Article Figures & Data
Figures
- Figure 1.
TMS experiment procedure. A, Random dot motion task: after a fixation cross and a period of random motion, coherent motion (here: upward, coherence 70%) was displayed for 2000 ms or until response (the same task was used in the EEG experiment). B, Response setup in TMS experiment: participants held one button (up) between their thumb and index finger (pinch) and one in the palm of their hand (down), attached to a cylinder (grasp); EMG electrodes were placed on the ADM and FDI. C, Example EMG traces from a single trial (here, a hard speed trial, where the responding muscle is the FDI and the nonresponding muscle is the ADM). D, To create model predictions which are comparable to MEP data, accumulation values from both the correct accumulator (corresponding to the responding muscle) and the incorrect accumulator (corresponding to the nonresponding muscle) are sampled at simulated TMS times. E, Illustrative real MEP amplitudes (from the speed/easy condition) collated from all participants. F, MEPs and simulations (data not shown) are then z-scored per muscle, participant, and session (note that latencies were normalized by the median, not maximum, EMG RT for each participant). G, Real and simulated continuous signals can be created for each muscle (responding, nonresponding), using a Gaussian smoothing kernel. H, However, to remove nonspecific processes, the same smoothing is applied to the difference between simultaneously recorded MEPs (responding minus nonresponding).
- Figure 2.
Behavioral results for both the TMS experiment (A) and the EEG experiment (B): reaction time (left) and accuracy scores (right) for each condition. Top left panel shows both EMG RT (bars) and button RT (dashed lines). Error bars indicate 95% confidence interval; **p <.001.
- Figure 3.
Neural and modeling results. Top, Neural data. Bottom, Model comparison. Left: TMS experiment. Right, EEG experiment. A, Stimulus-locked (left) and response-locked (right) MEP signal for each condition. Each panel shows both the MEP signal associated with the responding muscle (dark) and the nonresponding muscle (light). Shaded areas indicate 95% confidence intervals. B, CPP: stimulus-locked (left) and response-locked (right) CPP wave form for each condition. The bottom right of the panel shows the topography of the ERP, averaged over the stimulus-locked time interval of 0–1000 ms. Electrodes used to generate CPP waveforms are highlighted. C, Stimulus-locked (left) and response-locked (right) MEP-average signal (responding minus nonresponding muscle). D, Stimulus-locked (left) and response-locked (right) model predictions made by the free-excursion variant of the best-supported model. E, Stimulus-locked (left) and response-locked (right) model predictions made by the forced-excursion variant of the best-supported model. F, Stimulus-locked (left) and response-locked (right) CPP; note that the CPP here is a pooled average rather than a grand average and therefore differs from B. Additionally, the wave form has been low-pass filtered with a cutoff of 5 Hz to assist comparison with model predictions. G, Stimulus-locked (left) and response-locked (right) model predictions (correct and incorrect accumulator summed) made by the free-excursion variant of the best-supported model. H, Stimulus-locked (left) and response-locked (right) model predictions (correct and incorrect accumulator summed) made by the forced-excursion variant of the best-supported model.
- Figure 4.
Model fit for the TMS experiment (A) and the EEG experiment (B): quantiles estimated from behavioral data (circles) and model 2 simulations (crosses and lines) for easy (top) and hard (bottom) decisions. For each condition, correct (thick) and incorrect (thin) quantiles are displayed separately. Note that the model fit is identical for the forced-excursion and the standard free-excursion race model.
Tables
Model Number A Sz vcorrect vincorrect Ter STer σ Number of parameters TMS experiment EEG experiment AIC BIC AIC BIC Model 1 Free Fixed Fixed Fixed Fixed Fixed Fixed 9 44,868 44,933 62,398 62,466 Model 2 Free Free Fixed Fixed Fixed Fixed Fixed 10 44,859 44,932 62,389 62,464 Model 3 Free Fixed Fixed Fixed Free Fixed Fixed 10 44,865 44,937 62,404 62,479 Bayesian Information Criterion (BIC) and AIC values for each model and each experiment (best BIC and AIC values in bold). The terms “fixed” and “free” here relate specifically to changes across speed/accuracy instructions, as accumulation rate (V) was always free to vary between difficulty conditions.
- Table 2.
Estimated parameter values for the best-supported model (model 2) when expressed with both free and forced-excursion in both experiments
Parameters TMS experiment EEG experiment Free excursion Forced excursion Free excursion Forced excursion Accuracy Speed Accuracy Speed Accuracy Speed Accuracy Speed SZ 0.447 0.523 0.447 0.586 0.319 0.541 0.319 0.664 A 1 0.893 1 1 0.815 1 Ter 0.382 0.382 0.257 0.257 STer 0.374 0.374 0.229 0.229 σ2 0.499 0.499 0.558 0.785 0.785 0.964 vcorrect Easy 1.280 1.28 1.433 2.475 2.475 3.038 Hard 0.634 0.634 0.710 1.350 1.350 1.656 vincorrect Easy 0.098 0.098 0.109 0.253 0.253 0.310 Hard 0.004 0.004 0.005 0.054 0.054 0.066 The response boundary A in the accuracy condition was set to 1 as a scaling parameter. Parameters are not comparable across experiments, as the TMS fit is to data normalized to the median RT of each participant.
In this issue
