Habituation of Central and Electrodermal Responses to an Auditory Two-Stimulus Oddball Paradigm

Comments 1: This is an interesting paper examining SCR and P300 habituation in an auditory oddball task.

 Unfortunately, the view of the P300 as including only P3a and P3b is outdated and indefensible.  Some of the literature negating that view is cited but not integrated into the MS.  One citation of importance in this regard is listed in the references but not mentioned in text.

Response 1: We thank the reviewer for his/her comment. We recognize that there are consistent findings negating the view of P300 as a single entity, and that such component is instead described as a Late Positive Complex (LPC), which includes P3a, P3b, and the novelty P3 (nP3). Often, these components are followed by a late and broadly distributed Slow Wave (SW) component (1). In our original analysis, we attempted to identify such components by means of visual inspection on the time course of grand-average ERPs and their associated scalp map distribution, and the guidance from previous literature. However, we were able to find only P3a and P3b components, but neither nP3 nor SW could be identified. This may be related to the temporal overlapping of LPC components, which hinders their individual contribution to the observed responses (2).

To address the reviewer’s concern, we reviewed our analysis by applying the method of temporal PCA to isolate and identify the contribution of individual LPC components and investigate habituation to target stimuli. Temporal PCA decomposes ERP responses into a set of orthogonal (i.e., uncorrelated) components, whose temporal course and spatial distribution are described by factor loadings and factor scores, respectively. Specifically, factor loadings represent the time course of each component, and are fixed across channels, conditions and subjects. Instead, factor scores quantify the contribution of each component to the amplitude of each response (2).

Operationally, we concatenated average ERPs to target stimuli in the (90-470)ms time range across subjects and for each of the three stimulation blocks, and we applied temporal PCA decomposition through the erpPCA functions implemented by Kayser J and Tenke CE (http://psychophysiology.cpmc. columbia.edu/software/) (3). The input data consisted of 1140 trials (i.e., channels × blocks × subjects) by 39 time points. The covariance matrix underwent Kaiser normalization, and factors were rotated through unrestricted Varimax rotation. Following factors’ extraction order, they were associated to established components through visual inspection of their peak latency, polarity and scalp map. Finally, for each component contributing the LPC, we extracted the corresponding peak’s amplitude factor scores for each subject and for each stimulation block. Particularly, we selected channels of major interest based on the maximums of the scalp topography (i.e., factor scores) associated to each component (2).  

We report the results of temporal PCA analysis in Figure 1. In particular, we show the scalp topography of factor scores for each identified component and across stimulation blocks, together with time course of each factor (i.e., loadings) and their peak latency. The first 9 factors accounted for about the 94.8% of the total variance, with a maximum of 38.7% for the first factor and a minimum of 1.15% for the ninth factor. We did not consider any further factor since they explained a variance lower than 1%.

We found two subcomponents of the N1, i.e., N1-3 and the N1-1, at a latency of 100ms and 130ms respectively (1,4), followed by a P2 component peaking at 150ms and centered at the vertex Cz. Concerning the LPC, we found both an early P3a (eP3a) and a late P3a (lP3a), peaking at 200ms and 250ms, respectively (5). The eP3a was distributed towards central and frontal regions around Cz and Fz, whereas lP3a was shifted towards central and parietal regions around Cz and Pz. Furthermore, we observed a P3b component with latency of 310ms and a topography focused around Pz, followed by a frontal nP3 peaking at 340ms around Fz (1,5). Finally, we found two subcomponents of the SW, namely the positive SW (+SW) at 370ms and the negative SW (-SW) at 430ms. The latter subdivision of SW is in line with (5,6). Both +SW and -SW topographies showed a maximum and a minimum, respectively, in the frontal region around Fz.

Figure 1 Results of the temporal PCA decomposition on subject-average ERPs. Top: topographical grand-average voltage distribution (i.e., scores) of each identified factor across the three stimulation blocks. Bottom: time-course of each factor (i.e., loadings) in the (90-470)ms range, and its peak latency in ms.

We tested for the occurrence of habituation to target stimuli over stimulation blocks through a permutation-based one-way repeated-measures ANOVA (10000 permutations, α=0.05) on the factor scores of each LPC component, with the three stimulation blocks as the main effect of interest. Multiple comparisons were controlled with the false discovery rate (FDR) method. We conducted post-hoc comparisons through multiple paired t-tests, and we adjusted p-values with the Bonferroni correction.

We observed a significant effect for the stimulation blocks on the amplitude of lP3a component at Cz (F2,40 = 4.17, p<0.05), and for the amplitude of P3b component at Pz (F2,40 = 5.02, p<0.05). Post-hoc analysis highlighted a decrease in the amplitude of responses to target stimuli over time, with both lP3a and P3b showing a higher amplitude in the first stimulation block, with respect to the third. As depicted in Figure 1, the grand-average scalp distribution of -SW showed a similar trend at Fz, but without any statistical significance. Likewise, we did not find any significant effect for the stimulation blocks on the factor scores of eP3a, nP3, and +SW.

In light of these changes, in the new version of the manuscript, we revised both the introduction and discussion sections.

We added a Temporal PCA subsection to the Materials and Methods section, which we report here for the ease of the reviewer:

“We applied temporal PCA to investigate the occurrence of habituation to target stimuli on individual ERP components. Temporal PCA decomposes ERP responses into a set of orthogonal (i.e., uncorrelated) components, whose temporal course and spatial distribution are described by factor loadings and factor scores, respectively. Factor loadings represent factors’ (i.e., components) time course, and are fixed across channels, conditions and subjects. Instead, factor scores represent how much a factor contributes to the amplitude (i.e., voltage) of each observation [35]. Operationally, we further epoched subject-average ERPs to target stimuli in the (90-470)ms time range for each stimulation block, and we applied temporal PCA decomposition through the erpPCA functions implemented by Kayser J and Tenke CE (http://psychophysiology.cpmc. columbia.edu/software/) [36].

The input data matrix consisted of 1140 trials (i.e., channels × blocks × subjects) by 39 time points. Prior to svd decomposition, the covariance matrix underwent Kaiser normalization. Factors were rotated through unrestricted Varimax rotation. Following their extraction order, we identified factors in terms of their peak latency, polarity and scalp map. Finally, for each component contributing the LPC, we extracted the corresponding peak’s amplitude factor scores for each subject and for each stimulation block. Particularly, we selected channels of major interest based on the maximums of the scalp topography (i.e., factor scores) associated to each component [35].”

We also changed the EEG statistical analysis subsection as follows:

“We then investigated for the occurrence of habituation to target stimuli over time. To this aim, we performed a permutation-based one-way repeated-measures ANOVA (10000 permutations, α=0.05, corrected with FDR) on the factor scores of each LPC component identified through temporal PCA (see Section 2.4), with the three stimulation blocks as the effect of interest. We conducted post-hoc comparisons through multiple paired t-tests, and we adjusted p-values with the Bonferroni correction.”

Results were updated to account for the new findings following temporal PCA decomposition and habituation investigation on PCA components. We report here the edited section for the ease of the reviewer:

“The results of habituation analysis on LPC components’ amplitude (i.e., factor scores) over stimulation blocks are resumed in Table 1. We observed a significant effect for the stimulation blocks on the amplitude of lP3a component at Cz (F2,40 = 4.17, p<0.05), and for the amplitude of P3b component at Pz (F2,40 = 5.02, p<0.05). Post-hoc analysis highlighted a decrease in the amplitude of responses to target stimuli over time, with both lP3a and P3b showing a higher amplitude in the first stimulation block, with respect to the third. As depicted in Fig.9, the grand-average scalp distribution of -SW showed a similar trend at Fz, but without any statistical significance. Likewise, we did not find any significant effect for the stimulation blocks on the factor scores of eP3a, nP3, and +SW.”

Comments 2: With SCR, it is not clear on what basis the proximal rather than distal sites were chosen to record from.  There is no mention of the electrode paste used or the exciting voltage.

Response 2: We thank the reviewer for the comment. We have chosen proximal sites for SCR recording as this site has been reported to be less susceptible to movement artifacts and provide a wider area for electrodes fixing compared to distal sites (7). Concerning the electrode paste, we have used pre-gelled Ag/AgCl electrodes. The applied exciting voltage was 0.5V.  

We have modified Section 2.2 of the manuscript, which we report here for the ease of the reviewer:

“EEG and SC were acquired synchronously using a DSI-24 system from Wearable Sensing at the sampling frequency of 300Hz. The headset was composed of 19 dry electrodes placed according with the 10-20 international system. Electrode impedance was always kept below 1MΩ in accordance with the manufacturer guidelines. All channels were referenced to Pz. SC was recorded via a dedicated auxiliary channel in the EEG cap applying an exciting voltage of 0.5V. A pair of pre-gelled Ag/AgCl electrodes were placed on the proximal phalanx of the first and second fingers of the non-dominant hand, respectively. Proximal sites were chosen to minimize the presence of movement artifacts in the recordings [30].”

Comments 3: With EEG, there is no mention of the number of accepted epochs.

Response 3: We thank the reviewer for the comment. We believe the reviewer is referring to the number of clean epochs retained after EEG preprocessing. In this regard, we retained an average of 295 epochs (min: 280; max: 300) for the standard stimuli, and an average of 74 epochs (min: 70; max: 75) for the target stimuli.

We have updated Section 2.4 of the manuscript accordingly, which we report here for the ease of the reviewer:
“We epoched the EEG data in the (-200, 1000)ms range with respect to each stimulus’ onset (i.e., 0ms), and after a visual inspection we removed epochs containing residual artifact activity. An average number of 295 epochs (minimum: 280; maximum: 300) and 74 epochs (minimum: 70; maximum: 75) was retained over the subjects for the standard and target stimuli, respectively. We performed baseline correction of clean epochs by subtracting the average potential in the pre-stimulus interval (i.e., from -200ms to 0ms), and we estimated subject-average ERPs by grouping epochs in accordance with the stimulus type (standard/target) and stimulation blocks, for a total of 6 conditions (standard/target stimuli over three blocks).”

Comments 4: They state on P 5. L211 that they used SOAs between 5 and 6 s; at L 239 it states “between 4s and 5s”.
Response 4: We thank the reviewer for having pointed at the discrepancy regarding the specified SOAs. In the section 1.5 we have reported a wrong SOA interval. The correct interval is specified in section 2.1, which is between 4s and 5s.

We have corrected the typo in section 1.5, and we report here the edited part of the manuscript of the ease of the reviewer:
“Since we also had to study in parallel the effects on SCRs (phenomena that are overall slower and more delayed than ERPs), the only change we made to the standard oddball paradigm was to use longer stimulus-onset asynchronies (SOAs) than usual (i.e., between 4 and 5 s), to allow us the resolution of the individual SCRs without causing their temporal summation.”

Comments 5: Stimulus frequencies are stated as 10kHz for standards and 15kHz for targets. These are rather higher than is common, and there is no stated check on their audibility. They were not counterbalanced, so that any effect claimed as S/T is confounded with frequency differences (and potential reduced hearing level).  I did not see a mention of stimulus duration, rise/fall times, or intensity.
Response 5: We thank the reviewer for the comment. There is a typo in the specification of the stimulus frequencies adopted. Indeed, instead of 10kHz for standards and 15kHz for targets, the correct values were 1.0kHz for standards and 1.5kHz for targets. Stimuli duration was 200ms, rise and fall times were equally set to 5ms, and tone intensity was set to 60 dB SPL.

We have included the aforementioned information in Section 2.1 of the manuscript, which we report here for the ease of the reviewer:
“The experimental protocol consisted of 3 min of initial rest, followed by an auditory oddball task. The task consisted of three consecutive blocks of 125 auditory stimuli each, with 80% standard stimuli (frequency: 1.0kHz) and 20% target stimuli (frequency: 1.5kHz). All stimuli were monaural tones with a duration of 200ms, rise and fall times of 5msec, and an intensity of 60dB SPL. The presentation of the stimuli (i.e., standard/target) was pseudo-random, meaning that it was possible to present two target stimuli in a row. SOAs were randomly chosen between 4s and 5s, to allow for a reasonable recovery of stimulus-specific (i.e., evoked) SCRs.”

Comments 6: The number of retained epochs for analysis is not apparent.
Response 6: We thank the reviewer for the comment. As already reported in Response 3, the number of average retained epochs was 295 (min: 280; max: 300) for the standard stimuli, and  74 (min: 70; max: 75) for the target stimuli, respectively.

Comments 7:  Fig. 8 legend refers to blue and red lines, but the figure has no blue lines.
Response 7: We thank the reviewer for the comment. We have corrected the caption of Fig.8 which we report here for the ease of the reviewer:
“Grand-average ERP responses to the standard (red) and target (green) stimuli, and the

difference between them (difference wave; yellow), evaluated at a) Fz, b) Cz, and c) Pz channels. Responses are plotted in the (-200, 1000)ms interval with respect to the stimulus’ onset (black vertical line). The gray shaded areas indicate significant differences between responses to standard and target stimuli (p<0.05).”

Comments 8:  Fig. 9: “P3a” topography is atypical.
Response 8: We thank the reviewer for the comment. In the revised version of the submitted manuscript, we have performed temporal PCA to isolate LPC components to isolate their individual contribution to observed responses in terms of both time course and scalp topography. Please see Response 1 and Figure 1 for details about the methodology followed and the new scalp topographies.

Comments 9:  Discussion is verbose and many statements are unsupported, reflecting the misuse of P3a/P3b components and their outdated understanding of the P300 family.
Response 9: We thank the reviewer for the comment. We improved our discussion section according with our new analyses of LPC components through temporal PCA. Please see Response 1 for details about methodology and changes made to Discussion section.

Comments 1: A number of studies have shown that physiologic hormonal fluctuations in estrogen and progesterone during the menstrual cycle can affect central and autonomic regulation of psychophysiological functions, including the ERP (Fleck, K. M., & Polich, J. (1988). P300 and the menstrual cycle. Electroencephalography and clinical neurophysiology, 71(2), 157–160. https://doi.org/10.1016/0168-5597(88)90076-7), SCR (Gómez-Amor, J., Martínez-Selva, J. M., Román, F., & Zamora, S. (1990). Electrodermal activity in menstrual cycle phases: a comparison of within- and between-subjects designs. International journal of psychophysiology : official journal of the International Organization of Psychophysiology, 9(1), 39–47. https://doi.org/10.1016/0167-8760(90)90006-y) orienting reactions and attention  (Ronca, F., Blodgett, J. M., Bruinvels, G., Lowery, M., Raviraj, M., Sandhar, G., Symeonides, N., Jones, C., Loosemore, M., & Burgess, P. W. (2024). Attentional, anticipatory and spatial cognition fluctuate throughout the menstrual cycle: Potential implications for female sport. Neuropsychologia, 108909. Advance online publication. https://doi.org/10.1016/j.neuropsychologia.2024.108909

Response 1: We thank the reviewer for the comment. In the literature there are contrasting results regarding a potential effect of the menstrual cycle on ERP responses. While some authors found an effect for the specific cycle period on the amplitude of N1 and P2 components (1), others did not observe any consistent change over the course of the menstrual cycle (2,3). Particularly, in the auditory discriminant paradigm cited by the reviewer (2), it has been reported no difference in either the amplitude or the latency for any of the key components involved in the task (i.e., N1, P2, N2, and P3).
Similarly, there are no consistent findings on the effect of menstrual cycle on electrodermal activity (EDA). On the one hand, some studies reported no differences in EDA associated with menstrual cycle phases (4–7), whereas other studies reported increased EDA during particular menstrual periods (8,9). In the latter case, it has been reported an effect of the menstrual cycle on both the skin conductance response (SCR) amplitude evoked by stimuli and the skin conductance level (SCL). While SCL dynamics were not of interest in our study, we extracted evoked SCRs according with a methodology based on the cvxEDA algorithm (10). This algorithm allows to recover SCRs based on the estimation of the underlying sympathetic sudomotor nerve activity, thus providing normalized responses which mitigate between-subject differences due to potential confounding factors (10). Additionally, our statistical analyses follow of a within-subject design. Accordingly, observed group-level effects are based on intra-subject differences between experimental conditions, further mitigating the risk of inter-subjects confounding factors (9).
Finally, there are also conflicting results in literature on the association between the menstrual cycle period and attention. Specifically, while some authors reported a decrease of attention due to menstrual cycle (11), others did not find any significant effect (12).

In this light, we believe that further investigations should be necessary to confirm an effect of the menstrual cycle period on ERP and SCR responses.

We have updated Section 4.4 Limitations of our manuscript to account for the potential confounding effect of the menstrual cycle period on our findings, which we report here for the ease of the reviewer:

“It is important to recognize that several factors, not all easily controllable or accounted for in this study, can influence central and autonomic dynamics. Such factors may include stress, sleep quality, and hormonal fluctuations [55,56]. A specific instance is the menstrual cycle, during which some studies have observed differences in the amplitude and latency of ERP components [56] (e.g., N1, P2, N2, P3) and in the amplitude of evoked SCRs [57,58]. However, variability in study designs and task choices has led to contrasting results in the literature, underscoring the need for more comprehensive investigations [59,60]. To address these limitations, our statistical analyses followed a within-subject design, which helps mitigate potential confounding effects of the menstrual cycle on the observed results. Additionally, the analysis of evoked SCRs was based on the cvxEDA algorithm. This method estimates the underlying sympathetic nervous activity, reducing the impact of confounding factors and inter-subject variability, thereby enhancing the reliability of our results.”      
       

Comments 2: When almost half of the participants are women of reproductive age, the change in autonomic and central nervous system parameters (ERP and SCR) during the menstrual hormonal cycle must be taken into account.
Therefore, the female’s results should be reanalyzed according to the phases of the ovarian-hormonal cycle and re-presented in the improved paper.

Response 2: We thank the reviewer for the comment. We amended the lack of control on the potential confounding effect of the menstrual cycle period on the observed responses in Section 4.4 Limitations. Please see Response 1 for a detailed answer to the concern.

Comments 3: In this way the conclusions could change.

Response 3: We extended Section 4.4 Limitations of the manuscript (please see Response 1).

Comments 4:  References may be appropriate if the important issues listed above are addressed and cited.

Response 4: We have addressed the reviewer’s concern stated at Comment 1 by updating the Limitations subsection of the manuscript together with the addition of references to relevant state-of-the-art research.

4. Response to Comments on the Quality of English Language

Point 1: instead «Twenty healthy volunteers (age 24 ± 4, 10 females) participated TO the study» it would be more correct to write “Twenty healthy volunteers (age 24 ± 4 years, 10 women) participated IN the study”.

Response 1: We thank the reviewer for the comment. We have updated the sentence according to the reviewer’s suggestion.