Using subjective expectations to model the neural underpinnings of proactive inhibition

Abstract Proactive inhibition – the anticipation of having to stop a response – relies on objective information contained in cue‐related contingencies in the environment, as well as on the subjective interpretation derived from these cues. To date, most studies of brain areas underlying proactive inhibition have exclusively considered the objective predictive value of environmental cues, by varying the probability of stop‐signals. However, by only taking into account the effect of different cues on brain activation, the subjective component of how cues affect behavior is ignored. We used a modified stop‐signal response task that includes a measurement for subjective expectation, to investigate the effect of this subjective interpretation. After presenting a cue indicating the probability that a stop‐signal will occur, subjects were asked whether they expected a stop‐signal to occur. Furthermore, response time was used to retrospectively model brain activation related to stop‐expectation. We found more activation during the cue period for 50% stop‐signal probability, when contrasting with 0%, in the mid and inferior frontal gyrus, inferior parietal lobe and putamen. When contrasting expected vs. unexpected trials, we found modest effects in the mid frontal gyrus, parietal, and occipital areas. With our third contrast, we modeled brain activation during the cue with trial‐by‐trial variances in response times. This yielded activation in the putamen, inferior parietal lobe, and mid frontal gyrus. Our study is the first to use the behavioral effects of proactive inhibition to identify the underlying brain regions, by employing an unbiased task‐design that temporally separates cue and response.

probability to minimize the impact of the potential confound, these results needed to be replicated in an unbiased design. Therefore, we now used a single stop-signal probability level of 50% to allow the unbiased disentangling of brain activations related to subjective expectations and the processing of objective stop-signal probabilities.
To replicate our previous results, we investigated the effect of stop-signal probability and expectation during the cue and stimulus-response period in predefined regions of interest (ROIs).
These ROIs were based on activation patterns from a previous study in which sample of healthy participants performed the delayed-response stop-signal anticipation task (Zandbelt et al., 2013b).
For the cue period, we looked at the striatum, SMA, left PMd and midbrain. For the stimulusresponse period the ROIs included the rIFG and rIPC. Mean activation level, expressed as percentage of signal change, was calculated per participant for each ROI. Identical to the previous study, pairedsampled t-tests were used to investigate differences in activation between conditions. Imaging data Cue period Figure 5 shows activation in the ROIs during the cue period. Activation in the striatum t(24) = 3.7; p = 0.001, SMA t(24) = 5.4; p < 0.001 and PMd t(24) = 5.3; p < 0.001 was higher in trials with 50% stopsignal probability compared to trials with 0% probability, regardless of expectation. No such effect was observed in the midbrain region t(24) = 1.739; p = 0.095. Next, we looked at activation in these regions for trials with 0% vs. 50% stop-signal probability while subjects did not expect a stop-signal. These analyses showed a significant difference in activation for the striatum t(24) = -2.4; p = 0.022, SMA t(24) = -4.2; p < 0.001 and PMd t(24) = -3.0; p < 0.01. Again, no difference was found in the midbrain region t(24) = -1.2; p = 0.30.
Finally, we looked at the difference in trials where subjects did versus did not expect a stop-signal to occur, in the context of 50% probability. For trials in which subjects expected a stop to occur, activity in the striatum t(24) = -2.2; p = 0.035, SMA t(24) = -2.3; p = 0.028 and PMd t(24) = -3.9; p < 0.001 was higher than when they did not expect one. Again, there were no significant differences in the midbrain region t(24) = 1.1; p = 0.27.
Stimulus and response period Figure 6 shows activations in the ROIs during the stimulus and response period. Similar to previous studies (Zandbelt et al., 2013;Vink et al., 2015), we found heightened activation in both the rIFG t(24) = 4.5; p < 0.001 and rIPC t(22) = 6.6; p < 0.001 when stop-signal probability was 50% compared to 0%, regardless of stop-signal expectation.
In addition, we examined the effect of stop-signal expectation by examining a subset of trials where subjects indicated not expecting a stop-signal. There was significantly more activation in both the rIFG t(24) = -4.2; p < 0.001 and rIPC t(22) = -6.3; p < 0.001 in trials with 50% compared to 0% stop probability. Finally, we examined the effect of subjective expectation in the context of 50% stopsignal probability. There was no significant effect for stop-signal expectation on activation in the rIFG t(24) = 1.2; p = 0.23 and rIPC t(22) = -0.4; p = 0.68.  conditions. Center coordinates for rIFG 50,10,29, and rIPC 59,-40,30. *P < 0.05.