The interplay of music, movement, and culture on rhythm processing
Culture-Driven Neural Plasticity and Imprints of Body-Movement Pace on Musical Rhythm Processing
Abstract
Recommendation: posted 18 June 2024, validated 19 June 2024
Leongómez, J. (2024) The interplay of music, movement, and culture on rhythm processing. Peer Community in Registered Reports, . https://rr.peercommunityin.org/PCIRegisteredReports/articles/rec?id=646
Recommendation
Level of bias control achieved: Level 6. Data collection commenced during the later part of Stage 1 peer review; however, since no changes to the design were made after this point, the risk of bias due to prior data observation remains zero and the manuscript therefore qualifies for Level 6.
List of eligible PCI RR-friendly journals:
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.
Reviewed by anonymous reviewer 1, 07 Jun 2024
I have no further suggestions to improve this study, though I do still have some concern that the sample size will be too small.
Evaluation round #2
DOI or URL of the report: https://zenodo.org/doi/10.5281/zenodo.10221480
Version of the report: 2.1
Author's Reply, 06 Jun 2024
Decision by Juan David Leongómez, posted 23 May 2024, validated 23 May 2024
Dear Dr. Guérin,
Thank you for the revised version of your RR. As you will see, the reviewers made some important comments that need careful attention to achieve Stage 1 in-principle acceptance. In particular:
Reviewer #1 makes several points, including a critical concern about the potentially overestimated effect sizes. This could affect the statistical power, error rate, and precision of your results. This point must be addressed for the study to have reliable results, especially considering that, as noted in your response, “registered reports necessitate a careful justification of the required sample size for each research hypothesis”.
Reviewer #2, Dr. Anne Keitel, remains concerned that the study might measure an auditory effect rather than a movement effect. However, considering time limitations, she is okay with highlighting this conceptual issue in the discussion sections.
Please carefully consider the points raised by the reviewers.
Regarding my own comments, I think keeping mixed-model ANOVAs as the main statistical tests but adding linear mixed models (LMMs) as complementary analyses is a good alternative. However, I want to point out that, while it is true that estimating power for linear mixed models may be difficult without previous data, simulations offer a great alternative (you obviously do not need to do this given that ANOVAs will be the main statistical tests, but I want to mention at least one example included in previous reviews for PCI-RR that could be relevant: the round #2 review by Lisa DeBruine for How does perceptual and contextual information influence the recognition of faces?. The actual review and code for the simulation are available in Rmd and HTML formats here).
I look forward to receiving your revised submission along with a detailed response to the reviewers' comments.
Reviewed by anonymous reviewer 1, 28 Mar 2024
Reviewed by Anne Keitel, 16 May 2024
I would like to thank the authors for their clarifications and thorough responses to my questions.
Most of my comments have been addressed, but I am still worried about the possibility that the authors measure an auditory effect, rather than a movement effect. The senior author’s own seminal study on auditory imagery (Nozaradan et al., 2011) provides evidence for the argument that, even without practicing movements, participants can imagine a certain beat during listening, which leads to peaks in the EEG power spectrum. The worst-case scenario is that movement does not add anything, and the measured effect is purely based on participants’ auditory exposure to the imposed beat. There might not be a “multisensory” or “holistic” effect (page 10, lines 238,241), or even additive effect of movement, there might simply be none. This is very pessimistic thinking, but it would be good to provide any empirical evidence (or indication) that movement indeed might have an effect on subsequent “shaping of auditory information” within this research context.
I absolutely understand that it is not feasible to double the participant numbers, but a small pilot study would have helped to alleviate the worry that the main aim of the research (“to capture direct neuroscientific evidence for the shaping of auditory information by the pace of previous movement”) cannot be addressed unambiguously. However, I don’t want to stand in the way of this interesting project to get started, and so I’m okay with the solution that this conceptual issue will be highlighted in the respective discussion sections. All the best for your study!
Evaluation round #1
DOI or URL of the report: https://zenodo.org/doi/10.5281/zenodo.10221480
Version of the report: 1.2
Author's Reply, 15 Mar 2024
Decision by Juan David Leongómez, posted 13 Feb 2024, validated 14 Feb 2024
Dear Dr. Guérin,
My sincere apologies for the delay in providing feedback on your RR. Securing reviewers proved challenging, especially towards the year's end. However, once suitable individuals were found, their evaluations were detailed and promptly completed. I believe these reviews will greatly assist you in revising this Stage 1 plan.
Although both reviewers provided substantial recommendations for enhancement, their overall assessment of the proposed studies is positive. Reviewer #1 highlights several important concerns, including acknowledging the concept of active sensing, considering recent evidence on metre perception, clarifying the state of knowledge on brain oscillatory interactions, distinguishing between related theories, reconsidering movement and brain activity as the gold standard for measuring beat perception, refining participant exclusion criteria, addressing potential confounds related to metronome use, clarifying technical details, and considering Bayesian ANOVAs to address potential issues with sample size and interpretation of null findings.
Dr. Anne Keitel, the second reviewer, suggests adding a control condition without movement to rule out alternative interpretations, replacing pure tone stimuli with real drum sequences for better ecological validity, providing clearer information on participant samples and stimuli in abstracts and introductions, clarifying inclusion criteria regarding cultural exposure, considering self-report questionnaires for music exposure, and justifying the chosen alpha value of .02.
My own reading was mostly concentrated on the RR and statistical analyses because I am not a specialist in the topical area. The primary issues that I would like you to consider are outlined below:
- Most hypotheses are proposed to be tested by modelling the experiment as either pairwise t-tests or Mixed-model ANOVA. However, this approach, which only models fixed effects, has some limitations. I believe fitting mixed models (Bolker, 2015) that include random effects, such as participants, may be preferable, as it would allow for the generalisation of results to a wider population of participants (Barr et al., 2013; DeBruine & Barr, 2021). Linear mixed models could be fitted with random intercepts per participant, or even random intercepts and random slopes between sessions for each participant. To implement this in R, you could use packages such as brms (Bürkner, 2017) if you decide to follow Reviewer #1's suggestion of using a Bayesian approach. Alternatively, packages such as lme4 (for general or generalised models; Bates et al., 2015) or lmerTest (only general —normal/Gaussian— models; Kuznetsova et al., 2017) could be used if you decide to maintain a frequentist approach. Regardless of the chosen approach, whether frequentist or Bayesian, post-hoc comparisons can be tested using packages like emmeans (Lenth, 2024).
For example, for the Movement Condition × Session models, you could use one of these options:
# For a frequentist approach
## Load package
library(lmerTest)
## Fit model
model <- lmer(variable ~ Movement_Condition*Session + (1 + Session | Participant), data = XXX)
## See model results in ANOVA-type table
anova(model)
## Plot the estimated marginal means
ggplot(data.frame(emmeans::emmeans(model, pairwise~Session|Movement_Condition),
aes(x = Movement_Condition, emmean, color=Session)) +
geom_point(position = position_dodge(width = 0.5)) +
geom_errorbar(aes(ymin=emmean-SE, ymax=emmean+SE), width=0.4, position = position_dodge(width = 0.5))
# For a Bayesian approach (without specifying priors)
## Load package
library(brms)
## Fit model
model <- brm(variable ~ Movement_Condition*Session + (1 + Session | Participant), data = XXX, family = gaussian())
## See model results
summary(model)
## Plot the estimated marginal means
plot(marginal_effects(model))
# For post-hoc comparisons between sessions (regarless of whether you used lme4, lmertest or brms to fit the models)
## Load package
library(emmeans)
## Contrast between sessions
emmeans(model, pairwise~Session)
## Contrast between sessions by movement condition
emmeans(model, pairwise~Session | Movement_Condition)
### Note: Please keep in mind that, by defect, emmeans uses Tukey to adjust p-values. If you prefer to use Bonferroni correction, simply add the parameter 'adjust = "bonferroni"'.
- On page 27 (line 740) you state that “Normality will be checked using the Shapiro–Wilk test” and that “if violated, the data will be normalised using a transformation that will be contingent on data distribution curves”. I want to point out that when fitting general models, such as an ANOVA or a linear mixed model, the assumption of normality pertains to the residuals of the model, not necessarily the dependent variable itself. For this reason, I believe this approach may not be the most appropriate.
To test the normality of the residuals (and other assumptions of a model fit) in R, packages such as performance (Lüdecke et al., 2021), have very convenient functions, including check_model().
# Load package
library(performance)
# Check normality of residuals
check_normality(model)
# Visual check of model assumptions
check_model(model)
If the assumption of normality of residuals is violated, there are several options. First, you could use a function such as check_distribution(model) (also from the package performance), to get some idea of alternative distributions to then fit a generalised linear model. Alternatively, and perhaps preferably, you can use bootstrap techniques. This is useful because (1) bootstrapping the models could really help in dealing with issues such a (potential) non-normal residual distribution (see e.g., J. Fox, 2016 Chapter 21), but also because (2) having CIs would help in assessing effects (even in the absence of p values).
Minor points
- As mentioned by Dr. Keitel, I wonder what the rationale was behind setting an alpha of 0.02. To clarify, I am not against this decision, but I believe it should be explained in the text.
- In a few places, you mention that some computations were performed using RStudio (e.g., lines 440 and 724). However, you should cite R instead, as it is the software that performs all computations. RStudio is a very useful IDE that aids in interacting with R, but while all your code can be run in R (regardless of whether you are using RStudio or a different IDE or interface, or even just directly from a command line), your code cannot be executed in RStudio without R.
- On page 18 (line 474), you cite Figure 2, Panel A. However, unlike Fig. 1, Fig. 2 does not include panel labels, so perhaps these are to be added if you want to refer to a specific panel.
References
- Barr, D. J., Levy, R., Scheepers, C., & Tily, H. J. (2013). Random effects structure for confirmatory hypothesis testing: Keep it maximal. Journal of Memory and Language, 68(3), 255–278. https://doi.org/10.1016/j.jml.2012.11.001
- Bates, D., Mächler, M., Bolker, B., & Walker, S. (2015). Fitting linear mixed-effects models using lme4. Journal of Statistical Software, 67(1), 1–48. https://doi.org/10.18637/jss.v067.i01
- Bolker, B. M. (2015). Linear and generalized linear mixed models. In G. A. Fox, S. Negrete-Yankelevich, & V. J. Sosa (Eds.), Ecological Statistics: Contemporary theory and application (pp. 309–333). Oxford University Press. https://doi.org/10.1093/acprof:oso/9780199672547.003.0014
- Bürkner, P.-C. (2017). brms: An R Package for Bayesian Multilevel Models Using Stan. Journal of Statistical Software, 80, 1–28. https://doi.org/10.18637/jss.v080.i01
- Bürkner, P.-C. (2018). Advanced Bayesian Multilevel Modeling with the R Package brms. The R Journal, 10(1), 395–411. https://doi.org/10.32614/RJ-2018-017
- Bürkner, P.-C. (2021). Bayesian Item Response Modeling in R with brms and Stan. Journal of Statistical Software, 100, 1–54. https://doi.org/10.18637/jss.v100.i05
- DeBruine, L. M., & Barr, D. J. (2021). Understanding Mixed-Effects Models Through Data Simulation. Advances in Methods and Practices in Psychological Science, 4(1), 2515245920965119. https://doi.org/10.1177/2515245920965119
- Fox, J. (2016). Bootstrapping Regression Models. In Applied Regression Analysis and Generalized Linear Models (3rd ed., pp. 587–606). Sage. https://us.sagepub.com/sites/default/files/upm-binaries/68018_Fox_Chapter_21.pdf
- Kurz, A. S. (2019). Statistical Rethinking with brms, ggplot2, and the tidyverse (version 1.0.1). https://bookdown.org/ajkurz/Statistical_Rethinking_recoded/
- Kuznetsova, A., Brockhoff, P. B., & Christensen, R. H. B. (2017). lmerTest Package: Tests in Linear Mixed Effects Models. Journal of Statistical Software, 82(13), 1–26. https://doi.org/10.18637/jss.v082.i13
- Lenth, R. (2024). emmeans: Estimated Marginal Means, aka Least-Squares Means (R package version 1.10.0) [Computer software]. https://CRAN.R-project.org/package=emmeans
- Lüdecke, D., Ben-Shachar, M., Patil, I., Waggoner, P., & Makowski, D. (2021). performance: An R Package for Assessment, Comparison and Testing of Statistical Models. Journal of Open Source Software, 6(60), 3139. https://doi.org/10.21105/joss.03139
Reviewed by anonymous reviewer 1, 20 Dec 2023
This registered report proposes to test whether movement can influence the recorded brain activity to a subsequent rhythm presented without any concurrent movement. The underlying assumption is that such pre-stimulus movement and the associated beat perception will modulate the brain activity (and clapping) to the rhythm alone and result in beat-related brain activity. Furthermore, an interaction between the movement type and the participants’ cultural background will be tested for. I have several suggestions below to improve the paper and the study itself.
In the Introduction, the idea of active sensing is introduced. This idea goes back at least to J.J. Gibson and the ecological approach to psychology and should therefore be acknowledged. This includes rhythmic sensing because Gibson talked about how animals locomote (which is rhythmic) and thereby sense their environment below their feet and with their visual system. Furthermore, when you say “the way movement might shape perception is less straightforward”, I’m not sure I agree with this because as we move in our environment we hear our own footsteps and people may even use echolocation, which has been studied in various animals, most notably bats of course.
Line 132 “typically takes the form of a metre” Recent evidence from Nave-Blodgett et al., JEP:General2021a and JEP:HPP2021b shows that the extent Westerners actually perceive metre (rather than just a beat) is possibly over-stated, even for highly trained musicians, and definitely for children. How this is for people in other cultures is still unknown. This passage therefore could be re-phrased as “typically takes the form of a beat and sometimes multiple nested beats, or a metre”, perhaps with some acknowledgement for the cultural limitation of our knowledge on metre perception at this point in time. Subsequently, when referring to “metre perception”, it might be appropriate sometimes to refer to “beat and metre perception” so as not to presuppose that people are always perceiving the multiple levels of a metre. It might also be useful to distinguish being able to distinguish one meter from another based on perceiving a single beat, as in the case of telling whether music is in 3/4 or 6/8 meter vs. actually perceiving the metre, which requires perceiving multiple beat levels.
Line 165 “Oscillatory interactions between the auditory and motor areas of the brain would be crucial for metre perception to arise”. This implies that there is strong evidence for this, whereas my reading of the literature is that there is only circumstantial evidence. Certainly, motor areas have been shown to be active during rhythm/beat perception tasks and there is functional connectivity evidence, but of course this does not imply actual connectivity or interaction and whether any interaction is indeed oscillatory in nature as the cited model/theory suggests is in need of empirical evidence. So please make it clearer what the state of knowledge is on this topic.
Line 168 This “active sensing framework” sounds pretty similar to Dynamic attending. Are they the same or can your draw any important distinctions? Also, here you talk about “motor delta oscillations”, attention, and presumably there is some auditory processing of those events you mention. But is the motor oscillation really a different thing than the attention or just re-descriptions of each other on brain vs. behavior levels, e.g., see the Premotor theory of attention (https://doi.org/10.1016/B978-012375731-9/50035-5)? As for ASAP, is the possibility that “cortical motor planning regions would thus be entrained” any different than the just mentioned motor oscillations or attention entrainment? These theories are not sounding very distinct from each other, which is not necessarily your fault but this fact should be acknowledged or clearer distinctions between the theories should be explained. You go part of the way there when you say “each of them presupposes a strong role of motor production”, but I think the similarities are greater than that, at least as you have described them above.
Line 192 “only constitute an indirect way to capture the internal representation” This seems to imply that movement or movement-related brain activity is the gold standard/most direct way for measuring internal representation of beat and meter. I strongly disagree and would argue that there is no gold standard. In the case of movement, many things affect it besides perceived beat/meter, e.g., various kinematic constraints and prior learning unrelated to beat perception. In the case of brain activity, both auditory and motor activity are mostly driven by physical properties of sound and movement (plus a lot of noise), respectively, rather than being pure measures of perceived beat. One might argue that at least in somewhat musically sophisticated adult humans, the most direct way to measure beat perception is to ask them what the beat is, e.g., name the time signature or use a probe beat as in beat alignment tasks, but I would still not call this anything like a gold standard.
Line 303 “12-element rhythms to a three-beat metres” Is there any reason to believe this is a true metre in the sense I described above with listeners perceiving two or more beat levels at the same time? Or is the 12 element rhythm just perceived with 3 or 4 beat beats per rhythmic cycle? If the latter, I would not refer to this as a metre so the psychological phenomenon involved is clear.
Line 325 “statistical-learning processes” can you clarify what statistic is being learned? It seems more so that a simple beat structure is being perceived (perhaps learned in some sense) and being carried over for the short term while a rhythm is presented that is not entirely inconsistent with the preceding movement pattern. This bears little resemblance to statistical learning as described in many studies, but perhaps a connection can be made more explicit if you think it’s a useful way to think about it.
Line 403 “free of sensory, cognitive, and motor dysfunctions” This is too vague. Will you exclude people with ADHD, any type of autism spectrum disorder, personality disorder that affects cognition? I would caution against excluding too many disorders because you will end up studying a highly non-representative population. Plus it may be challenging to apply similar exclusion criteria in the two cultural groups. You might want to administer a basic IQ or executive function test that is cross-culturally validated so you can be somewhat convinced that you don’t have any major confounds between group and cognitive function.
Line 493 “a metronome-like acoustic pulse will be added to the auditory stimulus and will serve as a cue to the beat from the targeted metre” If I understand correctly, the metronome will tell the participants which beat to move to. But in the subsequent rhythm during which brain activity or movement is recorded, how will you know whether it is the movement per se that is having an influence on subsequent brain activity or movement as opposed to the metronome. We know from the Nave et al., 2022 study you cite earlier than even in the absence of movement, just perceiving the beat one way or another can influence subsequent brain activity.
Line 535 In what sense are the ER2A earphones magnetically shielded? Do you just mean the electronics are not right next to the head, or is the electronic portion somehow shielded by the manufacturer or is there custom shielding?
It’s a little confusing about when exactly they will be doing clapping, tempo change detection, and whole body movement, and how this relates to the experimental design. Please clarify.
Line 578 Why no electrodes near the eyes to detect eye blinks and horizontal eye movements?
“trials showing excessive artefacts will be rejected” How will you define excessive artefacts? It seems likely that there will be plenty of artefacts in each trial given how long the trials are so wouldn’t this result in all trials being rejected?
Will any baseline correct be applied before or after filtering of the EEG data? I suppose it might not matter for the frequency-based analysis but for display purpose I’m guessing you still do it.
Given that a number of interactions are being tested and they can sometimes require pretty large sample sizes, I wonder if Bayesian ANOVAs would be useful instead of frequentist so you can provide evidence in favor of both non-null and null hypotheses? Otherwise, null findings will be ambiguous about whether you were underpowered or actually have favorable evidence for a true null finding.
Reviewed by Anne Keitel, 07 Feb 2024
The registered report by Guerin and colleagues presents an interesting and important set of studies on rhythm processing, and is overall well written and thought-through. The hypotheses and their alternatives, and the rationale and feasibility, are sound. I have two methodological points (one major) and some requests for clarification.
Major point
1) The main question of the work is whether movement affects subsequent rhythm processing. Yet, any positive result (i.e. change of rhythm processing after the movement session) cannot be unambiguously attributed to the movement. Instead, an alternative interpretation would be that simply being exposed to the specific metre (by hearing the metre-specific drum superimposed on the sequence for ~20 minutes) might lead to altered rhythm processing afterwards. To rule out this alternative interpretation, I suggest adding a control condition with no movement (and simply listening to the sequence with the drum cue for the same duration as the movement session is). This would strengthen the paradigm quite dramatically, in my opinion. I think it would be sufficient to add the control condition (and therefore additional participants) to only one sample – perhaps the Western-enculturated sample as they might be easier to recruit.
Other points
1) The second methodological concern is that the used auditory stimuli (pure tone sequences) are quite far from being ecologically valid and might even be a bit unpleasant to listen to (although I could not find the stimuli, maybe they are okay to listen to). Would it be possible to replace the pure tones with real drum sequences? This could be completely equivalent to using pure tones, just using a single drum (like a snare or other tom). Making the stimuli resemble real musical rhythms would improve the ecological validity and perhaps even the compliance of participants. In any case, it would be good to mention the reasons for using pure tones.
2) The abstracts and introductions mention Stage #2, but I think this is still Stage #1, as no research has taken place?
3) It took me a while to figure out which study would use what type of participant sample and stimuli. The first time this is clearly stated is in the Hypotheses section. Before, it was unclear which study uses a Western sample vs. African sample, what music stimuli would be used, and whether this would differ between studies. Could it be stated in every abstract and introduction what the specific sample will be, and that the same (African) rhythmic pattern will be used in both studies? This would help to avoid confusion.
4) The first paragraph of the Stage 2 [sic] #2 Introduction is quite redundant with the #1 Introduction, although this might be deliberate.
5) The inclusion criteria are a bit unclear (starting line 407). Could you clarify whether it is sufficient for inclusion if participants themselves have not lived in the respective countries? It sounds like it is sufficient if only the parents have lived in the country for 15 years, which begs the question whether individuals who for example grew up in Western countries have experienced enough exposure to African music to be influenced by its culture (and vice versa).
6) As the inclusion criteria are a bit arbitrary, it would be good to ensure and show that participants have been exposed to the relevant music/metre throughout their lives. Could the extend of exposure to Western and African music be added as a self-report questionnaire?
7) Could you please justify where the alpha value of .2 (lines 721 & 725) comes from? Is this based on experience or a specific recommendation?