Evidence for non-selective response inhibition in uncertain contexts revealed by combined meta-analysis and Bayesian analysis of fMRI data

Abstract

Response inhibition is typically considered a brain mechanism selectively triggered by particular "inhibitory" stimuli or events. Based on recent research, an alternative non-selective mechanism was proposed by several authors. Presumably, the inhibitory brain activity may be triggered not only by the presentation of "inhibitory" stimuli but also by any imperative stimuli, including Go stimuli, when the context is uncertain. Earlier support for this notion was mainly based on the absence of a significant difference between neural activity evoked by equiprobable Go and NoGo stimuli. Equiprobable Go/NoGo design with a simple response time task limits potential confounds between response inhibition and accompanying cognitive processes while not preventing prepotent automaticity. However, previous neuroimaging studies used classical null hypothesis significance testing, making it impossible to accept the null hypothesis. Therefore, the current research aimed to provide evidence for the practical equivalence of neuronal activity in the Go and NoGo trials using Bayesian analysis of functional magnetic resonance imaging (fMRI) data. Thirty-four healthy participants performed a cued Go/NoGo task with an equiprobable presentation of Go and NoGo stimuli. To independently localize brain areas associated with response inhibition in similar experimental conditions, we performed a meta-analysis of fMRI studies using equal-probability Go/NoGo tasks. As a result, we observed overlap between response inhibition areas and areas that demonstrate the practical equivalence of neuronal activity located in the right dorsolateral prefrontal cortex, parietal cortex, premotor cortex, and left inferior frontal gyrus. Thus, obtained results favour the existence of non-selective response inhibition, which can act in settings of contextual uncertainty induced by the equal probability of Go and NoGo stimuli.

Figure 1

Flow chart of study selection in the meta-analysis.

Figure 2

Experimental design of the Go/NoGo task. S1—first stimulus (preparatory cue). S2—second stimulus (imperative stimulus). A—images of animals, P—images of plants. (A) First experiment: “A-A Go”, “A-P NoGo”. (B) Second experiment: “A-P Go”, “A-A NoGo”. Red boxes highlight trials that were compared to test the hypothesis of selective and non-selective response inhibition.

Figure 3

Results of the ALE meta-analysis. The result of the meta-analysis of 20 fMRI studies using equal probability Go/NoGo tasks (“50/50% Go/NoGo blocks > 100% Go-control blocks” contrast). Figure was created using MRIcroGL 1.2.2 software (https://www.nitrc.org/projects/mricrogl).

Figure 4

Results of classical voxel-wise and Bayesian parameter inference. (A) Classical NHST inference with an FWE-corrected voxel-wise threshold of 0.05. (B) BPI with the effect size threshold γ = 1 prior SD_θ = 0.1%, LogOdds > 3 (PP > 0.95). The “NoGo > Go” effect was not detected by any of the methods. Red colour depicts the “Go > NoGo” effect. Green colour depicts the “NoGo = Go” effect (practical equivalence of neuronal activity in Go and NoGo trials). White colour depicts (A) non-significant voxels and (B) “low-confidence” voxels. Figure was created using Mango 4.1 software (https://ric.uthscsa.edu/mango).

Figure 5

LBRs for the right hand (“L SM minus R SM” BOLD difference). (A) Sensorimotor ROIs in the left and right hemispheres (blue and violet color correspondingly). (B) Mean LBR (and 95% confidence intervals) for the Go, NoGo, and Ignore trials. Figure was created using MRIcroGL 1.2.2 software (https://www.nitrc.org/projects/mricrogl).

Figure 6

Three-way overlap between the ALE meta-analysis and Bayesian analysis. (A) Three-way overlap between (1) inhibitory-related brain areas according to the ALE meta-analysis (“50/50% Go/NoGo blocks > 100% Go-control blocks”), (2) brain areas with practically equivalent neuronal activity in Go and NoGo trials (“NoGo = Go”) and (3) brain areas activated in equiprobable Go and NoGo trials (“Go + NoGo > Ignore”). (B) Violin plots of the mean PSC in the revealed clusters. DLPFC dorsolateral prefrontal cortex, FEF frontal eye field, PMC premotor cortex, IPL inferior parietal lobule, TPJ temporoparietal junction, Ins insula, IFG inferior frontal gyrus, BA Brodmann area, L left, R right. Figure was created using MRIcroGL 1.2.2 software (https://www.nitrc.org/projects/mricrogl).