Recommendation

Does typing on a smartphone involve the same neural mechanisms as writing by hand?

ORCID_LOGO based on reviews by Todd Richards and Dorothy Bishop
A recommendation of:
toto

Cerebral laterality as assessed by functional transcranial Doppler ultrasound in left-and right-handers: A comparison between handwriting and writing using a smartphone

Abstract
Keywords
Submission: posted 01 November 2022
Recommendation: posted 30 November 2023, validated 04 December 2023
Cite this recommendation as:
Schwarzkopf, D. (2023) Does typing on a smartphone involve the same neural mechanisms as writing by hand?. Peer Community in Registered Reports, . https://rr.peercommunityin.org/PCIRegisteredReports/articles/rec?id=332

Recommendation

Language production is associated with a distinct lateralised pattern of brain activation biased toward the left cerebral hemisphere. This also applies to writing. It has also been shown to be modulated by handedness, with less pronounced lateralisation in left-handers. However, in recent decades the use of handwriting has declined significantly while the use of smartphones has exploded. To date, no study has explored whether the same neural correlates of written language production found for handwriting also hold for typing on a smartphone.
 
In the current study, Samsouris et al. (2023) will use functional transcranial Doppler ultrasound (fTCD) to measure blood flow velocity within cerebral hemispheres to investigate this question. This technique is particularly suited for this purpose because it provides better control for the movement confounds associated with a writing task and the technical challenges of using a smart device than other neuroimaging techniques like fMRI or M/EEG. The authors hypothesise that there will be no difference in left cerebral lateralisation for handwriting and typing on a smartphone. They also expect to replicate previous findings of weaker lateralisation in left-handers in written language production when typing on a smartphone. To isolate the effect of written language production, both these conditions will be corrected for their corresponding motor component using control conditions without a linguistic component.
 
The Stage 1 manuscript was evaluated over 6 rounds of in-depth review by the recommender and two expert reviewers, before issuing in-principle acceptance.
 
URL to the preregistered Stage 1 protocol: https://osf.io/j7egz
 
Level of bias control achieved: Level 6. No part of the data or evidence that will be used to answer the research question yet exists and no part will be generated until after IPA.
 
List of eligible PCI RR-friendly journals:
 
References
 
1. Samsouris, C., Badcock, N. A., Vlachos, F., & Papadatou-Pastou, M. (2023). Cerebral laterality as assessed by functional transcranial Doppler ultrasound in left-and right-handers: A comparison between handwriting and writing using a smartphone. In principle acceptance of Version 7 by Peer Community in Registered Reports. https://osf.io/j7egz
Conflict of interest:
The recommender in charge of the evaluation of the article and the reviewers declared that they have no conflict of interest (as defined in the code of conduct of PCI) with the authors or with the content of the article.

Evaluation round #5

DOI or URL of the report: https://osf.io/zert6/?view_only=c607e7486fa64301a24605be27d06bda

Version of the report: 4th Revision

Author's Reply, 24 Nov 2023

Decision by ORCID_LOGO, posted 01 Oct 2023, validated 02 Oct 2023

Dear authors

Your manuscript has now been reviewed by one of the previous reviewers. Rather than spending further time on waiting or recruiting a further reviewer, I decided to go ahead now. The reviewer's comments merit another revision to address questions about the statistical sensitivity.

Your manuscript is close to receiving in-principle acceptance but please be aware that Stage 2 acceptance is contingent on sufficient reliability being observed. While RRs are never rejected based on their main results, but they can be rejected if prospectively defined quality checks or positive controls fail to such an extent that the research question can’t be answered. You may formally achieve this by setting the reliability test as a positive control at Stage 1 and passing it a condition at Stage 2.

It could be useful to require a Bayes Factor Design Analysis to establish the probability of observing results in favour of H1 vs H0 (see e.g. https://link.springer.com/article/10.3758/s13423-017-1230-y). BFDAs are very useful for establishing the futility or otherwise of a Bayesian design.

Best wishes
Sam

Reviewed by ORCID_LOGO, 19 Sep 2023

Thanks to the authors for their considered responses to my comments.

I have to say that I remain unconvinced that this study is of any theoretical interest, but I am also aware that subjective judgements about importance of a research question are not regarded as a valid criterion of evaluation. The more applied point, that smartphone typing could substitute for handwriting in remote studies is reasonable, but it seems a bit of a stretch to then argue one could do studies using remote EEG, when this has not been demonstrated to be valid for assessment of laterality. Laterality assessment with EEG is complicated by choice of reference electrode.  

 

It is good to hear that the authors are already engaged in a study using typewriting. This needs to be referenced in the current paper, I think, as it provides important context. I would like to see a brief account of task analysis for the smartphone typing vs other tasks, including handwriting and typing.

 

The authors deal with the issue of participant unfamiliarity with single-thumb typing by using a pretest. This is a good solution. They also suggest restricting handedness to those with a clear preference and excluding those in the middle of the handedness distribution – this is in the spirit of what I suggested, though it should be noted that Mazoyer et al argued it was only extreme left-handers (scoring -100) who had high rates of atypical laterality – and these are in the minority. Unfortunately, both of these restrictions of participants will make it more difficult to recruit them.

 

If I’ve understood the method correctly, they will be using split half reliability on a task with 10 trials, so will be correlating LIs from odd and even trial means based on 5 trials. These are very unlikely to be reliable.  If you then subtract a control task, the reliability will fall even further.  

 

I’m not experienced with using Bayesian analyses, though I appreciate that they can be designed to distinguish between results that are inconclusive rather than null or positive. My concern is that, given the weak association between handedness and laterality, and the likely low reliability of the motor-corrected LIs obtained in the study, the result will be inconclusive, even if the maximum sample size is attained.

 

I think the authors should be allowed to go ahead and do this study if they want, but I see it as very high risk of being inconclusive, and involving a great deal of work, with major issues around recruitment, for an uninterpretable result. I would be very glad to be proved wrong, but would strongly advise the authors to do some simulations of possible datasets before putting a great deal of time into this study.

 

 

Minor points

Abstract: I always recommend including sample size in the Abstract, as it is an important detail when evaluating a study.

P 2 para 2 ‘laying down’ -> ‘lying down’

P 2 para 2, Altzheimer’s disease -> Alzheimer’s disease

P 3, 3rd sentence would benefit from some rewording, e.g. “In FTCD, ultrasound probes are positioned over the temporal windows of the participants on either side of the head to measure the blood flow velocity in the middle cerebral arteries (MCAs). “

Also ‘temporal window’ won’t be familiar to many readers, so need to explain it is a thin region in the temporal bone.

 

P 6, para 2 : “ absence of a motor component in oral language” ; this will seem odd to readers, as there is a speech motor component in oral language. It was unclear whether this referred to the silent nature of the word generation task, or just to the lack of a manual motor component in speech.

 

P 8, point 2: “total number of written words” (rather than “amount”)

 

P 8, exclusion criterion 2: “Consumption of medication with the potential to affect the central nervous system in the last six months” – I wonder if the authors are making life more difficult for themselves than they need. Given the very high rates of consumption of psychotropic drugs, both licit and illicit, this could rule out a lot of people. And some would include caffeine, painkillers and nicotine in such a definition! I’m not sure that such an exclusion is required, and if it is, I’d be more specific, and take a much shorter time scale unless one was talking about a drug that was known to have a long-term effect.

 

Immediately below Fig 1: change to “which we will provide”

 

P 13. “If the correlation is below 60%”. Does this mean r = .6?

 


Evaluation round #4

DOI or URL of the report: https://osf.io/gs5ch?view_only=65f12bc394dc4659a387e0dc71da8bc7

Version of the report: 3rd Revision

Author's Reply, 08 Sep 2023

Decision by ORCID_LOGO, posted 12 Feb 2023, validated 12 Feb 2023

Dear authors

First, let me apologise for the delay in getting back to you. With the December closure and summer in southern hemisphere, this time of year is usually particularly slow.

We have now had comments on your manuscript from two expert reviewers. As you can see, both reviewers were a bit confused by the fact that there have been several rounds of triage to meet RR criteria, which is partly independent from the scientific review. I will discuss this with the PCI:RR management to see if this can be clarified more in future submissions.

Either way, you see detailed and constructive comments from the reviewers that should be addressed before we can recommend in-principle acceptance of Stage 1 and data collection can commence. One reviewer raises a good point about the use of thumbs when typing with a smartphone. Thumbs were probably the primary mode for older pre-smartphone devices. However, when using modern touchscreen devices this may not be that common. Just checking how I type on my own phone, I realised that I use the index finger, and I usually rely a lot on the swype functionality rather than tapping individual letters. Of course, that's just me and this does not invalidate your research question - but I agree with the reviewer that this deserves more thought. It would probably be worth screening for participants who use the thumb of their dominant hand to type? Or alternatively the reviewer also makes suggestions.

The reviewer also has concerns about the prediction of a null result. I realise that your use of Bayes Factors will address this to some extent. However, the reviewer is correct that the sample size may be insufficient. Please consider and report the likelihood of obtaining conclusive results with your minimal and maximal sample size. I also agree with the reviewer that it would be important to support this null prediction with a positive control or outcome-neutral sanity check (e.g. the reliability test the reviewer proposes).

If you choose to submit a revision, please include a version with changes tracked/highlighted as this will expedite the review. Also please include a detailed response letter addressing each comment.

Best regards
Sam Schwarzkopf

Reviewed by , 21 Dec 2022

I can see that this proposal has already been through several rounds of review and meaningful adjustment. My only suggestion would be to add a hypothesis and study goal regarding the correlation/relationship between the behavioral handwriting skill/scores and the ultrasonography signal using your equipment. There is a lot of evidence that behavioral language skills do correlate with functional brain imaging as shown in these articles

https://pubmed.ncbi.nlm.nih.gov/29923097/

https://www.frontiersin.org/articles/10.3389/fnhum.2012.00119/full#:~:text=Notably%2C%20some%20high%20correlations%20between%20fMRI%20brain%20activations,uncontrolled%20underlying%20factors%2C%20such%20as%20age%20%28Lazic%2C%202010%29.

There is already a plan in place to study handwriting skills and obtain scores which could be used in this brain-behavior correlational analysis.

 

Reviewed by ORCID_LOGO, 12 Feb 2023


Evaluation round #3

DOI or URL of the report: https://osf.io/zert6/?view_only=ecc03e1a965b4d70bd1f7883fc14f709

Version of the report: 2nd Revision

Author's Reply, 29 Nov 2022

Decision by ORCID_LOGO, posted 23 Nov 2022, validated 24 Nov 2022

Dear authors

I am very sorry but I'm afraid I still need to ask you to make some further changes and/or clarify a few parts. I know this must be tedious but it is important that we have the manuscript ready for reviewers who will likely be unfamiliar with the processes and guidelines of PCI:RR. If you can address all these points quickly I expect we can invite external reviewers. At this stage I would also suggest not having tracked changes but simply refer to where the changes are in your response letter - it would be ideal to have a clean version of the manuscript for reviewers (or alternatively upload a clean and a tracked version - but the main weblink will need to be to the clean version that we will send to reviewers).

Direct link to the manuscript

Thanks for updating the link. However, this now points to the overall project rather than the latest version of the manuscript. This is better than before but I'm willing to bet it will still be confusing for many reviewers. Please use the direct link to the manuscript. For the present version this would be:
https://osf.io/57w2s/?view_only=ecc03e1a965b4d70bd1f7883fc14f709

However, for a further revision it will be yet another address of course.

Design table

Thanks for the modifications. Condensing the text there certainly helped with clarity (and it will be even clearer once tracked changes have all been accepted). However, there now seems to be an error in your interpretation of Hypothesis 2 - unless I'm getting very confused. Hypothesis 2 compares LI_typing_corrected between left- and right-handers, using an indep. sample t-test. In the Interpretation column, you interpret a null effect (no difference between left- and right-handers) as evidence that the lateralisation is unrelated to the medium - but isn't the exact opposite the case? If there is no difference between left- and right-handers for typing even though you'd expect to find this for handwriting, that would indicate that lateralisation does depend on the medium?

Moreover, in this context I wonder if you shouldn't be testing the interaction contrast here - that is, whether the Right-Lefthander_handwriting is greater than Right-Lefthander_typing. However, this is probably something to debate with your reviewers from this field.

I must concede that I find the whole concept generally quite confusing (e.g. talking about the absence of evidence for a reduction in the left/right laterality between left and right handers...). I suspect Doppler ultrasound and handedness experts will be familiar with this but I wouldn't necessarily count on that - however, I'll also let reviewers be the judge of that.

General changes required

You should have also received a generic list of required changes to the manuscript. As far as I can tell some of these have not yet been incorporated:

1) Data must be available to readers after recommendation, either in the text or through an open data repository such as Zenodo (free), Dryad (pay) or some other institutional repository. Data must be reusable, thus metadata or accompanying text must carefully describe the data;
Please add a data availability statement.

2) Details on quantitative analyses (e.g., data treatment and statistical scripts in R, bioinformatic pipeline scripts, etc.) and details concerning simulations (scripts, code) must be available to readers in the text, as appendices, or through an open data repository, such as Zenodo, Dryad or some other institutional repository. The scripts or code must be carefully described so that they can be reused;
Include this also in your availability statement.

3) Details on experimental procedures must be available to readers in the text or as appendices;
As for as I can tell this is all fine, so there shouldn't be anything to add - reviewers may of course have further questions about the methods though.

4) Authors must have no financial conflict of interest relating to the article. The article must contain a "Conflict of interest disclosure" paragraph before the reference section containing this sentence: "The authors of this article declare that they have no financial conflict of interest with the content of this article.";
Please include the COI disclosure before the reference section.

5) This disclosure has to be completed by a sentence indicating that some of the authors are PCI recommenders: “X is a recommender at PCI Registered Reports.”
Please include if this applies to you (since you submitted your manuscript anonymously I cannot ascertain if it does).

Best regards,
Sam Schwarzkopf


Evaluation round #2

DOI or URL of the report: https://osf.io/dvp4a?view_only=d7aaee6e65bb48d0b2c1fd95bd3d2f36

Version of the report: 1st Revision

Author's Reply, 23 Nov 2022

Decision by ORCID_LOGO, posted 20 Nov 2022, validated 21 Nov 2022

Dear authors

Thanks for your resubmission. I believe this is almost ready for being sent out to peer review. However, there are still a few things to fix and consider before that:

1. Design table

I appreciate that you kept in the table due to the requirements for authors. (My apologies for this oversight; I'm familiar enough with the recommender guidelines but not the author guidelines it seems!). I personally prefer flowcharts for this purpose and for relatively straightforward designs I don't think this adds very much, but obviously you should follow the requirements. However, the table as it stands in the current version seems to contain errors:

  • Under Hypothesis 1 Analysis Plan and Rationale, you still mention a prior scaled with r=0.2 instead of the new r=0.707.
  • Under Hypothesis 2 Question, Hypothesis and Analysis Plan, the table still mentions handwriting and typing as separate t-tests, even though you removed the former.

  • In your response letter you said you changed the orientation of the table (which I agree would be a good idea) but it is still the same.

  • Obviously you could also keep this orientation, but then I'd definitely suggest further condensing the text (and this seems wise to me in either case):

    The sentences in the table make sense as part of the manuscript text but to make the table easily digestible you could simplify this much more. Is it necessary to say "We will perform a Bayesian samples t-test..." or would it suffice to say "Bayesian dependent samples t-test..."? Similarly, I don't think we need statements like "This choice is justified because...".

    Or under Hypothesis the phrasing "There will be an absence of difference..." sounds like a bit of a double (or even triple?) negative. I appreciate that Bayesian tests for the null lend themselves to this sort of awkwardness. But in the interest of clarity you could simply say "There is no difference..." or even more directly "Cerebral laterality during handwriting is not different from writing with a smartphone" (the fact that your outcome measure is the linguistic component after motion correction is established in the methods and so I don't think this even needs to be mentioned here).

    Obviously, these last few points are mere suggestions. You can write your manuscript however you like but I believe it is advisable to make it clearer for reviewers/readers.

2. Link to your preregistered manuscript

Please ensure that the link in the system points to the latest version of the manuscript (ideally one with the last round of tracked changes). The current link points to the original version of the manuscript. It took me some time of clicking through to find the latest version. I am fairly familiar with using the OSF platform but I am sure many reviewers won't be. It also wouldn't be obvious to anyone coming in at this stage that this isn't the final version.

3. Bayes Factor criterion

I appreciate that you have updated the prior definition for your Hypothesis 1 test. The justification makes sense and all this is fine as it is. However, please note that a criterion of BF10=3 and 1/3 is insufficient for the evidence threshold set by many RR-friendly journals. Of course there are still many RR-friendly journals that do not have a strict minimum and you could still submit there when the Stage 2 manuscript is accepted (and in fact you don't need to submit to any journal). If you do want to meet those stricter thresholds, you probably need to rethink the sample sizes.

Best regards
Sam Schwarzkopf


Evaluation round #1

DOI or URL of the report: https://osf.io/dvp4a?view_only=d7aaee6e65bb48d0b2c1fd95bd3d2f36

Version of the report: 1

Author's Reply, 17 Nov 2022

Decision by ORCID_LOGO, posted 06 Nov 2022, validated 06 Nov 2022

Dear authors

Thank you for your Stage 1 submission. You propose an interesting study and I am curious to see the results. However, we regularly triage manuscripts before sending them out to reviewers to ensure RR criteria are met from the outset. You have done a commendable job in preparing your manuscript, but there are nevertheless a few points I'd suggest addressing before review by experts in your field:

1. Statistical approach / Power analysis

You propose to use a default Bayes Factor with a Cauchy prior scaled with r=0.2 for Hypothesis 1. This hypothesis is set to confirm the null. By using a small scaling factor you are concentrating probability density for the alternative hypothesis near small values and this may very well make sense for what you are testing here - but it also limits your ability to actually confirm the null. It is important to ensure sufficient sensitivity to reliably detect hypothesised effects - or the lack thereof.

Using the original Rouder implementation of the Bayesian t-test with your minimum sample size of 32 your BF10 would be 0.4891 even for a perfect null effect (d=0). Even with your maximum sample size of 48 it is still 0.4332, nowhere near your criterion of 1/6 of confirming the null. While this calculation may be different in whatever implementation of Bayesian inference you're using (e.g. my understanding is that JASP has undergone some changes since Rouder's paper was published), this is the kind of information that is required to ensure adequate power.

If experiment is not sensitive enough to obtain conclusive results, you either need to adjust the prior or increase your planned sample sizes.

2. Hypothesis 2

This hypothesis strikes me as a "main effect" of handedness. This should be a 1-df test that ignores the writing mode. Rather than two independent t-tests you effectively only need one for the sum (or average) of handwriting and typing. However, since you are interested in whether or not left-handers show weaker lateralisation than right-handers, isn't the crucial contrast here an interaction? Specifically you want to test whether the difference between left- and right-handers is weaker for typing than handwriting, and you hypothesise that this is a null effect. If this is incorrect, please clarify.

3. Exclusion criteria

Please provide more detail/information about how exclusions/rejections of participants will work.How will you detemine "Inadequate ultrasound penetration of the skull"? A similar question applies to the second criterion about noisy data: what determines whether epochs are "accepted"?

4. Hypothesis table

This is perhaps a matter of taste. Certainly any journal you may eventually submit this to will probably want to copyedit this in any case. However, I found that table very difficult to digest. A lot of text is awkwardly squeezed into narrow columns. Words are broken across lines. Generally it is very hard to understand. I realise that such tables are often used in RR manuscripts but I would reduce the amount of text here considerably. Move most of the prose into the manuscript itself (where it should be read) and keep this to the essentials and/or abbreviated statements. Instead of a table, it might perhaps in fact be better to have a box diagram outlining the various possible outcomes? This will make it a lot easier for reaers (and reviewers!) to understand. [Note added by Managing Board: please keep the design table in the manuscript as it is a required component, but abbreviate it as much as possible to improve readability. The idea of adding a box diagram or flowchart is a good idea]

5. Quantifying handedness with the pegboard

You have a thorough design for quantifying participants' handedness. However, the way you calculate the handedness index from the pegboard task is different from the other two measures (Edinburgh Handedness Inventory and the Hand Preference task), and even of the opposite sign. This is definitely a matter of taste, but for clarity I wonder if you could consider changing how the pegboard index is calculated to make it directly comparable to the other two tasks (at least by changing the sign so (LH-RH)+(LH+RH)). This would make it easier for readers to understand and might also reduce potential for confusions and errors for experimenters. But as I said, this is just a suggestion.

User comments

No user comments yet