Competing basal ganglia pathways determine the difference between stopping and deciding not to go
- University of Pittsburgh, United States
- University of Pittsburgh and Carnegie Mellon University, United States
- Carnegie Mellon University, United States
Figures
Figure 1
Conceptual framework of stopping and deciding not to go as separate but dependent processes.
(A) The organization of corticobasal ganglia pathways. Execution signals are relayed via the direct pathway (green connections) that result in a disinhibition of thalamic signals to cortex form the …
Figure 2
Reactive stopping task and behavioral results.
(A) Timeline of a go (left) and stop trial (right) in the reactive experiment. (B) Mean observed probability of stopping in the baseline (dark, solid) and caution (light, dotted) conditions; dots …
Figure 3
Comparison of reactive stopping models.
Fits of the three reactive models (Figure 1C–E) to behavioral data in the baseline condition, shown against (A) the histogram of RTs for correct (top) and incorrect (i.e., responses made on stop …
Figure 4 with 1 supplement
Comparison of modulation models in reactive task.
Goodness-of-fit measures for execution modulation models in reactive task. Bars show the estimated Akaike information criterion (AIC) and Bayesian information criterion (BIC) for the drift, onset, …
Figure 4—figure supplement 1
Model predictions of behavior in reactive task.
Example fits for modulation models on the execution process inthe baseline and caution conditions of the reactive task. Predicted RT distributions and stop accuracy are shown for the (A) drift-rate …
Figure 5 with 1 supplement
Proactive no-go decision task and behavioral results.
(A) Timeline of correct high (left) and low (right) go probability trials in the proactive task. The low probability example shows a trial in which a stop signal was presented. (B) Mean probability …
Figure 5—figure supplement 1
Subject-wise correlation of reactive and proactive control.
Correlation between the average PSEs in the reactive stopping curves and proactive no-go curves collapsed across baseline and caution conditions. Each point represents a single subject. PSE: point …
Figure 6 with 1 supplement
Comparison of modulation models in proactive task.
Goodness-of-fit measures for execution process modulation models in proactive task. Same plotting conventions as in Figure 4. As with the reactive experiment, the drift modulation model provided …
Figure 6—figure supplement 1
Model predictions of behavior in proactive task.
Predicted RT distributions and no-go probability curves in the proactive task for the drift modulation (green, A), onset modulation (blue, B), onset and drift modulation (yellow, C), and boundary …
Figure 7
Simulated interaction between trial outcome and response expectation on BOLD activation.
Time-course of BOLD and mean activity in the proactive task predicted by the (A) drift modulation, (B) onset modulation, and (C) boundary modulation models. Time courses (left column) reflect the …
Figure 8 with 1 supplement
Observed interaction between trial outcome and response expectation on BOLD activation.
Contrast maps for the comparison of no-go responses, modulated by go trial probability, against modulated go responses in the proactive task (center panel). Warm colors show areas where the …
Figure 8—figure supplement 1
In-scanner EMG protocol.
(A) Example of modified EMG set-up using the in-house Siemens physiological monitoring unit for monitoring cardiac signals. The recording sensors were placed to estimate the electrical vector from …
Tables
Table 1
Reactive model parameter estimates and fit statistics. In the top panel, best fit parameter estimates for boundary height (a), onset delay (tr), execution drift rate (ve), braking drift rate (vb), …
| Model | a | tr | ve | vb | sso | xb | χ2 | AIC | BIC |
| DPM | 0.534 | 0.174 | 1.266 | −0.990 | 0.878 | 0.0028 | −122.40 | −128.018 | |
| Ind-RM | 0.250 | 0.338 | 1.127 | 1.269 | 1.52 | 0.0075 | −106.652 | −112.270 | |
| Int-RM | 0.445 | 0.220 | 1.195 | 3.023 | 0.197 | 1.474 | 0.0069 | −104.379 | −111.815 |
| Drift | 0.536 | 0.178 | B: 1.289 C: 1.243 | −0.984 | 0.877 | 0.0051 | −273.459 | −273.301 | |
| Onset | 0.531 | B:0.171 C: 0.180 | 1.236 | −0.960 | 0.893 | 0.0054 | −271.651 | −271.492 | |
| Drift and onset | 0.538 | B: 0.173 C: 0.178 | B: 1.269 C: 1.247 | −0.989 | 0.858 | 0.0063 | −260.793 | −263.941 | |
| Bound | B: 0.525 C: 0.551 | 0.178 | 1.268 | -0.984 | 0.878 | 0.0057 | −269.994 | −269.835 |
-
DPM: dependent process model; Ind-RM: independent race model; Int-RM: interactive race model.
Table 2
Proactive model parameter estimates and fit statistics. Best fit parameter estimates for boundary height (a), onset delay (tr), execution drift rate (ve), and dynamic bias gain (xb) are listed for …
| Model | a | tr | ve | xb | P0 | P20 | P40 | P60 | P80 | P100 | cXχ2 | AIC | BIC | |
| Drift | 0.487 | 0.292 | 1.563 | ve | 1.411 | 1.562 | 1.683 | 1.761 | 1.880 | 1.925 | 0.0022 | −122.65 | −130.08 | |
| Onset | 0.628 | 1.42 | 0.641 | tr | 0.182 | 0.161 | 0.134 | 0.117 | 0.084 | 0.076 | 0.0095 | −99.38 | −106.81 | |
| Drift and onset | 0.06 | 1.468 | ve | 0.831 | 0.970 | 0.968 | 0.979 | 0.932 | 1.079 | 0.0033 | −104.65 | −122.77 | ||
| tr | 0.515 | 0.506 | 0.492 | 0.479 | 0.451 | 0.463 | ||||||||
| Bound | 0.272 | 0.914 | 0.913 | a | 0.379 | 0.344 | 0.305 | 0.281 | 0.246 | 0.236 | 0.0099 | −98.65 | −106.09 |
Additional files
-
Supplementary file 1
Table of significant clusters for the no-go parametric minus go parametric contrast shown in Figure 8A.
Coordinates are centers of mass for the cluster in MNI-space. N is the number of voxels in each cluster. Values in the left six columns show average condition-wise (general linear model) GLM coefficients and standard deviation across subjects is in parentheses.
- https://doi.org/10.7554/eLife.08723.017
