Testing Beat Perception Without Sensory Cues To The Beat: The Beat-Drop Alignment Test (BDAT) Part 2

Results

Preliminary analyses

A total of 104 individuals completed Study 1 in full, and 21 more responded to 29 out of 30 stimuli, so their data were included in the analysis. There were no missing data on demographics, Gold-MSI questionnaire, or test difficulty rating. Difficulty ratings varied from 1 to 7 (M = 4.28, SD = 1.59), where 1 meant very easy and 7 meant extremely difficult. Difficulty ratings correlated with the level of musical training, with musically trained participants rating the test as easier than musically untrained participants (r = −.406, p < .001). Raw test scores ranged from 7 to 30 correct out of 30 trials (M = 18.74, SD = 3.96). Across participants, the correlation between difficulty ratings and raw test scores was significant (r = −.216, p = .016), indicating that higher perceived difficulty ratings were associated with poorer performance on the test.

Music has been shown to have a certain influence on the brain and has a significant role in training and improving memory. Music training can provide many benefits, from intellectual development in infancy to cognitive function in older adults.

First of all, music training can improve people's sense of hearing and music perception. The purpose of hearing is to make people perceive sounds more keenly and thus better distinguish the information in the sounds. Music sensitivity can promote people's understanding and appreciation of music, enhance their love for music, and thus enhance their memory.

Secondly, music training also plays an important role in brain development. Research shows that music training can promote the nerve cells in the cerebral cortex to generate more synaptic connections, thereby enhancing the brain's memory and thinking ability. It can be said that music training can make the brain more flexible, thereby improving people's thinking ability and memory.

Also, music training can promote people's mental health. Studies have found that music can relax people, reduce negative emotions such as anxiety and depression, and thus improve people's mental state. A relaxed and positive attitude can make people more focused and memorize information effectively.

To sum up, music training has a very close relationship with memory. Music training can not only improve people's hearing and music sensitivity but also promote brain development and mental health. Therefore, you might as well spend more time practicing music to make your brain more flexible and your memory more efficient! It can be seen that we need to improve our memory. Cistanche deserticola can significantly improve memory because Cistanche deserticola is a traditional Chinese medicinal material with many unique effects, one of which is to improve memory. The efficacy of minced meat comes from the various active ingredients it contains, including acid, polysaccharides, flavonoids, etc. These ingredients can promote brain health in various ways.

Click know ways to improve brain function

participant response reliability was checked by investigating whether any of the participants provided all/nearly all ‘ON’ or all/nearly all ‘OFF’ answers; no such participant was identified, suggesting that the test was taken seriously.

Gold MSI Music Training scores varied from 7 to 46 (M = 27.41, SD = 12.85; scoring range 7 to 49), Gold MSI Perceptual Abilities scores varied from 22 to 62 (M = 47.60, SD = 8.51; scoring range 9 to 63). These values resemble published norms drawn from a sample of nearly 150,000 participants from the general population (M = 26.52, SD = 11.44 for Musical Training and M = 50.20, SD = 7.86 for Perceptual Abilities; Müllensiefen et al., 2014). A highly significant correlation was found between BDAT raw scores and Gold-MSI self-report questionnaire scores on subscales of Musical training and Perceptual Abilities (r = .324, p < .001; r = .297, p < .001 respectively).

Responses to individual tracks

The average response accuracy for each of the 30 tracks was computed by counting the number of correct responses to that track across all probe positions tested, divided by the total number of responses to that track. All accuracies were above the chance level of 0.5 (range = .53–.744, M = .63, SD = .48). However, track averages differed when calculated for ON and OFF conditions separately. In the ON condition overall accuracy was higher (M = 0.67, SD = .47), and only one track (Track 3) had below chance performance (M = .42, SD = .50). In the OFF condition overall accuracy was lower than in the ON condition (M = .58, SD = .49), with three tracks showing below-chance accuracy levels (Track 12, M = 0.48, SD = 0.50; Track 27, M = 0.48, SD = 0.50; Track 30, M = 0.49, SD = 0.50; for accuracy data on individual tracks, see figures in Supplementary Information). Because below-chance averages can be indicative of a stimulus bias, Track 3 was excluded from further analysis of the ON condition, and Tracks 12, 27, and 30 were excluded from further analyses of the OFF condition.

Effect of participant group: Differences between musically trained and general public.

Study 1 had a sample of 40 musicians, 19 dancers, and 64 individuals from the general public. Two people preferred not to disclose this information and were therefore removed from the analysis of the effect of the group. A nonsignificant Levene’s test showed no violation of the equality of variances assumption, F(2, 120) = .326, p = .723. A one-way analysis of variance (ANOVA) detected a significant effect of group, F(2, 120) = 6.04, p = .003; however, analysis of contrasts showed no significant difference between dancers and musicians, t(120) = −1.12, p = .265, who were therefore pooled into a ‘musically trained’ group (N = 59) and compared with the general public (N = 64). Musically trained people differed from the general public, t(120) = 2.88, p = .005, with the former scoring higher (M = 19.92, SD = 4.09 vs. M = 17.64, SD = 3.59; Fig. 2, cf. Supplementary Information Tables S1–S4).

Explanatory item response modeling

For modeling item difficulty at the level of the individual trial, we employed the approach outlined in Harrison and Müllensiefen (2018) and Larrouy-Maestri, Harrison & Müllensiefen (2019), based on the explanatory item response modeling framework laid out in De Boeck and Wilson (2004). The item response model took the form of a generalized mixed-effect model with a logit link function (also known as a mixed-effect logistic regression model) and modified asymptotes (Harrison & Müllensiefen, 2018). This model reproduces a four-parameter logistic IRT model in which the guessing, discrimination, and inattention parameters are constrained not to vary within the item bank (Magis & Raîche, 2012). In addition to calibrating the difficulty of the individual items, the explanatory item response model also allows one to quantify the relationship between item difficulty and the structural item features (probe displacement, probe direction, strength of target beat), which is useful for investigating the test’s assumptions and contributing to its construct validity. Generalized mixed-effects models were computed separately for ON and OFF conditions.

ON condition

The model for the items in the ON condition included the binary response accuracy (0 or 1) as the dependent variable, metrical strength as the only fixed effect (binary, with Target Beat 4 coded as weak and Target Beat 3 coded as strong), and participant ID as well as track as a random effect. The model was fit using the functions gamer () from the R package lme4 and logit.2asym() from the R package psyche. A summary of the model is given in Table 1, which shows a significant effect for strong beats, with the performance being better on strong beats (76.5% correct) than on weak beats (57.5% correct). The model for the ON condition achieved a prediction accuracy of 71.2%.

OFF condition The model for the items in the off condition also used response accuracy as a dependent variable, with metrical strength, probe direction (binary, with categories before and after target beat), probe displacement (numerical variable with seven levels, ranging from 15% to 45% displacement), and interaction between metrical strength and direction as fixed effects, and with track and participants as random effects. The model summary in Table 2 shows that all fixed effects except metrical strength make a significant contribution. However, because the interaction term Probe Direction × Metrical Strength is significant, we also kept the main effect of metrical strength in the model. The model for the OFF condition items achieved a prediction accuracy of 71.3%.

The plot of the mean response accuracy across the seven levels of probe displacement in Fig. 3 shows an approximately linear trend, which confirms that a transform for the linear predictor displacement is not necessary.

Discussion

Study 1 aimed to (1) determine if participants could do the BDAT without experiencing it as overly difficult, (2) identify any problematic test items, (3) investigate whether the BDAT is sensitive to the degree of musical training, (4) obtain data for estimating IRT parameters for the subsequent construction of an adaptive version of the BDAT, and (5) see if metrical strength influenced accuracy in judging alignment of a probe sound with a beat, which had not been examined before with real music.

We found that participants could do the BDAT and rated it as moderately difficult (mean of 4.3 on a scale of 1–7). As expected, participants with extensive musical training (dancers and musicians) performed significantly better than the general public, with mean raw scores about 13% higher than the general public (Fig. 1). A significant association between participant test scores and Gold-MSI scores indicated that participants with a higher level of musical training and higher self-reported perceptual abilities were better at this beat perception task. The correlation levels were comparable, though slightly lower, than that found between Gold-MSI Musical Training and CA-BAT performance in previous work, r(195) = .454, p < .001 (Harrison & Müllensiefen, 2018). Such a moderately-sized correlation was expected because beat perception abilities are assumed to depend moderately on musical training (i.e., less strongly associated with musical training than, for example, melodic discrimination ability; Harrison & Müllensiefen, 2018). This could be interpreted as an argument towards beat perception abilities being influenced by genetic predispositions (cf. Niarchou et al., 2022.

Both explanatory item response models showed acceptable predictive accuracy. The model for items in the ON condition showed that metrical strength was a significant factor determining response accuracy, with better performance when the probe was placed on the strong beat. This result is expected in light of the work of Palmer and Krumhansl (1990), who found that probe events placed on strong beats were rated as fitting an ongoing rhythmic pattern better than probe events on weak beats.

For the OFF condition, response accuracy was significantly higher with larger probe displacement, as expected, with a roughly proportional association between response accuracy and degree of displacement. Probe displacement direction was also a significant determinant of response accuracy: Accuracy was higher when the probe was placed ahead of rather than after the target beat location. This is the opposite pattern of results to that found in an earlier Beat Alignment Test study with 15 professional musicians (Van Der Steen et al., 2014) but accords with musical intuition since certain musical styles favor rhythmic delays rather than musical events ahead of implied beat locations (e.g., delayed snare in swing; Butterfield, 2010). It also agrees with two reported timing psychophysics results. First, the “filled duration” illusion causes listeners to perceive an empty interval as shorter than a subdivided interval (Repp, 2008; Repp & Bruttomesso, 2009); this would lead listeners to perceive slightly late events following empty intervals as on time, as we observed. Second, listeners experience “perceptual acceleration” during isochronous sequences, showing a bias toward reporting the last interval as slightly short (Li et al., 2016); this, too, would lead listeners to perceive late events as on time.

Metrical strength of the beat was a non-significant factor on its own in the model for OFF items, but its interaction with direction was significant: participants performed better when the probe came late on the weak beat than when it came late on the strong beat. Again, this could be interpreted as listeners being more tolerant of slightly delayed musical events especially when they come after strong beats, as this is a musical feature in certain styles where such delays can create the impression of a ‘laid back’ feel. Another possible explanation is that probes falling late on the weak beat were the closest to the point at which the rhythm reenters and could be more easily recognized as off the beat by comparing their timing to the timing of rhythm reentry.'

Study 2

Study 2 aimed to conduct a correlative investigation of the relationships among three scores: the BDAT, the CA-BAT, and self-reported musical sophistication as quantified by the Gold-MSI questionnaire (Müllensiefen et al., 2014). Our goal was to understand better the aspect(s) of beat perception quantified by BDAT performance.

Methods

Participants

A total of 103 participants were recruited for this online study. Two participants reported hearing impairments and were therefore excluded from the analysis. In the remaining sample of 101 participants, 52 identified as female, 47 as male, and two chose not to disclose this information. Participants were ages 21 to 63 years (M = 29.91, SD = 7.21).

Participants were recruited through social media and email invitations. All participants provided informed consent to participate in the study.

Materials

A three-part battery was constructed for Study 2, consisting of an adaptive version of BDAT, the Computerised Adaptive Beat Alignment Test (CA-BAT; Harrison & Müllensiefen, 2018), and the full version of the Gold-MSI questionnaire (Müllensiefen et al., 2014), including all five subscales as detailed below.

BDAT Study 2 used the same musical clips as Study 1, with the stimulus set expanded to include tempo variation as an additional random factor. Slight variation in tempo ensured that participants had to identify the tempo every time they listened to a new clip. Thus, their perceived tempo as well as the phase of the beat were based entirely on the clip they were hearing and not on a developed expectation that the tempo would always be fixed.

There were five tempo variations of each clip, which produced 4,500 files. Tempo varied in even steps ±5% from the original 125 bpm, producing 119 bpm, 122 bpm, 125 bpm, 128 bpm, and 132 bpm; 5% tempo steps provided five perceptually distinct tempi but did not distort the musical material.

Following the track-reliability analysis of Study 1, Study 2 eliminated the usage of Track 3 in the ON condition and Tracks 12, 27, and 30 in the OFF condition.

For Study 2, an adaptive version of BDAT was created based on the explanatory IRT models for ON and OFF conditions computed in Study 1. In this adaptive version, on each new trial, the difficulty of the item presented was matched to participant performance that was estimated dynamically after each trial. This was based on the item difficulty parameters as computed for all items in Study 1 (see Harrison & Müllensiefen, 2018, for details of item difficulty computations). In addition, the standard deviations of the participant random intercepts were extracted as constant discrimination parameters for ON and OFF items. Similarly, the parameters for lower and upper asymptotes were used as constants across all items. In IRT models, person abilities, as well as item difficulties, are defined on the same metric, a z-score scale typically ranging from −4 to 4. This means that, for example, a participant with an ability score of 1 is one standard deviation above the population mean. The first item of the adaptive test was always chosen to have a difficulty level of 0, matching the average ability of the participant sample from Study 1.

CA-BAT The purpose of this test is to evaluate the listener’s beat perception ability using an adaptive version of the beat alignment test (BAT) which tailors the difficulty level of the test to each participant by adapting to their previous responses (Harrison & Müllensiefen, 2018). This maximizes the amount of information that each successive item gives about the participant’s actual ability, which in turn ensures shorter testing time and thus improves the test’s efficiency. Similar to the BDAT, the CA-BAT uses naturalistic music stimuli. In contrast to the BDAT, the stimuli were not created specifically for the test (see Harrison & Müllensiefen, 2018, for details), meaning that minor temporal fluctuations are probable within items. However, this is mitigated by using a 2-AFC paradigm. For each trial, the participant is presented with two versions of a musical track, both overlaid with a metronomic probe track. The ON version of the track has the probe in time with the musical beat locations while the OFF version has the probe track displaced from the musical beat locations. The participant’s task is to identify the ON track. A person's ability score for CA-BAT is computed following the same principles as for the adaptive BDAT, described above.

Gold-MSI Study 2 employed the full version of the questionnaire, which assessed all five subscales (Active Engagement, Perceptual Abilities, Musical Training, Emotion, Singing Abilities) as well as the General Musical Sophistication (GMS) scale that draws on items from all five subscales. Altogether, the Gold-MSI self-report questionnaire was composed of 41 questions.

Procedure

Study 2 was conducted online using an interface based on the open-source psychTestR package (Harrison, 2020). All participants provided their consent for taking part in the study. They were asked to wear headphones for the entire duration of the test and adjust their volume to a comfortable level. As in Study 1, participants were asked not to tap or otherwise move to the beat of the music.

The battery started with the collection of demographic information and was followed by the BDAT (25 items), GoldMSI questionnaire (41 questions), and CA-BAT (25 items).3 Before each test participants were presented with a training phase which included instructions, two example stimuli (one for Condition ON and one for Condition OFF), and two practice items. The reported testing time was 25–30 minutes. Completion of the test led to the display of task performance as pseudo-IQ scores in a numerical as well as graphical format (bell curve with a mean of 100 and a standard deviation of 15), alongside the General Musical Sophistication (GMS) score.

Results

BDAT ability scores varied from −3.37 to 2.53 (M = 0.04, SD = 1.17), CA-BAT ability scores varied from −2.44 to 2.07 (M = 0.27, SD = 0.78), and GMS scores varied from 1.45 to 6.45 (M = 4.23, SD = 1.18). For comparison, published GMS norms vary from 1 to 7, M = 4.53, SD = 1.15 (Müllensiefen et al., 2013). Significant correlations were found between BDAT and CA-BAT scores (r = .23, p = .023; Fig. 4) as well as between BDAT and GMS scores (r = .55, p < .001). Indeed, BDAT scores were highly significantly correlated with all the dimensions of the Gold-MSI questionnaire, the strongest correlations being with Musical Training and overall GMS score.

CA-BAT scores were also significantly correlated with all of the Gold-MSI dimensions except Singing Abilities, though in most cases less strongly than BDAT scores. The highest correlation was found with the dimension of Active Engagement (Table 3).

Discussion

The purpose of Study 2 was to characterize relationships among BDAT performance, CA-BAT performance, and Gold-MSI questionnaire scores. Inspecting the correlations between Gold-MSI dimensions and BDAT scores shows that the overall GMS score, and musical training in particular, partially predicted performance on the BDAT’s covert beat continuation and comparison task. Regression analysis showed that the Gold-MSI subscales collectively accounted for around one-third of the variation in BDAT scores (R2 = .34), indicating that while beat perception ability as indexed by the BDAT was influenced by (or possibly influenced) musical training and sophistication, it also seemed to reflect skill or ability that was not solely based on training or engagement. However, the modest correlation between BDAT and CABAT scores suggested that this skill or ability was not identical to that measured by the CA-BAT, which was less correlated with the GMS score and most of its subfactors, including musical training.

Overall, these results indicate that the aspect of beat perception tested by the BDAT is more closely related to general music skills than the aspect tested by the CA-BAT. To make such a conclusion with confidence, it would be essential to estimate the degree to which variability in each test reflects measurement error vs. a consistent attribute of the participant and to observe the changes in both scores with musical training interventions.

General Discussion

This study developed and explored a new test of musical beat perception which does not rely on synchronized movement to the beat. The primary innovation of the BDAT is that it tests musical beat perception in the absence of any sensory cues to the beat. In the BDAT, the listener hears a few bars of beat-based music and then judges if a single probe event is on or off the beat during a “beat drop” when all rhythmic cues to the beat have been removed. Thus, the BDAT requires the listener to continue a beat percept formed while hearing rhythmic music through a beat-drop bar until the music resumes. The BDAT was created to provide a focused test of the capacity to covertly continue a beat, which cannot be directly investigated by tests like the BAT, which allows for the use of local acoustic cues in judging timing. Indeed, the results of our study of the BDAT suggest that the BDAT is testing aspects of beat perception that are partly independent of those tested by the BAT, as we discuss below.

Among other tests of rhythm perception, the BDAT has several advantages. It uses a variety of realistic musical materials composed in the style of electronic dance music (created specifically for the test), is quick to administer, and has beats at unambiguous locations (since the music was composed using a MIDI time grid). Furthermore, it can be used to study beat perception as a function of metrical position (strong vs. weak beats), direction of probe misalignment (early vs. late), and degree of probe displacement from beats.

The current study examined the performance of the BDAT in two experiments. The first experiment showed that the BDAT was not rated as highly difficult by participants and that individuals with a high degree of musical training scored significantly better on the test. However, overall performance was generally low, on average around 60%–70% correct. This is not surprising given that each trial of the BDAT has a single probe sound, unlike the BAT, in which there is an entire metronomic train of probe sounds. Furthermore, when the probe sound was misaligned, it was often very close to a beat location, making the misalignment difficult to detect (Fig. 2). Restricting off-beat probes to larger displacements should result in higher overall BDAT performance scores in future work.

A novel finding of experiment 1 is that accuracy in judging when a probe event is on the beat differed substantially depending on whether the probe was on a strong versus weak metrical position (Beat 3 vs. 4) in the beat drop bar, with accuracy about 20% higher on the strong beat (76.5% vs. 57.5% correct). Interestingly, the music for our study was not composed in a way to acoustically emphasizes strong versus weak beat positions, suggesting a potentially significant role for top-down metrical expectations in shaping our results. Our finding aligns with research by Palmer and Krumhansl (1990), who had listeners listen to metronomic sequences of equal-loudness events and imagine different meters, and who found that probe events placed on metrically strong beats were rated as fitting better with the rhythm than events on weak beats. However, the differences in their probe ratings were subtle compared with the large effects seen in the current study. Further work is needed to determine if the large accuracy difference we see on strong versus weak beats is due to metrical expectations (cf. Iversen et al., 2009), or simply reflects the fact that our weak beat position was later in the beat drop bar than the strong beat position. Due to this design, any internally maintained pulse might be diminished and/or less precise at the time of the weak versus strong beat simply by the greater time elapsed since the cessation of rhythmic cues to beat structure (Cannon, 2021). In future work it would also be interesting to determine if the metrical effect we see only emerges after a certain age, reflecting the development of metrical knowledge of the culture’s prevailing musical patterns (Nave-Blodgett et al., 2021).

The second experiment in our study used data on item difficulty taken from the first experiment to eliminate a few overly difficult stimuli and to create an adaptive version of the BDAT, in which item difficulty increased as the experiment progressed. Experiment 2 also introduced slight intertrial tempo variation between stimuli to force listeners to infer the tempo of the beat on each trial rather than forming an experiment-wide tempo prior that could influence task performance. All participants in this experiment were also tested on a version of the CA-BAT (the computerized-adaptive BAT) and were given the full Gold-MSI questionnaire to measure musical sophistication, including five subscales (Active Engagement, Perceptual Abilities, Musical Training, Emotion, Singing Abilities). This experiment found that performance on the BDAT and CA-BAT showed only a modest correlation and that the extent to which the two tests correlated with subscales of the Gold MSI and with general musical sophistication (GMS) differed substantially. For example, self-reported singing abilities correlated with the BDAT performance but not CA-BAT performance. Indeed, correlations between BDAT performance and all subscales of the Gold-MSI (and with GMS) were substantially higher than were correlations with CA-BAT performance (Table 3), suggesting that performance on the BDAT may be tied to a wide range of musical abilities.

One musical ability likely to be relevant to the BDAT is musical imagery during the beat-drop bar. Neuroimaging research has shown that detailed imagery for musical patterns involves a complex network of brain regions aside from regions involved in beat processing (Cannon & Patel, 2021; Regev et al.,2021). Furthermore, musical imagery abilities and pitch imitation abilities appear to be related (Greenspon et al., 2017), which could help explain why self-related singing abilities correlate with BDAT performance if listeners engage in musical imagery during the beat drop. Another relevant musical ability might be a motoric continuation of a beat: Although participants were asked not to move, subtle movements and/or motor imagery may have played roles in the performance. In future work, it would be interesting to determine if BDAT performance is associated with individual differences in auditory imagery abilities (e.g., as measured by the Bucknell Auditory Imagery Scale; Halpern, 2015), and in the continuation phase of a synchronization-continuation task.

Another factor that could help explain the relatively low correlation between BDAT and CA-BAT performance is the fact that the latter involves a memory component since two musical clips must be compared to determine which has onbeat beeps. The BDAT, like the original BAT, only requires listening to each clip once. Another difference could be the additional auditory processing required in the CA-BAT to compare the musical rhythm to a concurrently presented series of beeps. More generally, it appears that the BDAT provides a distinctive test of beat perception, engaging some different cognitive processes than the CA-BAT. Some of these processes may be specific to the continuation of a beat (as opposed to the initial recognition of the beat).

The BDAT offers a range of possible applications in the study of human auditory rhythmic processing. Researchers using the BDAT can decide whether to use BDAT stimuli that are uniform in tempi (as in Experiment 1) or slightly different in tempi from trial to trial (as in Experiment 2), depending on their goals. Furthermore, the BDAT stimuli, which are freely available, may prove useful in a range of studies of beat perception, including neural studies aimed at studying oscillatory neural dynamics during beat perception. (To facilitate such work, the online data archive for this paper includes versions of the BDAT stimuli without probe sounds). Due to the novel beat-drop design of the BDAT, any beat-related neural oscillations during the beatdrop bar cannot be due to stimulus-driven brain activity, and this could help test existing models of the causes of beat-related neural oscillations in the brain (Breska & Deouell, 2017; Doelling & Assaneo, 2021; Tal et al., 2017).

By placing beat drops of predictable duration at musically appropriate times and filling the gap with appropriate nonrhythmic musical content, these stimuli induce a strong expectation of the return of the rhythm. BDAT stimuli may thus have some advantages for studying perceptual and neural oscillations involved in rhythm perception compared with stimuli used in the past, which examine activity after the sudden end of a rhythmic stimulus (Hickok et al., 2015; Stupacher et al., 2013; van Bree et al., 2021). The BDAT stimuli may also prove useful in future neural studies of musical imagery if such imagery is indeed one cognitive tool that participants use to do the task. Finally, like one current use of the BAT, the BDAT may prove to be a useful tool for studies with patients with movement disorders or neurodegenerative diseases. Based on our findings we feel that the BDAT is a viable and novel instrument for exploring beat perception, and merits further study and development.

Open Access
This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution, and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third-party material in this article are included in the article's Creative Commons license unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license.

References

1.Allman, M. J., & Meck, W. H. (2012). Pathophysiological distortions in time perception and timed performance. Brain, 135(3), 656–677.

2.Bégel, V., Benoit, C.-E., Correa, A., Cutanda, D., Kotz, S. A., & Bella, S. D. (2017). "Lost in time" but still moving to the beat. Neuropsychologia, 94, 129–138.

3.Breska, A., & Deouell, L. Y. (2017). Neural mechanisms of rhythm-based temporal prediction: Delta phase-locking reflects temporal predictability but not rhythmic entrainment. PLOS Biology, 15(2), Article e2001665.

4.Breska, A., & Ivry, R. B. (2018). Double dissociation of single-interval and rhythmic temporal prediction in cerebellar degeneration and Parkinson’s disease. Proceedings of the National Academy of Sciences of the United States of America, 115(48), 12283–12288.

5. Butterfield, M. (2010). Participatory discrepancies and the perception of beats in jazz. Music Perception, 27(3), 157–176.

6. Cannon, J. (2021). Expectancy-based rhythmic entrainment as continuous Bayesian inference. PLOS Computational Biology, 17(6), Article e1009025.

7.Cannon, J. J., & Patel, A. D. (2021). How to beat perception co-opts motor neurophysiology. Trends Cognitive Science, 25(2), 137–150.

8.Cochen De Cock, V., Dotov, D. G., Ihalainen, P., Bégel, V., Galtier, F., Lebrun, C., Picot, M. C., Driss, V., Landragin, N., Geny, C., Bardy, B., & Dalla Bella, S. (2018). Rhythmic abilities and musical training in Parkinson's disease: Do they help? NPJ Parkinson's Disease, 4, Article 8.

9.Dalla Bella, S., Farrugia, N., Benoit, C. E., Bégel, V., Verga, L., Harding, E., & Kotz, S. A. (2017). BAASTA: Battery for the Assessment of Auditory Sensorimotor and Timing Abilities. Behavior Research Methods, 49(3), 1128–1145.

10.Dalla Bella, S., Farrugia, N., Benoit, C. E., Bégel, V., Verga, L., Harding, E., & Kotz, S. A. (2017). BAASTA: Battery for the Assessment of Auditory Sensorimotor and Timing Abilities. Behavior Research Methods, 49(3), 1128–1145.

11.Doelling, K. B., & Assaneo, M. F. (2021). Neural oscillations are a start toward understanding brain activity rather than the end. PLOS Biology, 19(5), Article e3001234.

12. Fiveash, A., Bella, S. D., Bigand, E., Gordon, R. L., & Tillmann, B. (2022). You got rhythm, or more: The multidimensionality of rhythmic abilities. Attention, Perception, & Psychophysics, 84, 1370–1392.

13.Fujii, S., & Schlaug, G. (2013). The Harvard Beat Assessment Test (H-BAT): A battery for assessing beat perception and production and their dissociation. Frontiers in Human Neuroscience, 7, 771.

14.Grahn, J. A., & Brett, M. (2009). Impairment of beat-based rhythm discrimination in Parkinson’s disease. Cortex, 45(1), 54–61.

15.Greenspon, E. B., Pfordresher, P. Q., & Halpern, A. R. (2017). Pitch imitation ability in mental transformations of melodies. Music Perception: An Interdisciplinary Journal, 34(5), 585–604.

For more information:1950477648nn@gmail.com