• No results found

A role for the stop-signal : an alternative analysis for the stop-signal task

N/A
N/A
Protected

Academic year: 2021

Share "A role for the stop-signal : an alternative analysis for the stop-signal task"

Copied!
12
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

A Role for the Stop-Signal: An Alternative Analysis for the Stop-Signal

Task

Carmen Wolvius

July 16, 2015

Abstract

The stop-signal task is a very popular task when investigating response inhibition. The established analysis in analyzing this task makes use of the general linear model (GLM). This analysis, however, does not take one of the important events, the stop-signal, into account. In order to make a better characterization of the go and the stop process, we therefore propose to add a parameter to the analysis, taking the stop-signal into account. First, we compared the already established analysis with an alternative analysis in a simulation study. Second, we reanalyzed existing data from Jahfari et al. (2011) with the alternative analysis. We show that the go-parameter can be estimated more accurately with the alternative analysis.

Keywords. Stop-Signal task, fMRI, General linear model

Introduction

Response inhibition has been studied extensively over the last 10 years. The solid findings of brain regions regarding response inhibition and the reliability of inhibition tasks have made response inhibition a popular research subject. To study response inhibition researchers use the stop-signal task, which consist of trials in which a simple task has to be performed (go trials) and trials in which an action that is initiated, has to be inhibited (go-stop trials). The stop-signal task has proven to be a useful tool, because it requires the participant to withhold a response that has already been triggered (Rubia, Smith, Brammer, & Taylor, 2003). In the stop-signal task, participants are presented with a stimulus, for example an arrow pointing right or left. Participants respond to the stimulus with a key-press on the right or the left of the keyboard. Occasionally, the go-stimulus (arrow) is followed by a stop-signal, usually a tone, which instructs the participants to withhold their response to the go-stimulus (Verbruggen & Logan, 2008). To study response inhibition the go trials are compared with the go-stop trials. A majority of the papers concerning this subject found a difference between the fMRI BOLD signal of these conditions and they indicated the right inferior frontal gyrus (rIFG) and the presupplementary motor area (pre-SMA) to be involved in the response inhibition process (Rubia et al., 2003; Aron & Poldrack, 2006; Li, Huang, Constable, & Sinha, 2006; Li, Yan, Sinha, & Lee, 2008; Jahfari et al., 2011).

The processes that underly response inhibition are mostly attributed to the horse race model. The horse race-model assumes that an initiated response can be inhibited when the stop process manages to catch up with the initiated process. When it does, the initiated process can be stopped before it is executed (Logan & Cowan, 1984). So, in response inhibition there are two processes: the process of the initial response and the suppressive process, suppressing the initial response. From the horse race model, we expect to see those processes as two separate patterns of activation in the brain. For example, a brain area that shows activity during the initial response, which is then suppressed during response inhibition and an area that only shows increased activity during response inhibition. The latter has been identified by most studies as the pre-SMA and the right-IFG (Rubia et al., 2003; Aron & Poldrack, 2006; Li et al., 2006; Li, Yan, et al., 2008; Jahfari et al., 2011). A likely area for showing activation to the initial response would be the motor cortex, for this reacts to the go-stimulus. This has been observed by Hampshire, Chamberlain, Monti, Duncan, and Owen (2010), who observed a negative BOLD response, compared to the baseline, in the motor cortex during response inhibition.

Most of the studies that observed brain activation during response inhibition have used a general linear model (GLM) to describe the observed activation pattern. In this analysis the hemodynamic response function (HRF) is time locked on the go-stimulus, which means that the HRF is expected to start at the onset of the go-stimulus. During a go-stop trial, the HRF is expected to start at the onset of the go-stimulus as well. The stop-signal is expected to influence the ongoing HRF. Any observed changes in a go-stop trial are attributed to the stop-signal. The amplitude of the HRF of the go-trial is compared with the amplitude of the HRF of the go-stop trial in the follow-up contrast (Verbruggen & Logan, 2008). Figure 1a shows the signal that could be expected with this analyses.

When, however, you consider the horse race-model, you would expect two different patterns of activation to be present in a go-stop trial. Which means that in each go-stop trial two effects can be observed. First an effect triggered by the go signal, which is followed by an effect triggered by the stop signal. Given the slow character of the BLOD

(2)

0

5

10

15

20

−1.0

−0.5

0.0

0.5

1.0

a: Established Analysis time (vol) BOLD

0

5

10

15

20

b: Alternative Analysis time (vol)

Figure 1: HRF of the stop-signal task. It shows five go-stimuli of which two are followed by a stop-signal. Figure 1A shows the signal with only the onset of the go-stimuli as parameters. Figure 1B shows the signal with the onset of the go-stimuli and the onset of the stop-signals as parameters.

response in fMRI, the reaction to the stop-signal is mixed with the initial reaction to the go-stimulus. We propose to add a parameter that is time locked on the onset of the stop-signal. By adding this parameter the effect that is brought about by the stop-signal can be studied more efficiently. Figure 1b shows how the signal in figure 1a could look like, when a parameter is added for the stop-signal.

A search of the Web of Science database with the entry ‘stop-signal task’ showed 1414 articles concerning this subject. We studied the 20 most cited articles that used and analyzed the stop-signal task, with a healthy population, and found that none of these articles accounted for the onsets of the stop-signal. The onset of the stop-signal was not used in the GLM analysis and no reasons were mentioned for not using or adding the stop-signal as a parameter. Given that these articles are cited most by other researchers using the stop- signal task, we conclude that the onset of the stop-signal is hardly ever used in analysis and should therefore be looked into. A summary of these articles can be found in Appendix B.

The alternative analysis might be useful for other experimental tasks as well. For example, priming research uses tasks that have more than one event per trial, like the stop-signal task. The stimuli in a priming task are always preceded by a priming stimulus. In the field of priming research there also does not seem to be consensus on whether to add the priming stimulus to the analysis. For example, Naccache and Dehaene (2001) seem to model their task after the onset of the priming stimulus, while Aron et al. (2003) time locked their model on the onset of the priming stimulus.

In this study, we investigate whether we get a better estimate of the inhibition process when the stop-signal is taken into account in the analysis. First, we conducted a simulation study, where we compared the established analysis with an alternative analysis, in which the stop-signal was taken into account as well. Second, we reanalyzed an existing dataset from participants conducting a stop-signal task from Jahfari et al. (2011). This dataset was analyzed with the established analysis and is reanalyzed with both the established and the alternative analysis. In the simulation study both analyses were compared on several factors, including different signal-to-noise ratios (SNRs) and different amplitude proportions of the HRFs.

(3)

0 20 40 60 80 100 120 140 160 180 200 −2 −1 0 1 2 3 time (vol) BOLD go stop both

Figure 2: Example of a simulation trial. The simulated signal (black dotted line) is a combination of the onset of the go-stimuli convolved with a double gamma function (green), the onset of the stop-signals convolved with a double gamma function (red) and different sources of noise.

Simulation Study

Methods

We generated data based on a simplification of the stop-signal task used by Jahfari et al. (2011). Each run consisted of 200 time points and 42 trials. One fourth of the trials were go-stop trials, in which the go-stimulus was followed by a stop-signal. Each run had sequence of SSD values, starting with 250 ms between the go-stimulus and the onset of the stop-signal. The sequence of SSDs was based on participant behavior, which means that it could only increase or decrease with 50 ms per trial. This way the adaptive nature of the task was included in the simulated signal. The BOLD response was simulated by convolving the timing function of each condition, the go-stimulus and the stop-signal, with a double gamma function. To observe the significance of possible outcomes we varied the SNR to be: 0.1, 0.2, 0.5, 1 and 2. This was calculated by dividing the average magnitude of the signal by the standard deviation of the noise. Finally, in order to make the data most realistic, different noise sources were added, namely: white, temporal, low-frequency, physiological and task-related noise. Example series of a trial run are shown in Figure 1.

We simulated the activation patterns of two different voxels that we expect to find in the brain during a stop-signal task. Namely, the brain areas that show activation during response inhibition, like the pre-SMA and right-IFG: the Inhibition condition, and the brain areas that show activation during the go-trial and deactivation during response inhibition, like the motor cortex: the Motor Cortex condition. We brought this about by varying the effect sizes of the go- and the stop-parameter. To simulate the activation of the pre-SMA and right-IFG we lowered the response to the go stimulus to 0 or 1, and heighten the response to the stop-signal, because we expect the response to the stop-signal to be stronger than the response to the go-stimulus. To simulate the activation of the motor cortex, the response to the go stimulus was held constant and the response to the stop-signal was simulated with a negative effect size. We simulated a suppression of the response to the go stimulus. Table 1 represents an overview of the conditions. All data are generated using neuRosim (Welvaert, Durnez, Moerkerke, Verdoolaege, & Rosseel, 2011). The simulation study was conducted in R (R Development Core Team, 2010) and each cell of the simulation design was replicated a 1000 times.

All datasets were analyzed with a GLM including the onsets of the parameters, convolved with a double gamma function (the red and green line in figure 1). To analyze the established analysis, only the onsets of the go-stimuli were included, while in the analysis of the alternative analysis, also the onsets of the stop-signals were included.

As mentioned above we used a mixture of different noise sources: white, temporal, low-frequency, physiological and task-related noise. The neuRosim package enabled us to weight the different noise functions. For the first simulations we used a mixture of noise based on an example from Welvaert et al. (2011). This mixture of noise mainly contained

(4)

Table 1: Conditions of the Simulation Study. The inhibition condition represent the brain areas that show activation during response inhibition, like the pre-SMA and right-IFG. The effect size of the Go is lower than the effect size of the Stop. The Motor Cortex conditions represent the brain areas that show activation during the go-trial and deactivation during response inhibition. The effect size of the Go is the same in all conditions and the effect size of the Stop is negative. Every condition is simulated with 5 different signal-to-noise ratios (SNR).

Inhibition condition 1 Inhibition condition 2 Motor Cortex condition 1 Motor Cortex condition 2 Motor Cortex condition 3 SNR Go Stop SNR Go Stop SNR Go Stop SNR Go Stop SNR Go Stop

0.1 0 1 0.1 1 2 0.1 1 -0.5 0.1 1 -1 0.1 1 -1.5 0.2 0 1 0.2 1 2 0.2 1 -0.5 0.2 1 -1 0.2 1 -1.5 0.5 0 1 0.5 1 2 0.5 1 -0.5 0.5 1 -1 0.5 1 -1.5 1.0 0 1 1.0 1 2 1.0 1 -0.5 1.0 1 -1 1.0 1 -1.5 2.0 0 1 2.0 1 2 2.0 1 -0.5 2.0 1 -1 2.0 1 -1.5

Table 2: Results of the Simulation Study

(a) Motor Cortex Condition 2: the effect size of the go is 1 and of the stop is -1. The estimated beta values of the parameters from the established analysis (Go 1) and from the alternative analysis (Go 2 and Stop) are shown in the table, just as the standard deviation (SD) and the R-squared values of the different analyses.

SNR Go 1 SD R2 Go 2 SD R2 Stop SD 0.1 0.66 0.33 0.03 0.97 0.34 0.11 -1.00 0.27 0.2 0.67 0.17 0.07 0.98 0.17 0.29 -0.99 0.13 0.5 0.69 0.08 0.15 1.00 0.08 0.66 -0.99 0.06 1 0.69 0.05 0.19 1.00 0.05 0.80 -0.99 0.04 2 0.69 0.05 0.20 1.00 0.05 0.84 -1.00 0.03

(b) Motor Cortex Condition 2 with adjusted noise parameters. In this simulation white noise and temporal noise were weighted less and task-related noise was weighted more, to see whether different noise weights lead to difference in the estimated values. The effect size of the go is again 1 and of the stop is -1.

SNR Go 1 SD R2 Go 2 SD R2 Stop SD 0.1 0.70 0.30 0.03 0.99 0.31 0.10 -0.94 0.24 0.2 0.71 0.16 0.06 1.01 0.17 0.23 -0.95 0.13 0.5 0.73 0.10 0.11 1.02 0.10 0.41 -0.95 0.07 1.0 0.74 0.08 0.13 1.03 0.08 0.46 -0.94 0.06 2.0 0.74 0.08 0.13 1.03 0.08 0.47 -0.94 0.06

white, temporal and task-related noise and contained less low-frequency and physiological noise. However, there does not seem to be consensus about the different sources of noise and their influence on the data. Welvaert and Rosseel (2014) showed with a literature study that a large number of fMRI simulation studies ignore information about noise sources in acquiring fMRI data and hardly compensate for noise sources in their design. For this reason little is known about the different sources of noise and in what way they can be used best in a simulation design. In a simulation study, Welvaert et al. showed that physiological noise is the largest source of bias when it correlates with the data (Welvaert & Rosseel, 2012). This concerns especially task-related noise. Because of these findings, we additionally observed the effect of adding task-related noise. We repeated the simulations of Inhibition condition 1 and Motor cortex condition 1 and 2 with different weight parameters, see table 1. For the noise mixture we used the weights that were used to indicate the influence of the task-related noise in Welvaert and Rosseel (2012). This mixture contained more task-related and low-frequency noise and less white and temporal noise.

Results and Conclusion

The estimated beta values and the R-squared values of every condition were compared between the analyses. Table 2a shows the results of the Motor Cortex condition 2 where the effect size of the go was simulated to be 1 and the effect size of the stop was set to -1. The table shows that the go-parameter in the established analysis is underestimated as compared to true values, whereas in the alternative analysis, the go parameter is estimated adequately. This effect can be observed for every value of SNR. Figure 2 shows the bias of the estimated go-parameters of the established and the alternative analysis. The error bars represent the standard deviation of every run, which decreases once the SNR increases. The figure shows that there is a lot of bias between the go-parameter of the established analysis and the true effect size, while the go-parameter the alternative analysis approaches the true effect size closely. The R-squared values show that, once SNR increases, the alternative analysis explains considerably more variance than the established analysis does. Results of other conditions show similar behavior. In the inhibition conditions, the go-parameter of the established analysis is much higher that the true effect size. Again in the alternative analysis the estimations are quite accurate and the R-squared values show an increase in explained variance. Results of all the conditions can be found in Appendix A.

Table 2b shows the estimated parameters of the Motor Cortex condition 2 with a different mixture of noise. This mixture contained more task-related and low-frequency noise and less white and temporal noise. The parameters of the alternative analysis (Go 2 and Stop) in table b differ more from the true effect sizes than the parameters of the alternative analysis in table a.The table shows that the estimates of the alternative analysis are less accurate compared to the true values that the results in table 2a: the estimates of the go parameter increased and the estimates of the stop parameter decreased. The estimates of the established analysis show an increase as well. The other conditions showed similar results and can be found in Appendix A.

The results from the simulation study indicate that leaving the stop parameter out of the analysis introduces bias. The behavior of the HRF that sets in from the onset of the stop trial seems to influences the go HRF. In the condition

(5)

−0.50 −0.25 0.00 0.25 0.0 0.5 1.0 1.5 2.0 SNR Bias Analysis Go Alternative Go Established

Figure 3: Results of the simulation study: Motor Cortex condition 2. The figure shows the bias of the estimated Go parameters of the established and alternative analysis. It shows how much the estimated parameters deviate from the true effect size for each value of SNR. The error bars represent the standard deviation of each run, that consisted of 1000 iterations.

with different noise weights, the go parameter of the alternative analysis is biased somewhat as well. However, the bias in the alternative analysis is essentially less than the bias in the established analysis.

Data reanalysis

Methods

We reanalyzed data from Jahfari et al. (2011). For the analysis we used 5 participants. For details of the data acquisition we redirect to their article. For the analysis we used FEAT analysis (for FMRI Expert Analysis Tool) version 6.00, part of FSL [for Functional MRI of the Brain (FMRIB) Software Library; www.fmrib.ox.ac.uk/fsl]. The images were realigned to compensate for small head movements. Translational movement parameters never exceeded 1 voxel in any direction for any subject or session. The data were filtered in a temporal domain using a high-pass filter with a 100 s cutoff frequency to correct for baseline drifts in the signal. Finally, the functional data were prewhitened using FSL. The functional datasets were registered into three dimensional space using the participants individual high-resolution anatomical images. The individual three-dimensional image was used to normalize the functional data into Montreal Neurological Institute (MNI) space by linear scaling (affine transformation) (Jahfari et al., 2011). The data were analyzed with a GLM for both analysis. All events present in the task were modeled after convolution with a HRF including the Go trials to which the participants did not respond or made errors. Stop trials were divided in correctly inhibited and responded events. Temporal derivatives were included as covariates to improve statistical sensitivity.

The software that was used by Jahfari et al. (2011) has had updates and modifications, therefore we analyzed the data with the established and alternative analyses, instead of just comparing the results. Because of the different stimuli in the stop-signal task of Jahfari et al. (2011) the established analysis had 8 different EV. These were modeled at the onset of the go stimulus. For the alternative analysis we added the onset of the stop-signal as a separate events for every condition that had a stop-signal, which let to a total of 12 EVs. We are only interested in the go trials and the successfully inhibited go-stop trials, because those trials were used in the simulations and can be compared between the analysis at higher level. Therefore, we computed the contrast: (1) go null, (2) gostop null, (3) go -go-stop and (4) -go-stop - go. In the higher level analysis the data of the participants were combined. We contrasted the go and the go-stop trials to see if the clusters resembled typical findings. Finally, we contrasted the established and the alternative analysis with a t-test to see whether the difference in analysis caused the difference in the observed clusters.

Results and Conclusion

For the contrast go-stop > go we expected to see activation in the pre-SMA and the right-IFG, and for the contrast go > go-stop we expected to see activation in the motor cortex, as was explained in the introduction. Because we only analyzed data of 5 participants, we used a cluster threshold of 2.3, whereas Jahfari et al. (2011) used a value of 3.1 in the original analysis. For the established analysis four clusters were found significant for the contrast go-stop > go, and one cluster was significant for go > go-stop. The results are shown in table 3 and figure 4a. For the alternative analysis only one cluster was significant for the contrast go-stop > go, which is shown figure 5a. The significant

(6)

Table 3: Activated Locations during go-stop versus go trials according to the MNI Atlas coordinates (mm). Cluster thresholding with z > 2.3

Anatomical area Cluster Size

(mm2)

x y z Maximum

Ef-fect Size Established analysis

go > go-stop

Left Cerebral Cortex/ Insular Cor-tex

346 -44 2 0 3.46

go-stop > go

Left Cerebral Cortex/ Left Superior Frontal Gyrus

10787 -2 34 56 6.76

Superior Frontal Gyrus posterior division/ Right Superior Temporal Gyrus posterior division

4476 68 -34 4 6.53

Left Supramarginal Gyrus, poste-rior division

4274 -64 -48 40 5.97

Precuneous Cortex 685 -12 -60 34 4.00

Alternative analysis go > go-stop

Left Cerebral White matter 190 -12 -20 -10 3.43

clusters do not match the areas of the original analysis (Jahfari et al., 2011). However, when observing the results uncorrected, we did observe activation in the pre-SMA and right-IFG for go-stop > go and the motor cortex for go > go-stop. In the original analysis the results were averaged over 20 participants, which might lead to different clusters. Given the small number of subjects, this seems justifiable. We contrasted the results of the contrasts go-stop> go and go > go-stop between the established analysis and the alternative analysis. Figure 5b shows that there is more activation with the established analysis for the contrast go-stop > go. Which we could have expected, given that we did not find any significant clusters in the go-stop > go contrast in the alternative analysis.

These results indicate that the additional parameters in the alternative analysis explained the variance caused by the stop-signal. The variance that was left to analyze with the alternative analysis was mainly caused by the go-signal. This did not lead to any significant clusters because the go and the go-stop trial have similar variance when the variance of the stop-signal is not included. Comparing go and go-stop in a contrast would therefore not lead to much difference as is shown with the comparative analysis in figure 5a.

Discussion

The simulation study clearly outlined the importance of the stop-signal parameter. When the stop-signal is neglected in the analysis of the stop-signal task, bias is introduced: The HRF of the stop-signal is included in the estimation of the HRF of the go stimulus, which leads to an under- or overestimation of the go HRF. When the stop-signal is added to the analysis the HRFs of the go stimulus and the stop-signal can be estimated separately, the estimation of the go HRF is no longer biased.

Reanalyzing the data of Jahfari et al. (2011) with the established analysis let to a higher amount of active clusters than reanalysis with the alternative analysis. Assuming that this difference is in fact present, we could conclude that neglecting the stop-signal in the established analysis let to an overestimation of the go HRF in the go-stop trials. Which could have lead to more active clusters. We observed this difference when contrasting the established analysis with the alternative analysis. Figure 5a shows that contrasting go with go-stop in the alternative analysis leads to little activation and figure 5b shows a reasonable difference between the two analysis. This difference suggests the variance caused by the stopsignal is explained by the additional parameter and therefore does not show in the go -go-stop contrasts. With the alternative analysis the stop signal could be studied separate from the go stimulus, by contrasting the additional parameter to the baseline. The results from the simulation study indicate that analyzing the stop signal separately would give a better estimation of the stop signal than when a go trial is merely contrasted with a go-stop trial and the reanalysis of the data indicated that the variance can in fact be separated and therefore could be studied separately without the influence of the go stimulus. Given the solid findings that have been made with the established analysis, using the alternative analysis could specify the inhibitory process even more and make it possible to observe the brain areas that are involved in the stopping process more closely.

The small amount of participants that were used in this study might have let to a low amount of power in the analysis, which might have lead these results to be different from the original results. Due to the scope of this research only 5 participants were reanalyzed to test the results of the simulation study. In order to observe the actual implications of changing the analysis of the stop-signal task data from a stop-signal study, such as Jahfari et al. (2011), should be analyzed with both analyses for all the participants.

(7)

(a) Areas of significantly increased activation for the contrast go > go-stop in the established analysis. The areas shown in the figure are more active during a go trial, compared to a go-stop trial. The clusters are determined with P < 0.00791.

(b) Areas of significantly increased activation for the contrast go-stop > go in the established analysis. The areas shown are more active during a go-stop trial, compared to a go trial. The clusters are determined with a maximum P < 0.0001.

Figure 4

(a) Areas of significantly increased activation for the contrast go > go-stop in the alternative analysis. The areas shown are more active during a go trial, compared to a go-stop trial.The clusters are determined with P < 0.00557.

(b) Areas of significantly increased activation when contrasting the established analysis with the alternative analysis. The figure shows the areas from the previously specified contrast go-stop > go (figure 4) in the established analysis that are more active compared to the previously specified contrast go-stop > go of the alternative analysis. It shows the difference between the analyses.

Figure 5

There are a few of issues that should be taken into account when interpreting the results. As was mentioned in the method section, not much is known about the different sources of noise that are present in fMRI research. A fair amount of noise was used in the simulation study, however, the weights that were given to the different sources were relatively arbitrary. Welvaert and Rosseel (2012) showed that task-related noise causes the largest amount of bias and should therefore be payed attention to in simulation studies. However, there is no perspective on the other noise sources.

Another issue that should be taken into account is the signal-to-noise ratio (SNR). Except for a few articles that reported their SNRs to be robust or low, we could not discover what SNRs are realistic with respect to the stop-signal task. SNRs are not often reported in fMRI research and it is a topic that is still under a lot of discussion. There are different ways to determine a SNR, so a single SNR value is hard to interpret. Furthermore, the SNR value that is necessary to detect an expected signal is subject to debate (Welvaert & Rosseel, 2013). The values that we used in the simulation study might therefore not be realistic and could be too low or too high. When in reality the SNR is much higher, the established analysis might proof just as efficient as the alternative analysis and when it would be much lower, the alternative analysis could be subject to a lot of bias as well. Future research of the stop-signal task should consider the SNR and its report.

The analysis introduced in this study enables a more efficient estimation of the behavior of the parameters in the stop-signal task. The question remains how a better estimation of the go and stop parameter adds to the theoretical comprehension of response inhibition. In response inhibition researchers study the effect or changes in the behavior of the go parameter and compare these changes with behavioral results. It would be interesting to see whether observing the go and stop parameters separately, by adding a stop parameter to the analysis, has implications on observing these behavioral results. Further research on these topics should be done to see what the actual implications are on studying response inhibition, when using the alternative analysis in the stop-signal task.

(8)

Appendix

Results Simulation Study

Table 4: Results Inhibition Condition 1: Effect size of the Go = 0 and the effect size of the Stop =1. The estimated beta values of the parameters from the established analysis (Go 1) and from the alternative analysis (Go 2 and Stop) are shown in the table, just as the standard deviation (SD) and the R-squared values of the different analyses.

SNR Go 1 SD R Go 2 SD R Stop SD 0.1 0.30 0.39 0.01 -0.02 0.41 0.07 1.02 0.31 0.2 0.28 0.20 0.01 -0.03 0.21 0.20 1.00 0.16 0.5 0.31 0.09 0.03 -0.002 0.09 0.55 1.01 0.07 1.0 0.31 0.06 0.04 -0.004 0.06 0.74 1.01 0.05 2.0 0.31 0.05 0.05 -0.001 0.05 0.81 1.00 0.04

Table 5: Results Inhibition Condition 2: Effect size of the Go = 1 and the effect size of the Stop =2, see table 4

SNR Go 1 SD R Go 2 SD R Stop SD 0.1 1.53 1.43 0.01 0.91 1.48 0.03 2.00 1.18 0.2 1.58 0.74 0.02 0.95 0.76 0.09 2.03 0.57 0.5 1.62 0.29 0.11 0.99 0.30 0.37 2.02 0.24 1 1.62 0.15 0.20 0.99 0.15 0.68 2.00 0.12 2 1.62 0.09 0.25 1.00 0.09 0.87 2.00 0.06

Table 6: Results Motor Cortex Condition 1: Effect size of the Go = 1 and the effect size of the Stop =-0.5, see table 4.

SNR Go 1 SD R Go 2 SD R Stop SD 0.1 0.80 0.54 0.02 0.96 0.55 0.03 -0.50 0.42 0.2 0.82 0.26 0.05 0.97 0.27 0.09 -0.50 0.21 0.5 0.83 0.11 0.21 0.99 0.11 0.33 -0.50 0.09 1.0 0.84 0.07 0.37 1.00 0.07 0.58 -0.50 0.05 2.0 0.84 0.05 0.45 1.00 0.05 0.71 -0.50 0.04

Table 7: Results Motor Cortex Condition 2: Effect size of the Go = 1 and the effect size of the Stop = -1, see table 4

SNR Go 1 SD R Go 2 SD R Stop SD 0.1 0.66 0.33 0.03 0.97 0.34 0.11 -1.00 0.27 0.2 0.67 0.17 0.07 0.98 0.17 0.29 -0.99 0.13 0.5 0.69 0.08 0.15 1.00 0.08 0.66 -0.99 0.06 1 0.69 0.05 0.19 1.00 0.05 0.80 -0.99 0.04 2 0.69 0.05 0.20 1.00 0.05 0.84 -1.00 0.03

(9)

Table 8: Results Motor Cortex Condition 3: Effect size of the Go = 1 and the effect size of the Stop = -1.5, see table 4 SNR Go 1 SD R Go 2 SD R Stop SD 0.1 0.53 0.14 0.04 0.99 0.14 0.54 -1.50 0.11 0.2 0.53 0.08 0.06 1.00 0.08 0.78 -1.49 0.06 0.5 0.53 0.05 0.07 1.00 0.05 0.89 -1.49 0.04 1 0.53 0.04 0.07 1.00 0.05 0.92 -1.50 0.03 2 0.53 0.04 0.07 1.00 0.04 0.92 -1.50 0.03

Table 9: Results Inhibition condition, different noise parameters: Effect size of the Go = 1 and the effect size of the Stop = 2. In this simulation white noise and temporal noise were weighted less and task-related noise was weighted more, to see whether different noise weights lead to difference in the estimated values.

SNR Go 1 SD R Go 2 SD R Stop SD 0.1 1.40 1.31 0.01 0.75 1.36 0.03 2.08 1.06 0.2 1.55 0.68 0.03 0.90 0.69 0.10 2.08 0.52 0.5 1.64 0.27 0.10 1.00 0.28 0.36 2.06 0.22 1 1.66 0.16 0.18 1.02 0.16 0.62 2.06 0.12 2 1.67 0.10 0.22 1.02 0.10 0.76 2.06 0.08

Table 10: Results Motor Cortex condition 1, different noise parameters: Effect size of the Go = 1 and the effect size of the Stop = -0.5, see table 9

SNR Go 1 SD R Go 2 SD R Stop SD 0.1 0.83 0.46 0.019 0.96 0.48 0.03 -0.43 0.36 0.2 0.85 0.24 0.05 0.98 0.25 0.08 -0.43 0.20 0.5 0.88 0.12 0.16 1.02 0.13 0.22 -0.44 0.10 1.0 0.89 0.09 0.21 1.02 0.09 0.30 -0.44 0.07 2.0 0.89 0.08 0.24 1.03 0.08 0.33 -0.44 0.06

Table 11: Results Motor Cortex condition 2, different noise parameters: Effect size of the Go = 1 and the effect size of the Stop = -1, see table 9

SNR Go 1 SD R Go 2 SD R Stop SD 0.1 0.70 0.30 0.03 0.99 0.31 0.10 -0.94 0.24 0.2 0.71 0.16 0.06 1.01 0.17 0.23 -0.95 0.13 0.5 0.73 0.10 0.11 1.02 0.10 0.41 -0.95 0.07 1.0 0.74 0.08 0.13 1.03 0.08 0.46 -0.94 0.06 2.0 0.74 0.08 0.13 1.03 0.08 0.47 -0.94 0.06

(10)

Table 12: Studies using GLM model to analyse fMRI data from the stop/signal-task

H

Study N GLM parameters Onset Citations

Aron and Poldrack (2006) 18 go, stop succes, stop respond, nuis. go signal 581

Rubia et al. (2003) 20 go, stop succes, stop respond go signal 383

Rubia et al. (2001) 15 go/no go task 1, go/no go task 2, stop task 1, stop task 2, stop task 3*

go signal 482

Li et al. (2006) 24 go succes, go error, stop succes, stop

re-spond, go succes RT, stop succes SSD, stop respond SSD

go signal 214

Duann, Ide, Luo, and Li (2009) 60 go succes, go error, stop succes, stop re-spond

go signal 144

Li, Yan, et al. (2008) 30 go succes, go error, stop succes, stop re-spond, realign.

go signal 112

Xue, Aron, and Poldrack (2008) 15 go, stop succes, stop respond (late respond and early respond)

go signal 99

Chevrier, Noseworthy, and Schachar (2007)

go left hand, go right hand, stop succes, stop respond

go signal 99

Chikazoe et al. (2009), 2009 22 certain go succes, uncertain go succes, stop succes, error

go signal 89

Li, Yan, Bergquist, and Sinha (2007)

40 go succes, go error, Pre-G, stop succes, stop respond, Pre-ss, Pre-se

go signal 71

Jahfari et al. (2011) 16 go none, go low, go high, stop succes low, stop succes high, stop respond low, stop respond high

go signal 70

Curtis, Cole, Rao, and D’Esposito (2005)

12 catch, go, stop succes, stop respond go signal 67

Ramautar, Slagter, Kok, and Rid-derinkhof (2006)

16 go, stop succes, stop respond go signal 63

Li, Huang, et al. (2008) 40 go succes, go error, stop succes high RT, stop succes low RT, stop respond high RT, stop respond low RT

go signal 59

Chao, Luo, Chang, and Chiang-Shan (2009)

65 go succes, go error, stop succes, stop re-spond

go signal 57

Zheng, Oka, Bokura, and Yam-aguchi (2008)

18 go, stop succes, stop respond go signal 54

Zandbelt and Vink (2010) 24 go, stop succes, stop respond go signal 53

Congdon et al. (2010) pooled

data

go, stop succes, stop respond go signal 50

Padmala and Pessoa (2010) 35 go succes (baseline ), stop succes, stop re-spond, nuissance

go signal 45

Sagaspe, Schwartz, and Vuilleumier (2011)

12 go neutral, go emotional, stop succes neu-tral, stop succes emotional, stop respond neutral, stop respond emotional

go signal 35

Li et al. (2009) 60 go succes, go error, stop succes, stop

re-spond

go signal 25

Two different tasks were used: Go/No go task and the Stop/signal task, both had one stimulus and one motor controlled. The stimulus control Stop/signal task had stop-signal frequecy of 30% and 50%

(11)

References

Aron, A. R., & Poldrack, R. A. (2006). Cortical and subcortical contributions to stop-signal response inhibition: role of the subthalamic nucleus. The Journal of Neuroscience, 26 , 2424–2433.

Aron, A. R., Schlaghecken, F., Fletcher, P., Bullmore, E. T., Eimer, M., Barker, R., . . . Robbins, T. (2003). Inhibition of subliminally primed responses is mediated by the caudate and thalamus: evidence from functional mri and huntingtons disease. Brain, 126 , 713–723.

Chao, H. H., Luo, X., Chang, J. L., & Chiang-Shan, R. L. (2009). Activation of the pre-supplementary motor area but not inferior prefrontal cortex in association with short stop-signal reaction time–an intra-subject analysis. BMC Neuroscience, 10 , 75.

Chevrier, A. D., Noseworthy, M. D., & Schachar, R. (2007). Dissociation of response inhibition and performance monitoring in the stop-signal task using event-related fmri. Human Brain Mapping , 28 , 1347–1358.

Chikazoe, J., Jimura, K., Hirose, S., Yamashita, K.-i., Miyashita, Y., & Konishi, S. (2009). Preparation to inhibit a response complements response inhibition during performance of a stop-signal task. The Journal of Neuroscience, 29 , 15870–15877.

Congdon, E., Mumford, J. A., Cohen, J. R., Galvan, A., Aron, A. R., Xue, G., . . . Poldrack, R. A. (2010). Engagement of large-scale networks is related to individual differences in inhibitory control. Neuroimage, 53 , 653–663. Curtis, C. E., Cole, M. W., Rao, V. Y., & D’Esposito, M. (2005). Canceling planned action: an fmri study of

countermanding saccades. Cerebral Cortex , 15 , 1281–1289.

Duann, J.-R., Ide, J. S., Luo, X., & Li, C.-S. R. (2009). Functional connectivity delineates distinct roles of the inferior frontal cortex and presupplementary motor area in stop-signal inhibition. The Journal of Neuroscience, 29 , 10171–10179.

Hampshire, A., Chamberlain, S. R., Monti, M. M., Duncan, J., & Owen, A. M. (2010). The role of the right inferior frontal gyrus: inhibition and attentional control. Neuroimage, 50 , 1313–1319.

Jahfari, S., Waldorp, L., van den Wildenberg, W. P., Scholte, H. S., Ridderinkhof, K. R., & Forstmann, B. U. (2011). Effective connectivity reveals important roles for both the hyperdirect (fronto-subthalamic) and the indirect (fronto-striatal-pallidal) fronto-basal ganglia pathways during response inhibition. The Journal of Neuroscience, 31 , 6891–6899.

Li, C.-S. R., Huang, C., Constable, R. T., & Sinha, R. (2006). Imaging response inhibition in a stop-signal task: neural correlates independent of signal monitoring and post-response processing. The Journal of Neuroscience, 26 , 186–192.

Li, C.-S. R., Huang, C., Yan, P., Paliwal, P., Constable, R. T., & Sinha, R. (2008). Neural correlates of post-error slowing during a stop-signal task: a functional magnetic resonance imaging study. Journal of Cognitive Neuroscience, 20 , 1021–1029.

Li, C.-S. R., Yan, P., Bergquist, K. L., & Sinha, R. (2007). Greater activation of the default brain regions predicts stop-signal errors. Neuroimage, 38 , 640–648.

Li, C.-S. R., Yan, P., Sinha, R., & Lee, T.-W. (2008). Subcortical processes of motor response inhibition during a stop-signal task. Neuroimage, 41 , 1352–1363.

Li, C.-S. R., Zhang, S., Duann, J.-R., Yan, P., Sinha, R., & Mazure, C. M. (2009). Gender differences in cognitive control: an extended investigation of the stop-signal task. Brain Imaging and Behavior , 3 , 262–276.

Logan, G. D., & Cowan, W. B. (1984). On the ability to inhibit thought and action: A theory of an act of control. Psychological review , 91 , 295.

Naccache, L., & Dehaene, S. (2001). The priming method: imaging unconscious repetition priming reveals an abstract representation of number in the parietal lobes. Cerebral cortex , 11 , 966–974.

Padmala, S., & Pessoa, L. (2010). Interactions between cognition and motivation during response inhibition. Neu-ropsychologia, 48 , 558–565.

Ramautar, J. R., Slagter, H. A., Kok, A., & Ridderinkhof, K. R. (2006). Probability effects in the stop-signal paradigm: the insula and the significance of failed inhibition. Brain Research, 1105 , 143–154.

Rubia, K., Russell, T., Overmeyer, S., Brammer, M. J., Bullmore, E. T., Sharma, T., . . . others (2001). Mapping motor inhibition: conjunctive brain activations across different versions of go/no-go and stop tasks. Neuroimage, 13 , 250–261.

Rubia, K., Smith, A. B., Brammer, M. J., & Taylor, E. (2003). Right inferior prefrontal cortex mediates response inhibition while mesial prefrontal cortex is responsible for error detection. Neuroimage, 20 , 351–358.

Sagaspe, P., Schwartz, S., & Vuilleumier, P. (2011). Fear and stop: a role for the amygdala in motor inhibition by emotional signals. Neuroimage, 55 , 1825–1835.

Verbruggen, F., & Logan, G. D. (2008). Response inhibition in the stop-signal paradigm. Trends in cognitive sciences, 12 , 418–424.

Welvaert, M., Durnez, J., Moerkerke, B., Verdoolaege, G., & Rosseel, Y. (2011). neurosim: An r package for generating fmri data. Journal of Statistical Software, 44 , 1–18.

Welvaert, M., & Rosseel, Y. (2012). How ignoring physiological noise can bias the conclusions from fmri simulation results. Journal of neuroscience methods, 211 (1), 125–132.

(12)

Welvaert, M., & Rosseel, Y. (2013). On the definition of signal-to-noise ratio and contrast-to-noise ratio for fmri data. Welvaert, M., & Rosseel, Y. (2014). A review of fmri simulation studies. PloS one, 9 , e101953.

Xue, G., Aron, A. R., & Poldrack, R. A. (2008). Common neural substrates for inhibition of spoken and manual responses. Cerebral Cortex , 18 , 1923–1932.

Zandbelt, B. B., & Vink, M. (2010). On the role of the striatum in response inhibition. PloS One, 5 , e13848. Zheng, D., Oka, T., Bokura, H., & Yamaguchi, S. (2008). The key locus of common response inhibition network for

Referenties

GERELATEERDE DOCUMENTEN

Therefore, and because of the continuous pressure on conventional public transport services, this study explores which and how different factors (i.e. service attributes and

Goal: The research goal is to provide academic insight in the perception of SMEs towards alternative financing in order to determine which factors influence the

To compare different SSRT estimation methods, we ran a set of simulations which simulated performance in the stop-signal task based on assumptions of the independent race model: on

To compare different SSRT estimation methods, we ran a set of simulations which simulated performance in the stop-signal task based on assumptions of the independent race model:

The w lines following 1c, 2c, and 3c in the listing show the minimum column widths specified by the ‘w’ keys in the format; 35 pt is 7 times TABLE’s default column width unit of 0.5

The LaTeX package decision-table provides a command \dmntable, which allows for an easy way to generate decision tables in the Decision Model and Notation (DMN) format. 1 ) This

Collaboration with cooperating organisations after the Stop-reaction, the assistance agencies (if necessary, in the case of underlying (parenting or psychosocial) problems) is,

In the literature, comparison studies have often shown that equivalence scales based on income evaluation and satisfaction data represent higher scale economies than the