Abstract
Objectives
Several studies have investigated the feasibility of using electrophysiology as an objective tool to efficiently map cochlear implants (CIs). A pervasive problem when measuring event-related potentials (ERPs) is the need to remove the direct-current (DC) artifact produced by the CI. Here we describe how DC artifact removal can corrupt the response waveform and how the appropriate choice of stimulus duration may minimize this corruption.
Design
ERPs were recorded to a synthesized vowel /a/ with a 170- or 400-ms duration.
Results
The P2 response, which occurs between 150 and 250 ms, was corrupted by the DC artifact removal algorithm for a 170-ms stimulus duration but was relatively uncorrupted for a 400-ms stimulus duration.
Conclusions
To avoid response waveform corruption from DC artifact removal, one should choose a stimulus duration such that the offset of the stimulus does not temporally coincide with the specific peak of interest. While our data have been analyzed with only one specific algorithm, we argue that the length of the stimulus may be a critical factor for any DC artifact removal algorithm.
INTRODUCTION
Cochlear implants (CIs) can be highly effective at restoring speech understanding to individuals with severe-to-profound hearing loss. Some individuals can score above 80% in speech perception tasks administered in quiet (Gifford et al. 2008), but such positive results are not realized in every CI user. This variability in performance is a result of multiple factors, including those that are biological, surgical, or device-related in nature. Device-related factors include the proper programming of the device, which needs to be tailored to each individual user. Proper programming that maximizes speech understanding is performed by an audiologist; this includes determining the frequency-to-electrode allocations, stimulation rates, and stimulation levels. Proper CI programming often involves behavioral responses, such as reporting comfortable stimulation levels. However, not everyone can participate in behavioral testing (e.g., infants), and behavioral measures may fail to yield the parameters that would result in maximum speech understanding.
As an alternative to behavioral measures, CI program parameters may be optimized with electrophysiological (EEG) objective measurements (Friesen et al. 2010; Gilley et al. 2006; Gordon et al. 2016; Lopez Valdes et al. 2014; Scheperle et al. 2015; Viola et al. 2012). Another advantage of using an objective measure like EEG is that it has the potential for increasing the efficiency of the CI programming process, thus addressing the time constraints currently imposed upon CI audiologists. The event-related potentials (ERPs) P1, N1, and P2 (which normally occur at around 50, 100, and 200 ms, respectively) have been investigated in CI users (Friesen et al. 2009; Ponton et al. 1996), and are well suited for consideration in the programming process because their latency and amplitude are known to be modulated both by the physical features of the acoustical stimulus and by attention and age (Tremblay et al. 2003; Wunderlich et al. 2006).
Despite the promise of using EEG to optimize CI programming, barriers do exist. A main problem related to collecting EEG recordings from CI users is the presence of substantial recording artifacts, which originate primarily from two sources. The first source arises from the radio frequency transmission generated by the transmitter and receiver that pass information from the external sound processor to the internal device (Hofmann and Wouters 2010; Friesen & Picton, 2010). The second source arises from the electrical pulses delivered by intracochlear electrodes of the internal device. Several algorithms have been used to remove these recording artifacts, including Independent Component Analysis (ICA; Sandmann et al. 2015; Viola et al. 2012) and multivariate regression analysis (Mc Laughlin et al. 2013). Here we show that the artifact removal algorithm adopted by McLaughlin et al. (2013) can be affected by parameters related to the stimulus. Specifically, our results suggest that the duration of the stimulus should be carefully considered to avoid corruption of ERP peaks of interest. More generally, it is possible that any DC offset removal algorithm that introduces a substantial discontinuity in the stimulus waveform near ERP peaks of interest may similarly distort the waveform.
MATERIALS AND METHODS
Participants
Ten adults (53 – 77 years old; mean ± SD, 62 ± 8.04 years), all post-lingually deafened, were recruited from the Maryland, Washington D.C., and Virginia areas. Nine participants used a Cochlear sound processor (Freedom or N5), while one (CBY) used an Advanced Bionics CII sound processor. Bilateral implanted users were tested in their better ear. All procedures were reviewed and approved by the Institutional Review Board (IRB) of the University of Maryland. Participants gave informed consent and were paid for their time. See Table, Supplemental Digital Content 1, for more detailed information about each participant.
Data collection and analysis
Stimuli and recording
ERPs were elicited in response to a vowel /a/ synthesized at a 20-kHz sampling rate with a Klatt-based synthesizer (Klatt 1980). Two different durations were used: 170 and 400 ms (see Figure, Supplemental Digital Content 2). Stimuli were presented unilaterally at a 1-Hz rate at the participant’s self-reported most comfortable loudness level. Stimuli were delivered via direct audio input to the participant’s clinical sound processor, which was set to their everyday program including typical sound processing features like automatic gain control. Participants with Cochlear devices had a stimulation rate of 900 pulses/s/electrode; the participant with the Advanced Bionics device had a rate of 3712 pulses/s/electrode. Using a Biosemi Active Two acquisition system (Biosemi B.V., Amsterdam, Netherlands), EEG data were recorded at a 2048-Hz sampling rate with an antialiasing filter (low-pass filter with a −3 dB point at 1/5th of the selected sampling frequency). Data were recorded with an all-pass filter with the fixed amplifier gain imposed by the Biosemi system. The recordings were performed with a 32-channel cap organized according to the 10–20 International system (top of the head ground, earlobe contralateral to the CI stimulated as reference) with two additional electrodes used to track horizontal and vertical ocular artifacts of the eye contralateral to the ear that was tested.
Data analysis
Data were converted into MATLAB format (MathWorks, version R2011b) by using the function pop_biosig available in EEGLab (Delorme et al. 2004). Data analyses were limited to electrode Cz. Raw data were digitally bandpass-filtered offline from 0.03 to 30 Hz (forward-backward 4th-order Butterworth filter) to eliminate the high-frequency transmission artifact generated by the CI (Mc Laughlin et al. 2013). Ocular artifacts were reduced using a regression-based electrooculography reduction method (Schlögl et al. 2007) in all participants except CBV (due to recording noise from the eye electrodes in that participant).
In order to remove the DC artifact or “pedestal” elicited by the current generated by the intracochlear electrodes, two different averages were computed: the average from the electrode of interest (Cz) and the average from the electrode that best represented the envelope of the stimulus recording artifact from the intracochlear electrodes. Because the presence of the DC pedestal resulted in larger amplitudes than what are normally recorded in individuals tested without CIs, the artifact rejection limits were adjusted up to ±800 μV if necessary to achieve 500 sweeps used in the average response. The stimulus recording artifact (usually recorded from the electrode most adjacent to the CI and chosen to best represent the features of the auditory stimulus) and the time domain of the neural response (Cz) were used as the arguments of a second-degree bivariate polynomial1 (Mc Laughlin et al. 2013) and were combined in a Matrix M to estimate the DC pedestal. As an alternative, a different matrix M was built by using the time domain of the neural response (Cz) and the envelope of the stimulus recording artifact as calculated by (Mc Laughlin et al. 2013): the waveform of the speech stimulus was first rectified and then low-passed filtered at 35 Hz using a zero-phase 2nd order Butterworth filter. The envelope was then decimated down to 2048 Hz and finally bandpass-filtered at 0.03 and 30 Hz using the same filter applied to the EEG data. An orthogonal-triangular decomposition was applied to each matrix M in order to find the least square solution of the bivariate polynomial. The estimated DCA for each electrode was then subtracted from each ERP, leading to the final ERP. See Figure, Supplemental Digital Content 3, for a flow chart summarizing the key steps of our analysis; see Figure, Supplemental Digital Content 4, for the topographical representation of the mean amplitude of the ERPs from each electrode recorded for pre-DC-pedestal artifact (A) and post-DC-pedestal artifact (B) removal for each participant; see Figure, Supplemental Digital Content 5, for results from electrode Cz from each participant. see Figure, Supplemental Digital Content 6, for results from electrode Cz for peaks P1, N1 and P2.
RESULTS
Figure 1 shows the stimulus recording artifact for each participant tested with the 170- and 400-ms stimuli (panels A and B, respectively). There was a wide range of DC pedestal amplitudes with negative and positive polarities. An example of the negative amplitude of the estimated DC pedestal can been seen for participant CAX in Figure 2. The waveforms have been standardized and negative polarities (six participants) were inverted for visualization purposes in Figure 1.
Figure 1.
Stimulus recording artifact for each participant tested with the (A) 170- and (B) 400-ms stimuli extracted of the electrode that best represented the stimulus recording artifact. Vertical lines represent the onset (0 ms) and offset (170 and 400 ms) of the stimuli. The seemingly thick line results from superimposing the DC pedestals from all participants. This DC pedestal creates a positive (or negative) DC offset that lasts for the duration of the stimulus.
Figure 2.
ERPs from electrode Cz for a single participant (CAX) elicited by the 170-ms (solid black) and the 400-ms (solid red) stimulus. The estimated DC pedestal for the respective stimuli are represented by dashed lines. The main peaks have been labeled P1-N1-P2. The offset of the 170-ms stimulus created a negative deflection in P2 near 180 ms. The same DC removal analysis artifact was not seen with the 400-ms stimulus.
The duration of the DC pedestal corresponded to the duration of the stimulus. For example, in Figure 2 the DC pedestals had durations that corresponded to the length of the stimulus. This resulted in an abnormal increase of the magnitude of the ERP amplitude, consistent with data reported by Mc Laughlin et al. (2013). Figure 2 also shows the effect of different stimulus duration on P2 peak morphology for CAX; Figure 3 shows the results from all the participants. The stimulus recording artifact rejection algorithm removed or minimized the DC pedestal in the participants tested, but created an additional “DC removal analysis artifact” at the onset and offset of the stimulus. The introduction of this DC removal analysis artifact can be best observed in Figs. 2 and 3 at peak P2, which is known to occur between 150 and 250 ms in normal-hearing adult participants. At an individual participant level, it is clear from the figures that the DC removal analysis artifact affects the waveform at P2; however, some participants are more affected than others. The participant most affected were CAX, CBG, CAY, CAO, and CAQ.
Figure 3.
ERPs from electrode Cz elicited by the 170-ms (black) and the 400-ms (red) stimulus for individual participants. The P2 peak was often corrupted with the 170-ms stimulus, while the late offset of the 400-ms stimulus prevented DC removal analysis artifact from altering the shape of P2.
We also sought to quantify the potential deleterious effects of the DC removal analysis artifact on response morphology using the different stimulus durations at the group level. Specifically, we observed an extremely sharp peak or spike in the time region of P2 for the 170-ms but not the 400-ms vowel. To quantify the slope, we calculated the first derivative of the response at the offset of the stimulus (150 – 180 ms), took the absolute value of the derivative function, and calculated the sum of this function. A paired t-test revealed significantly higher derivative sums for the 170-ms compared to the 400-ms stimulus (t[9] = 2.28, p = 0.048). Conversely, when the first derivative was calculated around the expected latency of peak P1 (30 – 70 ms) and N1 (70 – 150 ms), no significant differences were found between the 170- and 400-ms stimuli (t[9] = 1.03, p = 0.328 and t[9] = 0.48, p = 0.64, respectively). The time window used for P1 was based on standard latency reported in the literature; the window was also visually inspected for each individual to ensure it coincided with P1. We expanded the time window of N1 to accommodate the wide range of latencies of the participants. Means and standards deviations of the amplitude and latency for each peak are reported in Figure, Supplemental Digital Content 6.
DISCUSSION
The results of this study highlight how the offset of the stimulus may distort the morphology of evoked neural responses when removing the DC pedestal stimulus recording artifact that occurs in people with CIs. Although our analysis used the McLaughlin et al. (2013) method, DC removal analysis artifact could potentially be a problem for any DC artifact removal method, including ICA (Viola et al. 2012), and we therefore recommend caution when interpreting evoked-response morphology in the temporal window immediately following the offset of the stimulus recording artifact. In our case, we were interested in analyzing the P1-N1-P2 complex in response to the vowel /a/. The use of the 170-ms stimulus led to corruption of P2 because the stimulus offset occurred when this peak is typically elicited, 150–250 ms after stimulus onset. Specifically, the waveform distortion and DC removal analysis artifact are best observed in Figures 2 and 3 in most of the participants (CAX, CBG, CAY, CAO, and CAQ), which show the introduction of a steep positive (or negative) deflection and two short and sharp peaks in contrast to the expected single long and smooth peak. In participant CBV, the DC removal analysis artifact was so dominant that P2 was completely replaced by a single spike-like peak. Note that the morphology of P1 and N1 was preserved, as their latencies were temporally removed from the onset and offset of the stimulus such that the DC pedestal removal only affected their amplitudes. Although we used only one stimulus and specific removal algorithm, the DC removal analysis artifact may be problematic for any stimulus for which the onset or offset overlaps with a particular peak of interest.
Our observation appears to contrast with McLaughlin et al. (2013)’s conclusions, that is “the DC estimation procedure robustly attenuates the artifact even when neural response and stimulus offset overlap in time (p. 90, McLaughlin et al. (2013)).” The difference in conclusions between the two studies is easily explained; the McLaughlin et al. (2013) study focused on peak attenuation and did not look at morphology. In our study, we found that the DC removal analysis artifact associated with a 170-ms stimulus might indeed elicit a peak at around 170 ms, thus disrupting the expected P2. The differences in P2 morphology are unlikely to be caused by the use of low sampling frequency recording, as McLaughlin et al. (2013) showed that their method can be implemented even with most commonly used low-rate acquisition systems. It is important to point out that even in McLaughlin et al. (2013), the morphology of the ERP elicited in response to a 100-ms stimulus differs from the one elicited using a 300-ms stimulus (see Figure 4, p. 87, McLaughlin et al., 2013). However, in that case the offset of the stimulus was so late that it had only a limited impact on the “tail” of P2, making the differences of the ERP to different stimuli harder to visualize. In our case, the artifact was more pronounced, as we used a stimulus whose offset (170 ms) degraded P2 at its very origin, thus significantly compromising the morphology of this peak in its entirety”
Note that substantial DC removal analysis artifact was not observed in all of the participants (participants CBA, CBM, CCA, and CAQ, Fig. 3); nonetheless, it remains an important issue that needs to be taken into consideration when performing cortical recordings and data analysis. We strongly advise choosing the length of the stimulus to be much longer than the ERP of interest to minimize the possibility that the offset of stimulus recording artifact produces a DC removal analysis artifact that occurs at the time of specific peaks of interest. If P1 is the ERP of interest, we would recommend avoiding stimuli whose offset is in the [20 – 80 ms] range; if N1 is the ERP of interest, we would recommend avoiding stimuli whose offset is in the [80 – 150 ms] range; if P2 is the ERP of interest, we would recommend avoiding stimuli whose offset is in the [150 – 250 ms] range. Although we used a simple stimulus, we believe the effects generalize to any stimulus, for which the onset or offset overlaps with a particular peak of interest.
Supplementary Material
Supplemental Digital Content 1. Table that provides additional information about participants. Doc
Information about participants
Supplemental Digital Content 2. Figure that illustrates stimulus waveforms. Tif
Vowel /a/ stimulus waveforms used for the experiment. Top: 170 ms, Bottom: 400 ms.
Supplemental Digital Content 3. Figure displays a flow chart of the analysis. tif
Flow chart with the key steps taken to analyze the data.
Supplemental Digital Content 4. Figure that shows topographic waveforms. tif
The analysis for the data band-pass filtered between 0.03 and 30 Hz was extended to all the 32 electrodes for the 170 ms /a/ to display the scalp map. The topographical representation of the ERPs pre and post DCA removal was also calculated by using the function topoplot from EEGlab (Delorme et al. 2004) in order to visualize the amplitude pre and post DCA removal. Topographical representation of the mean amplitude (μV) from each electrode recorded is displayed prior to removal of the stimulus recording artifact (A) and after removing the artifact (B) removal for each subject. Results show that (Mc Laughlin et al. 2013)’s algorithm significantly reduces the amplitude of the response in the vast majority of the electrodes. For subject CBV the scale of the DC stimulus recording artifact prior to removal had a wider range because of the substantially high amplitude of the artifact.
Supplemental Digital Content 5. Figure that displays amplitudes pre- and post DCA correction. Tif
Mean amplitude of the ERPs from electrode Cz pre DCA (black) and post DCA (red). The inset shows the results for subject CBV, whose pre DCA response was much higher than that of the other subjects. The mean amplitude of the ERP recorded from Cz was lower after DCA removal in all the subjects, but CAY, where the effect of the CI artifact was negligible.
Supplemental Digital Content 6. Figure that displays peaks amplitudes for the two stimulus durations. Tif
Mean ± SD of the amplitude (A) and latency (B) for each peak analyzed. The time windows used are: [30 – 70] ms for P1, [70 – 150] ms for N1 and [150 – 250] ms for P2. P1 and N1 do not show any remarkable difference between 170- and 400-ms stimuli, while the amplitude and latency of P2 are smaller and longer, respectively.
Acknowledgments
We want to thank Blake Johnson, David Meng, Joann Tang, Colette McKay, and Aline Gesualdi Manhaes for their helpful comments on the data analysis and Maureen Shader for her help in collecting data. We want to also thank Advanced Bionics and Cochlear Ltd. for the equipment to perform this study, and technical support. This work was supported by NIH Grant R01-AG051603 (M.J.G.), T32-DC000046 (Center of Comparative Evolutionary Biology of Hearing training grant award to A.P.), the International Graduate research fellowship awarded by the University of Maryland (A.P.), a seed grant from the College of Behavioral and Social Sciences at University of Maryland-College Park (Dean’s Research Initiative, M. J. G.), and a seed grant from the Brain and Behavior Initiative (BBI) at the University of Maryland-College Park (M. J. G.).
Footnotes
The second-degree bivariate polynomial was chosen because no substantial differences were noted when using either the second- or fourth-degree bivariate polynomial and this led us to opt for the use of the simplest model.]
AUTHORS CONTRIBUTION
A.P., H.I.-B., S.A., and M.J.G. designed the experiments; A.P. and S.A. collected the data; A.P. analyzed the data; A.P., H.I.-B., S.A., and M.J.G. wrote the paper.
Financial disclosures/Conflicts of interest:
We have no conflict of interest to report. This study was funded by UMCP Department of Hearing and Speech Sciences, NIH-NIDCD Grant T32DC000046, The Bionics Institute, the International Graduate research fellowship awarded by the University of Maryland, and NIH-NIDCD Grant R01 AG051603 (Goupell). Cochlear Ltd. and Advanced Bionics provided equipment and technical support. The Bionics Institute acknowledges the support it receives from the Victorian Government through its Operational Infrastructure Support Program. HIB was supported by NHMRC Early-Career Fellowship #1069999.
References
- Delorme A, Makeig S. EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. J Neurosci Methods. 2004;134:9–21. doi: 10.1016/j.jneumeth.2003.10.009. [DOI] [PubMed] [Google Scholar]
- Friesen LM, Picton TW. A method for removing cochlear implant artifact. Hear Res. 2010;259:95–106. doi: 10.1016/j.heares.2009.10.012. [DOI] [PubMed] [Google Scholar]
- Friesen LM, Tremblay KL, Rohila N, et al. Evoked cortical activity and speech recognition as a function of the number of simulated cochlear implant channels. Clin Neurophysiol. 2009;120:776–782. doi: 10.1016/j.clinph.2009.01.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Gifford RH, Shallop JK, Peterson AM. Speech recognition materials and ceiling effects: considerations for cochlear implant programs. Audiol Neurootol. 2008;13:193–205. doi: 10.1159/000113510. [DOI] [PubMed] [Google Scholar]
- Gilley PM, Sharma A, Dorman M, et al. Minimization of cochlear implant stimulus artifact in cortical auditory evoked potentials. Clin Neurophysiol. 2006;117:1772–1782. doi: 10.1016/j.clinph.2006.04.018. [DOI] [PubMed] [Google Scholar]
- Gordon KA, Abbasalipour P, Papsin BC. Balancing current levels in children with bilateral cochlear implants using electrophysiological and behavioral measures. Hear Res. 2016;335:193–206. doi: 10.1016/j.heares.2016.03.013. [DOI] [PubMed] [Google Scholar]
- Klatt DH. Software for a cascade/parallel formant synthesizer. J Acoust Soc Am. 1980;67:971–995. [Google Scholar]
- Lopez Valdes A, Mc Laughlin M, Viani L, et al. Objective assessment of spectral ripple discrimination in cochlear implant listeners using cortical evoked responses to an oddball paradigm. PLoS One. 2014;9:e90044. doi: 10.1371/journal.pone.0090044. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Mc Laughlin M, Lopez Valdes A, Reilly RB, et al. Cochlear implant artifact attenuation in late auditory evoked potentials: a single channel approach. Hear Res. 2013;302:84–95. doi: 10.1016/j.heares.2013.05.006. [DOI] [PubMed] [Google Scholar]
- Ponton CW, Don M, Eggermont JJ, et al. Maturation of human cortical auditory function: differences between normal-hearing children and children with cochlear implants. Ear Hear. 1996;17:430–437. doi: 10.1097/00003446-199610000-00009. [DOI] [PubMed] [Google Scholar]
- Sandmann P, Plotz K, Hauthal N, et al. Rapid bilateral improvement in auditory cortex activity in postlingually deafened adults following cochlear implantation. Clin Neurophysiol. 2015;126:594–607. doi: 10.1016/j.clinph.2014.06.029. [DOI] [PubMed] [Google Scholar]
- Scheperle RA, Abbas PJ. Peripheral and Central Contributions to Cortical Responses in Cochlear Implant Users. Ear Hear. 2015;36:430–440. doi: 10.1097/AUD.0000000000000143. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Schlögl A, Keinrath C, Zimmermann D, et al. A fully automated correction method of EOG artifacts in EEG recordings. Clin Neurophysiol. 2007;118:98–104. doi: 10.1016/j.clinph.2006.09.003. [DOI] [PubMed] [Google Scholar]
- Tremblay KL, Piskosz M, Souza P. Effects of age and age-related hearing loss on the neural representation of speech cues. Clin Neurophysiol. 2003;114:1332–1343. doi: 10.1016/s1388-2457(03)00114-7. [DOI] [PubMed] [Google Scholar]
- Viola FC, De Vos M, Hine J, et al. Semi-automatic attenuation of cochlear implant artifacts for the evaluation of late auditory evoked potentials. Hear Res. 2012;284:6–15. doi: 10.1016/j.heares.2011.12.010. [DOI] [PubMed] [Google Scholar]
- Wunderlich JL, Cone-Wesson BK, Shepherd R. Maturation of the cortical auditory evoked potential in infants and young children. Hear Res. 2006;212:185–202. doi: 10.1016/j.heares.2005.11.010. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplemental Digital Content 1. Table that provides additional information about participants. Doc
Information about participants
Supplemental Digital Content 2. Figure that illustrates stimulus waveforms. Tif
Vowel /a/ stimulus waveforms used for the experiment. Top: 170 ms, Bottom: 400 ms.
Supplemental Digital Content 3. Figure displays a flow chart of the analysis. tif
Flow chart with the key steps taken to analyze the data.
Supplemental Digital Content 4. Figure that shows topographic waveforms. tif
The analysis for the data band-pass filtered between 0.03 and 30 Hz was extended to all the 32 electrodes for the 170 ms /a/ to display the scalp map. The topographical representation of the ERPs pre and post DCA removal was also calculated by using the function topoplot from EEGlab (Delorme et al. 2004) in order to visualize the amplitude pre and post DCA removal. Topographical representation of the mean amplitude (μV) from each electrode recorded is displayed prior to removal of the stimulus recording artifact (A) and after removing the artifact (B) removal for each subject. Results show that (Mc Laughlin et al. 2013)’s algorithm significantly reduces the amplitude of the response in the vast majority of the electrodes. For subject CBV the scale of the DC stimulus recording artifact prior to removal had a wider range because of the substantially high amplitude of the artifact.
Supplemental Digital Content 5. Figure that displays amplitudes pre- and post DCA correction. Tif
Mean amplitude of the ERPs from electrode Cz pre DCA (black) and post DCA (red). The inset shows the results for subject CBV, whose pre DCA response was much higher than that of the other subjects. The mean amplitude of the ERP recorded from Cz was lower after DCA removal in all the subjects, but CAY, where the effect of the CI artifact was negligible.
Supplemental Digital Content 6. Figure that displays peaks amplitudes for the two stimulus durations. Tif
Mean ± SD of the amplitude (A) and latency (B) for each peak analyzed. The time windows used are: [30 – 70] ms for P1, [70 – 150] ms for N1 and [150 – 250] ms for P2. P1 and N1 do not show any remarkable difference between 170- and 400-ms stimuli, while the amplitude and latency of P2 are smaller and longer, respectively.