged (indicated by the asynchrony of dotted lines and filled triangles). Wave Pa has a prolonged latency as well so that the peaks are found a few milliseconds later than the gray vertical lines. Apart from these differences in amplitude and latency, the four curves closely resemble each other.

V Pa

Pa

400 nV

V

2.

dB SL1 60

80 nV

5

2

-1 0

50

100

150

200

Time (ms) Fig. 1 – Auditory evoked potential elicited by a series of eight tone pulses presented at 16-ms intervals (pulse occurrence times indicated by black vertical lines). Response waveforms (ipsilateral channel) are shown for −1, 2, 5, and 60 dB SL1 (dB values relative to the behavioral threshold for a single pulse). Each pulse elicited a clear wave V (peaks marked by filled triangles); at the highest level, this response had a latency of 6 ms (dotted vertical lines). The dominant response component was the middle-latency wave Pa, with a latency of about 26 ms (gray vertical lines) at the highest level. Sound level has a tremendous effect on the response amplitudes (note that the upper curve has a different scale), and there is also a clear effect on the response latencies. However, the basic pattern of the response appears to be largely independent of level.

208

BR A I N R ES E A RC H 1 3 8 5 ( 2 01 1 ) 2 0 6 –21 6

The Pa and other relatively slow response components can be largely removed by high-pass filtering, leaving the ABR. Fig. 2 shows how the ABR waveform changes as a function of sound level. The figure also allows one to compare the response to the first pulse (gray curve) with the response to the last pulse (black curve). Wave V dominates the response throughout the whole

V 50 nV

I

dB SL1 60

range of stimulus levels. Other response components such as wave I (marked by an arrow) cannot be identified near threshold, and therefore they will not be considered here. The first and the last pulse elicit similar responses. The differences that can be noticed at higher levels (in the response to the last pulse, the amplitude of wave I is reduced and wave V is delayed by about 0.3 ms) can be attributed to adaptation. A more detailed view on the two lowest levels is provided in Fig. 3, where the response to the first pulse (gray curve) is compared with the responses to the subsequent seven pulses (thin black curves). The curves are relatively noisy, but apart from that, they show no obvious difference. Averaging the responses to the eight pulses yields a significant improvement of the signal-to-noise ratio. This mean response (white curve on black background) will serve as reference in the analyses that follow. The amplitude of the wave V in the mean response will be referred to as the mean wave V amplitude.

2.2.

50

40

30

20

15 10 5 2 -1 0

5

10

15

Time (ms) Fig. 2 – Auditory brainstem response as a function of sound level. A gray curve shows the response to the first pulse of the series, a black curve the response to the last pulse (ipsilateral channel). Wave V, the dominant response component for levels below 60 dB, can be traced down to the lowest level, 1 dB below the behavioral threshold for a single pulse. Wave I (marked by an arrow) cannot be identified near threshold, owing to its significantly lower amplitude. At higher sound levels, the wave V in response to the last pulse is delayed and reduced in amplitude, which can be attributed to adaptation. But near threshold, there is no obvious difference between the responses to first and last pulse.

Quantitative analyses of peak amplitudes

Fig. 4a allows one to quantitatively consider the question as to whether the amplitude of the first wave V (ordinate) significantly deviates from the mean wave V amplitude (abscissa). The lowest five sound levels (−1 to 15 dB) are each represented by error crosses, composed of error bars for both abscissa and ordinate. The error bars intersect at a point corresponding to the mean values, and they extend one standard error of the mean (SEM) in each direction. The horizontal error bars are shorter than the vertical ones because averaging eight responses reduces the SEM by the factor √8. The level dependence of the error-bar lengths results from the fact that the number of averaged responses increased with decreasing sound level. By means of one-sample t-tests (with a hypothesized mean of zero) it was verified that all the mean values considered in the figure significantly differ from zero: the smallest t-value, obtained for the ipsilateral wave V at 2 dB SL1, was 3.61 (p ≈ 0.0003). The dotted line represents the hypothetical case that the response to the first pulse and the mean response have exactly the same amplitude. The data points for the two recording channels clearly scatter around this line. A least squares fit gave the constant of proportionality as 1.00± 0.07 for the ipsilateral channel (black line) and 0.95± 0.07 for the contralateral channel (gray line). Fig. 4b is analogous, but the ordinate now represents the amplitude of the Pa in response to the first pulse of the series. One-sample t-tests showed again that all the mean values significantly differ from zero: the smallest t-value, obtained for the contralateral Pa at −1 dB SL1, was 3.20 (p ≈ 0.0014). A least squares fit gave the constant of proportionality as 3.67± 0.16 for the ipsilateral channel (black line) and 4.19 ± 0.19 for the contralateral channel (gray line). The figure suggests a somewhat stronger activation of the contralateral Pa generator, at least at higher stimulus levels. Quantitative results for the Pa responses to the subsequent pulses would not be meaningful because some interference with responses of longer latency, elicited by previous pulses, cannot be excluded. However, the bottom curve in Fig. 1 suggests that temporal integration has no major effect on the Pa, if there is an effect at all. Fig. 5 shows the mean wave V amplitude as a function of sound level (error bars, representing twice the SEM). Fitting the

209

BR A I N R ES E A RC H 1 3 8 5 ( 2 01 1 ) 2 0 6 –2 16

a

2 dB SL1

60

ipsilateral contralateral

Amplitude of first wave V (nV)

50

40

15 dB SL 1 30

10

5 20

2 -1

10

0

b

Amplitude of first Pa (nV)

200

100

-1 dB

15 dB SL 1

10

5

2 -1 0 0

10

20

30

40

50

5 nV

Mean wave-V amplitude (nV)

0

5

10

15

Time (ms) Fig. 3 – Detailed look at the auditory brainstem responses recorded at −1 dB SL1 (below threshold) and 2 dB SL1. The gray curve shows the response to the first pulse of the series; the thin black curves show the responses to the other seven pulses. The average of these eight curves is represented by a white curve on black background.

data in the range − 1 to 10 dB to a theoretical function derived for the compound action potential of the auditory nerve (Lütkenhöner, 2008) resulted in the black curve. Data and model are in good agreement. The deviation at 15 dB SL1 can

Fig. 4 – Mean wave V amplitude versus amplitude of response to first stimulus. Error crosses, composed of orthogonal error bars, indicate the standard errors of the mean (ipsilateral channel in black, contralateral channel in gray). (a) Mean wave V amplitude (abscissa) versus amplitude of wave V in response to the first pulse of the series (ordinate). The two amplitudes do not significantly differ, suggesting that temporal integration takes place more centrally than the generation of wave V. (b) Mean wave V amplitude (abscissa) versus amplitude of wave Pa in response to the first pulse of the series (ordinate). The data are consistent with the idea that the two amplitudes are roughly proportional to each other.

be attributed to the fact that additional compressive components, which are not accounted for by the model, become relevant at higher levels. Consistent results were obtained by reanalyzing data of a previous experiment (Lütkenhöner and Seither-Preisler, 2008), in the figure shown as dots. The level dependence of the response amplitude near-threshold changes, according to the model, from exponential (dashed curve) to linear (dotted line). There is no indication of a hard threshold, i.e. the existence of a sound level where the response amplitude becomes zero.

210

BR A I N R ES E A RC H 1 3 8 5 ( 2 01 1 ) 2 0 6 –21 6

60

ipsilateral contralateral

Mean wave-V amplitude (nV)

50

40

30

20

10

0 -5

0

5

10

15

Level (dB SL) Fig. 5 – Mean wave V amplitude as a function of sound level. The results of the present experiment (error bars, representing twice the standard error of the mean) are supplemented by results of a previous experiment (Lütkenhöner and Seither-Preisler, 2008), shown as dots. The black curve represents a theoretical amplitude function (Lütkenhöner, 2008). Data and model contradict the idea of a hard threshold.

2.3.

Behavioral thresholds

The behavioral threshold for a single tone pulse (corresponding to 0 dB sensation level; see section 4.3 for methodological details) corresponded to 28 dB peak-to-peak equivalent sound pressure level. The threshold for the series of eight pulses was 4.4 dB lower than the threshold for a single pulse. Under the common assumption that the threshold level is linearly related to the logarithm of stimulus energy (here proportional to the number of pulses), this estimate corresponds to a threshold reduction by 4.9 dB per tenfold increase in the number of pulses.

3.

Discussion

The key result of this study is that the auditory evoked response to a just perceptible pulse series does basically not depend on whether single pulses are below or above behavioral threshold. This finding has not only implications for auditory temporal integration at threshold but also raises the question as to how information about subthreshold events is encoded and conveyed in the auditory pathways. These two aspects will be discussed in detail below. First of all, however, previous work will be briefly reviewed, to provide a basis for the interpretation of the current data.

3.1.

Brief review of previous work

3.1.1.

Temporal integration in sensory systems

Temporal integration can be found in sensory organs as simple as the noctuid moth ear, which contains only two receptor cells.

Tougaard (1996) measured the thresholds for single clicks, pairs of clicks with a separation of 1–20 ms, and trains of up to eight clicks at separations of 1–2 ms, and he found the results fully consistent with the assumption of a leaky energy integrator with a time constant of about 4 ms. The integration was attributed to the passive electric properties of the receptor cell membrane. In the highly sophisticated sensory systems that have evolved in vertebrates, the time constants are generally much longer. But there are also exceptions, and temporal integration does not always serve the purpose of reducing the threshold for the detection of weak signals. Stüttgen and Schwarz (2010) compared in rats the psychometric detection performance for whisker stimuli composed of pulses with the firing properties of concomitantly recorded barrel cortex neurons. To match neurometric and psychometric performance, barrel cortical signals needed to be filtered by a leaky integrator with a time constant of only 5–8 ms. Thus, the whisker system does not seem to make use of temporal integration to improve detectability (the integration seems to be necessary to extract features that carry information about the spatial configurations being palpated). The vestibular system is exceptional as well in that temporal integration of velocity and angular velocity or even double integration of acceleration is critical for encoding the motion of the head relative to the outside world (see e.g., Israel and Berthoz, 1989; Angelaki and Cullen, 2008; Klier et al., 2008). The best-studied integrators are brainstem nuclei belonging to the oculomotor system; they integrate velocity signals from the inner ear and produce maintained, graded, compensatory eye movements (Koulakov et al., 2002). Threshold reduction by means of temporal integration has been most thoroughly investigated for sight, hearing, and touch. The temporal limit for complete integration in the rod visual system is of the order of 100 ms (Barlow, 1958; Saunders, 1975; Zuidema et al., 1981; Sharpe et al., 1993). Thresholds in the human auditory system continue to improve up to stimulus durations of 500 ms, as already noted above. Comprehensive reviews of the pertinent literature can be found elsewhere (Algom and Babkoff, 1984; Eddins and Green, 1995). In the sense of touch, temporal integration is mainly a feature of the channel associated with the Pacinian corpuscles (Bolanowski et al., 1988; Checkosky and Bolanowski, 1992; Gescheider et al., 1999). Increasing the duration of a vibrotactile stimulus from 10 to 1000 ms leads to a threshold improvement similar to that in the auditory system (Verrillo, 1965; Gescheider et al., 1999; Gescheider et al., 2002). By contrast, no temporal integration seems to occur in the channel associated with the rapidly adapting mechanoreceptive fibers (NP-I channel in the psychophysical work by Bolanowski et al., 1988), where a tactile sensation apparently requires only a single spike in a single fiber (Hensel and Bowman, 1960; Johansson and Vallbo, 1979). In the human taste system, Bagla et al. (1997) found temporal integration for stimulus durations ranging from 200 to 1500 ms, in a manner analogous to that seen in the sensory systems considered before. Stimulus duration appears to have only a relatively small effect on electrogustometric thresholds, though (Loucks and Doty, 2004). As to nasal chemesthesis, temporal integration with a roughly linear relationship between the logarithm of stimulus duration and the logarithm of concentration has

BR A I N R ES E A RC H 1 3 8 5 ( 2 01 1 ) 2 0 6 –2 16

been reported (Wise et al., 2007). This finding is consistent with an earlier report according to which the common chemical sense behaves more like a total mass detector than a concentration detector (Cometto-Muniz and Cain, 1984). Temporal summation was shown also for pain evoked by electrical (Arendt-Nielsen et al., 1994), thermal (Nielsen and Arendt-Nielsen, 1998), and mechanical stimuli (Nie et al., 2009). The physiological mechanisms underlying temporal integration are, for most sensory channels, not well established, implying uncertainty also as to the site of integration. The visual system constitutes an exception in that temporal integration is known to take place already in the retina. Thousands of rod photoreceptors are pooled in a specialized circuitry, and only a few photon absorptions in the pool are required for near-threshold sensations (Okawa and Sampath, 2007). It is not clear yet to what extent further summation of quantal signals occurs more centrally in the visual pathway (Saszik et al., 2002). At least in cat visual cortex, neurons with long time constants that might represent temporal integrators were found (Duysens et al., 1991). By contrast, an exclusively central origin of temporal integration is assumed for the sense of touch, more precisely the sensory channel associated by the Pacinian corpuscles (Bolanowski et al., 1988). But as in the case of the auditory system, conclusive electrophysiological evidence is still lacking.

3.1.2. Physiological correlates of auditory temporal integration at threshold Zwislocki (1960) opposed the idea of a peripheral origin of auditory temporal integration, in the organ of Corti, by arguing that first-spike latencies in excess of 100 ms for nearthreshold stimuli of long duration were not observed in firstorder neurons. Regarding the amazing auditory sensitivity to binaural time differences, he considered it unlikely that longterm integration takes place before the pathways ascending from the two ears meet. However, more recent studies challenge this argument. Auditory nerve fibers with very low spontaneous rate can indeed show the expected long firstspike latencies (Heil and Irvine, 1997). Moreover, temporal integration functions derived from such data are remarkably similar to those found at the perceptual level (Heil and Neubauer, 2003; Heil et al., 2008). This led Heil and Neubauer (2001; 2003) to propose that the integration might take place in the first synapse of the auditory pathway, between the inner hair cells and the distal dendrites of the auditory nerve fibers. However, this hypothesis is not conclusive in all respects. First, the assumption of temporal integration is not essential to explain the first-spike latencies of low-spontaneous-rate fibers (Krishna, 2002; 2006). Second, the most sensitive auditory nerve fibers have a high spontaneous rate (Liberman, 1978), and, for such fibers, the first spike after the onset of a near-threshold stimulus can almost certainly be attributed to spontaneous activity. A possible physiological correlate of the behavioral phenomenon of temporal integration is found by determining thresholds based on a spike-counting criterion. Thresholdduration functions for single auditory nerve fibers (Clock Eddins et al., 1998) as well as chopper and primary-like units in the anteroventral cochlear nucleus (Clock et al., 1993) closely

211

resemble those measured psychophysically. Similar observations have been published much earlier by Gersuni (1965). However, such investigations do not allow one to characterize the nature or location of the integrator, which remains to be clarified at higher levels of the auditory nervous system (Clock Eddins et al., 1998; Meddis, 2006). Clues as to the site of temporal integration in the auditory pathways can be obtained from electrical stimulation studies. Applying series of current pulses through a scala tympani electrode, Beitel et al. (2000) recorded threshold responses for single neurons and multineuronal clusters in the central nucleus of the inferior colliculus (ICC) and primary auditory cortex. Amongst other things, their results suggest that the output from ICC may be integrated spatially across neurons and temporally across pulses. Gerken et al. (1991) implanted permanent electrodes in the cochlear nucleus and the ICC of cat. Then they behaviorally measured detection thresholds for series of up to 16 current pulses. For both stimulation sites, the animals showed temporal integration on a time scale of hundreds of milliseconds, resembling the temporal integration for acoustic stimuli. This led the authors to conclude that a major part of temporal integration in the auditory pathways takes place in the inferior colliculus or higher. After they had produced a sound-induced hearing loss with a mean permanent threshold shift for acoustic stimuli of 48.5 dB, the mean slope of the temporal integration function for electrical stimulation dropped from 7.6 dB to 1.5 dB per factor of 10 pulses. The latter value is roughly consistent with Donaldson et al. (1997), who typically obtained a threshold reduction by 0.5 dB per doubling of the number of electrical pulses, although for cochlear implant patients. The reduced slope of the temporal integration function after cochlear damage has to be considered in conjunction with another finding: stimulation hypersensitivity, which is indicated by lower thresholds. Gerken (1993) proposed that hypersensitivity results from a reduction in inhibition rather than a reduction of internal neural noise. The fact that short-duration electrical stimulation results in a greater hypersensitivity (Gerken et al., 1991) may largely explain the reduced slope of the temporal integration function after cochlear damage.

3.2. Peripheral versus central origin of auditory temporal integration Notwithstanding all methodological differences, the present study bears reference to the electrical stimulation studies just reviewed. What the studies have in common is that the input to the presumed temporal integrator is a series of synchronized neural activations. A related view has been expressed by Gerken et al. (1991), who argued that the central effects produced by brief acoustic stimuli and electrical pulses are similar (if presented at a sufficiently low rate). The electrical stimulation studies render it unlikely that temporal integration takes place more peripherally than inferior colliculus. The present study is absolutely consistent with that view: The putative origin of wave V of the ABR is the termination of the lateral lemniscus in the inferior colliculus (Møller, 2007) so that the failure to find evidence of temporal integration for that wave suggests an integrator site that is located more centrally in the auditory pathways.

212

BR A I N R ES E A RC H 1 3 8 5 ( 2 01 1 ) 2 0 6 –21 6

Major temporal integration effects can also be excluded for wave Pa. As this wave is presumably generated more centrally than primary auditory cortex (Lütkenhöner et al., 2003), auditory temporal integration appears to be a phenomenon that is essentially based on higher cortical processing. This experiment was not designed to study later AEP waves such as the N100, with a typical latency of 100 ms. The N100 would, in all probability, have dominated the response if the pulse series had been presented at a much slower rate. Previous experiments, mostly using a tone or noise burst as the stimulus, suggest that this wave may indeed reflect temporal integration. The level dependence of the N100 latency was found to be roughly consistent with the assumption that there is almost perfect temporal integration of a neural response proportional to sound pressure (Lütkenhöner and Klein, 2007). At higher levels, the N100 amplitude appears to increase with stimulus duration up to a duration of about 30–40 ms (Onishi and Davis, 1968; Joutsiniemi et al., 1989; Gage and Roberts, 2000); for stimuli with slow onset, integration times even up to 50–70 ms have been reported (Biermann and Heil, 2000). Temporal integration effects on the N100 were also described for series of clicks (Forss et al., 1993), i.e., for stimuli that resemble the one used in the present experiment. The results of this study are contradictory to Heil and Neubauer (2003), who proposed a temporal integrator in the first synapse of the auditory pathway. In their model, which may be applied to higher levels of the auditory pathways as well, the integrand is proportional to the envelope of sound pressure (or, in Neubauer and Heil, 2004, sound pressure diminished by a constant baseline value), and a neural spike is elicited when the integral exceeds a certain threshold. For a tone pulse series as used in the present study this would mean that, near threshold, the occurrence of a spike is critically dependent on the number of immediately preceding tone pulses. More specifically, the model predicts that, for each n with 1 < n ≤ 8, there is a certain sound level range where the first spike is elicited after the nth tone pulse. In the anuran auditory midbrain, a comparable integrator was indeed found, although integrating information rather than energy (Alder and Rose, 1998). For a population of neurons with somewhat different sensitivities one would expect a sound level range where the number of responding neurons, and so also the amplitude of the compound action potential, increases with the tone pulse number. However, the current data do not support these model predictions: Near threshold, the responses to the eight tone pulses appear to have essentially the same amplitudes and latencies. It has to be conceded that possible integration processes with time constants of a few milliseconds (perhaps taking place at a peripheral level) would have been completely overlooked in the present study, where the interval between successive tone pulses was 16 ms. Whether or not the human auditory system exhibits such short-term integration is questionable, though: short-term effects found by Krumbholz and Wiegrebe (1998) appear to be due to mechanical interaction at a level prior to the mechanical to neural transduction. Conclusions consistent with the present study were drawn by Rupp et al. (2002), although their experiment vastly differs from the present one. They studied middle-latency auditory evoked fields (AEF) elicited by clicks and up- and down-chirps. The up-chirps (characterized by an increasing instantaneous

frequency) were designed to compensate for the propagation delay along the human cochlea. The down-chirps were timereversed copies of the up-chirps. The potentials elicited by the three stimuli could be explained by the convolution of a (modelbased) peripheral spike probability function with a neural unit response. The authors concluded that, in the path which generates the middle-latency AEF and for the stimuli studied, there is little, if any, temporal integration prior to primary auditory cortex.

3.3. Flow of information about subthreshold events in the periphery At the lowest sound level studied, each of the eight pulses was below the behavioral threshold. If the audibility of the pulse series is the result of temporal integration in the cortex, the fundamental question arises as to how information about subthreshold events is conveyed from the periphery to higher centers. The neural code at the various levels of the auditory pathways is only partially understood (Eggermont, 2001; King and Nelken, 2009), but there is undoubtedly a strong stochastic component for near-threshold stimuli. Organisms often pay a high metabolic and structural price to reduce noise at the first stage of sensory processing, because noise levels set perceptual thresholds for later processing stages (Faisal et al., 2008). However, there are also potential benefits of noise (Faisal et al., 2008; Deco et al., 2009), and the auditory system apparently has evolved sophisticated strategies to use noise to its advantage. Hearing is classically thought to be limited by thermal noise (Ashmore, 2008). Hair cells are very efficient detectors, operating close to the physical limits: without a stimulus, their intracellular voltage basically mirrors the Brownian motion of their bundles, and the voltage caused by a barely perceptible stimulus has about the same magnitude (Denk and Webb, 1992). The Brownian motion of the hair cell bundles is not necessarily detrimental to the sensitivity of the auditory system. Through a phenomenon called stochastic resonance it may, somewhat counter-intuitively, serve to enhance the sensitivity of mechanoelectrical transduction (Jaramillo and Wiesenfeld, 1998). The most obvious manifestation of noise in the auditory system is the high spontaneous activity of the most sensitive auditory nerve fibers (Liberman, 1978). In the following, this aspect is identified as being pivotal for the flux of subthreshold information in the auditory pathways. A near-threshold stimulus modulates the spontaneous firing rate of auditory nerve fibers without noticeably increasing the mean firing probability (Rose et al., 1971; Joris and Yin, 1992; Dreyer and Delgutte, 2006). In principle, the modulation amplitude can be arbitrarily small, which means that a hard threshold does not exist. Regarding the neural representation of a specific stimulus this implies that it is basically irrelevant whether the stimulus is just above or below the perceptual threshold. The carrier of information is, in both cases, the spontaneous activity so that fundamental differences as to how the encoded information travels from the periphery to the cortex, finally resulting in a kind of cortical image, are not to be expected. This concept applies to any kind of near-threshold stimulus so that it seems reasonable to suppose that the main conclusions of this study have general validity.

BR A I N R ES E A RC H 1 3 8 5 ( 2 01 1 ) 2 0 6 –2 16

3.4.

Some concepts for central integration

The data suggest that a hard threshold does not exist (see Fig. 5). Moreover, they are consistent with the idea that each of the eight tone pulses imprinted basically the same image on the cortical areas responsible for the generation of wave Pa. Starting from this idea, a rough concept of temporal integration may be developed. If a single pulse is presented at a level below the behavioral threshold, the cortical image may be assumed to be so noisy that, when read out and processed by higher cortical circuits, it is most likely scored as background noise. A sequence of such noisy images, by contrast, has a structure that these circuits can exploit. The nature of this processing and whether the integration is accomplished by single neurons (Loewenstein and Sompolinsky, 2003) or neural networks (Koulakov et al., 2002; Okamoto and Fukai, 2009) can only be speculated about. A clue about the integration performed is given by the fact that the perceptual thresholds for the single pulse and the pulse series differed by 4.4 dB, which corresponds to about 1.5 dB per doubling of the number of stimuli. This is exactly the result predicted by the multiple-look theory (Swets et al., 1959; Viemeister and Wakefield, 1991), according to which decision making on the basis of N independent observations increases the internal signal-to-noise ratio by the factor √N. With increasing sound level, the sequence of cortical images can be assumed to become less noisy so that higher cortical processing can accomplish more than just signal detection, making it possible to estimate, for example, the pitch or the duration of the stimulus. These ideas are in harmony with an approach taken by Eddins and Green (1995), who denied that the auditory system is a simple one and hypothesized that different parts of the nervous system process auditory information in different ways. They proposed that, in any particular psychophysical task, the physiological information is selected to maximize the performance in that task. The conclusion that auditory temporal integration at threshold is essentially a cortical phenomenon does by no means devalue the fundamental importance of subcortical processing in general. But as to temporal integration, it makes perfect sense that the required processing is left to higher cortical circuits. That way, the auditory system can elude the problems discussed as the resolution–integration paradox (de Boer, 1985; Green, 1985; Viemeister and Plack, 1993). The paradox refers to the fact that basic time constants in the auditory system differ by orders of magnitude. While temporal integration has a typical time constant of several hundred milliseconds (Algom and Babkoff, 1984; Florentine et al., 1988), time constants for temporal acuity range from fractions of a millisecond to 30 ms (Eddins and Green, 1995), and some aspects of binaural hearing even require a time resolution in the range of a few microseconds (Gerstner et al., 1996). Temporal integration at more peripheral levels in the auditory pathways would necessitate separate channels for sound encodings with and without temporal integration. This study suggests that evolutionary pressure has resulted in a more parsimonious and flexible solution: the auditory system apparently has established a central integration mechanism that is about as efficient as the peripheral one in the visual system.

3.5.

213

Formation of a pitch percept

When presented at a level well above behavioral threshold, a pulse series elicits a pitch sensation at a frequency that is the reciprocal of the inter-pulse interval (62.5 Hz for the pulse series used in the present study). Such periodicity pitch perception clearly requires some kind of temporal integration, and, as already supposed for stimulus detection at the behavioral threshold, one would expect this integration to become noticeable in AEP components that arise from sites located more centrally than the integrator. Thus, the failure to find evidence of temporal integration in the data collected in this study suggests a cortical site not only for signal detection at threshold, but also for the formation of a pitch percept. A consistent conclusion has been reached by Patterson et al. (2002), although for a quite different stimulus: pitch-bearing noise. In their opinion, the role of subcortical processing is confined to the extraction of time interval information from the neural firing pattern in the auditory nerve. Such a preprocessing step would appear to be dispensable if the stimulus is a pulse series, as in the present study.

3.6.

Limitations of this study

While the ideas presented above are rather general, the experiment actually performed had a limited scope, and apart from that it was hitting the limits of what is possible. It is self-evident that some caveats are to be considered under such circumstances. The fact that only a single subject was investigated is probably of minor importance, because the results are consistent with those obtained in the preparatory study (Lütkenhöner and Seither-Preisler, 2008), in which two control subjects were investigated as well. What is more critical is that only a single stimulus was studied. The representation of a stimulus in the auditory pathways undergoes multiple complex transformations so that conclusions drawn for one condition do not necessarily apply to another. Thus, although we are not aware of any data that fundamentally question the basic concept developed here, it appears quite conceivable that other experimental conditions may require extensions and modifications. Another point is that focusing on the amplitudes of ABR wave V and the early cortical response Pa might not fully do justice to the data. The data analysis could, of course, be modified and extended in various ways: for example by considering other shape parameters and by applying data transformations (such as filtering, differentiation, integration, Fourier transformation). But without a hypothesis-based (or data-driven) clue as to how to proceed, such approaches run the risk of becoming arbitrary. To identify aspects that might have been overlooked in the analysis, it is useful to revisit the most elementary representation of the data, which is Fig. 1. As to waves V and Pa, it appears doubtful that another approach to the data would lead to fundamentally different conclusions. But it cannot be ruled out that there is relevant information in the slower activity. A prominent phenomenon in the 60-dB curve is the positive deflection with a latency of about 200 ms. The deflection might be a manifestation of AEP wave P200; the small amplitude (compared to other experiments) could be explained by the relatively fast repetition rate of the pulse series. The wave

214

BR A I N R ES E A RC H 1 3 8 5 ( 2 01 1 ) 2 0 6 –21 6

cannot be reliably identified near threshold, though, and at − 1 dB it is clearly missing. Another potentially relevant phenomenon is the slower positive wave that interferes with the waves V and Pa. The peak latency of this wave appears to increase with decreasing stimulus level: from about 40 ms at 60 dB to about 100 ms at −1 dB SL1. The latency increase may well be related to temporal integration. But such an interpretation requires caution, because it is difficult to distinguish the response to the current pulse series from the baseline variation caused by responses to previous pulse series (cf. Lütkenhöner, 2010). The problem is especially clear at −1 dB, where the slow activity resembles a sine wave with a period corresponding to the reciprocal of the repetition rate of the tone pulse series. A possible objection of a quite different nature is that temporal integration might depend on activity that is not visible in the recorded AEP. Melcher and Kiang (1996) showed that, in cat, only two of the multiple parallel output pathways of the cochlear nucleus significantly contribute to the ABR. Thus, it might be the case that auditory temporal integration depends on one of the other channels. As to wave V of the ABR, this possibility cannot be excluded. However, the time window considered here comprises almost 250 ms, and the idea that a hypothetical subcortical integrator has no noticeable effect whatsoever on the AEP in this time range appears not very plausible. For a system as complex as the human brain, one can, of course, never exclude with certainty that an observed phenomenon is due to a hidden mechanism. However, according to the law of parsimony that is ascribed to William of Occam (see e.g., Wildner, 1999), plurality should not be posited without necessity (“pluralitas non est ponenda sine necessitate”).

4.

Experimental procedures

4.1.

Stimulus

A series of eight Gaussian-shaped tone pulses with a 4-kHz carrier frequency and a full width at half maximum of 0.5 ms was presented at a rate of 4/s. The interval between two successive tone pulses (onset to onset) was 16 ms. The sound level was set relative to the behavioral threshold for a single pulse.

4.2.

Experiment

Considering the experiences gained in a previous study, where a single tone pulse was steadily presented at 16-ms intervals (Lütkenhöner and Seither-Preisler, 2008), it was clear from the outset that recording robust near-threshold responses to the stimulus described above would require an incredible measuring time, rendering the intended experiment virtually impossible. This necessitated adopting an unconventional approach, with the author being the only subject. Taking advantage of the fact that only passive listening was required during the measurement, the author temporarily did part of his deskwork (e.g., reviewing manuscripts or examining doctoral theses) in a double-walled anechoic room while simultaneously serving as a subject. The net time for the electrophysiological measurements exceeded 125 h, and completing the experiment took 61 measuring sessions distributed over nearly 10 months.

Each session began with a careful estimation of the threshold for a single pulse using a two-interval two-alternative forcedchoice procedure that was combined with a transformed up– down method converging to p = 0.794 (Levitt, 1971). In the subsequent electrophysiological measurements, the sound level was adjusted relative to this threshold (referred to as 0 dB SL1, where SL stands for sensation level), considering the range from −1 to 60 dB SL1. The sound level was typically kept constant for 1000 s. The temporal order in which the sound levels were tested was randomized. Only a subset of levels could be investigated in each session. Auditory evoked potentials were recorded between vertex and right and left earlobe, respectively, using Ag–AgCl electrodes. These channels are referred to as the ipsilateral channel (same side as the stimulation) and the contralateral channel. The recorded potential was amplified by 70 dB, filtered between 0.1 Hz (firstorder RC filter) and 3,000 Hz (eighth-order Bessel filter), and sampled at a rate of 10,000/s. The data chunks that were stored on disk (epochs) comprised a time window of 245 ms, starting 12 ms before stimulus onset. The total number of stimulus repetitions was level dependent: 652,000 at −1 dB, 284,000 at 2 dB, 234,000 at 5 dB, 147,000 at 10 dB, 106,000 at 15 dB, and more than 56,000 for higher levels. Further methodological details such as a precise explanation of the threshold estimation procedure can be found in a previous article (Lütkenhöner and Seither-Preisler, 2008).

4.3.

Data processing

After having completed the experiment, all further processing was done with MATLAB (The MathWorks Inc., Natick, MA, USA). Three versions of the data were created by applying different filters. Bandpass filters were used to pre-process the data for the analysis of the ABR (passband 200–1500 Hz) and the cortical component Pa (passband 20–200 Hz). For the overview plot in Fig. 1, the data were low-pass filtered at 1500 Hz. The signal was passed through the filter (fourth-order Butterworth) in both the forward and the reverse direction (Matlab routine FILTFILT). The filtered data were averaged over all sessions, with the artifact rejection threshold set at 100 μV. In addition, the standard error of the mean was calculated for each time point. Peak amplitudes were measured against a baseline that corresponded to the mean potential in the full 245-ms recording time window. Under the given circumstances (in particular because of the bandpass filtering), this baseline is essentially identical with the prestimulus baseline (but the latter would be more sensitive to residual noise in the data). The linear fits in Fig. 4 were done using a custom MATLAB implementation of the FITEXY algorithm by Press et al. (1992), which accounts for errors in both x and y. The original algorithm was modified such that the line could be forced to pass through the origin (assumption of a proportional relationship between x and y).

Acknowledgment The author thanks Ms. Katrin Zobel for carefully performing the extremely time-consuming experiment.

BR A I N R ES E A RC H 1 3 8 5 ( 2 01 1 ) 2 0 6 –2 16

REFERENCES

Alder, T.B., Rose, G.J., 1998. Long-term temporal integration in the anuran auditory system. Nat. Neurosci. 1, 519–523. Algom, D., Babkoff, H., 1984. Auditory temporal integration at threshold: theories and some implications of current research. In: Neff, W.D. (Ed.), Contributions to Sensory Physiology. Academic Press, New York, pp. 131–159. Angelaki, D.E., Cullen, K.E., 2008. Vestibular system: the many facets of a multimodal sense. Annu. Rev. Neurosci. 31, 125–150. Arendt-Nielsen, L., Brennum, J., Sindrup, S., Bak, P., 1994. Electrophysiological and psychophysical quantification of temporal summation in the human nociceptive system. Eur. J. Appl. Physiol. 68, 266–273. Ashmore, J., 2008. Cochlear outer hair cell motility. Physiol. Rev. 88, 173–210. Bagla, R., Klasky, B., Doty, R.L., 1997. Influence of stimulus duration on a regional measure of NaCl taste sensitivity. Chem. Senses 22, 171–175. Barlow, H.B., 1958. Temporal and spatial summation in human vision at different background intensities. J. Physiol. 141, 337–350. Beitel, R.E., Snyder, R.L., Schreiner, C.E., Raggio, M.W., Leake, P.A., 2000. Electrical cochlear stimulation in the deaf cat: comparisons between psychophysical and central auditory neuronal thresholds. J. Neurophysiol. 83, 2145–2162. Biermann, S., Heil, P., 2000. Parallels between timing of onset responses of single neurons in cat and of evoked magnetic fields in human auditory cortex. J. Neurophysiol. 84, 2426–2439. Bolanowski Jr., S.J., Gescheider, G.A., Verrillo, R.T., Checkosky, C.M., 1988. Four channels mediate the mechanical aspects of touch. J. Acoust. Soc. Am. 84, 1680–1694. Burkard, R.F., Don, M., 2007. The auditory brainstem response. In: Burkard, R.F., Eggermont, J.J., Don, M. (Eds.), Auditory Evoked Potentials - Basic Principles and Clinical Applications. Lippincott Williams & Wilkins, Philadelphia, pp. 229–253. Checkosky, C.M., Bolanowski, S.J., 1992. Effects of stimulus duration on the response properties of Pacinian corpuscles: implications for the neural code. J. Acoust. Soc. Am. 91, 3372–3380. Clock Eddins, A., Salvi, R.J., Wang, J., Powers, N.L., 1998. Threshold-duration functions of chinchilla auditory nerve fibers. Hear. Res. 119, 135–141. Clock, A.E., Salvi, R.J., Saunders, S.S., Powers, N.L., 1993. Neural correlates of temporal integration in the cochlear nucleus of the chinchilla. Hear. Res. 71, 37–50. Cometto-Muniz, J.E., Cain, W.S., 1984. Temporal integration of pungency. Chem. Senses 8, 315–327. de Boer, E., 1985. Auditory time constants: a paradox? In: Michelsen, A. (Ed.), Time resolution in auditory systems. Springer, Berlin, pp. 141–158. Deco, G., Rolls, E.T., Romo, R., 2009. Stochastic dynamics as a principle of brain function. Prog. Neurobiol. 88, 1–16. Denk, W., Webb, W.W., 1992. Forward and reverse transduction at the limit of sensitivity studied by correlating electrical and mechanical fluctuations in frog saccular hair cells. Hear. Res. 60, 89–102. Donaldson, G.S., Viemeister, N.F., Nelson, D.A., 1997. Psychometric functions and temporal integration in electric hearing. J. Acoust. Soc. Am. 101, 3706–3721. Dreyer, A., Delgutte, B., 2006. Phase locking of auditory-nerve fibers to the envelopes of high-frequency sounds: implications for sound localization. J. Neurophysiol. 96, 2327–2341. Duysens, J., Gulyas, B., Maes, H., 1991. Temporal integration in cat visual cortex: a test of Bloch's law. Vis. Res. 31, 1517–1528. Eddins, D.A., Green, D.M., 1995. Temporal integration and temporal resolution. In: Moore, B.C.J. (Ed.), Hearing. Academic Press, San Diego, pp. 207–242.

215

Eggermont, J.J., 2001. Between sound and perception: reviewing the search for a neural code. Hear. Res. 157, 1–42. Faisal, A.A., Selen, L.P., Wolpert, D.M., 2008. Noise in the nervous system. Nat. Rev. Neurosci. 9, 292–303. Florentine, M., Fastl, H., Buus, S., 1988. Temporal integration in normal hearing, cochlear impairment, and impairment simulated by masking. J. Acoust. Soc. Am. 84, 195–203. Forss, N., Mäkelä, J.P., McEvoy, L., Hari, R., 1993. Temporal integration and oscillatory responses of the human auditory cortex revealed by evoked magnetic fields to click trains. Hear. Res. 68, 89–96. Gage, N.M., Roberts, T.P., 2000. Temporal integration: reflections in the M100 of the auditory evoked field. NeuroReport 11, 2723–2726. Gerken, G.M., 1993. Alteration of central auditory processing of brief stimuli: a review and a neural model. J. Acoust. Soc. Am. 93, 2038–2049. Gerken, G.M., Solecki, J.M., Boettcher, F.A., 1991. Temporal integration of electrical stimulation of auditory nuclei in normal-hearing and hearing-impaired cat. Hear. Res. 53, 101–112. Gerstner, W., Kempter, R., van Hemmen, J.L., Wagner, H., 1996. A neuronal learning rule for sub-millisecond temporal coding. Nature 383, 76–81. Gersuni, G.V., 1965. Organization of afferent flow and the process of external signal discrimination. Neuropsychologia 3, 95–109. Gescheider, G.A., Berryhill, M.E., Verrillo, R.T., Bolanowski, S.J., 1999. Vibrotactile temporal summation: probability summation or neural integration? Somatosens. Mot. Res. 16, 229–242. Gescheider, G.A., Bolanowski, S.J., Pope, J.V., Verrillo, R.T., 2002. A four-channel analysis of the tactile sensitivity of the fingertip: frequency selectivity, spatial summation, and temporal summation. Somatosens. Mot. Res. 19, 114–124. Green, D.M., 1985. Temporal factors in psychoacoustics. In: Michelsen, A. (Ed.), Time resolution in auditory systems. Springer, Berlin, pp. 122–140. Heil, P., Irvine, D.R., 1997. First-spike timing of auditory-nerve fibers and comparison with auditory cortex. J. Neurophysiol. 78, 2438–2454. Heil, P., Neubauer, H., 2001. Temporal integration of sound pressure determines thresholds of auditory-nerve fibers. J. Neurosci. 21, 7404–7415. Heil, P., Neubauer, H., 2003. A unifying basis of auditory thresholds based on temporal summation. Proc. Natl Acad. Sci. USA 100, 6151–6156. Heil, P., Neubauer, H., 2004. Auditory thresholds re-visited. In: Collet, L. (Ed.), Auditory Signal Processing: Psychophysics, Physiology and Modeling. Springer, New York, pp. 454–470. Heil, P., Neubauer, H., Brown, M., Irvine, D.R., 2008. Towards a unifying basis of auditory thresholds: distributions of the first-spike latencies of auditory-nerve fibers. Hear. Res. 238, 25–38. Hensel, H., Bowman, K.A., 1960. Afferent impulses in cutaneous nerves in human subjects. J. Neurophysiol. 23, 564–578. Israel, I., Berthoz, A., 1989. Contribution of the otoliths to the calculation of linear displacement. J. Neurophysiol. 62, 247–263. Jaramillo, F., Wiesenfeld, K., 1998. Mechanoelectrical transduction assisted by Brownian motion: a role for noise in the auditory system. Nat. Neurosci. 1, 384–388. Johansson, R.S., Vallbo, A.B., 1979. Detection of tactile stimuli. J. Physiol. 297, 405–422. Joris, P.X., Yin, T.C., 1992. Responses to amplitude-modulated tones in the auditory nerve of the cat. J. Acoust. Soc. Am. 91, 215–232. Joutsiniemi, S.L., Hari, R., Vilkman, V., 1989. Cerebral magnetic responses to noise bursts and pauses of different durations. Audiology 28, 325–333.

216

BR A I N R ES E A RC H 1 3 8 5 ( 2 01 1 ) 2 0 6 –21 6

King, A.J., Nelken, I., 2009. Unraveling the principles of auditory cortical processing: can we learn from the visual system? Nat. Neurosci. 12, 698–701. Klier, E.M., Hess, B.J., Angelaki, D.E., 2008. Human visuospatial updating after passive translations in three-dimensional space. J. Neurophysiol. 99, 1799–1809. Koulakov, A.A., Raghavachari, S., Kepecs, A., Lisman, J.E., 2002. Model for a robust neural integrator. Nat. Neurosci. 5, 775–782. Krishna, B.S., 2002. A unified mechanism for spontaneous-rate and first-spike timing in the auditory nerve. J. Comput. Neurosci. 13, 71–91. Krishna, B.S., 2006. Comment on "Auditory-nerve first-spike latency and auditory absolute threshold: a computer model" [J. Acoust. Soc. Am. 119, 406–417 (2006)]. J. Acoust. Soc. Am. 120, 591–593. Krumbholz, K., Wiegrebe, L., 1998. Detection thresholds for brief sounds–are they a measure of auditory intensity integration? Hear. Res. 124, 155–169. Levitt, H., 1971. Transformed up-down methods in psychoacoustics. J. Acoust. Soc. Am. 49, 467–477. Liberman, M.C., 1978. Auditory-nerve response from cats raised in a low-noise chamber. J. Acoust. Soc. Am. 63, 442–455. Loewenstein, Y., Sompolinsky, H., 2003. Temporal integration by calcium dynamics in a model neuron. Nat. Neurosci. 6, 961–967. Loucks, C.A., Doty, R.L., 2004. Effects of stimulation duration on electrogustometric thresholds. Physiol. Behav. 81, 1–4. Lütkenhöner, B., 2008. Threshold and beyond: modeling the intensity dependence of auditory responses. J. Assoc. Res. Otolaryngol. 9, 102–121. Lütkenhöner, B., 2010. Baseline correction of overlapping event-related responses using a linear deconvolution technique. Neuroimage 52, 86–96. Lütkenhöner, B., Klein, J.S., 2007. Auditory evoked field at threshold. Hear. Res. 228, 188–200. Lütkenhöner, B., Seither-Preisler, A., 2008. Auditory brainstem response at the detection limit. J. Assoc. Res. Otolaryngol. 9, 521–531. Lütkenhöner, B., Krumbholz, K., Lammertmann, C., SeitherPreisler, A., Steinsträter, O., Patterson, R.D., 2003. Localization of primary auditory cortex in humans by magnetoencephalography. Neuroimage 18, 58–66. Meddis, R., 2006. Auditory-nerve first-spike latency and auditory absolute threshold: a computer model. J. Acoust. Soc. Am. 119, 406–417. Melcher, J.R., Kiang, N.Y., 1996. Generators of the brainstem auditory evoked potential in cat: III. Identified cell populations. Hear. Res. 93, 52–71. Møller, A.R., 2007. Neural generators for auditory brainstem evoked potentials. In: Burkard, R.F., Eggermont, J.J., Don, M. (Eds.), Auditory Evoked Potentials - Basic Principles and Clinical Applications. Lippincott Williams & Wilkins, Philadelphia, pp. 336–354. Nelken, I., Ulanovsky, N., Las, L., Bar Yosef, O., Anderson, M., Chechnik, G., Tishby, N., Young, E.D., 2004. Transformation of stimulus representations in the ascending auditory system. In: Pressnitzer, D., de Cheveigné, A., McAdams, S., Collet, L. (Eds.), Auditory Signal Processing: Psychophysics, Physiology and Modeling. Springer, New York, pp. 265–274. Neubauer, H., Heil, P., 2004. Towards a unifying basis of auditory thresholds: the effects of hearing loss on temporal integration reconsidered. J. Assoc. Res. Otolaryngol. 5, 436–458. Neubauer, H., Heil, P., 2008. A physiological model for the stimulus dependence of first-spike latency of auditory-nerve fibers. Brain Res. 1220, 208–223.

Nie, H., Graven-Nielsen, T., Arendt-Nielsen, L., 2009. Spatial and temporal summation of pain evoked by mechanical pressure stimulation. Eur. J. Pain 13, 592–599. Nielsen, J., Arendt-Nielsen, L., 1998. The importance of stimulus configuration for temporal summation of first and second pain to repeated heat stimuli. Eur. J. Pain 2, 329–341. Okamoto, H., Fukai, T., 2009. Recurrent network models for perfect temporal integration of fluctuating correlated inputs. PLoS Comput. Biol. 5, e1000404. Okawa, H., Sampath, A.P., 2007. Optimization of single-photon response transmission at the rod-to-rod bipolar synapse. Physiology (Bethesda) 22, 279–286. Onishi, S., Davis, H., 1968. Effects of duration and rise time of tone bursts on evoked V potentials. J. Acoust. Soc. Am. 44, 582–591. Patterson, R.D., Uppenkamp, S., Johnsrude, I.S., Griffiths, T.D., 2002. The processing of temporal pitch and melody information in auditory cortex. Neuron 36, 767–776. Pratt, H., 2007. Middle-latency responses. In: Burkard, R.F., Eggermont, J.J., Don, M. (Eds.), Auditory Evoked Potentials - Basic Principles and Clinical Applications. Lippincott Williams & Wilkins, Philadelphia, pp. 463–481. Press, W.H., Teukolsky, S.A., Vetterling, W.T., Flannery, B.P., 1992. Numerical recipes in C. Cambridge University Press, Cambridge, The art of scientific computing. Recanzone, G.H., Sutter, M.L., 2008. The biological basis of audition. Annu. Rev. Psychol. 59, 119–142. Rose, J.E., Hind, J.E., Anderson, D.J., Brugge, J.F., 1971. Some effects of stimulus intensity on response of auditory nerve fibers in the squirrel monkey. J. Neurophysiol. 34, 685–699. Rupp, A., Uppenkamp, S., Gutschalk, A., Beucker, R., Patterson, R.D., Dau, T., Scherg, M., 2002. The representation of peripheral neural activity in the middle-latency evoked field of primary auditory cortex in humans. Hear. Res. 174, 19–31. Saszik, S.M., Robson, J.G., Frishman, L.J., 2002. The scotopic threshold response of the dark-adapted electroretinogram of the mouse. J. Physiol. 543, 899–916. Saunders, R.M., 1975. The critical duration of temporal summation in the human central fovea. Vis. Res. 15, 699–703. Sharpe, L.T., Stockman, A., Fach, C.C., Markstahler, U., 1993. Temporal and spatial summation in the human rod visual system. J. Physiol. 463, 325–348. Stüttgen, M.C., Schwarz, C., 2010. Integration of vibrotactile signals for whisker-related perception in rats is governed by short time constants: comparison of neurometric and psychometric detection performance. J. Neurosci. 30, 2060–2069. Swets, J.A., Shipley, E.F., McKey, M.J., Green, D.M., 1959. Multiple Observations of Signals in Noise. J. Acoust. Soc. Am. 31, 514–521. Tougaard, J., 1996. Energy detection and temporal integration in the noctuid A1 auditory receptor. J. Comp. Physiol. A 178, 669–677. Verrillo, R.T., 1965. Temporal summation in vibrotactile sensitivity. J. Acoust. Soc. Am. 37, 843–846. Viemeister, N.F., Plack, C.J., 1993. Time analysis. In: Yost, W.A., Popper, A.N., Fay, R.R. (Eds.), Human psychophysics. Springer, New York, pp. 116–154. Viemeister, N.F., Wakefield, G.H., 1991. Temporal integration and multiple looks. J. Acoust. Soc. Am. 90, 858–865. Wildner, M., 1999. In memory of William of Occam. Lancet 354, 2172. Wise, P.M., Toczydlowski, S.E., Wysocki, C.J., 2007. Temporal integration in nasal lateralization of homologous alcohols. Toxicol. Sci. 99, 254–259. Zuidema, P., Verschuure, H., Bouman, M.A., Koenderink, J.J., 1981. Spatial and temporal summation in the human dark-adapted retina. J. Opt. Soc. Am. 71, 1472–1480. Zwislocki, J., 1960. Theory of temporal auditory summation. J. Acoust. Soc. Am. 32, 1046–1060.