AbstractAlpha wave bands are prevalent in most electroencephalogram (EEG) data. Alpha waves can be induced by closing the eyes and relaxing. Thus, alpha waves are usually associated with sleepiness (Teplan, 2002). To obtain stronger and less noisy data, EEG experiments must take place in a dark and quiet room and participants must remain as still as possible. Additionally, to increase the signal to noise ratio, EEG experiments must have many trials. This combination proves to be a great environment for sleepiness to set in. Sleepiness might become a problem, however, as it introduces alpha which can distort the signal. Therefore, this project aims to investigate if alpha power increases over time in a cognitive experiment. The data analyzed in this project was obtained from an opensource database in which the original experiment was investigating perceptual awareness (Benwell, et al., 2017). Results indicated that alpha power increased significantly over experiment time. To ensure this effect was solely seen with alpha, a repeated measures ANOVA as well as two ANCOVA tests were conducted to compare alpha and gamma modulation. Implications for these results and future research directions are discussed in further detail.The Prevalence of Alpha in EEG Data over Experiment TimeAlpha rhythms, in the brain, are some of the most prevalent rhythms seen in electroencephalogram (EEG) data. Alpha oscillations, in fact, were one of the first visible oscillations when EEG waveforms were first recorded (Bazanova & Vernon, 2014).  Typically, alpha oscillates from about 8 to 13 Hz and is best seen in posterior and occipital regions of the brain (Niedermeyer, 2004; Sterman and Egner, 2006; Teplan, 2002; Treder et al., 2011). Among the literature there are two views associated with alpha activity. The more classical view is founded on the idea that alpha is, simply, an idling state for several cortical areas (Bazanova & Vernon, 2014). The, modern view suggests that alpha has more functional purposes than originally thought (Basar, 1997). While this paper is aimed at understanding the potential problem for researchers who are experiencing frequent alpha modulation, it is important to address the overall research climate surrounding alpha activity.The alpha waveform has been easily spotted in EEG data for almost a century. Along with the identification of alpha came an important characteristic—the suppression of alpha. Researchers realized while plotting alpha, that the waveform vanishes as soon as the participants open their eyes. This suppression of alpha soon became known as the Berger effect (Bazanova & Vernon, 2014; Teplan, 2002). Due to this observation, it is now well known that alpha can be induced by closing the eyes and/or relaxing. Alpha can, therefore, also be associated with drowsiness (Teplan, 2002). When alpha is induced, the residual waveforms are created due to the synchronization of oscillating neurons (Cohen, 2014). When alpha is suppressed, the oscillating neurons desynchronize and alpha waveforms vanish; this has often been used as a marker for mental activation (Klimesch, Sauseng, & Hanslmayr, 2007).Even though this waveform can be easily seen in an idle state, newer developments have shown possibilities for alpha to have more functionality. Schurmann and Basar (2001), for example, have set out a list of different functional correlates that have been associated with alpha. They mention that alpha can be induced, evoked or spontaneous. Further, they point to the idea that alpha is a waveform that needs a considerable amount of research to fully understand. Increasingly more research has emerged describing different potential functions of alpha (Bartsch, Hamuni, Miskovic, Lang, & Keil, 2015; Basar, 1997; Klimesch et al., 2007; Schurmann & Basar, 2001). Specifically, there are arguments emerging that synchronization of alpha waves are possibly a demonstration of inhibition in cognitive tasks (Klimesch et al., 2007). If this is true, then it might be that alpha does not only occur when one is drowsy but also when one is engaged in a specific mental task. Thus, it has been argued that different alpha frequency bands might mean different things for cognitive function (Bazanova & Vernon, 2014).The Problem While the above findings are interesting and could possibly change the entire field of research revolving around alpha modulation, there is still an inherent problem of alpha distorting data for EEG researchers. Unless researchers are specifically interested in the alpha waveform, they usually see alpha as a source of noise rather than a usable signal (Luck, 2014). Due to the nature of EEG experiments (e.g., dark cold room, many trials, holding still), sleepiness is a very real possibility, especially in the later trials. If participants are becoming drowsy during an experiment, alpha activity might increase and cause large problems for their signal to noise ratio in multiple areas. For example, participants who are falling asleep might move their heads or experience twitching muscles. Further, if they are closing their eyes they might be missing response times entirely. As such, this paper sets out to see if this problem truly exists; the current study aims to investigate if alpha power increases over time in a cognitive experiment. It is hypothesized that a positive correlation will be found between alpha power and time.Method Data for this study was obtained from Benwell et al. (2017) on the website Open Science Framework. The materials and method used in the study are explained below.Participants14 participants (seven female, two left-handed, M=23.79 years, SD=3.17) were recruited for the study. All participants had normal or corrected-to-normal vision and no history of neurological or psychiatric disorders. The study was approved by the Ethics Committee of the College of Science and Engineering at the University of Glasgow. Written informed consent was gathered from each participant before beginning the experiment.Stimuli and Experimental Procedure Experimental Procedure. The experiment took place over two separate sessions conducted on two separate days. The first session consisted of a threshold assessment and allowed participants to become familiar with the stimuli. In the second session participants were given a threshold reassessment and then participated in a forced choice discrimination task while EEG was recorded. Stimuli. Gaussian patches (see Figure 1 for example) were presented to each participant. The stimuli were either white or black circle patches that had a Gaussian envelope and were presented on a gray background in the upper right visual field. In the EEG forced choice discrimination task, half of the patches were lighter, and half were darker compared to the background. Before the task was given, each of the black and white Gaussian patches were adjusted to either present as light or dark gray. Based on the participant’s threshold assessment, six luminance levels (three lighter and three darker than the gray background) were identified and presented. EEG Experiment Task.  During the EEG session, participants viewed the Gaussian patches and completed a forced choice discrimination task. Each trial (Figure 1) began with a black fixation cross that lasted for 400 ms. The fixation was followed by an acoustic warning (150 ms). The acoustic warning was then followed by a 1000 ms interval. A light or dark Gaussian patch were then presented for 30 ms each in the upper right visual field. The fixation cross was then presented again for 1000 ms which was followed by a response screen where participants had to judge how bright the stimuli were. Using their right index and middle fingers, they had to respond 1 on the keypad for “lighter” and 2 for “darker”. If participants did not see any stimulus they were instructed to guess. After their first response, they were then asked to rate the quality of their perception based on the perceptual awareness scale (PAS; Ramsoy and Overgaard, 2004). The categories were as follows: 0—no experience of stimulus, 1—a brief glimpse, 2—an almost clear experience and 3—a clear experience. Participants’ responses were given by pressing “0”, “1”, “2”, and “3” on the keypad. The experimental task was given in 10 blocks and each block consisted of 80 trials. 10 trials were assigned for each stimulus type (i.e., three lighter and three darker) and 20 trials were assigned to present no stimulus. Therefore, participants completed a total of 800 trials. The order of trials was randomized within each block. The threshold assessment and EEG task were programmed in MATLAB, using the Psychophysics Toolbox (Brainard, 1997; Pelli, 1997).EEG Acquisition and AnalysisEEG data was collected using 61 Ag/AgCl pellet pin scalp electrodes that were attached according to the 10-10 International System. Two BrainAmp MR Plus units (Brain Products GmbH) were used to record the data (electrode impedance<10k?, 1000Hz sampling rate). A MATLAB script was created using ERPLAB (http://erpinfo.org/erplab/) and EEGLAB (http://sccn.ucsd.edu/eeglab) functions to analyze the data. Inferential statistics were conducted using JASP (https://jasp-stats.org/). The preprocessing analysis began by re-referencing to the average reference. The data was filtered with a high-pass filter with a 0.1 Hz half-amplitude cutoff and a low-pass filter with a 30 Hz half-amplitude cutoff (IIR-Butterworth). Independent component analysis (ICA; Bell and Sejnowski, 1995) was used to identify and remove eye blinks and muscle artifacts. ICA was used in order to follow the original authors’ procedure. The data was segmented into epochs of 5s that started -2.5s before stimulus onset and all epochs were corrected to baseline. The data was then down sampled to 250 Hz and visually inspected for any remaining artifacts. About 5% of the trials, on average per participant, were rejected due to artifacts. Average alpha power was taken from each epoch using the Pz electrode. The Pz electrode was used because it has been shown in alpha oscillation research that alpha is most prevalent in the parietal and occipital regions of the brain (Bazanova & Vernon, 2014; Niedermeyer, 2004; Sterman and Egner, 2006; Teplan, 2002; Treder et al., 2011). To examine if an effect was found among other oscillating waveforms, average gamma power (30-80 Hz) was also taken from each epoch using the Pz electrode.  Results To discover if there was a relationship with alpha power and time a Pearson’s correlation analysis was run between average alpha power and epoch time. The results demonstrated a significant positive correlation, indicating alpha power increased with time (r=0.062, p<.001; Table 1, Figure 2). To investigate if this effect was present among other frequency bands a correlation analysis was also run between average gamma power and epoch time. These results also demonstrated a significant positive correlation, suggesting that gamma power also increased with time (r=0.042, p<.001; Table 1, Figure 3).To understand how each frequency band covaried with time, a repeated measures ANOVA was run. The results demonstrated that there was a main effect among the different waveforms (F(1,8496)=76539.77, p<.001; Table 2) and an interaction between frequency band and time demonstrating that alpha and gamma covary with time differently (F(1,8496)=16.37, p<.001; Table 2). As post-hoc tests, two separate ANCOVAs were run between the different wavebands and time to examine effect sizes. The ANCOVA for alpha power proved to be significant (F(1,8496)=32.88, p<.001, ?2=0.004; Table 3). The ANCOVA for gamma power also proved to be significant, however, with a smaller effect size (F(1,8496)=15.19, p<.001, ?2=0.002; Table 4).Discussion The analysis of alpha power demonstrated a positive correlation between alpha power and time in the cognitive experiment. This suggests that the proposed hypothesis was correct in that alpha modulation does increase over experiment time. To understand if this effect was solely significant for alpha, the relationship between both waveforms and time was investigated. Ultimately, it was found that alpha and gamma covary with time differently. Further analysis demonstrated a positive significant correlation for gamma and experiment time as well. While a significant positive correlation was found for gamma, the effect size was smaller than that of alpha power indicating that the strength of correlation with time is higher for alpha power. The finding that alpha increases with time poses a serious problem for researchers who conduct EEG experiments and are not investigating alpha modulation. Presence of alpha has been linked to relaxation and drowsiness (Teplan, 2002). As such, these results suggest that the participants, over time in this experiment, were becoming increasingly sleepier. Obviously if participants are becoming sleepy, the researchers will have less usable data either due to more noise or missed responses. These findings have important implications for researchers to consider when designing their experiment. For example, it might be beneficial to keep trial times and the number of trials short. If that is not a possibility, researchers might need to consider giving frequent breaks to the participants where they can move around, eat snacks or do anything to help themselves wake up and focus.Future DirectionsThere are three main areas in which this research needs to be investigated further. The first requires a larger scale study that investigates many different types of experiments. It should be noted that this particular experiment consisted of 800 trials. For a participant, that is rather lengthy. Shorter studies might reveal a lesser effect of alpha power over time therefore suggesting alpha modulation is only a problem for much longer studies. If there is an effect in these shorter studies, that might even be more problematic for EEG researchers and force them to reconsider their entire experiment configuration. Likewise, investigations of experiments that allow participants adequate break time, snacks or the ability to move might prove to show interesting results as well. For example, we may see in those experiments less alpha modulation because participants are less likely to be falling asleep.The second area of further research involves analyzing behavioral data along with the EEG data. With this combination further analyses can be conducted to see if there is a correlation among alpha power, time and behavioral results. For example, participants who are falling asleep might demonstrate increased alpha activity as well as diminished behavioral results (i.e., inaccurate responses, delayed responses, missed responses). This type of analysis, therefore, could provide further evidence that alpha modulation over time is a problem for EEG researchers and require them to reconfigure experiments to account for such problems.Finally, the third area of further research that should be conducted is examining the effect at different electrode sites. This study looked at alpha power from a single electrode (Pz). It might benefit the research to investigate this question among different parietal and occipital electrodes to see if an effect is present. Further, it might be interesting to look at a group of electrodes and see if the same effect is present.Conclusion Overall, this research demonstrated that alpha power increases over time in a cognitive experiment. This suggests that participants might become increasingly drowsy and non-responsive over time in an experiment. This is detrimental, however, to researchers who need those critical trials in the end of the experiment. Further research should aim to investigate if this effect is related to poor behavioral responses and if an effect is seen in different experiments and electrodes. ReferencesBartsch, F., Hamuni, G., Miskovic, V., Lang, P. J., & Keil, A. (2015). Oscillatory brain activity in the alpha range is modulated by the content of word prompted mental imagery. Psychophysiology, 1-9.Basar, E. (1997). Preface. International Journal of Psychophysiology, 26, 1-3.Bazanova, O. M., & Vernon, D. (2014). Interpreting EEG alpha activity. Neuroscience and Biobehavioral Reviews, 44, 94-110.Bell, A.J., & Sejnowski, T. (1995). An Information-Maximization approach to blind separation and blind deconvolution. Neural Computation, 7(6), 1129-1159. doi:10.1162/neco.1995.7.6.1129.Benwell, C. S. Y., Tagliabue, C. F., Veniero, D., Cecere, R., Savazzi, S., & Thut, G. (2017). Prestimulus EEG power predicts conscious awareness but not objective visual performance.Brainard, D. H. (1997). The Psychophysics Toolbox. Spatial Vision, 10(4), 433-436.http://dx.doi.org/10.1163/156856897X00357Cohen, M. X. (2014). Analyzing Neural Time Serires Data: Theory and Practice. Cambridge, Massachusetts: MIT Press.Klimesch, W., Sauseng, P., & Hanslmayr, S. (2007). EEG alpha oscillations: The inhibition-timing hypothesis. Brain Research Reviews, 53, 63-88. Luck, S. J. (2014). An Introduction to the Event-Related Potential Technique (2nd ed.): MIT Press.Pelli, D. (1997). The VideoToolbox software for visual psychophysics: Transforming numbers into movies. Spatial Vision, 10, 437-442. doi:10.1163/156856897X00366.Ramsoy, T.Z., Overgaard, M. (2011). Introspection and subliminal perception. Phenomenology and Cognitive Sciences, 3(1), 1-23.Schurmann, M., & Basar, E. (2001). Functional aspects of alpha oscillations in the EEG. International Journal of Psychophysiology, 39, 151-158.Teplan, M. (2002). Fundamentals of EEG measurement. Measurement Science Review, 2(2), 1-11.Appendix ATable 1Correlation Matrix of Alpha Power and Gamma Power with TimePearson Correlations   Pearson’s rpAlpha Power–Gamma Power0.138***< .001Alpha Power–Time0.062***< .001Gamma Power–Time0.042***< .001* p < .05, ** p < .01, *** p < .001Table 2Repeated Measures ANOVA of Frequency Band and TimeWithin Subjects Effects Sum of SquaresdfMean SquareFpFrequency Band1.160e?+611.160e?+676539.77< .001Frequency Band ? Time248.01248.0216.37< .001Residual128729.78496 15.15Note.  Type III Sum of SquaresTable 3ANCOVA of Alpha Power and TimeANCOVA – Alpha PowerCasesSum of SquaresdfMean SquareFp?²Time945.11945.0932.88< .0010.004Residual244208.8849628.74Note.  Type III Sum of SquaresTable 4ANCOVA of Gamma Power and TimeANCOVA – Gamma PowerCasesSum of SquaresdfMean SquareFp?²Time71.75171.75015.19< .0010.002Residual40130.8084964.723Note.  Type III Sum of SquaresFigure 1Experimental Design (Benwell et al., 2017)Figure 2Scatter Plot of Alpha Power Correlation with TimeFigure 3Scatter Plot of Gamma Power Correlation with Time Get Help With Your EssayIf you need assistance with writing your essay, our professional essay writing service is here to help!Find out more