Reduced risk-taking behavior during frontal oscillatory theta band neurostimulation

Background Most of our decisions involve a certain degree of risk regarding the outcomes of our choices. People vary in the way they make decisions, resulting in different levels of risk-taking behavior. These differences have been linked to prefrontal theta band activity. However, a direct functional relationship between prefrontal theta band activity and risk-taking has not yet been demonstrated. Objective We used noninvasive brain stimulation to test the functional relevance of prefrontal oscillatory theta activity for the regulatory control of risk-taking behavior. Methods In a within-subject experiment, 32 healthy participants received theta [4-8 Hertz (Hz)], gamma (30-50 Hz), and sham transcranial alternating current stimulation (tACS) over the left prefrontal cortex (lPFC). During stimulation, participants completed a task assessing their risk-taking behavior as well as response times and sensitivity to value and outcome probabilities. Electroencephalography (EEG) was recorded before and immediately after stimulation. Results Theta-band, but not gamma-band or sham, tACS led to a significant reduction in risk-taking behavior, indicating a frequency-specific effect of prefrontal brain stimulation on the modulation of risk-taking behavior. Moreover, theta-band stimulation led to increased response times and decreased sensitivity to reward values. EEG data analyses did not show an increase in power in the stimulated frequencies. Conclusion These findings provide direct empirical evidence for the effects of prefrontal theta-band stimulation on behavioral risk-taking regulation.

5 47 uncertainty, rather than risk, which is a different economic construct (16), since the probabilities 48 are not explicit to participants. 49 The present study aims at investigating the functional relationship between frontal theta-50 band oscillations and risk-taking behavior. Although previous studies (9,10) have shown a 51 correlation between resting state frontal theta band asymmetry and risk-taking behavior, no direct 52 causal relationship has thus far been shown. We therefore applied tACS to the left DLPFC in theta-53 band (6.5 Hz), and gamma-band (40 Hz) frequency as well as sham stimulation while participants 54 performed a risk-taking task. We chose gamma-band tACS as a control frequency since it has not 55 been linked to risk-taking behavior thus far. We also implemented a new behaviorally-controlled 56 risk-taking protocol paired with financial incentives for more robust measures of risk-taking 57 behavior.

58
To monitor possible power changes in the stimulated frequencies and to investigate their 59 functional relationship with the behavioral results, we implemented EEG measurements before 60 and immediately after the transcranial brain stimulation. We hypothesized that, compared to sham 61 and gamma-band stimulation, theta-band stimulation to the left DLPFC decreases risk-taking 62 behavior, confirming the central regulatory role of theta-frequencies on the electrophysiological 63 mechanism underlying the modulation of risk-taking behavior (9,17).  When a participant chooses a color, the choice is highlighted and the position of the token 147 is revealed (Fig 2). Therefore, in this same example, as the participant chose blue and the token 148 was hidden behind a blue box, the participant gets 100 points (as indicated in the white text on the 149 right).

150
To gain more insight on the different types of trials, we divided them into three clusters 151 according to the differences (or contrasts) in expected values offered by the two options (Pink and 152 Blue), which could capture the difficulty of making a choice in the trial. The lower this difference 153 the more difficult it is for a subject to make a choice. This led to the division of trials in these 154 clusters: low, medium and high contrast. In our analysis, we excluded trials with no difference in 155 expected value, since this group of options includes less trials than the remaining clusters and 156 would not allow balanced analyses. Trials with one strictly dominant option, meaning (for 157 simplicity) trials where the options have differences in expected value > |65| were excluded. This 158 exclusion was made since these were considered non-informative, because these choices are 159 considered obvious and would hardly be affected by any environmental or intrinsic factor. In total 160 204 out of 250 trials were analyzed per session. The cluster division can be seen in detail in the S1 161 Table. 162 163 Transcranial alternating current stimulation 164 We aimed at stimulating the left DLPFC. A small circular (diameter: 2.1cm, thickness: 235 the probability is below 50% and safe if its probability is above 50%. To allow a more refined 13 237 from -2 to 2. The choice of a higher probability was classified with a negative score and that of a 238 lower probabilities received a positive score. In simple terms, these scores indicate that options 239 with a higher level of uncertainty have positive scores, while safer options have negative scores.
240 These probability scores can be seen in Table 1. 241

289
To look for differences in theta and gamma power before and after the stimulation 290 protocols, the power spectra were averaged for the before and post stimulation measurements.  Table).

385
Further contrasts showed significant effects of both theta and gamma-band stimulation.
386 Of interest, there was a significant effect of theta-band stimulation (and not gamma) on asymmetry 387 change (pre-post) when compared to sham t(2)= 2,528, p=.012. However, this effect is mainly 388 driven by a decrease in asymmetry in the sham condition, which could be observed in Fig 7, 389 indicating that the decrease is due to the task execution and not due to the stimulation. According to our analyses, the asymmetry went from a negative score in the first minute 391 after the task to a positive score when 3 minutes are analyzed. This indicates that during the first 392 minute after the task execution there was a higher theta power in the left hemisphere, which 393 changed during the following minutes, since the analyses including 3 minutes after task indicate a 394 higher theta power in the right hemisphere. More details can be seen in Fig 7.

401
We also conducted a partial correlation analysis between the frontal theta asymmetry,

411
The inclusion of asymmetry of theta-power in any of the electrodes in the regression 412 models used to analyze the behavioral results did not improve the fit of these models and therefore 414 Gamma-band entrainment 415 The effects of gamma-band stimulation were investigated using a 3 (stimulation 416 condition: theta, gamma and sham) by 2 (time: before and after stimulation) by 6 (theta power 427 being fundamental for the modulation of risk-taking behavior (12). We therefore expected theta-428 band stimulation to cause a reduction in risk-taking behavior and that this effect is frequency 429 specific.

430
As predicted, we were able to effectively reduce risk-taking behavior in healthy 431 participants using theta-band tACS over the left DLPFC, compared to sham and gamma-band 432 stimulation. These findings confirm the functional relationship between theta-band frequencies 433 and risk-taking behavior regulation, being a fundamental part of the electrophysiological 434 mechanism responsible for this modulation. Theta band tACS leads to a significant decrease of 437 we observed a significant reduction on value sensitivity due to theta-band (and not gamma) 438 stimulation, meaning that participants opted for lower values after theta-band stimulation 439 compared to the results obtained in the sham or gamma conditions. These results are in line with 440 previous studies, where participants became more risk-averse after non-invasive brain stimulation 441 with reduced sensitivity to value (22,33-35). However, our study was able to show that also this 442 effect is frequency specific. Therefore, it is expected that theta-frequencies would play a 443 fundamental role in the reduction of value sensitivity, meaning the recruitment of DLPFC as 444 executive control to modulate the VMPFC response to the value (36).

445
The stimulation did not affect the probabilities chosen by the participants, indicating that 446 the choice of probabilities might be regulated by a different electrophysiological mechanism. Even 447 though our results indicate that probabilities and value are evaluated independently in our brain, 448 behaviorally and in terms of neurological activity these processes are at least strongly correlated 449 (2,29,37). This means that both inputs are considered (bet value and its probabilities) in order to 450 inform the decision process, which justifies the use of standard deviation as an estimation of risk.
451 Our approach considers the option's expected value (meaning the bet's probabilities and value) to 452 estimate risk, which is in our perspective a more naturalistic evaluation of risk. Our findings 453 indicate that participant's reductions on risk-taking behavior were mainly driven by a reduction on 454 the average value sensitivity.

455
Although we did not have a specific hypothesis regarding the response time, it is 456 interesting to notice that theta stimulation increased response time compared to sham and gamma