Changes in Functional Connectivity of Electroencephalography While Learning to Touch-Type

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This is a paper that generally reads well (see my comments below) are the study addresses a topic that has a reasonable level of novelty and contribution because while the analysis method [electroencephalography] has been the subject of many published studies the use-case [typing] is not typical. However, while the article is initially interesting there are issues which arise when reading the manuscript. I have comments:

• The introduction needs major revision in three dedicated sections:
  • Introduction: setting out the background and motivation for the study, a brief overview of the proposal (avoid duplication), the claimed contribution c/w all limitations and assumptions, and the paper structure.
  • Related Research: setting out the consideration of the related research considered  c/w a comprehensive analysis and also the research method adopted for the literature review.
  • Problem Statement: detailing the problem being addressed in this study including why? the topic is relevant, significant, and of interest.
• All equation numbering must be referenced in the related text. Moreover, the author needs to check all references to equations - for example see line 111 - and ensure clarity.
• The M&M section needs revision in two dedicated sections:
  • Preliminaries: where a comprehensive discussion is provided setting out all the preliminaries and process which apply to this study c/w all any assumptions and limitations.
  • Proposed Method (the author to provide an appropriate section heading): where all the details relating to the proposed method c/w all aspects of the process required to enable reproducibility to be evaluated.
• The current Results and Discussion sections are not logically structure and must be revised in two sections:
  • Results: where all the results c/w the analysis and conclusions are presented in detail with appropriate figures and related captions.
  • Discussion: where all the threads introduced in the study and all related research considered are introduced and considered.
  • Open Research Questions (ORQ) and Future Work: in all such studies there will be ORQ drawn from both the study and the related research. ORQ must be considered c/w directions for future research and proposed solutions.
    • ORQ would logically form a sub-section in the new discussion section.
• The current conclusion section needs revision limited to closing observations - all substantive aspects should be located in the discussion section.
• I missed a suitable consideration of the practical managerial significance (PMS) which is essential given the [at least in part] applied research aspect of this study. PMS must be considered including an illustrative scenario-based evaluation to show the potential utility of the proposal set out in this paper. 
  • In considering PMS I thought of the potential for the proposal to generalize to other domains of interest where keyboards are a feature. It would be interesting [and increase the potential audience for the study] is the potential for the proposal to be applied to other keyboards such as a piano?  The author may wish to consider this.
• Turning to the referencing, this paper is not intended to be a systematic review and on this basis the referencing is adequate for the study. However, in the References section all citations must include the DOI. 

In summary, I found this paper to be potentially interesting. However, the manuscript is flawed in that it lacks a suitably logical structure and narrative flow. Additionally, at only 12 pages including reference the manuscript is short [not always a bad thing] and requires revision and extension in what will in effect be a new manuscript which may then be re-reviewed and the proposal fully evaluated. Once suitably revised in my view the paper may be publishable and will be of interest to the intended audience. 

Author Response

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the "diff.pdf" file

The introduction needs major revision in three dedicated sections:
Introduction: setting out the background and motivation for the study, a brief overview of the proposal (avoid duplication), the claimed contribution c/w all limitations and assumptions, and the paper structure.Related Research: setting out the consideration of the related research considered  c/w a comprehensive analysis and also the research method adopted for the literature review.
Problem Statement: detailing the problem being addressed in this study including why? the topic is relevant, significant, and of interest.

The introduction has been revised accordingly.

All equation numbering must be referenced in the related text. Moreover, the author needs to check all references to equations - for example see line 111 - and ensure clarity.

It is not clear what the reviewer meant with this request. The referred example seem right and equations are all numbered and properly referred to when needed. Please add more details in case further editing is required.

The M&M section needs revision in two dedicated sections:
Preliminaries: where a comprehensive discussion is provided setting out all the preliminaries and process which apply to this study c/w all any assumptions and limitations.
Proposed Method (the author to provide an appropriate section heading): where all the details relating to the proposed method c/w all aspects of the process required to enable reproducibility to be evaluated.
The current Results and Discussion sections are not logically structure and must be revised in two sections:
Results: where all the results c/w the analysis and conclusions are presented in detail with appropriate figures and related captions.
Discussion: where all the threads introduced in the study and all related research considered are introduced and considered.

All this sections have been arranged to keep the proposed structure. 

Open Research Questions (ORQ) and Future Work: in all such studies there will be ORQ drawn from both the study and the related research. ORQ must be considered c/w directions for future research and proposed solutions.
ORQ would logically form a sub-section in the new discussion section.
The current conclusion section needs revision limited to closing observations - all substantive aspects should be located in the discussion section.
I missed a suitable consideration of the practical managerial significance (PMS) which is essential given the [at least in part] applied research aspect of this study. PMS must be considered including an illustrative scenario-based evaluation to show the potential utility of the proposal set out in this paper. 

 

ORQ and PMS have been added in the conclusions.

In considering PMS I thought of the potential for the proposal to generalize to other domains of interest where keyboards are a feature. It would be interesting [and increase the potential audience for the study] is the potential for the proposal to be applied to other keyboards such as a piano?  The author may wish to consider this.

 

No, we have not considered broadening our study to other keyboards, yet it sounds very interesting and worth pursuing. The study is interested in learning in general, and results can be compared for different tasks (as in the case of the reference dealing with violin lessons).

Turning to the referencing, this paper is not intended to be a systematic review and on this basis the referencing is adequate for the study. However, in the References section all citations must include the DOI. 

All DOIs were added.

 

In summary, I found this paper to be potentially interesting. However, the manuscript is flawed in that it lacks a suitably logical structure and narrative flow. Additionally, at only 12 pages including reference the manuscript is short [not always a bad thing] and requires revision and extension in what will in effect be a new manuscript which may then be re-reviewed and the proposal fully evaluated. Once suitably revised in my view the paper may be publishable and will be of interest to the intended audience.

I hope that the revision properly addresses the reviewer concerns.

Reviewer 2 Report

Comments and Suggestions for Authors

The paper complements a previous study published by the authors using the same database, aiming to further develop the previous analysis. In the earlier results, no significant involvement of the alpha rhythm was observed, which motivated the present study. However, the objectives and contributions of the paper could be described in greater detail in the Introduction.

The authors chose to use partial directed coherence as the functional connectivity feature. To broaden the scope of the analysis, it would be beneficial to include additional functional connectivity measures, even simpler ones such as cross-correlation, coherence, or mutual information.

Author Response

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the "diff.pdf" file

The paper complements a previous study published by the authors using the same database, aiming to further develop the previous analysis. In the earlier results, no significant involvement of the alpha rhythm was observed, which motivated the present study. However, the objectives and contributions of the paper could be described in greater detail in the Introduction.

Introduction was edited accordingly.

The authors chose to use partial directed coherence as the functional connectivity feature. To broaden the scope of the analysis, it would be beneficial to include additional functional connectivity measures, even simpler ones such as cross-correlation, coherence, or mutual information.

Cross-correlation and coherence do not provide directed information, hence their inclusion would not broaden the infromation already obtained through partial directed coherence, but overlap with it. In regards to mutual information, its interpretation still poses numerous methodological challenges, such as the impact of the choice of bandwidth for band-pass filters, the consequences of the application of the Hilbert transformation on signals with closely spaced frequency components, short-time and weak disturbances, and the difficulty to detect cross-frequency coupling. The addition of an analysis based on mutual information could be seen as future work (already mentioned in the revised conclusions) but it is out of our scope at this time.

Reviewer 3 Report

Comments and Suggestions for Authors

Dear Author,

I appreciate the opportunity to review the manuscript “Changes in Functional Connectivity of Electroencephalography while Learning to Touch-type”. This study re-analyzes an existing EEG dataset from a Colemak touch-typing learning experiment using Partial Directed Coherence (PDC) to quantify directed functional connectivity in alpha, beta and gamma frequency bands across three difficulty levels. The author argues that a PSD-based analysis in the original work failed to capture alpha involvement in learning and proposes that PDC reveals new insights into connectivity changes and network topology during skill acquisition.

The study is interesting and potentially relevant. However, there are substantial methodological and interpretive flaws that seriously limit the conclusions that can be drawn from this work. In its current form, I do not think the manuscript should be accepted. Below, I outlined the major concerns, followed by secondary issues.

Major Concerns

1.     Limited EEG montage

            The analysis relies on nine EEG electrodes (F3, Fz, F4, C3, Cz, C4, P3, POz, P4), which is an extremely sparse montage for any directed-connectivity method and especially insufficient for drawing conclusions about large-scale functional networks. With so few electrodes, the EEG cannot capture enough distinct neural sources, given that the signals are strongly shaped by volume conduction, where a single underlying source influences multiple electrodes. Further, the manuscript does not specify the referencing scheme (e.g., mastoids, average reference) and does not describe the use of any spatial filtering or other techniques to mitigate volume-conduction effects. PDC, like all Granger-based methods, is highly sensitive to instantaneous activity; without appropriate controls, the resulting directionality estimates can be distorted by reference choice and shared source contamination. To my understanding, this is a fundamental methodological limitation and must be explicitly acknowledged.

2.     Absence of artifact rejection

Although the Methods section specifies the sampling rate (256 Hz) and filtering parameters (0.1–50 Hz bandpass), it does not describe any procedures for artifact removal, such as eye-blink or ocular correction, EMG and muscle-artifact handling (which is especially important in a typing task), bad-channel detection, or re-referencing. Given that PDC measures are highly sensitive to noise and artifacts, this omission is a significant methodological limitation. PDC attempts to model temporal dependencies across electrodes; if those electrodes contain uncorrected EMG or movement contamination, the algorithm will simply model those artifacts as if they were neural interactions. This is particularly challenging in the beta and gamma ranges, where scalp EEG is easily overwhelmed by muscle activity. Without a rigorous artifact-cleaning pipeline, the observed connectivity patterns, especially in the gamma band, are more likely to reflect muscle activity than genuine neural communication.

3.     Conceptual confusion between power (ERD) and connectivity

The discussion interprets reduction in alpha-band connectivity as alpha-power desynchronization (ERD), borrowing conclusions directly from the ERD literature while linking decreased alpha to better performance. However, these are different measures. PDC reflects directed interactions between channels (a form of frequency-domain Granger causality), whereas ERD/ERS reflects local changes in spectral power. While both activities occur in the alpha band, they capture different aspects of neural activity. A drop in alpha connectivity between P3 and P4 is not the same as alpha ERD at those sites. Treating them as equivalent exaggerates the mechanistic implications of the findings. For the interpretation to be scientifically sound, the manuscript needs to clearly separate what PDC actually measures from what is being inferred by analogy to prior power-based work. Right now, that distinction is blurred.

4.     Use of gamma band without addressing EMG contamination

The manuscript attributes the observed gamma-band connectivity patterns to several neural processes, such as increased cognitive demand, a rise-and-fall trajectory with skill automatization, and even repetition suppression linked to memory encoding. However, the gamma band in this study is defined as 30–40 Hz and is measured during an overt motor task (typing) without any EMG or muscle-artifact correction. This frequency range is usually dominated by muscle activity. With no artifact correction, it is not possible to assume that the gamma-band PDC reflects genuine cortical gamma synchronization. The patterns could very easily be driven by muscle activity rather than neural processes. Without demonstrating that the gamma signal is clean or artifact-free, the conclusions drawn about gamma-band connectivity and its relationship to learning are on very shaky ground.

5.     Overinterpretation

The Discussion makes strong claims about large-scale brain networks. For example, increases in frontal-parietal alpha connectivity are interpreted as signs of DMN activation, reductions in parietal connectivity are presented as evidence that a “larger neural network” is engaged to support learning, and increases in frontocentral connectivity are tied to more efficient motor programming. In addition, channels with high in-degree are labeled as “information sinkholes” and linked to specific cognitive roles, for instance, suggesting that the right parietal electrode reflects upper-row typing demands, that Fz reflects “peak concentration,” or that C4 indexes “attentional suppression.”

These interpretations amount to strong reverse inference and are difficult to justify given the constraints of the dataset: a very sparse set of scalp electrodes, no source modeling, no clear artifact control, and no analyses that directly relate connectivity changes to behavior or cognitive performance. With this level of spatial resolution, it is far safer to interpret the findings as changes in directed interactions among scalp channels, patterns that may be consistent with certain functional ideas, but certainly cannot establish them.

6.     Statistical framework is underspecified

The manuscript explains that PDC values were kept only if they reached p < 0.05, that the first and last attempts were compared, and that the resulting connectivity differences were then filtered using a 5% cutoff. But several important details are missing. The manuscript never specifies how the PDC significance was tested. There is also no discussion of any correction for multiple comparisons, even though many frequencies, channel pairs, and task conditions are being tested. Additionally, the 5% threshold used to binarize the connectivity differences appears arbitrary, and it is not clear whether results would look different with a slightly higher or lower cutoff.

7.     Weak link between connectivity results and behavior

The manuscript briefly notes that participants improved their typing speed, for example, a 30% reduction in execution time during the easier lessons, but this behavioral information is never linked back to the connectivity findings in a meaningful way. There is no analysis showing whether changes in PDC relate to performance gains, whether participants who learned more quickly showed different connectivity patterns, or how much individual variability might be driving the group averages. Given the manuscript’s central claim related to learning, the argument would be much stronger if the connectivity measures were tested for their ability to predict or explain these behavioral improvements. As it stands, the behavioral results are mentioned only in passing, and the relationship between behavior and connectivity feels largely descriptive.

8.     Overstatements

The manuscript repeatedly suggests that the PDC analysis “proves” the involvement of alpha in learning or the method “brings light to understanding learning processes” more generally. Given the limitations mentioned above, these statements seem exaggerated. A more accurate and balanced framing would present this work as a methodological case study. It should not be suggested as a strong, general contribution to our understanding of the neural mechanisms of learning.

 

Secondary / Minor Issues

The manuscript contains numerous typos and formatting errors that should be addressed. The writing occasionally misuses terms, for example, using “phenomena” when “phenomenon” is intended, and includes non-idiomatic phrases such as “turned to be a limiting factor,” “associated to,” or “bring light,” which should be replaced with more standard scientific language. Some expressions, like “information sinkhole” or “peak concentration,” come across as informal metaphors rather than precise scientific descriptions. In addition, the figures are relied on heavily, yet the text provides no quantitative summaries, such as the number or proportion of connections that increase or decrease, and many captions repeat generic statements about arrow colors without highlighting the actual findings.

 

Comments for author File: Comments.pdf

Comments on the Quality of English Language

The manuscript contains numerous typos and formatting errors, and the writing quality varies considerably across sections. Several sentences are awkwardly phrased, which makes the narrative harder to follow and detracts from the clarity of the findings. A thorough language edit, focusing on grammar, style, consistency, and overall readability, would substantially strengthen the presentation and help ensure that the scientific contributions are communicated clearly.

Author Response

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the "diff.pdf" file

Major Concerns
1.     Limited EEG montage
            The analysis relies on nine EEG electrodes (F3, Fz, F4, C3, Cz, C4, P3, POz, P4), which is an extremely sparse montage for any directed-connectivity method and especially insufficient for drawing conclusions about large-scale functional networks. With so few electrodes, the EEG cannot capture enough distinct neural sources, given that the signals are strongly shaped by volume conduction, where a single underlying source influences multiple electrodes. Further, the manuscript does not specify the referencing scheme (e.g., mastoids, average reference) and does not describe the use of any spatial filtering or other techniques to mitigate volume-conduction effects. PDC, like all Granger-based methods, is highly sensitive to instantaneous activity; without appropriate controls, the resulting directionality estimates can be distorted by reference choice and shared source contamination. To my understanding, this is a fundamental methodological limitation and must be explicitly acknowledged.

 

We acknowledged the limitation that the sensor array poses to this study, hence the manuscript has been revised to show the potential of the functional connectivity analysis.

 

2.     Absence of artifact rejection
Although the Methods section specifies the sampling rate (256 Hz) and filtering parameters (0.1–50 Hz bandpass), it does not describe any procedures for artifact removal, such as eye-blink or ocular correction, EMG and muscle-artifact handling (which is especially important in a typing task), bad-channel detection, or re-referencing. Given that PDC measures are highly sensitive to noise and artifacts, this omission is a significant methodological limitation. PDC attempts to model temporal dependencies across electrodes; if those electrodes contain uncorrected EMG or movement contamination, the algorithm will simply model those artifacts as if they were neural interactions. This is particularly challenging in the beta and gamma ranges, where scalp EEG is easily overwhelmed by muscle activity. Without a rigorous artifact-cleaning pipeline, the observed connectivity patterns, especially in the gamma band, are more likely to reflect muscle activity than genuine neural communication.

 

Complete information about the artifact rejection that the system provides have been added.

3.     Conceptual confusion between power (ERD) and connectivity
The discussion interprets reduction in alpha-band connectivity as alpha-power desynchronization (ERD), borrowing conclusions directly from the ERD literature while linking decreased alpha to better performance. However, these are different measures. PDC reflects directed interactions between channels (a form of frequency-domain Granger causality), whereas ERD/ERS reflects local changes in spectral power. While both activities occur in the alpha band, they capture different aspects of neural activity. A drop in alpha connectivity between P3 and P4 is not the same as alpha ERD at those sites. Treating them as equivalent exaggerates the mechanistic implications of the findings. For the interpretation to be scientifically sound, the manuscript needs to clearly separate what PDC actually measures from what is being inferred by analogy to prior power-based work. Right now, that distinction is blurred.

 

The discussion, specially in regards to alpha band, has been clarified to make the disctintion clearer.

 

4.     Use of gamma band without addressing EMG contamination
The manuscript attributes the observed gamma-band connectivity patterns to several neural processes, such as increased cognitive demand, a rise-and-fall trajectory with skill automatization, and even repetition suppression linked to memory encoding. However, the gamma band in this study is defined as 30–40 Hz and is measured during an overt motor task (typing) without any EMG or muscle-artifact correction. This frequency range is usually dominated by muscle activity. With no artifact correction, it is not possible to assume that the gamma-band PDC reflects genuine cortical gamma synchronization. The patterns could very easily be driven by muscle activity rather than neural processes. Without demonstrating that the gamma signal is clean or artifact-free, the conclusions drawn about gamma-band connectivity and its relationship to learning are on very shaky ground.

 

Description of the EMG decontamination process has been included, hence I hope the reviewer can trust of the results.

5.     Overinterpretation
The Discussion makes strong claims about large-scale brain networks. For example, increases in frontal-parietal alpha connectivity are interpreted as signs of DMN activation, reductions in parietal connectivity are presented as evidence that a “larger neural network” is engaged to support learning, and increases in frontocentral connectivity are tied to more efficient motor programming. In addition, channels with high in-degree are labeled as “information sinkholes” and linked to specific cognitive roles, for instance, suggesting that the right parietal electrode reflects upper-row typing demands, that Fz reflects “peak concentration,” or that C4 indexes “attentional suppression.”

These interpretations amount to strong reverse inference and are difficult to justify given the constraints of the dataset: a very sparse set of scalp electrodes, no source modeling, no clear artifact control, and no analyses that directly relate connectivity changes to behavior or cognitive performance. With this level of spatial resolution, it is far safer to interpret the findings as changes in directed interactions among scalp channels, patterns that may be consistent with certain functional ideas, but certainly cannot establish them.

 

It is not the purpose of the paper to overinterpret the results, but to provide an example of how the tools of functional connectivity could aid in the understanding of underlying brain processes during learning. Our conclusions now state the limitations of the database and the need to use more channels and larger sample in future work.

6.     Statistical framework is underspecified
The manuscript explains that PDC values were kept only if they reached p < 0.05, that the first and last attempts were compared, and that the resulting connectivity differences were then filtered using a 5% cutoff. But several important details are missing. The manuscript never specifies how the PDC significance was tested. There is also no discussion of any correction for multiple comparisons, even though many frequencies, channel pairs, and task conditions are being tested. Additionally, the 5% threshold used to binarize the connectivity differences appears arbitrary, and it is not clear whether results would look different with a slightly higher or lower cutoff.

I dissagree: previous studies have confirmed that α=0.05 provides reasonable power for PDC confidence intervals and hypotesis testing under mild sample size. Hence, we strongly believe our significance tests are quite convincing. Please see DOI:10.1080/02664760701593065

7.     Weak link between connectivity results and behavior
The manuscript briefly notes that participants improved their typing speed, for example, a 30% reduction in execution time during the easier lessons, but this behavioral information is never linked back to the connectivity findings in a meaningful way. There is no analysis showing whether changes in PDC relate to performance gains, whether participants who learned more quickly showed different connectivity patterns, or how much individual variability might be driving the group averages. Given the manuscript’s central claim related to learning, the argument would be much stronger if the connectivity measures were tested for their ability to predict or explain these behavioral improvements. As it stands, the behavioral results are mentioned only in passing, and the relationship between behavior and connectivity feels largely descriptive.

At this point, and given the limitations of the database, it si not possible to pursue an analysis of individual variability and it has been listed as future work. For that purpose, it is defenitely needed a more dense array of sensors.

8.     Overstatements
The manuscript repeatedly suggests that the PDC analysis “proves” the involvement of alpha in learning or the method “brings light to understanding learning processes” more generally. Given the limitations mentioned above, these statements seem exaggerated. A more accurate and balanced framing would present this work as a methodological case study. It should not be suggested as a strong, general contribution to our understanding of the neural mechanisms of learning.

Exaggerated statements have been edited and a greater focus on the methodological framework is left.

Secondary / Minor Issues
The manuscript contains numerous typos and formatting errors that should be addressed. The writing occasionally misuses terms, for example, using “phenomena” when “phenomenon” is intended, and includes non-idiomatic phrases such as “turned to be a limiting factor,” “associated to,” or “bring light,” which should be replaced with more standard scientific language. Some expressions, like “information sinkhole” or “peak concentration,” come across as informal metaphors rather than precise scientific descriptions. In addition, the figures are relied on heavily, yet the text provides no quantitative summaries, such as the number or proportion of connections that increase or decrease, and many captions repeat generic statements about arrow colors without highlighting the actual findings.

Typos have been corrected. Descriptions on the figures have not been edited as the relevant description is alrteady in the specific discussion when a Figure is referred to. I prefer to keep captions simple and short.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

I will be brief:

• The introduction has not been suitably revised.
• The equation numbers are not referred to in the preceding text
• The open research questions and future work has not been suitably addressed
• The conclusion has many substantive matters which belong in the discussion section and/or related sections.

In summary, I can only conclude that the authors do not fully grasp the requirements for a manuscript to be logically structured with a clear narrative flow [the authors may disagree] and it is pointless to try to revise this manuscript and further reviews are clearly pointless.

Author Response

The author regrets that reviewer's concerns were not addressed in proper manner. Changes were made according to reviewer's vague comments in order to improve the manuscript as much as possible. The author definitely disagree with the structure the reviewer wishes to impose and final changes suggested by the editor in the discussion and conclusion might cover some of the concerns of the reviewer. In any case, the author thanks the reviewer for the time spent in the review. 

Reviewer 3 Report

Comments and Suggestions for Authors

Thank you for the opportunity to re-review this manuscript. The authors have comprehensively addressed my concerns, and I recommend acceptance of the manuscript in the current form.

Author Response

Thanks again for the comments provided.