- Email: [email protected]
The time-course of alpha neurofeedback training effects in healthy participants
ARTICLE IN PRESS
G Model BIOPSY 6817 1–4
Biological Psychology xxx (2013) xxx–xxx
Contents lists available at ScienceDirect
Biological Psychology journal homepage: www.elsevier.com/locate/biopsycho
The time-course of alpha neurofeedback training effects in healthy participants
1
2
3 4 5
Q1
Marian K.J. Dekker a,b , Margriet M. Sitskoorn a , Ad J.M. Denissen b , Geert J.M. van Boxtel a,∗ a b
Department of Cognitive Neuropsychology, Tilburg University, Warandelaan 2, Tilburg 5037 AB, The Netherlands Brain Body and Behaviour group, Philips Research Eindhoven, High Tech Campus 5, Eindhoven 5656 AE, The Netherlands
6
7
a r t i c l e
i n f o
a b s t r a c t
8 9 10 11
Article history: Received 4 April 2013 Accepted 27 November 2013 Available online xxx
12
17
Keywords: Neurofeedback training (NFT) Lower alpha activity Upper alpha activity Healthy participants
18
1. Introduction
13 14 15 16
19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39
The time-course of alpha neurofeedback training (NFT) was investigated in 18 healthy participants who received 15 sessions of training (eyes open), each consisting of three training periods (data are from Van Boxtel et al., 2012). Here we report on the within- and between-session training effects using multilevel analyses. Over sessions, total alpha power (8–12 Hz) increased up to the tenth session, after which low alpha power (8–10 Hz) remained at the same level, while high alpha power (10–12 Hz) decreased. Within each training session, total alpha power increased from the first to the second period, and then decreased again. This decrease, however, was caused by a decrease in high alpha power only; low alpha power remained up to the end of training. These effects are discussed in terms of attention and motivation, and suggest different trainability for low and high alpha power. © 2013 Published by Elsevier B.V.
In a recent study, Van Boxtel et al. (2012) introduced a novel neurofeedback training (NFT) system based on a commercially available audio headset with water-based electrodes mounted in the band between the ear-pads. They tested the system in healthy participants who received alpha (8–12 Hz) training in 15 daily sessions while listening to their favourite music, of which the quality was adapted as a function of the alpha power. Each session consisted of three 8-min periods of NFT, interspersed with cognitive tasks, intended to prevent the participants from falling asleep while at the same time allowing the study of the effects of alpha training on cognition. Total alpha power was indeed increased after training. Interestingly, a further increase was observed in a 3 month followup measurement without further training. The training appeared to be self-guided in that participants received no instructions and still managed to increase the alpha power. Here we report on the learning effects of the alpha NFT in that study. We wanted to know how alpha training progressed over sessions, but also between the three training periods within a single training session. If alpha power over sessions would reach an asymptote, this would suggest that the number of training sessions was appropriate or could be lower. In clinical applications
∗ Corresponding author. Tel.: +31 13 466 2492; fax: +31 13 466 2067. E-mail addresses: [email protected] (M.K.J. Dekker), [email protected] (M.M. Sitskoorn), [email protected] (A.J.M. Denissen), [email protected] (G.J.M. van Boxtel).
individuals often need repetitive sessions (20–60) for clear effects to occur at late stages of training (e.g., Ros & Gruzelier, 2010). Nonclinical applications often use fewer sessions (1–30; e.g., Myers & Young, 2011; Ros & Gruzelier, 2010; Vernon, 2005), and we wanted to make sure that 15 sessions were appropriate. Analysing the training data allowed us to check whether the design choices were appropriate, while at the same time demonstrating that learning actually occurred. In the original report (Van Boxtel et al., 2012) only total alpha power was analysed. However, as lower and upper alpha bands may be differentially involved in selective attention (e.g., Klimesch, Pfurtscheller, Mohl, & Schimke, 1990; Klimesch, Sauseng, & Hansmayr, 2007) and cognitive processing (e.g., Doppelmayr, Klimesch, Hodlmoser, Sauseng, & Gruber, 2005; Zoefel, Huster, & Sherrmann, 2010), we analysed lower (8–10 Hz) and upper (10–12 Hz) alpha bands separately. It is largely unknown whether lower and higher alpha bands differ in trainability, but a few isolated studies suggest that this may be the case (Aftanas & Golocheikine, 2001; Bazanova, 2012).
40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58
2. Methods
59
2.1. Participants
60
Eighteen healthy right-handed participants (6 males and 12 females; M age = 20.7 years, 18–25, S.D. 1.79 years; Group A in Van Boxtel et al., 2012) were participated.
0301-0511/$ – see front matter © 2013 Published by Elsevier B.V. http://dx.doi.org/10.1016/j.biopsycho.2013.11.014
Please cite this article in press as: Dekker, M. K. J., et al. The time-course of alpha neurofeedback training effects in healthy participants. Biol. Psychol. (2013), http://dx.doi.org/10.1016/j.biopsycho.2013.11.014
61 62 63
G Model BIOPSY 6817 1–4
ARTICLE IN PRESS M.K.J. Dekker et al. / Biological Psychology xxx (2013) xxx–xxx
2.2. Design and procedure
83
The NFT sessions took place in a normal office room. All participants underwent 15 NFT sessions on consecutive working days. One training session always consisted of the same sequence of tasks: 5 min baseline with eyes opened (EO), 5 min baseline with eyes closed (EC), 5 min cognitive task, 8 min neurofeedback training (NFT1), 5 min cognitive task, 8 min NFT2, 5 min cognitive task, 8 min NFT3, 5 min cognitive task. The NFT was always performed with eyes open. Training was given by auditory feedback, presented in the form of quality of participants’ favourite music that they listened to through a headset in which 4 water-based electroencephalography (EEG) electrodes (Ag/AgCl ring) were mounted (roughly at C3 and C4, referred to mastoids). The NFT procedure (see Van Boxtel et al., 2012 for details) resulted in a very intuitive feedback mechanism in which a person’s own well-known favourite music sounded thin and distant if alpha levels were low; and sounded rich and full, as the participants were used to hearing it, when alpha levels were high. Participants received no instructions for the training; they were not told that they were expected to increase the quality of the music.
84
2.3. Data analysis
65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82
(a) -.70 Alpha power (log10 µV2 )
64
Total alpha power M -.84 SEM .06
-.80 -.90
M -0.95 SEM .06
-1.00 -1.10 -1.20
Grou p A Grou p A (fit)
M -1.19 SEM .06
-1.30
(b) Alpha power (log10 µV2 )
2
1
4
7
10
13
Session
Low alpha power -.30 -.35
M -.37 SEM .03
-.40 M -.34 SEM .03
-.45 -.50
Group A Group A (fit)
M -.55 SEM .03
-.55 -.60
93
2.4. Statistical analysis
86 87 88 89 90 91
108
Linear mixed models (Pinheiro & Bates, 2000) were used for all analyses. For between-session comparisons of the alpha power, the eyes open (EO) baseline data were analysed with fixed withinsubjects factors session (15 levels) and electrode (C3, C4). For within-session training, the fixed within-subjects factors were period (3 levels), session (15 levels) and electrode (C3, C4). Polynomial contrasts were computed for the session factor; only linear and quadratic are reported here because no higher-order contrasts were found. We used an AR1 covariance structure with random intercept, which resulted in the best models, determined by Akaike’s information criterion. All tests were computed separately for low, high, and total alpha power. NFT responders versus non-responders were determined by positive and negative difference values for the alpha power between training sessions 1 and 15.
109
3. Results
94 95 96 97 98 99 100 101 102 103 104 105 106 107
110 111 112
113
In all analyses the fixed factor electrode was removed because the alpha power appeared not to differ between positions (F(1,290) = 0.97, p = 0.33).
(c) Alpha power (log10 µV2 )
92
The EO baseline period and the three NFT periods were used in the analyses. See Van Boxtel et al. (2012) for details about eye movement correction, filtering and segmenting of the EEG training data. The segments were transformed to a frequency domain using a Hanning window for tapering. The FFT power values were transformed to a log10 scale and all frequency components (0.25 Hz resolution) were averaged to form the frequency bands low alpha (8–10 Hz), high alpha (10–12 Hz) and total alpha (8–12 Hz).
85
1
4
7
10
13
Session
High alpha power
-.40
M -.50 SEM .04
-.45 -.50
M -.57 SEM .03
-.55 -.60 -.65
M -.64 SEM .04
-.70 1
4
7
10
13
Session
Fig. 1. Total alpha power (a), low alpha power (b), and high alpha power (c) on position C3 as a function of session, including the fitted quadratic functions by linear mixed models analysis. Error bars present SEM.
1 to 10 (F(1, 74) = 33.4, p < 0.001), and tended to remain unchanged thereafter (session 10–15: F(1, 42) = 3.3, p = 0.08). Fig. 1 also presents both low- (Fig. 1b) and high (Fig. 1c) alpha power as a function of session. There were similar linear and quadratic effects of session (low alpha power: linear F(1,82) = 28.7, p < 0.001, quadratic F(1,120) = 15.7, p < 0.001; high alpha power: linear F(1,104) = 4.9, p = 0.03, quadratic F(1,132) = 6.9, p = 0.01). Sixteen participants were found to be responders in low alpha power, and 13 in high alpha power. In both low- and high alpha power, a linear increase between session 1 and 10 was found (low alpha power: F(1,66) = 42.7, p < 0.001; high alpha power: F(1,78) = 14.4, p < 0.001), and this increase did not differ statistically between both sub-bands (F(1,126) = 3.3, p = 0.07). After session 10, the low alpha power remained unchanged (linear: F(1,45) = 1.56, p = 0.22), whereas the high alpha power decreased (linear: F(1,42) = 4.3, p = 0.045).
115 116 117 118 119
121 122 123 124 125 126 127 128 129 130 131 132 133 134 135
3.1. Alpha power between training sessions 3.2. Alpha power within training sessions
114
120
Fig. 1a shows the total (baseline) alpha power as a function of Session. As the figure suggests, there were linear (F(1,92) = 17.4, p < 0.001) and quadratic (F(1,125) = 12.9, p < 0.001) effects of sessions. Fifteen participants were found to be responders. Alpha power seemed to show a maximum around session 10. Statistically, the total alpha power indeed showed a linear increase from session
Similar to the baseline periods, the total alpha power of the three NFT periods averaged together showed a linear increase (F(1,182) = 11.4, p = 0.001) as fitted over the 15 training sessions (session 1: M −1.11, SEM .05; session 15: M −.97, SEM .06). This fitted linear increase, however, was limited to lower alpha power
Please cite this article in press as: Dekker, M. K. J., et al. The time-course of alpha neurofeedback training effects in healthy participants. Biol. Psychol. (2013), http://dx.doi.org/10.1016/j.biopsycho.2013.11.014
136
137 138 139 140 141
ARTICLE IN PRESS
G Model BIOPSY 6817 1–4
M.K.J. Dekker et al. / Biological Psychology xxx (2013) xxx–xxx
(a)
4. Discussion
Total alpha power within one session group A Alpha power (log10 µV2 )
-.80 -.85 -.90
M -.94 SEM .08
M -.98 SEM .06
-.95
M -.97 SEM .06
-1.00 -1.05 -1.10 1
(b)
2
3
Period
Low alpha powergroup within A one session
Alpha power (log10 µV2 )
-.30 -.35
M -.38 SEM .04
-.40 -.45 -.50
M -.41 SEM .04
M -.38 SEM .04
-.55 -.60 1
(c)
2
3
Period
High alpha powergroup within A one session
Alpha power (log10 µV2 )
-0.40 -0.45 -0.50
M -.57 SEM .03
M -.56 SEM .04
-0.55 M -.59 SEM .06
-0.60 -0.65 -0.70 1
2
3
Period
Fig. 2. Total alpha power (a), low alpha power (b), and high alpha power (c) on position C3 as a function of (NFT) tasks. The figure presents data of training session 15. Error bars present SEM.
142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161
3
(F(1,163) = 18.7, p < 0.001; session 1: M −.49, SEM −.02; session 15: M −.38, SEM .03), and was not significant in high alpha power (F(1,200) = 2.92, p = 0.09; session 1: M −.62, SEM .03; session 15: M −.58, SEM .03). Fig. 2a presents the total alpha power during the three NFT periods within a single training session (session 15). Alpha power increased from the first to the second periods (F(1,866) = 3.9, p = 0.046), but decreased between the second and third periods (F(1,852) = 6.8, p = 0.01). This pattern appeared to be constant over the 15 training sessions (period × linear effect of session; F(2,1390) = 2.76, p = 0.06). Fig. 2 presents both low (Fig. 2b) and high (Fig. 2c) alpha power of the three NFT periods within a training session. The increase between the first two periods, averaged over training sessions, was found in low alpha power as well (F(1,880) = 13.8, p < 0.001). Low alpha power did not change between the second and third NFT periods, however (F(1,859) = 2.43, p = 0.12). High alpha power was statistically indistinguishable between the first and second NFT periods (F(1,858) = 0.002, p = 0.97), and decreased from the second to the third periods (F(1,848) = 10.8, p = 0.001).
Clear indices of learning were observed with the novel NFT system. Alpha power increased both between and within training sessions. The results suggest that 10 training sessions are sufficient for this increase to develop, whereas the last 5 sessions did not result in further change. A similar pattern was found for training within a session; an increase from the first to the second training periods, and a decrease towards the end of the session. We interpret these results as time-on-tasks effects; it seems that the participants grew tired after some time and could not sustain attention, an effect which has shown in previous neurofeedback research (e.g. Gruzelier, 2013). Although the NFT was instructionless and perceived as relaxing, this might still have influenced the participant’s ability to regulate EEG alpha activity. We also found differences between alpha sub-bands. In contrast to baseline measures, the upper alpha power measured during the three NFT periods did not show any increase, neither within nor between sessions. We consider a ceiling effect in high alpha power unlikely, because lower alpha power showed larger values than high alpha power, and because training was done with eyes open. A possible explanation might be that every NFT period was preceded by a cognitive task, which might specifically have influenced high alpha power, which has been suggested to play an important role in cognitive performance (e.g., Doppelmayr, Klimesch, Hodlmoser, Sauseng, & Gruber, 2005; Zoefel, Huster, & Sherrmann, 2010). However, upper alpha power was also different from lower alpha power in the progression over sessions of baseline power, and the baseline period was not preceded by cognitive tasks. After the initial increase in alpha power in both sub-bands from session 1 to 10, upper alpha power decreased in the last 5 sessions, whereas lower alpha power remained stable. It is an intriguing possibility that training of the lower alpha band, often associated with decreased vigilance and increased internal attention (Aftanas & Golocheikine, 2001; Klimesch et al., 2007) may inherently be easier to accomplish, and that the effects may persist longer after the training. Bazanova (2012), who reported on differences in test performance between participants with individual alpha frequency in the high or low alpha band, suggested that such differences may be of neurophysiological origin. To summarise, we showed between- and within-session learning effects of the alpha NFT system previously reported by Van Boxtel et al. (2012). These effects were different for the lower and upper alpha bands in that power decreased at the end of each single session as well as at the end of the series of sessions, for the upper but not the lower alpha band.
References Aftanas, L. I., & Golocheikine, S. A. (2001). Human anterior and frontal midline theta and lower alpha reflect emotionally positive state and internalized attention: High resolution EEG investigation of meditation. Neuroscience Letters, 310, 57–60. Bazanova, O. M. (2012). Comments for current interpretation EEG alpha activity: A review and analysis. Journal of Behavioral and Brain Science, 2, 239–248. Doppelmayr, M., Klimesch, W., Hodlmoser, K., Sauseng, P., & Gruber, W. (2005). Intelligence related upper alpha desynchronization in a semantic memory task. Brain Research Bulletin, 66(2), 171–177. Gruzelier, J. H. (2013). Differential effects on mood of 12-15 (SMR) and 15-18 (beta1) Q2 Hz neurofeedback. International Journal of Psychophysiology (in press). Klimesch, W., Pfurtscheller, G., Mohl, W., & Schimke, H. (1990). Event-related desynchronization. ERD-mapping and hemispheric differences for words and numbers. International Journal of Psychophysiology, 8(3), 297–308. Klimesch, W., Sauseng, P., & Hanslmayr, S. (2007). EEG alpha oscillations, the inhibition-timing hypothesis. Brain Research Reviews, 53(1), 63–88. Myers, J. E., & Young, J. S. (2011). Brain wave biofeedback: Benefits of integrating neurofeedback in counseling. Journal of Counseling & Development, 90(1), 20–28. Pinheiro, J. C., & Bates, D. M. (2000). Linear mixed effects models: Basic concepts and examples. New York: Springer-Verlag.
Please cite this article in press as: Dekker, M. K. J., et al. The time-course of alpha neurofeedback training effects in healthy participants. Biol. Psychol. (2013), http://dx.doi.org/10.1016/j.biopsycho.2013.11.014
162
163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206
207
208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227
G Model BIOPSY 6817 1–4 4
228 229 230 231 232
ARTICLE IN PRESS M.K.J. Dekker et al. / Biological Psychology xxx (2013) xxx–xxx
Ros, T., & Gruzelier, J. H. (2010). The immediate effects of EEG neurofeedback on cortical excitability and synchronization. Neurofeedback and neuromodulation techniques and applications. Elsevier. Van Boxtel, G. J. M., Denissen, A., Jager, M., Vernon, D., Dekker, M. K. J., Mihaljovic, V., et al. (2012). A novel self-guided approach to alpha activity training. International Journal of Psychophysiology, 83(3), 282–294.
Vernon, D. J. (2005). Can neurofeedback training enhance performance? An evaluation of the evidence with implications for future research. Applied Psychophysiology and Biofeedback, 30, 347–364. Zoefel, B., Huster, R. J., & Herrmann, C. S. (2010). Neurofeedback training of the upper alpha frequency band in EEG improves cognitive performance. NeuroImage, 54(2), 1427–1431.
Please cite this article in press as: Dekker, M. K. J., et al. The time-course of alpha neurofeedback training effects in healthy participants. Biol. Psychol. (2013), http://dx.doi.org/10.1016/j.biopsycho.2013.11.014
233 234 235 236 237 238
G Model BIOPSY 6817 1–4
Biological Psychology xxx (2013) xxx–xxx
Contents lists available at ScienceDirect
Biological Psychology journal homepage: www.elsevier.com/locate/biopsycho
The time-course of alpha neurofeedback training effects in healthy participants
1
2
3 4 5
Q1
Marian K.J. Dekker a,b , Margriet M. Sitskoorn a , Ad J.M. Denissen b , Geert J.M. van Boxtel a,∗ a b
Department of Cognitive Neuropsychology, Tilburg University, Warandelaan 2, Tilburg 5037 AB, The Netherlands Brain Body and Behaviour group, Philips Research Eindhoven, High Tech Campus 5, Eindhoven 5656 AE, The Netherlands
6
7
a r t i c l e
i n f o
a b s t r a c t
8 9 10 11
Article history: Received 4 April 2013 Accepted 27 November 2013 Available online xxx
12
17
Keywords: Neurofeedback training (NFT) Lower alpha activity Upper alpha activity Healthy participants
18
1. Introduction
13 14 15 16
19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39
The time-course of alpha neurofeedback training (NFT) was investigated in 18 healthy participants who received 15 sessions of training (eyes open), each consisting of three training periods (data are from Van Boxtel et al., 2012). Here we report on the within- and between-session training effects using multilevel analyses. Over sessions, total alpha power (8–12 Hz) increased up to the tenth session, after which low alpha power (8–10 Hz) remained at the same level, while high alpha power (10–12 Hz) decreased. Within each training session, total alpha power increased from the first to the second period, and then decreased again. This decrease, however, was caused by a decrease in high alpha power only; low alpha power remained up to the end of training. These effects are discussed in terms of attention and motivation, and suggest different trainability for low and high alpha power. © 2013 Published by Elsevier B.V.
In a recent study, Van Boxtel et al. (2012) introduced a novel neurofeedback training (NFT) system based on a commercially available audio headset with water-based electrodes mounted in the band between the ear-pads. They tested the system in healthy participants who received alpha (8–12 Hz) training in 15 daily sessions while listening to their favourite music, of which the quality was adapted as a function of the alpha power. Each session consisted of three 8-min periods of NFT, interspersed with cognitive tasks, intended to prevent the participants from falling asleep while at the same time allowing the study of the effects of alpha training on cognition. Total alpha power was indeed increased after training. Interestingly, a further increase was observed in a 3 month followup measurement without further training. The training appeared to be self-guided in that participants received no instructions and still managed to increase the alpha power. Here we report on the learning effects of the alpha NFT in that study. We wanted to know how alpha training progressed over sessions, but also between the three training periods within a single training session. If alpha power over sessions would reach an asymptote, this would suggest that the number of training sessions was appropriate or could be lower. In clinical applications
∗ Corresponding author. Tel.: +31 13 466 2492; fax: +31 13 466 2067. E-mail addresses: [email protected] (M.K.J. Dekker), [email protected] (M.M. Sitskoorn), [email protected] (A.J.M. Denissen), [email protected] (G.J.M. van Boxtel).
individuals often need repetitive sessions (20–60) for clear effects to occur at late stages of training (e.g., Ros & Gruzelier, 2010). Nonclinical applications often use fewer sessions (1–30; e.g., Myers & Young, 2011; Ros & Gruzelier, 2010; Vernon, 2005), and we wanted to make sure that 15 sessions were appropriate. Analysing the training data allowed us to check whether the design choices were appropriate, while at the same time demonstrating that learning actually occurred. In the original report (Van Boxtel et al., 2012) only total alpha power was analysed. However, as lower and upper alpha bands may be differentially involved in selective attention (e.g., Klimesch, Pfurtscheller, Mohl, & Schimke, 1990; Klimesch, Sauseng, & Hansmayr, 2007) and cognitive processing (e.g., Doppelmayr, Klimesch, Hodlmoser, Sauseng, & Gruber, 2005; Zoefel, Huster, & Sherrmann, 2010), we analysed lower (8–10 Hz) and upper (10–12 Hz) alpha bands separately. It is largely unknown whether lower and higher alpha bands differ in trainability, but a few isolated studies suggest that this may be the case (Aftanas & Golocheikine, 2001; Bazanova, 2012).
40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58
2. Methods
59
2.1. Participants
60
Eighteen healthy right-handed participants (6 males and 12 females; M age = 20.7 years, 18–25, S.D. 1.79 years; Group A in Van Boxtel et al., 2012) were participated.
0301-0511/$ – see front matter © 2013 Published by Elsevier B.V. http://dx.doi.org/10.1016/j.biopsycho.2013.11.014
Please cite this article in press as: Dekker, M. K. J., et al. The time-course of alpha neurofeedback training effects in healthy participants. Biol. Psychol. (2013), http://dx.doi.org/10.1016/j.biopsycho.2013.11.014
61 62 63
G Model BIOPSY 6817 1–4
ARTICLE IN PRESS M.K.J. Dekker et al. / Biological Psychology xxx (2013) xxx–xxx
2.2. Design and procedure
83
The NFT sessions took place in a normal office room. All participants underwent 15 NFT sessions on consecutive working days. One training session always consisted of the same sequence of tasks: 5 min baseline with eyes opened (EO), 5 min baseline with eyes closed (EC), 5 min cognitive task, 8 min neurofeedback training (NFT1), 5 min cognitive task, 8 min NFT2, 5 min cognitive task, 8 min NFT3, 5 min cognitive task. The NFT was always performed with eyes open. Training was given by auditory feedback, presented in the form of quality of participants’ favourite music that they listened to through a headset in which 4 water-based electroencephalography (EEG) electrodes (Ag/AgCl ring) were mounted (roughly at C3 and C4, referred to mastoids). The NFT procedure (see Van Boxtel et al., 2012 for details) resulted in a very intuitive feedback mechanism in which a person’s own well-known favourite music sounded thin and distant if alpha levels were low; and sounded rich and full, as the participants were used to hearing it, when alpha levels were high. Participants received no instructions for the training; they were not told that they were expected to increase the quality of the music.
84
2.3. Data analysis
65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82
(a) -.70 Alpha power (log10 µV2 )
64
Total alpha power M -.84 SEM .06
-.80 -.90
M -0.95 SEM .06
-1.00 -1.10 -1.20
Grou p A Grou p A (fit)
M -1.19 SEM .06
-1.30
(b) Alpha power (log10 µV2 )
2
1
4
7
10
13
Session
Low alpha power -.30 -.35
M -.37 SEM .03
-.40 M -.34 SEM .03
-.45 -.50
Group A Group A (fit)
M -.55 SEM .03
-.55 -.60
93
2.4. Statistical analysis
86 87 88 89 90 91
108
Linear mixed models (Pinheiro & Bates, 2000) were used for all analyses. For between-session comparisons of the alpha power, the eyes open (EO) baseline data were analysed with fixed withinsubjects factors session (15 levels) and electrode (C3, C4). For within-session training, the fixed within-subjects factors were period (3 levels), session (15 levels) and electrode (C3, C4). Polynomial contrasts were computed for the session factor; only linear and quadratic are reported here because no higher-order contrasts were found. We used an AR1 covariance structure with random intercept, which resulted in the best models, determined by Akaike’s information criterion. All tests were computed separately for low, high, and total alpha power. NFT responders versus non-responders were determined by positive and negative difference values for the alpha power between training sessions 1 and 15.
109
3. Results
94 95 96 97 98 99 100 101 102 103 104 105 106 107
110 111 112
113
In all analyses the fixed factor electrode was removed because the alpha power appeared not to differ between positions (F(1,290) = 0.97, p = 0.33).
(c) Alpha power (log10 µV2 )
92
The EO baseline period and the three NFT periods were used in the analyses. See Van Boxtel et al. (2012) for details about eye movement correction, filtering and segmenting of the EEG training data. The segments were transformed to a frequency domain using a Hanning window for tapering. The FFT power values were transformed to a log10 scale and all frequency components (0.25 Hz resolution) were averaged to form the frequency bands low alpha (8–10 Hz), high alpha (10–12 Hz) and total alpha (8–12 Hz).
85
1
4
7
10
13
Session
High alpha power
-.40
M -.50 SEM .04
-.45 -.50
M -.57 SEM .03
-.55 -.60 -.65
M -.64 SEM .04
-.70 1
4
7
10
13
Session
Fig. 1. Total alpha power (a), low alpha power (b), and high alpha power (c) on position C3 as a function of session, including the fitted quadratic functions by linear mixed models analysis. Error bars present SEM.
1 to 10 (F(1, 74) = 33.4, p < 0.001), and tended to remain unchanged thereafter (session 10–15: F(1, 42) = 3.3, p = 0.08). Fig. 1 also presents both low- (Fig. 1b) and high (Fig. 1c) alpha power as a function of session. There were similar linear and quadratic effects of session (low alpha power: linear F(1,82) = 28.7, p < 0.001, quadratic F(1,120) = 15.7, p < 0.001; high alpha power: linear F(1,104) = 4.9, p = 0.03, quadratic F(1,132) = 6.9, p = 0.01). Sixteen participants were found to be responders in low alpha power, and 13 in high alpha power. In both low- and high alpha power, a linear increase between session 1 and 10 was found (low alpha power: F(1,66) = 42.7, p < 0.001; high alpha power: F(1,78) = 14.4, p < 0.001), and this increase did not differ statistically between both sub-bands (F(1,126) = 3.3, p = 0.07). After session 10, the low alpha power remained unchanged (linear: F(1,45) = 1.56, p = 0.22), whereas the high alpha power decreased (linear: F(1,42) = 4.3, p = 0.045).
115 116 117 118 119
121 122 123 124 125 126 127 128 129 130 131 132 133 134 135
3.1. Alpha power between training sessions 3.2. Alpha power within training sessions
114
120
Fig. 1a shows the total (baseline) alpha power as a function of Session. As the figure suggests, there were linear (F(1,92) = 17.4, p < 0.001) and quadratic (F(1,125) = 12.9, p < 0.001) effects of sessions. Fifteen participants were found to be responders. Alpha power seemed to show a maximum around session 10. Statistically, the total alpha power indeed showed a linear increase from session
Similar to the baseline periods, the total alpha power of the three NFT periods averaged together showed a linear increase (F(1,182) = 11.4, p = 0.001) as fitted over the 15 training sessions (session 1: M −1.11, SEM .05; session 15: M −.97, SEM .06). This fitted linear increase, however, was limited to lower alpha power
Please cite this article in press as: Dekker, M. K. J., et al. The time-course of alpha neurofeedback training effects in healthy participants. Biol. Psychol. (2013), http://dx.doi.org/10.1016/j.biopsycho.2013.11.014
136
137 138 139 140 141
ARTICLE IN PRESS
G Model BIOPSY 6817 1–4
M.K.J. Dekker et al. / Biological Psychology xxx (2013) xxx–xxx
(a)
4. Discussion
Total alpha power within one session group A Alpha power (log10 µV2 )
-.80 -.85 -.90
M -.94 SEM .08
M -.98 SEM .06
-.95
M -.97 SEM .06
-1.00 -1.05 -1.10 1
(b)
2
3
Period
Low alpha powergroup within A one session
Alpha power (log10 µV2 )
-.30 -.35
M -.38 SEM .04
-.40 -.45 -.50
M -.41 SEM .04
M -.38 SEM .04
-.55 -.60 1
(c)
2
3
Period
High alpha powergroup within A one session
Alpha power (log10 µV2 )
-0.40 -0.45 -0.50
M -.57 SEM .03
M -.56 SEM .04
-0.55 M -.59 SEM .06
-0.60 -0.65 -0.70 1
2
3
Period
Fig. 2. Total alpha power (a), low alpha power (b), and high alpha power (c) on position C3 as a function of (NFT) tasks. The figure presents data of training session 15. Error bars present SEM.
142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161
3
(F(1,163) = 18.7, p < 0.001; session 1: M −.49, SEM −.02; session 15: M −.38, SEM .03), and was not significant in high alpha power (F(1,200) = 2.92, p = 0.09; session 1: M −.62, SEM .03; session 15: M −.58, SEM .03). Fig. 2a presents the total alpha power during the three NFT periods within a single training session (session 15). Alpha power increased from the first to the second periods (F(1,866) = 3.9, p = 0.046), but decreased between the second and third periods (F(1,852) = 6.8, p = 0.01). This pattern appeared to be constant over the 15 training sessions (period × linear effect of session; F(2,1390) = 2.76, p = 0.06). Fig. 2 presents both low (Fig. 2b) and high (Fig. 2c) alpha power of the three NFT periods within a training session. The increase between the first two periods, averaged over training sessions, was found in low alpha power as well (F(1,880) = 13.8, p < 0.001). Low alpha power did not change between the second and third NFT periods, however (F(1,859) = 2.43, p = 0.12). High alpha power was statistically indistinguishable between the first and second NFT periods (F(1,858) = 0.002, p = 0.97), and decreased from the second to the third periods (F(1,848) = 10.8, p = 0.001).
Clear indices of learning were observed with the novel NFT system. Alpha power increased both between and within training sessions. The results suggest that 10 training sessions are sufficient for this increase to develop, whereas the last 5 sessions did not result in further change. A similar pattern was found for training within a session; an increase from the first to the second training periods, and a decrease towards the end of the session. We interpret these results as time-on-tasks effects; it seems that the participants grew tired after some time and could not sustain attention, an effect which has shown in previous neurofeedback research (e.g. Gruzelier, 2013). Although the NFT was instructionless and perceived as relaxing, this might still have influenced the participant’s ability to regulate EEG alpha activity. We also found differences between alpha sub-bands. In contrast to baseline measures, the upper alpha power measured during the three NFT periods did not show any increase, neither within nor between sessions. We consider a ceiling effect in high alpha power unlikely, because lower alpha power showed larger values than high alpha power, and because training was done with eyes open. A possible explanation might be that every NFT period was preceded by a cognitive task, which might specifically have influenced high alpha power, which has been suggested to play an important role in cognitive performance (e.g., Doppelmayr, Klimesch, Hodlmoser, Sauseng, & Gruber, 2005; Zoefel, Huster, & Sherrmann, 2010). However, upper alpha power was also different from lower alpha power in the progression over sessions of baseline power, and the baseline period was not preceded by cognitive tasks. After the initial increase in alpha power in both sub-bands from session 1 to 10, upper alpha power decreased in the last 5 sessions, whereas lower alpha power remained stable. It is an intriguing possibility that training of the lower alpha band, often associated with decreased vigilance and increased internal attention (Aftanas & Golocheikine, 2001; Klimesch et al., 2007) may inherently be easier to accomplish, and that the effects may persist longer after the training. Bazanova (2012), who reported on differences in test performance between participants with individual alpha frequency in the high or low alpha band, suggested that such differences may be of neurophysiological origin. To summarise, we showed between- and within-session learning effects of the alpha NFT system previously reported by Van Boxtel et al. (2012). These effects were different for the lower and upper alpha bands in that power decreased at the end of each single session as well as at the end of the series of sessions, for the upper but not the lower alpha band.
References Aftanas, L. I., & Golocheikine, S. A. (2001). Human anterior and frontal midline theta and lower alpha reflect emotionally positive state and internalized attention: High resolution EEG investigation of meditation. Neuroscience Letters, 310, 57–60. Bazanova, O. M. (2012). Comments for current interpretation EEG alpha activity: A review and analysis. Journal of Behavioral and Brain Science, 2, 239–248. Doppelmayr, M., Klimesch, W., Hodlmoser, K., Sauseng, P., & Gruber, W. (2005). Intelligence related upper alpha desynchronization in a semantic memory task. Brain Research Bulletin, 66(2), 171–177. Gruzelier, J. H. (2013). Differential effects on mood of 12-15 (SMR) and 15-18 (beta1) Q2 Hz neurofeedback. International Journal of Psychophysiology (in press). Klimesch, W., Pfurtscheller, G., Mohl, W., & Schimke, H. (1990). Event-related desynchronization. ERD-mapping and hemispheric differences for words and numbers. International Journal of Psychophysiology, 8(3), 297–308. Klimesch, W., Sauseng, P., & Hanslmayr, S. (2007). EEG alpha oscillations, the inhibition-timing hypothesis. Brain Research Reviews, 53(1), 63–88. Myers, J. E., & Young, J. S. (2011). Brain wave biofeedback: Benefits of integrating neurofeedback in counseling. Journal of Counseling & Development, 90(1), 20–28. Pinheiro, J. C., & Bates, D. M. (2000). Linear mixed effects models: Basic concepts and examples. New York: Springer-Verlag.
Please cite this article in press as: Dekker, M. K. J., et al. The time-course of alpha neurofeedback training effects in healthy participants. Biol. Psychol. (2013), http://dx.doi.org/10.1016/j.biopsycho.2013.11.014
162
163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206
207
208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227
G Model BIOPSY 6817 1–4 4
228 229 230 231 232
ARTICLE IN PRESS M.K.J. Dekker et al. / Biological Psychology xxx (2013) xxx–xxx
Ros, T., & Gruzelier, J. H. (2010). The immediate effects of EEG neurofeedback on cortical excitability and synchronization. Neurofeedback and neuromodulation techniques and applications. Elsevier. Van Boxtel, G. J. M., Denissen, A., Jager, M., Vernon, D., Dekker, M. K. J., Mihaljovic, V., et al. (2012). A novel self-guided approach to alpha activity training. International Journal of Psychophysiology, 83(3), 282–294.
Vernon, D. J. (2005). Can neurofeedback training enhance performance? An evaluation of the evidence with implications for future research. Applied Psychophysiology and Biofeedback, 30, 347–364. Zoefel, B., Huster, R. J., & Herrmann, C. S. (2010). Neurofeedback training of the upper alpha frequency band in EEG improves cognitive performance. NeuroImage, 54(2), 1427–1431.
Please cite this article in press as: Dekker, M. K. J., et al. The time-course of alpha neurofeedback training effects in healthy participants. Biol. Psychol. (2013), http://dx.doi.org/10.1016/j.biopsycho.2013.11.014
233 234 235 236 237 238