Skip to main content
. 2016 Nov 7;26(21):2893–2898. doi: 10.1016/j.cub.2016.08.016

Figure 3.

Figure 3

LATER Analysis of Inter-movement Intervals

Although the patients made self-paced alternating movements, it is licit to treat the inter-movement intervals as reaction times relative to an endogenous timing signal setting the individual rate of alternation. The observation catalytic of the LATER model—that reaction times show a linear relationship when plotted as their reciprocals against their cumulative (assumed Gaussian) distribution—can thus be tested on our data. Plotted here so transformed are the intervals for the electrodes where a significant effect of stimulation (in red) was observed (the manual task in DH, top, and the vocal task in LW, bottom) with time on a reciprocal scale as the abscissa and the Z score as the ordinate index of position within a Gaussian distribution. Maximum likelihood fits of the major components of the distributions and (only in DH where it was present) separately for the minor early components are given in dashed lines. According to the LATER model, stimulation-induced reversed procrastination predicts a change in the slope of the function, causing it to swivel around a fixed intercept, whereas acceleration of the competing processes predicts a shift to the left along the abscissa, leaving the slope unchanged. Model comparison using the BIC as the metric of modeling felicity indicated that swivel was better than shift (change in BIC = 4.82, substantial evidence). It was also better than both the unconstrained (change in BIC = 13.45, very strong evidence) and the null model (change in BIC = 32.38, very strong evidence). LATER analysis thus here supports reversed procrastination. See Supplemental Experimental Procedures for details. Note the discretization of timing data in DH is a consequence of the relatively sparse temporal sampling of standard clinical video recording (every 40 ms). LATER modeling was performed using Mike Shadlen’s Reciprobit Toolbox v.1.0. See also Figures S1 and S2.