The Effect of Binaural Beats on Synchronization to a Pacing Stimulus Final Project for Music 251 Roy Fejgin CCRMA, Stanford University Introduction Background Tapping experiments have been widely used to study sensorimotor synchronization (SMS) in humans. The topic is of interest since several important functions are involved, such as mental timekeeping, acting in response to a stimulus, and the coordination of the auditory, visual and motor systems . Moreover, SMS is essential to activities such as music playing and dancing, allowing the performer to stay in sync with other performers. The task often used in SMS experiments is tapping to a periodic beat (a “pacing stimulus”). Tapping seems to be a good choice since it is simple, and involves contact with a surface, which mimics the interaction with instruments such as keyboards and percussion. Binaural beats have been the topic of study since the early 20th century. The phenomenon occurs when two tones are presented to a subject, one to each ear. If the tones are at neighboring frequencies (with a difference of up to about 35 Hz), the subject perceives a beating effect – periodic variation of loudness -‐ at the frequency of the difference between the tone frequencies . For example, if one tone is at 440 Hz and the other at 447 Hz, beating is perceived at 7 Hz. It should be stressed that beating is perceived even though there is no beating in actual stimuli signals (unlike regular acoustic beating). Binaural beats have generated much interest due to some evidence they can entrain brainwaves. That is, exposure to binaural beat stimulation causes an increase in EEG brainwave activity at the beating frequency. Brainwave activity at particular frequency bands is correlated with specific states of consciousness. For example, the Theta band (4-‐8Hz) is with drowsiness or deep meditation; Beta activity (13-‐30Hz) is associated with waking state and active concentration. Therefore, there is reason to believe through the mechanism of entrainment we can use binaural beats to alter one’s state of consciousness. Problem Statement The purpose of this experiment is to explore the effect of binaural beats on tapping performance. To the best of my knowledge, previous tapping experiments have been carried out without regard for the subject’s state of consciousness. Assuming binaural beats can indeed alter one’s state of consciousness, we can use them to test tapping at various states. In this experiment, performance on two types of tapping tasks was measured. The first was paced tapping, meaning tapping along with a pacing stimulus. The second – unpaced tapping – involves first acquiring a tempo using paced tapping, then continuing to tap without the pacing stimulus. For both tasks, the key parameter measured was tap error, which was defined as the difference between the beat time and the subject’s actual tap time. Statistical measures such as the mean tap error and standard deviation were then computed from the tap error. The hypothesis was that subjects would exhibit different tapping performance depending on the frequency of the binaural beats to which they were exposed. Method Stimuli and Task Each subject performed three rounds of a task, each round time at a different binaural beat frequency. A round consisted of three sections. First, binaural beats mixed with white noise were played for 130 seconds. Next, with the beats and noise continuing, a pacing stimulus was introduced and the subject requested to tap to it (“paced tapping”). This went on for 40 seconds. Finally, the pacing stimulus was removed and the subject was requested to continue tapping (“unpaced tapping”). That section was 5 seconds long. The audio signal was then stopped for a few seconds while the operator adjusted the beat frequency, and then the next round was performed. The binaural beats were generated using two tones. The tone presented to the left ear was always at 440 Hz. The tone presented to the right ear was at higher by either 0 Hz, 6 Hz (later 5 Hz, see the section ‘experiment versions’), or 17 Hz. The first round was always at 0 Hz, that is, the tones presented to both ears were at the same frequency and thus generated no binaural beating. This was done in order to get a baseline for the subject’s performance before being exposed to any binaural beats. The order of the beat frequencies in the remaining rounds was randomized. The noise that was mixed with the binaural beats was white noise, low-‐pass filtered with a cutoff frequency of 2000 Hz. The purpose of the noise was twofold. First, it was meant to make the binaural beats more pleasant to listen to. Second, the presence of the noise made it harder for the subject to become aware of the beating -‐ or lack of beating during the 0 Hz round, allowing an effective control round. The pacing stimulus was a beep presented once per second (1 Hz). The beep was a 100 ms tone at 777 Hz. The main consideration in selecting that frequency was to ensure that the pacing stimulus would be easily audible in the presence of the noise and binaural tones. It was selected by trial and error. It was easy to hear because it is neither close to the frequencies of the binaural tones, nor a multiple of those frequencies. The subject tapped on to the pads of a MIDI controller. The MIDI events were sent to a computer running ChucK, where their timestamps were recorded. The ChucK program calculated the tap error by subtracting the tap timestamp from the stimulus timestamp, and then compensating for system latencies. The following sections go into detail regarding the experimental setup. Setup and Software Figure 1: Experiment Setup The setup consisted of a Macintosh laptop running ChucK, a MIDI controller, a microphone placed near the MIDI controller, and a pair of headphones. The MIDI controller employed was an M-‐Audio Axiom 25, which has pads as inputs. The pads turned out to be a convenient and accurate interface for capturing the taps, since they are designed for percussion. The taps generated MIDI events that were sent over a USB interface to the laptop. A microphone was placed on the tapping surface and connected to the laptop’s analog input. On the laptop, the ChucK program redirected the microphone input back out to the headphones. The purpose of the microphone was to give the subject auditory feedback for their taps. It would have otherwise been difficult for the subject to hear their taps in the presence of the loud sound (noise, binaural beats, and pacing stimulus) generated by the headphones. The laptop ran a ChucK program written for this experiment. The program executes a complete round (binaural beats -‐> paced tapping -‐> unpaced tapping) without need for user intervention. Between rounds, the operator saves the test results, select the next beat frequency, and restarts the test. When MIDI tapping events are received, the program records their timestamp. At the end of a round, the program processes all tap events and associates each event with the closest pacing stimulus (or, in the cased of unpaced tapping, with the time when the pacing would have occurred). When a pacing stimulus is associated with more than one tap, the program discards the taps as invalid. Finally, the program calculates some statistics, for example: the mean and STD of tap errors during the paced section, and the same for the unpaced section. System Latency and Jitter The tap timing error, the key parameter measured in the experiment, was defined as the difference between the pacing beat time and the tap time. In practice, however, it is not straightforward to measure this time difference due to multiple unknown (and variable) delays present in the system. Among these are: delay caused by the buffering of output audio samples in the OS; delay introduced by the USB interface; MIDI event processing latency; other events occurring in the system such as network interface IO, activity in other processes, etc. In order to estimate the overall latency and latency jitter in the system, the following technique was employed. A simplified version of the ChucK program used in the experiment was created. The simplified program generated a few beeps and recorded the timestamps of the input MIDI events. In addition, a microphone was placed where it could capture both the sound of the stimulus and sound the subject’s finger hitting the pad. On the laptop, the audio from the microphone was recorded to a .wav file. The resulting waveform was then visually examined in an audio editing program. By measuring the distance between peaks in the waveform, an accurate stimulus-‐response delay was determined, free of any error introduced by system delays. Comparing the measured delay to the stimulus-‐response delay observed by the ChucK program allowed calculating the system latency. The above procedure was repeated five times. The observed system latencies were all the range of 11 to 15 ms. A value of 13 ms was selected as the system latency estimate. When running the actual experiment, 13 ms were subtracted from all tap timestamps in order to compensate for the system latency. Experiment Versions After testing the first eleven subjects and looking at the preliminary results, it was clear that the results were not as expected. The expectation, based on the three people tested before the pilot, had been that average tap errors and STDs would be lower for the higher beat frequency. However, the error with 6 Hz beats was -‐42 ms when averaged across all taps and all subjects, as compared to -‐44 ms for 17 Hz beats. Given the large variations in tap errors among subjects and the small data set, this difference did not appear to be significant. At this point I re-‐examined the experiment in search of problems. I decided to make two adjustments: change the lower beat frequency from 6 to 5 Hz, and reduce the level of the filtered noise in the mix. Five more subjects were tested with the new parameters. The first change was made in an attempt to ensure that the binaural beat was at a frequency that would not periodically align with the tap frequency of 1 Hz. In retrospect, I do not believe that the change from 6 to 5 Hz ensures that. See the following sections for further examination of the topic. The second change, lowering of the noise level, was made in an attempt to ensure that the binaural beats are not masked by the noise to such and extent that their effect is lost. In the sections below, the first set of parameters is referred to as version 1 (v1) and the second as version 2 (v2). Results and Analysis Median of Average Tap Error (ms) Paced Tapping Mean Tap Error, Paced, Median Across All Subjects (v1+v2) 0.00 Binaural Beat Frequency 0 Hz (no beafng) 5 or 6 Hz 17 Hz -‐20.00 -‐40.00 -‐60.00 -‐34.85 -‐39.96 -‐49.70 Figure 2: Median of all subjects’ mean tap error, paced case. In the paced tapping task, the mean tap error was computed for each subject. Figure 2 shows the median of all subjects’ mean tap error. We see that the tap error magnitude with a 17 Hz binaural beat frequency was larger than the tap error with at lower beat frequencies (5 and 6 Hz). However, this result should not be interpreted as conclusive evidence that tapping accuracy is improved with low-‐frequency binaural beat stimulation. Figure 3 helps understand why. It shows the individual subjects’ results. We can see that five subjects performed better with 6 Hz, four performed better with 17 Hz, and one performed similarly with both beat frequencies. Thus there is no clear trend of better performance with any of the frequencies. For version 2, the results are more consistent: out of five subjects, four performed better with 5 Hz stimulation. However, given the small sample size in version 2 and the large variability of the tap errors, we should be cautious in drawing any conclusions from the last result. Surprisingly, the greatest difference found was between the tap error without any binaural beats (0 Hz -‐ baseline) and either of the tests with binaural beats. This finding does is not in agreement with the hypothesis: we would expect the baseline result to be somewhere in between the results for the two other tests, reflecting the random state of mental activity with which subjects entered the experiment. I suggest two possible explanations for this finding. First, it is possible that the binaural beats were actually acting as an additional pacing stimulus (and thus making tapping easier), since they complete an integral number of periods for every beat of the pacing stimulus. The change from 6 to 5 Hz between version 1 and version 2 did not fix that problem. A better solution would be to use a fractional frequency for the binaural beats, such as 5.1 Hz. A fractional frequency would ensure that the binaural beats do not complete an integral number of periods every 1 second (the rate of the pacing stimulus). As a result, the phase relationship between the binaural beat and the pacing stimulus would be constantly changing, so the binaural beat would no longer contribute to pacing. Another possible explanation for the better tapping accuracy with binaural beats is that perhaps, on average, the subjects entered the experiment at high levels of concentration, with considerable brainwave activity at frequencies above 17 Hz. If that were indeed the case, it would explain the results since both binaural beat stimulations would work in the same direction of reducing the dominant brainwave frequencies. It is worth noting that most subjects had been studying immediately before taking part in the experiment, so high levels of concentration would not be surprising. Figure 3: Mean Tap Error By Subject Unpaced tapping Though not the intended focus of this study, an interesting find was a consistent difference between paced and unpaced tapping performance. In paced tapping, all subjects showed a negative mean tap error. That is, on average they all tapped before the beat time. In unpaced tapping, on the other hand, some subjects tapped before the beat time (on average) while others after it, but in general they tended to tap later than during paced tapping: in 25 of 31 (81%) of all rounds executed with binaural tapping the mean tap error was smaller (more negative) for paced tapping. When including the baseline round (no binaural beat) the percentage drops to 64%. Figure 4 demonstrates these results for the 17 Hz case. Figure 4: Mean Tap Error: Paced versus Unpaced Other Observations It is interesting to note that both the two most accurate tappers play percussion instruments. Conclusions The experiment did not find conclusive evidence for an effect of binaural beats on tapping performance. However, I would like to point out some limitations of the current paradigm, and suggest ways in which the question could be further explored. First, this experiment was constrained by the limited availability of subject test time. Since all subjects were volunteers and were not compensated, the experiment had to be short. As a result, an entrainment time of only two minutes was used. Previous research on binaural beats employed much longer entrainment duration (such as 20 minutes, see for example ). Extending the entrainment time would important in order to ensure entrainment is actually taking place. Along the same line, monitoring the subject’s EEG would be very useful. Direct evidence of the level of entrainment would enhance our confidence in the results. Also, it would allow us to discern which subjects are susceptible to entrainment, and which are not. Naturally, the latter subjects should not be considered when evaluating the effect of entrainment on tapping accuracy. Lastly, the relationship between the binaural beat frequency and the pacing stimulus frequency should be given more attention. As described in the results section, use of a fractional binaural beat frequency with an integral pacing frequency could help minimize the contribution of the binaural beats to pacing. Acknowledgement I would like to thank Nori Jacoby for suggesting the latency measurement technique; Jonathan Berger and Jieun Oh for their helpful suggestions regarding the experiment design; Xiang Zhang for letting me use his MIDI controller in the experiment setup; Sean Colvin for suggesting to explore the relationship between the pacing and beat frequencies; and classmates and other people at CCRMA for voluntarily participating in the experiment. References  Patel, A. D., Iversen, J. R., Chen, Y., & Repp, B. H. (2005). The influence of metricality and modality on synchronization with a beat. Experimental Brain Research, 163, 226-238.  Repp, B. H. (2005). Sensorimotor synchronization: A review of the tapping literature. Psychonomic Bulletin & Review, 12(6), 969992.  Licklider, J. C. R., Webster, J. C., Hedlun, J. M. On the frequency limits of binaural beats. Journal of the Acoustical Society of America., 1950, 22, 468-473.  Russel, H., Turow, G. “Rhythmic sensory stimulation of the brain: The possible use of inexpensive sensory stimulation technologies to improve IQ test scored and behavior.” The Rhytmic Brain: Music, Ritual and Healing. Ed. Jonathan Berger and Gabe Turow. (forthcoming).
© Copyright 2018