When we increase the distance from the origin to the A bound the reaction time for the positive coherences covers a broader range. The fast times become longer, and the long times become much longer. The reaction times for the negative coherences also cover a broader range as A increases, but the change is much less dramatic. The fast reaction times do not change significantly, and the slow reaction times do not increase as dramatically as the positive coherences.
The choice plot also changes. The positive coherences change very little, but the slope on the negative side becomes steeper as A increases.
As expected, changing the B bound, has the opposite affect.
Increasing alpha causes the reaction times slopes to change from increasing as the coherence approaches zero to decreasing as the coherence approaches zero. It also causes the fastest times to get slower, but has little affect on the slow times.
Increasing alpha causes the choice probability plots to become shallower. Too much alpha and you don't get 0 or 1 as a probability at all.
Increasing T_r causes the fastest and slowest reaction times to become slower. The plot doesn't change, just shifts upwards.
Has no affect on the choice probability graph.
Here we can see the fundamental problem with this model. Allowing A and B to be unconstrained gives us greater freedom to fit the data on any plot, but it does not change the fact that trying to fit data on one plot, will move the fit farther from the data in the other plot. This is due to what we saw above, the only parameter that does not affect both plots is Tr, all of the rest change both plots, most often in wrong direction for the other plot.
Errors are the circles, filled circles are correct. There is a red square for the zero coherence, where there are neither errors or corrects. The error data points with a dot in the center are based on a very few data points, and are not statistically significant.
