Intensity and inter-stimulus-interval effects on human middle- and long-latency auditory evoked potentials in an unpredictable auditory context

Abstract

It is not known how Auditory-Evoked Responses (AERs) comprising Middle Latency Responses (MLRs) and Long Latency Responses (LLRs) are modulated by stimulus intensity and inter-stimulus interval (ISI) in an unpredictable auditory context. Further, intensity and ISI effects on MLR and LLR have never been assessed simultaneously in the same humans. To address this important question, thirty participants passively listened to a random sequence of auditory clicks of three possible intensities (65, 75, and 85 dB) at five possible ISI ranges (0.25 to 0.5 s, 0.5 to 1 s, 1 to 2 s, 2 to 4 s, 4 to 8 s) over four to seven one-h our sessions while EEG was recorded. P0, Na, Pa, Nb, and Pb MLR peaks and N1 and P2 LLR peaks were measured. MLRs P0 $(p=.005)$, Pa $(p=.021)$, and Pb $(p=<.001)$ were modulated by intensity, while only MLR Pb $(p=<.001)$ was modulated by ISI. LLR N1 and P2 were modulated by both intensity and ISI (all p values < .001). Intensity and ISI interacted at Pb, N1, and P2 (all p values < .001), with greater intensity effects at longer ISIs and greater ISI effects at louder intensities. Together, these results provide a comprehensive picture of intensity and ISI effects on AER across the entire thalamocortical auditory pathway, while controlling for stimulus predictability. Moreover, they highlight P0 as the earliest MLR response sensitive to stimulus intensity and Pb (~50 ms) as the earliest cortical response coding for ISIs above 250 ms and showing an interdependence between intensity and ISI effects.

1 |. INTRODUCTION

Human Auditory evoked responses (AERs) comprise a series of waves recorded non-invasively with Electroencephalography(EEG)orMagnetoencephalography (MEG) that reflect neural responses to auditory stimuli. Within AERs, auditory brainstem responses (ABRs) or brainstem auditory evoked responses (BAERs) arise from brainstem, collicular, and thalamic subcortical auditory processing centers and are measured within the first 10 ms post-sound (Jewett, 1970). Mid-latency responses (MLRs) include five waves (P0, Na, Pa, Nb, and Pb) that range in latency from ~12 to 70 ms after sound onset (Picton et al., 1974; Yvert et al., 2001). MLRs arise in the medial geniculate nucleus of the thalamus (MGN: Deiber et al., 1988; Kraus et al., 1982; McGee et al., 1992) and areas of the auditory cortex, including Heschl’s gyrus (HG: Godey et al., 2001), Planum Temporale (PT), and the lateral aspect of the Superior Temporal Gyrus (STG: Kileny et al., 1987; Yvert et al., 2005). Long-latency responses (LLRs) include N1(/N100) and P2(/P200) AERs, ranging from ~80 to 250 ms from sound onset. LLRs are generated in HG, PT, and lateral STG (Godey et al., 2001; Yvert et al., 2005) and likely also include contributions from auditory-related executive areas in parietal and frontal cortices (Alcaini et al., 1994; Anderer et al., 1998), as well as cingulate, insular, and other extra-temporal auditory-associated areas (Giard et al., 1994).

The amplitude of both MLRs and LLRs depends on the physical properties of the sound such as intensity, but also on contextual factors, such as the inter-stimulus interval (ISI) between sounds. While the dependence on sound intensity is mechanically inherited from the inner ear, the dependence on temporal context is not, and instead reflects properties of neural processing along the ascending auditory processing hierarchy such as the duration of contextual memory. However, despite decades of research, fundamental questions remain about the basic phenomenology of the effects of intensity and ISI on auditory evoked potentials across the ascending auditory pathway. How early in the auditory processing hierarchy, as reflected in MLR and LLR, do intensity and ISI effects arise? How do intensity and ISI interact? How do intensity and ISI affect MLRs and LLRs without stimulus predictability confounds? Answering these questions would further characterize prominent response properties of AERs while allowing for a better interpretation of their responsiveness without predictive effects potentially generated by highly repetitive, unchanging (and therefore predictable) stimulus contexts.

Although the sensitivity of LLRs to stimulus intensity is well documented (i.e., increased amplitude with greater stimulus intensity), there is less work regarding the psychophysics of MLRs and extant results are inconsistent. For example, using auditory clicks as stimuli, a MEG study reported intensity effects at the Na (Borgmann et al., 2001), while another using scalp EEG did not (Tucker et al., 2001). The opposite results were found in these two studies for the Pa wave, while both reported intensity effects at Pb. Results from the older literature converge in finding modulation by stimulus intensity in later (Pa, Nb, Pb) versus earlier (P0, Na) MLR peaks (Kupperman & Mendel, 1974; Thornton et al., 1977). It is unclear then, whether intensity information is coded from initial auditory processing in AER amplitude, and precisely how MLRs react to intensity changes. In contrast with MLRs, LLR intensity effects have been consistently shown for N1 and P2 in both EEG sensor (Harris et al., 2007; Juckel et al., 2003; Keidel & Spreng, 1965) and source-resolved activity in the auditory cortex measured with EEG (Hagenmuller et al., 2016; Schadow et al., 2007). Equivalent results in N1 and P2 were obtained with MEG (Wyss et al., 2014), and N1-P 2 amplitudes at different intensities have also been used to determine pure-tone audiometry thresholds (Lightfoot & Kennedy, 2006; Mahdavi & Peyvandi, 2007).

Regarding ISI effects in AERs (i.e. larger responses to sounds preceded by longer periods of silence), in the MLR, consistent evidence still is scarce, but modulations of amplitude by ISI have been reported for Pa (Tucker et al., 2002) and Pb (Erwin & Buchwald, 1986; Tucker et al., 2002), as well as in their magnetic counterparts (Onitsuka et al., 2003). ISI effects on LLR amplitudes were first reported with EEG by Davis et al. (1966), and have since been consistently found in both N1 and P2 (Herrmann et al., 2014, 2016; Pereira et al., 2014; Sussman et al., 2008), in the trough-to-peak N1-P2 complex (Muller-Gass et al., 2008; Shucard & Callaway, 1974), as well as in MEG N1m (Okamoto & Kakigi, 2014; Sams et al., 1993; Zygierewicz et al., 2012) and P2m (Rosburg et al., 2010). Evidence points to maximal N1 and P2 amplitudes with ISIs at ~10 s (Nelson & Lassman, 1968; Ritter et al., 1968).

Despite the extensive literature, it remains unclear which evoked potential is the first to be reliably modulated by both intensity and ISI, with inconsistent and sometimes contradictory results for the MLRs. How these two factors interact in their modulation of AERs is also uncertain, but intensity and ISI effects were rarely studied together in the same subjects (with a few exceptions, Roth et al., 1976; Zhang et al., 2009), thus leaving them non-comparable. Moreover, regularity of the auditory stimulation sequence is a confounding element in the aforementioned studies of intensity and ISI. When auditory stimuli are predictable, AERs amplitudes are smaller (Coffman et al., 2017; Kaufmann et al., 1982; Lange, 2009; Schwartze et al., 2013). In human event-related designs, it is common to use predictable repetitive stimulation paradigms (i.e., one intensity or one ISI in a block) due to the large number of trials needed to obtain MLR or LLRs.

In the current study, we examined intensity and ISI effects in the entire middle-and long-latency hierarchy of the sensory auditory information processing stream by measuring these phenomena simultaneously in the same participants and within the same trials, which has not been previously done. Furthermore, the stimulation paradigm employed here used entirely unpredictable intensities and ISIs, resulting in 15 possible combinations for each stimulus (3 intensities by 5 exponential ISI groupings with ISIs sampled randomly within each range). This allowed us to quantify intensity and ISI effects and their interaction as independent effects, and prevent any potential confound of stimulus regularity.

2 |. METHOD

2.1 |. Participants

Thirty healthy participants (19 males) were included in the study, ranging in age from 19 to 38 years (mean 24,03; SD 5.27). The target sample size was determined based on previous authors work with MLR amplitude modulations (Lopez-Caballero et al., 2016) and on studies with similar experimental designs (Herrmann et al., 2016; Pereira et al., 2014). Exclusion criteria were history of concussion or head injury with sequelae, psychiatric or neurological disorder, the history of alcohol or drug addiction in the last 5 years, or the use of medication affecting brain function or structure. Participants completed pure-tone audiometry (1000–4000 Hz) before the experiment started, ensuring mean hearing thresholds below 30 dB nHL at each ear and less than 15 dB threshold differences between ears. Mean hearing thresholds averaged across frequencies were of 7.8 and 7.7 dB on the left and right ear, respectively (see Figure S1 for extended hearing thresholds results). All procedures were approved by the University of Pittsburgh IRB and in accordance with the WMA Declaration of Helsinki Ethical Principles for Medical Research Involving Human Subjects. All participants provided written informed consent and were paid for participation.

2.2 |. Stimuli and procedure

The stimulus was a 1 ms biphasic click, created using MATLAB (The Mathworks, Inc., Natick, MA) delivered through ER-3A insert earphones (Etymotic Research, Inc., Elk Grove Village, IL, USA) using the Presentation^® software (Version 9.90).

The experiment was divided into four to seven identical one-hour sessions. Each session consisted of three blocks of 20 min each, with short (~2 min) breaks between blocks. During each session, participants sat in the MEG/EEG scanner in a magnetically shielded room while passively listening to a series of auditory clicks. They were instructed to watch a silent movie of their choice and ignore the sounds. Clicks were one of three intensities (65, 75, or 85 dB SPL). ISI on a given trial was randomly selected from a uniform distribution within one of five equally probable time ranges (0.25 to 0.5 s, 0.5 to 1 s, 1 to 2 s, 2 to 4 s or 4 to 8 s). Intensity and ISI were manipulated in a fully crossed 3-by-5 design, resulting in 15 intensity-ISI combinations (Figure 1a). The intended length of the study was of six sessions to retrieve, across all sessions, 600 trials of each intensity-ISI combination, ensuring enough remaining sweeps for visualization of MLRs after EEG cleaning procedures (which we set as 400 sweeps). However, eight participants completed an additional seventh session due to technical problems with the MEG acquisition system shortening previous sessions. Two other participants could only complete four and five sessions, respectively, due to MEG technical problems and the impossibility to record further sessions with them. Among them, only the one with five sessions was used for MLR analyses (sufficient surviving sweeps after EEG cleaning procedures), while both were used for LLR. The remaining twenty participants completed six sessions successfully, despite for three of them not enough sweeps survived to include them in MLR analyses, and no further sessions could be recorded. On average, 31.56 days $(SD=49.24)$ passed between sessions and 156.96 days $(SD=126.93)$ passed between the first and last session.

FIGURE 1.

(a) Example of the auditory sequence. Auditory clicks could be of three possible intensities (65, 75 and 85 dB), and the ISIs between them could fall within five possible time ranges (0.25 to 0.5 s, 0.5 to 1 s, 1 to 2 s, 2 to 4 s or 4 to 8 s). The intensity and ISI on each trial was random. (b) Example illustration of the subtraction procedure. Elongated averages (templates, green to red) for each click category (intensity and ISI) were computed to subtract from individual epochs (blue) in their overlapping time-window (red). In the example, the second click comes after 0.3 s from the first one.

2.3 |. EEG acquisition

Combined EEG and MEG acquisition was performed in a magnetically shielded room (Imedco AG, Hägendorf, Switzerland) using a low-impedance EEG cap (BrainCap MEG, Brain Vision, LLC, Morrisville, NC, USA) with 60 scalp electrode locations based on the 10–10 system and a whole-head MEG system (Elekta Neuromag, Helsinki, Finland). The acquisition system (Elekta Neuromag) was the same for EEG and MEG. Only scalp EEG activity is reported here. MEG results will be discussed in a separate report. Data were recorded with a sampling rate of 3000 Hz and an online bandpass filter of 0.1 to 1000 Hz. Additional EEG electrodes were placed in the left mastoid (online EEG reference), right mastoid, back of the neck (ground), below the left clavicle (ECG), and below the left eye (VEOG). Two bipolar electrodes were also placed on the outer canthi of both eyes (HEOG).

2.4 |. EEG analysis

EEG data were first down-sampled to 1500 Hz digitization rate and preprocessed using the EEGLAB toolbox (Delorme & Makeig, 2004) in MATLAB (The Mathworks, Inc., Natick, MA). The data were processed differently for MLRs and LLRs.

2.4.1 |. MLRs

For MLRs, a high-pass filter (10 Hz; 12 dB/oct) was applied (Kavanagh & Domico, 1987), and channels and data segments with excessive noise were removed by visual inspection. Next, Adaptive Mixture Independent Component Analysis (AMICA) was performed to remove artifacts associated with eye blinks/movements and heartbeat. Channels removed prior to AMICA were interpolated afterwards. After preprocessing, EEG data were analyzed using the Brainstorm software (Tadel et al., 2011). First, data were re-referenced to averaged mastoids. A low-pass filter (100 Hz; 24 dB/oct, see Figure S2a,b for comparison of 200 Hz and 100 Hz low-pass filtered data) was then applied, and epochs of −50 to 100 ms (baseline corrected) were extracted around click stimuli. Next, epochs with amplitudes larger than ±50 μV at any EEG channel were discarded. Finally, epochs from all sessions were averaged within each participant, separately for each of the 15 intensity-ISI conditions. Four participants with less than 400 surviving sweeps on average in these conditions were removed from further MLR analyses. Total number of sweeps per condition and participant, as well as signal-to-noise ratio values, can be found in Figures S3 and S4.

2.4.2 |. LLR

For LLR, a high-pass filter (0.5 Hz; 12 dB/oct) was applied (Rentzsch et al., 2008), and channels and data segments with excessive noise were removed by visual inspection. AMICA was then performed to remove eye movements and heartbeats from the signal, and channels removed prior to AMICA were interpolated afterwards. After preprocessing, AMICA-corrected data were analyzed using the Brainstorm software (Tadel et al., 2011). First, data were re-referenced to averaged mastoids and down-sampled to 500 Hz. Next, a low-pass filter (20 Hz; 24 dB/oct) was applied (Whitham et al., 2007), and epochs of −150 to 400 ms (baseline corrected) were extracted around click stimuli. Epochs with amplitudes larger than ±50 μV at any EEG channel were discarded. Finally, epochs from all sessions were averaged within each participant, separately for each of the 15 intensity-ISI conditions.

2.4.3 |. Subtraction of prior AERs in shortest ISI condition

Epochs preceded by the shortest ISI (0.25 to 0.5 s) for LLR analyses were distorted due to the continued response to the preceding click. This response overlap was corrected using a similar procedure as employed by Budd and Michie (1994) and Briley and Krumbholz (2013) (Figure 1b). Briefly, elongated epochs of 0.6 s post-stimulus were extracted and averaged for each of the 15 intensity-ISI conditions, following the same procedures as with original epochs and separately for each experimental session. We constructed a “template” of the response for each trial type, which was then subtracted from individual epochs based on the overlapping time-window and the preceding trial type. To ensure a smooth transition of the response to baseline in the template averages, we applied a half Hanning window in the last 0.2 s and assumed all activity past 0.6 s after click stimulus was equal to zero. For consistency, we performed this same procedure in epochs from all conditions, and in both MLR and LLR, despite only epochs from trials preceded by the shortest ISI in LLR appearing contaminated by the previous response.

2.5 |. Statistical analysis

Two-way repeated measures ANOVAs were performed separately for each MLR and LLR peak, measured at the FCz electrode (see EEG scalp topographies in Figure S5). The two factors were intensity (3 levels) and ISI (five levels), and the time windows were defined based on the grand-average waveforms (P0: 5 to 11 ms; Na: 15 to 21 ms; Pa: 27 to 33 ms; Nb: 37 to 43 ms; Pb: 51 to 61 ms; N1: 78 to 106 ms; P2: 150 to 182 ms). For each of the comparisons, degrees of freedom were corrected using the Huynh-Feldt epsilon. All analyses were carried out using IBM SPSS Statistics (Version 26).

3 |. RESULTS

3.1 |. Middle latency response

P0 (8ms), Na (18 ms), Pa (30 ms), Nb (40 ms) and Pb (56 ms) peaks were identified at typical MLR latencies (Borgmann et al., 2001; Picton et al., 1974; Yvert et al., 2001). Figure 2a displays MLR waveforms averaged across ISI (left) and click intensities (right), while Figure 2b shows MLRs for each intensity-ISI combination. Results of the two-way rmANOVA are summarized in Table 1.

FIGURE 2.

Grand-average MLR waveforms $(n=26)$ at FCz, with the standard error of the mean. Gray areas show the time windows of peaks where conditions significantly differed in rmANOVA ($p<.05$). (a) MLRs averaged across ISI (left), or across intensities (right). (b) MLRs of each intensity-ISI combination.

TABLE 1.

Results from two-way repeated measures ANOVA on each MLR and LLR peak

	dB				ISI				dB × ISI
	F (2,50/58)	p	ε	${\eta }_p^2$	F (4,100/116)	p	ε	${\eta }_p^2$	F (8,200/232)	p	ε	${\eta }_p^2$
P0	5.78	.005	1	.188	.40	.783	.89	.016	.67	.714	.82	.026
Na	.42	.651	.98	.017	.69	.596	.65	.027	.74	.655	.88	.029
Pa	4.64	.021	.81	.157	1.60	.184	.93	.600	.77	.608	.86	.030
Nb	.37	.683	.96	.015	4.19	.006	.84	.144	1.26	.267	.88	.048
Pb	41.45	<.001	.74	.624	33.64	<.001	.71	.574	4.05	<.001	.84	.139
N1	85.02	<.001	.62	.746	32.03	<.001	.85	.525	6.39	.001	.95	.181
P2	111.74	<.001	.67	.794	125.47	<.001	.66	.812	7.85	<.001	.95	.213

Note: Intensity (3 levels: 65, 75 and 85 dB) and ISI (five levels: 0.25 to 0.5 s, 0.5 to 1 s, 1 to 2 s, 2 to 4 s or 4 to 8 s) factors were used. Dependent variable was mean response amplitude over a time window around each peak. For each factor and their interaction, F and p values are presented along with epsilon sphericity index and effect sizes (${\eta }_p^2$). Degrees of freedom are displayed for MLR and LLR, respectively. p values below .05 are highlighted.

At P0 (8ms), amplitudes increased as click intensities became louder $(p=.005)$, but not as ISI became longer $(p=.783)$. At Na (18 ms), amplitudes were not affected by either click intensity $(p=.651)$ or ISI $(p=.596)$. At Pa (30 ms), amplitudes increased with louder clicks $(p=.021)$ but not with longer ISI $(p=.184)$. At Nb (40 ms), amplitudes were not modulated by click intensity $(p=.683)$, but they were by ISI $(p=.006)$. However, Nb amplitudes did not increase with longer ISIs. Instead, pairwise comparisons based on estimated marginal means revealed a significant Nb amplitude increase (more negative) in the 0.25–0.5 s ISI versus the 1–2 s ISI condition.

At Pb (56 ms), amplitudes progressively increased as clicks were louder $(p<.001)$ and ISIs were longer $(p<.001)$. The effects of intensity and ISI interacted $(p<.001)$; the effect of intensity was enhanced for longer ISIs (Figure 4e). To parse the effects of the interaction, rmANOVAs were conducted to assess modulations by intensity within each ISI level, revealing louder clicks produced significantly larger Pb responses regardless of which ISI preceded the stimulus: 0.25 to 0.5 s $\left(F_{\left(\mathrm{2,50}\right)}=14.16;p<.001;\epsilon =.92;{\eta }_p^2=.362\right)$, 0.5 to 1 s $\left(F_{\left(\mathrm{2,50}\right)}=5.35;p=.013;\epsilon =.81;{\eta }_p^2=.176\right)$, 1 to 2 s $\left(F_{\left(\mathrm{2,50}\right)}=23.16;p<0.001;\epsilon =.99;{\eta }_p^2=.481\right)$, 2 to 4 s $\left(F_{\left(\mathrm{2,50}\right)}=29.66;p<0.001;\epsilon =.91;{\eta }_p^2=.543\right)$ or 4 to 8 s $\left(F_{\left(\mathrm{2,50}\right)}=26.65;p<.001;\epsilon =.82;{\eta }_p^2=.516\right)$. Similarly, rmANOVAs were conducted to assess modulations by ISI within each intensity level, which revealed that shorter ISIs resulted in smaller Pb responses for all click intensities: 65 dB $\left(F_{\left(\mathrm{4,100}\right)}=6.78;p<0.001;\epsilon =.85;{\eta }_p^2=.213\right)$, 75 dB $\left(F_{\left(\mathrm{4,100}\right)}=10.33;p<0.001;\epsilon =1;{\eta }_p^2=.292\right)$ or 85 dB $\left(F_{\left(\mathrm{4,100}\right)}=26.76;p<0.001;\epsilon =.90;{\eta }_p^2=.517\right)$. Pairwise comparisons (Bonferroni corrected) between each individual intensity and ISI level were then performed. Significant differences are highlighted in Figure 4e. Overall, the intensity of the click modulated Pb to a greater extent than ISI and the most prominent shift in response amplitude occurred when changing from the 0.5–1 s to the 1–2 s ISI.

FIGURE 4.

Peak amplitudes as a function of ISI, separated by click intensities, for MLR P0, Na, Pa, Nb and Pb (a–e) and LLR N1 and P2 (f and g). Significant pair-wise comparisons performed at Pb, N1 and P2 ($p<.05$, Bonferroni corrected) are highlighted with (*) when between intensities and with (▲) when between ISIs.

3.2 |. Long latency response

All click intensity and ISI combinations elicited measurable LLRs, with identifiable N1 (92ms) and P2 (166ms) peaks (Figure 3) at typical latencies (Crowley & Colrain, 2004; Näätänen & Picton, 1987). LLR waveforms averaged across ISIs (left) and click intensities (right) are shown in Figure 3a. LLRs for each intensity-ISI combination separately are displayed in Figure 3b. LLR results from the two-way rmANOVA are summarized in Table 1.

FIGURE 3.

Grand-average LLR waveforms ($n=30$) at FCz, with the standard error of the mean. Gray areas show the time windows of peaks where conditions significantly differed in rmANOVA ($p<.05$). (a) LLRs averaged across ISIs (left), or across intensities (right). (b) LLRs of each intensity-ISI combination.

At N1 (92ms), more intense clicks $(p<.001)$ and longer ISIs $(p<.001)$ elicited larger responses. The enhancement of N1 amplitude with more intense stimuli was more pronounced for long versus short ISI, and response reduction with shorter ISI was also more pronounced for stronger versus weaker stimuli (intensity × ISI interaction: $p<.001$; Figure 4f). To parse the effects of the interaction, rmANOVAs were conducted to assess modulations by intensity within each ISI level, revealing louder clicks increased N1 amplitude for all ISIs: 0.25 to 0.5 s $\left(F_{\left(\mathrm{2,58}\right)}=36.28;p<.001\right.;\left.\epsilon =.95;{\eta }_p^2=.556\right)$, 0.5 to 1 s $\left(F_{\left(\mathrm{2,58}\right)}=46.05;p=<.001\right.;\left.\epsilon =.72;{\eta }_p^2=.614\right)$, 1 to 2 s $\left(F_{\left(\mathrm{2,58}\right)}=68.27;p<0.001;\right.\left.\epsilon =.98;{\eta }_p^2=.702\right)$, 2 to 4 s $\left(F_{\left(\mathrm{2,58}\right)}=40.97;p<0.001;\right.\left.\epsilon =.77;{\eta }_p^2=.586\right)$ or 4 to 8 s $\left(F_{\left(\mathrm{2,58}\right)}=80.89;p<.001;\right.\left.\epsilon =.78;{\eta }_p^2=.736\right)$. Similarly, rmANOVAs were conducted to assess modulations by ISI within each intensity level, revealing shorter ISIs reduced N1 amplitude for each of the stimulus intensities: 65 dB $\left(F_{\left(\mathrm{4,116}\right)}=6.104;p=<0.001\right.;\left.\epsilon =1;{\eta }_p^2=.174\right)$, 75 dB $\left(F_{\left(\mathrm{4,116}\right)}=17.96;p<0.001;\epsilon =.83\right.;\left.{\eta }_p^2=.383\right)$ or 85 dB $\left(F_{\left(\mathrm{4,116}\right)}=26.91;p<0.001;\epsilon =.87\right.;{\eta }_p^2=.481)$. Pairwise comparisons (Bonferroni corrected) between each individual intensity and ISI level were then performed. Significant differences are highlighted in Figure 4f. N1 responses were significantly larger at 0.25–5 s ISI compared to 0.5–1 s ISI, in both 65 and 75 dB click intensity conditions, contrary to the general reduction of response amplitudes with shorter ISIs.

At P2 (166ms), louder clicks $(p<.001)$ and longer ISIs $(p<.001)$ elicited larger responses. Intensity and ISI factors interacted $(p<.001)$, such that the enhancement of P2 with louder clicks was more pronounced for long versus short ISI, and P2 response reduction with shorter ISI was more pronounced for louder versus quieter stimuli (Figure 4g). To parse the effects of the interaction, rmANOVAs were conducted to assess modulations by intensity within each ISI level, revealing louder clicks increased P2 amplitude for all ISIs: 0.25 to 0.5 s $\left(F_{\left(\mathrm{2,58}\right)}=54.61;p<.001;\left.\epsilon =1;{\eta }_p^2=.653\right)\right.$, 0.5 to 1 s $\left(F_{\left(\mathrm{2,58}\right)}=38.89;p<.001;\epsilon =.79;\left.{\eta }_p^2=.573\right)\right.$, 1 to 2 s $\left(F_{\left(\mathrm{2,58}\right)}=69.26;p<0.001;\epsilon =.72;\left.{\eta }_p^2=.705\right)\right.$, 2 to 4 s $\left(F_{\left(\mathrm{2,58}\right)}=133.98;p<0.001;\epsilon =.83\right.;\left.{\eta }_p^2=.822\right)$ or 4 to 8 s $\left(F_{\left(\mathrm{2,58}\right)}=72.71;p<.001;\epsilon =.73;\right.\left.{\eta }_p^2=.715\right)$. Similarly, rmANOVAs were conducted to assess modulations by ISI within each intensity level, which revealed shorter ISIs reduced P2 amplitude for each stimulus intensity: 65 dB $\left(F_{\left(\mathrm{4,116}\right)}=39.29;p<0.001;\left.\epsilon =.94;{\eta }_p^2=.575\right)\right.$, 75 dB $\left(F_{\left(\mathrm{4,116}\right)}=95.61;p<0.001;\epsilon =.84\right.;\left.{\eta }_p^2=.767\right)$ or 85 dB $\left(F_{\left(\mathrm{4,116}\right)}=102.66;p<0.001;\epsilon =.71;\right.{\eta }_p^2=.780)$. Pairwise comparisons (Bonferroni corrected) between each individual intensity and ISI level were then performed. Significant differences are highlighted in Figure 4g. P2 amplitudes increased between all ISI levels at all click intensities, except between 2–4 s and 4–8 s, suggesting they stabilized at 2–4 s ISI (Figure 4g).

4 |. DISCUSSION

4.1 |. Summary of results

The present study examined the behavior of auditory middle- and long-latency evoked responses to changes in stimulus intensity and ISI while using an auditory stimulation sequence that did not contain any temporal or stimulus intensity predictability. Increased evoked response amplitudes with more intense stimuli were observed as early as the first MLR peak recorded (P0, 8 ms). Pb (~50 ms) was the earliest cortical response coding for ISIs above 250 ms and showing an interdependence between intensity and ISI effects. Greater amplitude enhancement with more intense clicks and longer ISIs, as well as an interaction between intensity and ISI factors, were observed at both N1 (92ms) and P2 (166ms) peaks. N1 increases in amplitude at the shortest ISI (0.25 to 0.5 s) and asymptotic P2 amplitudes after 2–4 s ISI were also observed.

4.2 |. Middle-latency response

Regarding effects of stimulus intensity at P0, our findings contrast with the general lack of findings for this EP in the literature. Most studies instead find intensity effects in MLR beginning at Na (Borgmann et al., 2001; Kupperman & Mendel, 1974; Thornton et al., 1977; Tucker et al., 2001). Only Madell and Goldstein (1972) reported a peak-to-peak amplitude increase between P0 and Na with more intense sounds. This could be due to earlier peaks being typically smaller, and thus varying over a more limited range of amplitudes (Zerlin & Naunton, 1974), as we found effect sizes of stimulus intensity more than three times larger at Pb than at Pa or P0. It could also be due to its subcortical origin (Yoshiura et al., 1996), requiring a larger amount of stimulus repetitions to reach similar signal-to-noise ratios than later MLR peaks. To that regard, evoked responses at earlier latencies than those of MLR, at the ABR, are shown to be sensitive to stimulus intensity (Dzulkarnain et al., 2020; Stapells & Oates, 1997; van den Honert & Stypulkowski, 1986). At Pa, we replicated the intensity effects observed with similar click intensities (Tucker et al., 2001) and other types of stimuli (Kupperman & Mendel, 1974). However, our results for Pa contrast with reports suggesting saturation of amplitude effects at 60 dB (Borgmann et al., 2001; Madell & Goldstein, 1972; Özdamar & Kraus, 1983). At Pb, we replicated findings of increased amplitudes with louder sounds from previous studies using clicks (Borgmann et al., 2001; Tucker et al., 2001) and other stimuli (Thornton et al., 1977). Despite stimulus intensity modulating the positive MLR peaks at P0, Pa and Pb, the peak negative amplitudes at Na and Nb were apparently unaffected by this stimulus feature. We speculate these results may be due to an extended underlying processing of stimulus intensity at P0 and Pa peaks, of positive polarity, overlapping, respectively, Na and Nb negative peaks.

Regarding effects of ISI at MLR, our results are consistent with previous literature showing ISI affecting Pb but not earlier peaks (Erwin & Buchwald, 1986; Goldstein et al., 1972; Mcfarland et al., 1975). Two other studies (Onitsuka et al., 2003; Tucker et al., 2002) found ISI modulations at Pa. In Onitsuka et al. (2003), ISIs used were between 0.4 s to 2 s, which are included among the ranges of ISIs investigated here. However, the different stimuli used (tone burst vs clicks), the signal recorded (MEG vs EEG), and the nature of the sound sequence (repetitive vs random) may account for the discrepancies. In Tucker et al. (2002), using clicks and measuring EEG, the differences found were between ~0.1 s and ~ 1 s ISI at 85 dB. Since the authors used ISIs below our shortest (0.25 s), our lack of ISI effects at Pa may indicate neural recovery times below 0.25 s at Pa′s neural generators (discussed in detail below).

4.3 |. Long-latency response

Our N1 and P2 amplitudes showed a clear dependence on stimulus intensity, with larger amplitudes with more intense stimuli. This is consistent with the previous EEG literature observing this phenomenon when using auditory clicks (Keidel & Spreng, 1965), as we did, but also with studies using pure tones (Harris et al., 2007).

Regarding ISI effects, both N1 and P2 amplitudes measured here increased with longer ISIs, as in previous studies (Davis et al., 1966; Herrmann et al., 2016). Previous literature reports maximal N1 and P2 amplitudes at ~6–10 s ISI (Nelson & Lassman, 1968; Ritter et al., 1968). Such results are in line with our findings at N1, with maximal amplitudes at the longest ISI condition (4–8 s). However, our P2 amplitudes did not grow past the 2–4 ISI. More recent studies manipulating ISI and measuring the P2 have not used ISIs beyond 6 s (e.g. Pereira et al., 2014; Rosburg et al., 2010; Sussman et al., 2008), which leaves the question open as to whether the ~6–10 s limit was replicated. Pereira et al. (2014) found different amplitudes at P2 between 3 and 6 s ISI, but Herrmann et al. (2016) found a similar stabilization of P2 after 2 s as the one we observed. Future studies on P2 response characteristics would benefit from revisiting this topic. N1 and P2 displaying different patterns of increase with longer ISI may reflect local circuit organizational differences between their different neural generators (Crowley & Colrain, 2004; Herrmann et al., 2016; Verkindt et al., 1994), but this would need further research to determine.

The N1 amplitude increase we observed at our shortest ISI (0.25–0.5 s) replicated those of previous studies using ISIs below 0.3 s (Budd & Michie, 1994; Heinemann et al., 2011; Loveless et al., 1989; McEvoy et al., 1997; Teder et al., 1993; Todd et al., 2000), even for other sensory modalities (Wang et al., 2008). This phenomenon, referred as N1 facilitation, has been hypothesized as caused by several factors: (1) an accelerated decay of inhibitory over excitatory postsynaptic potentials at high (<200 ms) stimulation rates (Loveless et al., 1989); (2) avoiding a latent inhibitory feedback from N1 generators if stimulus is presented before inhibition takes place (Sable et al., 2004); (3) a separate neural component with a time course overlapping N1 and P2 that is modulated exclusively by very short ISI (Wang et al., 2008). Some evidence supports this overlapping component may be Mismatch Negativity (Volosin et al., 2017, 2021). Our results best fit the third hypothesis, as P2 in addition to N1 followed a different pattern of results under our shortest ISI, reduced more prominently than for other ISIs.

4.4 |. Intensity effects precede ISI effects

In our study, the effects of Intensity emerged earlier than the effects of ISI. Specifically, while P0 (8ms) already showed increased peak amplitudes as a function of stimulus intensity, not until Pb (56 ms) did peak amplitudes increase as a function of ISI. Stimulus intensity is represented in the ear drum and affects the earliest neural responses in the auditory hierarchy. Studies with single unit recordings from cats (Rhode & Smith, 1986) and birds (MacLeod & Carr, 2007) point towards the cochlear nucleus as the first neural relay to process stimulus intensity. Thus it is reasonable to expect AERs of early latencies to be modulated by this feature, as seen at P0 in our study and in human ABRs (Dzulkarnain et al., 2020; Stapells & Oates, 1997; van den Honert & Stypulkowski, 1986). It is less clear how and at which stage within the auditory hierarchy ISI is encoded. ABR studies in humans demonstrate ISI modulates AER as early as 5 ms after sound onset (Burkard & Sims, 2001; Dzulkarnain et al., 2013, 2020; Scott & Harkins, 1978), which contrasts with our lack of effects until ~50 ms. However, such ABR studies used ISIs between 10 and 100 ms, far below our minimum ISI of 250 ms. A possible interpretation for our lack of ISI effects in the early (non-cortically generated) MLR waves could be neural adaptation. It is well documented that neurons from subcortical structures recover faster from adaptation than neurons from the auditory cortex (Fitzpatrick et al., 1999; Joris et al., 2004). For instance, single unit recordings from animal studies report cochlear nucleus neurons recovery times between 10 and 200 ms (Boettcher et al., 1990), while auditory cortical neurons have a much wider range of recovery times spanning from milliseconds to tens and possibly hundreds of seconds (Tsodyks & Markram, 1997; Ulanovsky et al., 2004). Since our shortest ISI was of 250 ms, it could be that neural generators of early MLR peaks had already recovered and thus did not show sensitivity to our ISI manipulations. In other words, this would suggest that neural generators of Pb are the first stage of the auditory hierarchy that has not yet recovered for ISIs above 250 ms and are thus capable of coding for longer ISIs.

Alternatively, our lack of ISI effects in the early MLR waves could be attributed to circuit complexity rather than single neuron response properties. As the cortex represents more complicated circuity responding to each sound, it is likely that the increased complexity of the circuit allows for prolonged effects (sensory memory) from prior stimuli. That sensory memory would thus be shorter at subcortical structures lacking such complexity, only allowing for the codification of very short ISIs (below 250 ms).

4.5 |. Interactions between effects of stimulus intensity and ISI

We observed a positive interaction between intensity and ISI for MLR Pb and LLRs N1 and P2, which was reported in a few previous studies (Roth et al., 1976; Zhang et al., 2009). Such positive interaction in a context free of confounds from stimulus predictability can be explained by a feedforward model of short-term presynaptic depression (see Teichert et al., 2016). In the model, a sound of a particular intensity will release a certain fraction of the readily releasable pool of vesicles, with more intense sounds releasing a larger fraction. In the model, the predicted AEP amplitude is proportional to the total number of vesicles released. For short ISIs, the readily releasable pool of vesicles is smaller, thus leading to a smaller overall effect size of intensity for short than for long ISIs.

4.6 |. Modulation by ISI: Habituation versus adaptation views

AER modulation by ISI can be interpreted with either the habituation or the neural adaptation hypotheses, which provide alternative but potentially complementary explanations (for a review, see Rosburg & Mager, 2020; Ruusuvirta, 2020). Habituation could be described as an active psychologically relevant process in living organisms that tends to reduce responses to innocuous predictable stimuli. Under this view, the reduced amplitude of evoked responses with shorter ISIs would be the result of a novelty-oriented loss (Barry et al., 1992; Budd et al., 1998; Thompson & Spencer, 1966). The habituation hypothesis would associate such response decrement with the building of a neuronal model or template of the stimulus due to its repeated exposure (Sokolov, 1963), an interpretation falling within the predictive coding framework (see Huang & Rao, 2011). According to the neural adaptation hypothesis (O’Shea, 2015), the phenomenon of AERs attenuation by ISI shortening would rather emerge from a passive and more basic neurophysiological process involving the inability of the neural population to respond repeatedly to the incoming stimuli (Ritter et al., 1968). Whether this inability is entirely explained by neuronal ‘fatigue’ or refractoriness at the single neuron level or mediated by inhibitory feedback of interneurons at the circuit level is of debate, we will refer to this concept as neural ‘adaptation’ rather than ‘refractoriness’.

A crucial feature of our stimulation paradigm was that it ensured no prediction of the incoming stimuli was possible and it is therefore free of confounds from a constant (utterly predictable) stimulus presentation rate. Pb, N1, and P2 suppression with shorter ISI can thus be explained here with adaptation alone (Rosburg & Mager, 2020) instead of top-down predictive processes. We also manipulated the intensity of the stimuli and found stronger modulations by ISI with louder stimuli, in contrast with the habituation hypothesis prediction that response reduction with repetitive stimuli is weaker with louder sounds (Rankin et al., 2009; Thompson & Spencer, 1966).

Previous research manipulating the predictability of the sound sequence addressed habituation versus adaptation hypotheses. Pereira et al. (2014) compared N1 and P2 attenuation with different ISIs when presenting stimuli repeatedly with and without occasional pitch deviant stimuli breaking the regularity. While habituation was theoretically stronger in the condition without deviant stimuli, ISI effects were identical in the two conditions, thus suggesting that habituation was not involved. Herrmann et al. (2016) found different patterns of attenuation of N1 and P2 between regular and irregular non-predictable sound sequences, both affected by the long temporal stimulation history. The authors fitted the data to a single-neuron computational model and concluded adaptation alone could explain their observed effects. More research replicating these findings and isolating pure adaptation from habituation is necessary to help disentangle which neural mechanisms are behind AER modulations by ISI, thus improving our interpretations on the significance of evoked responses.

5 |. CONCLUSIONS

This is the first study simultaneously measuring middle-and long-latency auditory evoked responses while manipulating stimulus intensity and ISI in an unpredictable auditory context. We proved both factors modulated AERs without stimulus predictability confounds, with stimulus intensity affecting evoked responses as early as 8 ms from stimulus onset. Moreover, we provided evidence of Pb (~50 ms) being the first stage of the auditory processing hierarchy that has not yet recovered for ISIs above 250 ms and that shows a positive interdependence between intensity and ISI modulations. Our findings also add into the discussion of the neurophysiological origin of AERs reduction with short ISIs, supporting the neural adaptation hypothesis.

Supplementary Material

sup fig 5

Figure S5 EEG scalp topographies for every MLR and LLR peak. The white dot represents the position of FCz electrode, where statistical analyses were conducted.

sup fig 4

Figure S4 (Top) LLR number of sweeps after amplitude threshold rejection for every experimental condition (numbers 11 to 93). First digit of the condition number indicates the ISI (1 = 0.25–0.5 s; 9 = 4–8 s) and the second digit indicates the intensity level (1 = 65 dB; 3 = 85 dB). (Bottom) LLR Signal-to-Noise Ratio calculated as the Log Root Mean Square (RMS) of the signal after the stimulus divided by the signal at the baseline. Values are provided for every experimental condition and the average.

sup fig 3

Figure S3 (Top) MLR number of sweeps after amplitude threshold rejection for every experimental condition (numbers 11 to 93). First digit of the condition number indicates the ISI (1 = 0.25–0.5 s; 9 = 4–8 s) and the second digit indicates the intensity level (1 = 65 dB; 3 = 85 dB). Participant’s under a minimum of 400 surviving (dashed line) sweeps on average across conditions were discarded from further analyses. (Bottom) MLR Signal-to-Noise Ratio calculated as the Log Root Mean Square (RMS) of the signal after the stimulus divided by the signal at the baseline. Values are provided for every experimental condition and the average.

sup fig 2

Figure S2 MLR waveforms with 200 Hz (left) or 100 Hz (right) low-pass filter for comparison. N reflects the number of subjects surviving amplitude threshold correction with more than 400 sweeps.

sup fig 1

Figure S1 Mean and standard deviation of hearing thresholds (dB) averaged across participants. Thresholds were measured with a pure-tone audiometry (1000–4000 Hz).

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.