EEG-Correlates of Emotional Memory and Seasonal Symptoms

Round 1

Reviewer 1 Report

Dear Authors,

I have reviewed the manuscript “EEG-correlates of emotional memory and seasonal symptoms” Manuscript ID: applsci-2527047 that has been submitted for publication in the: Applied Sciences (ISSN 2076-3417) and I have identified a series of aspects that in my opinion must be addressed to bring a benefit to the manuscript.

The article under review will be improved if the authors address the following aspects in the text of the manuscript:

1.     The abstract mentions statistical results (p-values) but does not elaborate on the practical significance of these findings. What do these differences in EEG power and memory performance mean in the context of the study's objective?

2.     In the abstract some sentences are quite long and complex, which may make it harder for readers to grasp the main points. Simplifying and clarifying certain sentences would improve readability.

3.     The authors should use 2 or more datasets, when a study incorporates data from multiple datasets, it can increase the generalizability of the findings. By analyzing diverse samples, the study's results may have broader applicability to different populations or contexts.

4.     Compare Results and Conclusions: Compare the results and conclusions of the latest research with those of the paper you are examining. Identify similarities, differences, and any new insights that the latest research might offer.

5.     The references need to be updated for the years 2022 and 2023, as this field has been recently raised.

DOI 10.7717/peerj-cs.492

https://link.springer.com/chapter/10.1007/978-3-030-91103-4_6

 

6.      Absence of Future Directions.

Author Response

Revision of

Manuscript ID: applsci-2527047
Type of manuscript: Article
Title: EEG-correlates of emotional memory and seasonal symptoms
Authors: Dagný Theódórsdóttir, Yvonne Höller *
Received: 11 July 2023
Submitted to section: Applied Neuroscience and Neural Engineering,
https://www.mdpi.com/journal/applsci/sections/neurosciences

 

 

Reviewer 1

 

Dear Authors,

I have reviewed the manuscript “EEG-correlates of emotional memory and seasonal symptoms” Manuscript ID: applsci-2527047 that has been submitted for publication in the: Applied Sciences (ISSN 2076-3417) and I have identified a series of aspects that in my opinion must be addressed to bring a benefit to the manuscript.

We thank the reviewer for this comprehensive review and the time taken to reply to us and explain all the necessary changes! We addressed all comments and have changes highlighted with “track changes” in the manuscript.

The article under review will be improved if the authors address the following aspects in the text of the manuscript:

1. The abstract mentions statistical results (p-values) but does not elaborate on the practical significance of these findings. What do these differences in EEG power and memory performance mean in the context of the study's objective?

We added the following sentence to the abstract: “This differential pattern of activation while viewing emotional pictures suggests a difference in emotional processing between the groups.“

Also, the last sentence of the abstract directly addresses the research aim of the study as described in the first sentence of the abstract:

First sentence(s):“The aim of this study was to investigate a potential all-year vulnerability of people with seasonal mood fluctuations. We compared behavioral and neurophysiological responses to emotional stimuli in summer between people who report seasonal symptoms in winter and those who do not.“

Please note that according to the comment nr. 2 of the reviewer we have split up this sentence in two sentences.

Last sentence: “The absence of behavioral effects but presence of differences in EEG activity suggests an all-year-long difference in processing of emotional contents in people who experience seasonal symptoms in winter.“

 

2. In the abstract some sentences are quite long and complex, which may make it harder for readers to grasp the main points. Simplifying and clarifying certain sentences would improve readability.

We have split up several sentences in two sentences, e.g.
„The aim of this study was to investigate a potential all-year vulnerability of people with seasonal mood fluctuations. We compared behavioral and neurophysiological responses to emotional stimuli in summer between people who report seasonal symptoms in winter and those who do not.“

„EEG power was overall higher in participants without elevated levels of seasonal symptoms (p=.043). This group difference interacted with emotional valence (p=.037), region of interest (p=.003), hemispheric differences (p=.027), frequency band (.032), and time-window (.018).“

 

3. The authors should use 2 or more datasets, when a study incorporates data from multiple datasets, it can increase the generalizability of the findings. By analyzing diverse samples, the study's results may have broader applicability to different populations or contexts.

We agree with this limitation, we had mentioned it already in the limitations section as follows:

“Furthermore, the investigation was conducted in Iceland, where the summers come with almost 24h of daylight. It is therefore possible that the results are not applicable to areas at a less extreme latitude.“

In order to make this more clear in accordance with the reviewer’s suggestion we added the following statement:

„Future research should therefore investigate diverse samples in order to improve the generalizability of the results.“

 

4. Compare Results and Conclusions: Compare the results and conclusions of the latest research with those of the paper you are examining. Identify similarities, differences, and any new insights that the latest research might offer.

We discuss our results in relation to previous research in the sections 4.1 and 4.2 of the article. We discuss similarities and differences between previous research and ours, for example: „So far, studies reporting a negative memory bias in patients with SAD found this effect in winter [47,50,59]. This negative memory bias is well known for patients with major depression [60]. However, there are some subtle differences between the effect of depression and SAD. It was reported that patients with SAD remember more positive pictures in summer than in winter and this seasonal change distinguishes them from healthy controls [48]. The very same study found no significant difference between healthy controls and patients in the seasonal change of negative or neutral words recall performance [48]. An interventional study found that treatment of SAD improves the memory for positive contents [47]. In another experiment, patients with SAD showed impaired recognition of positive stimuli and impaired capacity to suppress responses to negative stimuli in winter whereas healthy controls did not demonstrate such biases [59]. Our findings are in line with these prior reports. In our study that was conducted during remission phase in summer, the groups did not differ with respect to their responses to emotional stimuli. Thus, it is possible that people who report seasonal symptoms to occur during winter do not differ from people without such symptoms during summer with respect to emotional memory biases. It is also possible that differences in the stimulus material explain the missing effect. Common stimuli used in emotional memory experiments include emotional faces [51,59], pictures from the International affective picture system [14], affective words [47], or stories [50]. We used the OASIS picture database, which is relatively new [54]. It was acknowledged by the authors of the OASIS picture database that the pictures in this database are underrepresenting the low-arousal positive and low-arousal negative segment. While we balanced the three valence groups by arousal, that is, low-middle- high arousal stimuli where proportionate in all three emotional categories, we cannot rule out that the positive images were on average more arousing and, therefore, more likely recalled. Finally, in addition to previous reports on negativity bias in relation to seasonality [47,50,59] not all experimenters have found such an effect [61].”

And furthermore: “This is in line with previous reports that found a general difference in brain activation between patients with SAD and healthy controls that was independent of emotional reactivity [21,51]. The localization of this effect is not as exact in the EEG as in fMRI, but we narrowed down the region of interest to lobes that were reported previously to be important for emotional processing [62]. We found that the difference between people with low and high seasonality scores was particularly strong over the temporal lobe when viewing negative or positive pictures. Our findings resemble earlier results that found differential hemisphere x valence interactions in EEG band-power over the temporal lobe [63,64]. According to our data, the right hemisphere seems to play a more important role as it responds differentially in the two groups especially during viewing of negative pictures. Less clear but consistent effects can be found for positive pictures which are lateralized to the left hemisphere. This left-right differentiation for positive and negative stimuli is consistent to what was found in theta and alpha EEG band-power previously [63,64]. Furthermore, this finding supports the valence hypothesis, which assumes that the left hemisphere is dominant for positive emotions while the right hemisphere is dominant for negative emotions [65]. The finding that our results are significant on the right hemisphere, while they are significant only before but not after correcting for multiple comparisons on the left hemisphere supports the right hemisphere model of brain asymmetry in emotional processing, which states a dominant role of the right hemisphere [65]. Therefore, our findings support the claim that the valence hypothesis and the right hemisphere model of brain asymmetry in emotional processing are not mutually exclusive [66]. The enhanced group difference over the right hemisphere for negative stimuli is also in agreement with previously reported reduced lateralization of emotional processing in people at risk for depression [11]. It should also be noted that negative stimuli seem to be generally prioritized in processing [67] which might explain the enhanced difference for negative stimuli between people with and without seasonality symptoms. Another interesting dissociation in the alpha band is that group differences were detected in the early time-window (50-150 ms post-stimulus) for positive pictures but the later time-window (300-400 ms) for negative pictures. This effect is significant only before but not after correction for multiple comparisons in the gamma range, but with a similar pattern (early for positive pictures, both time windows for negative pictures). Earlier timing of positive stimulus processing vs. later timing of negative stimulus processing correlates in the alpha range was reported previously [63] and our results suggest that this timing is affected by seasonality. Gamma-band reactivity to emotional stimuli, especially of negative valence, differs between patients with depression and healthy controls [68]. A common assumption is that gamma-band involvement reflects the degree to which a stimulus is consciously processed [40]. An early process around 150 ms post-stimulus has been suggested to mirror early attention for emotional cues [69,70] whereas processes later than 300 ms have been indicated to mirror additional emotional processing that cannot be detected when presenting neutral stimuli [71]. This late gamma component coincides with the P300 in the event related potential and is strongly linked to the processing of unpleasant stimuli [41,44].“

 

5. The references need to be updated for the years 2022 and 2023, as this field has been recently raised.

We are very sorry but the suggested citation „DOI 10.7717/peerj-cs.492” is not even by far related to our research, it is about “applying survival analysis methods like cox regression on RNAseq microarray“, thus, genetic data, and not brain data, and also the methods are very similar. We do not see how we can cite this work in the present manuscript.

The same holds for the other recommended article https://link.springer.com/chapter/10.1007/978-3-030-91103-4_6 which is about AI methods for detection of ocular diseases. We do neither use AI methods in this article nor do we talk about ocular diseases.

 

6. Absence of Future Directions.

We agree we have to add a future direction. As mentioned earlier and in accordance with the reviewer’s comments, we added one sentence in section 4.3 (Limitations) “Future research should therefore investigate diverse samples in order to improve the generalizability of the results.“ In addition, we added: „Moreover, future research should investigate an all-year vulnerability in people who have a clinical diagnosis for SAD.“

 

 

Author Response File: Author Response.docx

Reviewer 2 Report

Authors examined electrophysiological correlates of emotional memory in relation to seasonal symptoms, especially not in a remission phase, i.e. in the season when individuals who experience seasonal fluctuations of mood feel relatively good. The manuscript is written clearly and concisely, and I have only minor comments:

Ln 27: Full name of the abbreviation DSM-V should be written.

 

Figure 2: The value of the axis should have a larger size.

Author Response

Authors examined electrophysiological correlates of emotional memory in relation to seasonal symptoms, especially not in a remission phase, i.e. in the season when individuals who experience seasonal fluctuations of mood feel relatively good. The manuscript is written clearly and concisely, and I have only minor comments:

We thank the reviewer for this very nice evaluation and the helpful suggestions. We incorporated them with „track changes“.

Ln 27: Full name of the abbreviation DSM-V should be written.

We added this information as follows at the first occurence: „ in the Diagnostic and Statistical Manual for Mental Disorders, 5^th edition (DSM-V)“

 

Figure 2: The value of the axis should have a larger size.

We assume the reviewer means that the labeling of the axes should have larger font size (?). We agree that this would be better but since this is a pre-generated plot from the analysis software this adaption is unfortunately not possible. But we added information on the ranges of the axes in the figure caption to ease interpretation for the reader as follows:

“Wavelet plot average for participants with low seasonality scores while learning negative pictures. The top-left plot shows frequency on the y-axis (0-48 Hz from bottom to top) and time on the x-axis (40ms before stimulus onset to 760ms after stimulus onset). The black cross in the top-left plot represents the time-frequency intersection (10 Hz / 100 ms) that is illustrated as a comparison with the group of high seasonality scores in the plots at the bottom and right side. The difference between the average of the two groups at 10 Hz is shown in the bottom left plot with the same time-range on the x-axis but activity in µV on the x-axis (0 to 7 µV from top to bottom). The plot on the top-right panel shows activity at 100 ms over all frequencies, with the same frequency range as the left top figure, and in µV on the x-axis(0 to 7 µV from left to right). In this figure, the black line represents the group with low seasonality scores and the red line represents the group with high seasonality scores. The topographic plot in the lower right corner shows the current density distribution at 10 Hz and 100ms.”

 

 

Author Response File: Author Response.docx

Reviewer 3 Report

The work “EEG-Correlates of emotional memory and seasonal symptoms” explores seasonal affective disorders in connection with EEG and memory in a group of 119 individuals. Measurements based on the Seasonal pattern assessment questionnaires were performed. Moreover, EEG records were analyzed, focusing on the alpha and gamma bands. Early and late time-windows were also considered in the statistics. The memory was analyzed using free-recall methodology, positive, neutral and negative pictures were counted for every individual according to the season. The results of the experiment were exposed in a clear way and the statistics are well described. However, I have pointed out the following points in the text that could be better elucidated.

 

(1) Many abbreviations are used in the manuscript (SAD, SPAQ, DSM, etc), a list of abbreviations could be annexed in the text to make the reading easier.

(2) In line 76 gamma band is associated with frequency range 35-120Hz. However, in line 186 it is assumed that a filter was used and extract frequencies in the range 0-48Hz. Please explain way part of gamma band was cut off in the analysis.

(3) Line 179 describes the 31 electrodes used in the original measurements. Line 201 show only 8 electrodes. It was not clear the reason for excluding electrodes in the statistical analysis.

(4) Line 197 make explicit the band frequencies of alpha and gamma bands. I understand that the statistical analysis show in table 1 used this band frequencies. What is the reason for using Morlet transformation? It is for producing Figure 2?

(5) Please provide the R-packages used to filter the data. The Fourier filter was done in R? The Morlet transform was performed in R?

(6) Line 220 show the word “score”. What score is that?

(7) The manuscript have only 2 figures. Moreover, figure 1 has just one panel. The result of remembered pictures according to season could also been show in the manuscript (even if the result is not significant).

(8) In table 1 all degrees of freedom present 2 numbers, the second is usually infinity. Why just seasonality factor has one df?

 

Author Response

The work “EEG-Correlates of emotional memory and seasonal symptoms” explores seasonal affective disorders in connection with EEG and memory in a group of 119 individuals. Measurements based on the Seasonal pattern assessment questionnaires were performed. Moreover, EEG records were analyzed, focusing on the alpha and gamma bands. Early and late time-windows were also considered in the statistics. The memory was analyzed using free-recall methodology, positive, neutral and negative pictures were counted for every individual according to the season. The results of the experiment were exposed in a clear way and the statistics are well described. However, I have pointed out the following points in the text that could be better elucidated.

We thank the reviewer very much for taking the time to review our manuscript and the helpful comments. We addressed them all and made them visible in the revised manuscript with “track changes”.

(1) Many abbreviations are used in the manuscript (SAD, SPAQ, DSM, etc), a list of abbreviations could be annexed in the text to make the reading easier.

We added a Table of Abbreviations in Appendix B:

Appendix B

Table of Abbreviations.

Abbreviation	Meaning
DSM	Diagnostic and Statistical Manual for Mental Disorders
EEG	Electroencephalogram
GSS	Global Seasonality Score
ICA	Independent Component Analysis
OASIS	Open Affective Standardized Image Set
SAD	Seasonal Affective Disorder
SPAQ	Seasonal Pattern Assessment Questionnaire
S-SAD	Subsyndromal Seasonal Affective Disorder

 

 

(2) In line 76 gamma band is associated with frequency range 35-120Hz. However, in line 186 it is assumed that a filter was used and extract frequencies in the range 0-48Hz. Please explain way part of gamma band was cut off in the analysis.

We would like to clarify that in the first instance the reviewer refers to this frequency range was used in a previous study:

“EEG-research on emotional memory has led to the conclusion that affective stimuli evoke increased gamma oscillations (35–120 Hz) in the neocortex and the amygdala, particularly stimuli with negative valence [37].“

The second instance the reviewer refers to is the data range in which we filtered, since we wanted to exclude a) line noise and b) muscle artefacts that appear above 48 Hz – we added this explanation to the respective sentence:

„ For preprocessing, common average re-referencing was performed, and data was filtered from 0.5-48 Hz with zero-phase shift Butterworth filters in order to exclude line noise (50 Hz) and muscle artefacts above that range.“

Finally, the third instance where we talk about gamma range (not mentioned by the reviewer) is the one where we indicate which frequency range we analysed, and these frequency ranges were chosen in accordance with classical definitions of frequency ranges of the human EEG power spectrum. We added now references to the frequency ranges mentioned there to support the rationale for the choice of these frequency ranges:

„ Then, data was band-pass filtered in the alpha (8-12 Hz [26]) and gamma range (35-45 Hz [39]).“

 

(3) Line 179 describes the 31 electrodes used in the original measurements. Line 201 show only 8 electrodes. It was not clear the reason for excluding electrodes in the statistical analysis.

While the 31 electrodes are just those included in the EEG cap of the system we used, the other electrodes are those chosen based on the literature (as described in the introduction). We added a hint to that in the line that mentions those 8 chosen electrodes:

„The electrodes used for statistical analysis were F3, F4, F7, F8, T7, T8, O1, and O2 to cover core electrodes over the regions of interest (frontal, temporal, and occipital).“

(4) Line 197 make explicit the band frequencies of alpha and gamma bands. I understand that the statistical analysis show in table 1 used this band frequencies. What is the reason for using Morlet transformation? It is for producing Figure 2?

Yes indeed, this was only for purpose of creating Figure 2. We detailed this in the methods:

“For the purpose of illustrating the appropriateness of the choice of the time-window and frequency range, for each picture category (negative, neutral, and positive), data was segmented from –500 ms to +1000 ms around stimulus presentation and submitted to a wavelet transform. We conducted wavelet analysis with Morlet complex wavelet (Morlet parameter c=5) for the frequency range 1-48 Hz in 1 Hz linear frequency steps. Wavelet Normalization was done by instantaneous amplitude (Gabor Normalization). We averaged the results for each picture category, over the participants with elevated seasonality and low seasonality scores.“

 

(5) Please provide the R-packages used to filter the data. The Fourier filter was done in R? The Morlet transform was performed in R?

The EEG analysis was not done with R, we explained that in the methods:

„We analyzed EEG-data with the software Brain Vision Analyzer (Brain Products GmbH, Gilching/Germany) from the picture learning condition in the following steps…“

(6) Line 220 show the word “score”. What score is that?

We thank the reviewer for pointing to this confusing use of the word score. Indeed, we refer to the GSS, the Global Seasonality score. We replaced all instances of score where GSS is more correct. E.g. in the following sentence: “The seasonality grouping variable categorized participants as having low symptoms if their seasonality score was 10 or lower, and high if their seasonality score was 11 or higher.“

 

(7) The manuscript have only 2 figures. Moreover, figure 1 has just one panel. The result of remembered pictures according to season could also been show in the manuscript (even if the result is not significant).

We thank the reviewer for this hint and updated figure 1 to include also the effect of seasonality. We updated the figure caption accordingly.

“Boxplots of number of remembered pictures (y-axis) per emotional valence category (x-axis), grouped by seasonality group (SAD low vs. high, colors red and blue). Boxes indicate the interquartile range; the center line indicates the median. Whiskers indicate the range from the minimum (0) to the largest value that does not exceed the 1.5 x of the interquartile range above the upper quartile. Values outside that range are considered as outliers and represented as circles.“

 

 

(8) In table 1 all degrees of freedom present 2 numbers, the second is usually infinity. Why just seasonality factor has one df?

We are sorry for this confusion – seasonality has also two df, the two values are separated by a comma. But in the original version of the manuscript we forgot the space between the comma and the second df value. We added this now, so now it reads like: 1, 513.27, where 1 is the first df and 513.27 is the second df.

 

 

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Accept in present form