A comment on the Revised Diffusion Model for Conflict tasks (RDMC)

Abstract

In conflict tasks, such as the Simon, Eriksen flanker, or Stroop task, a relevant and an irrelevant feature indicate the same or different responses in congruent and incongruent trials, respectively. The congruency effect refers to faster and less error-prone responses in congruent relative to incongruent trials. Distributional analyses reveal that the congruency effect in the Simon task becomes smaller with increasing RTs, reflected by a negative-going delta function. In contrast, for other tasks, the delta function is typically positive-going, meaning that congruency effects become larger with increasing RTs. The Diffusion Model for Conflict tasks (DMC; Ulrich et al., Cognitive Psychology, 78, 148-174, 2015) accounts for this by explicitly modeling the information accumulated from the relevant and the irrelevant features and attributes negatively- versus positively-sloped delta functions to different peak times of a pulse-like activation resulting from the task-irrelevant feature. Because the underlying function implies negative drift rates, Lee and Sewell (Psychonomic Bulletin & Review, 31(5), 1-31, 2024) recently questioned this assumption and suggested their Revised Diffusion Model for Conflict tasks (RDMC). We address three issues regarding RDMC compared to DMC: (1) The pulse-like function is not as implausible as Lee and Sewell suggest. (2) RDMC itself comes with a questionable assumption that different parameters are required for congruent and incongruent trials. (3) Moreover, we present data from a new parameter recovery study, suggesting that RDMC lacks acceptable recovery of several parameters (in particular compared to DMC). In this light, we discuss RDMC as not (yet) a revised version of DMC.

Keywords: Decay; Diffusion Model for Conflict tasks; Diffusion model; Inhibition; Parameter recovery.

Fig. 1

Left panel: DMC (Ulrich et al., 2015), right panel: RDMC (Lee & Sewell, 2024). Note. For DMC, the solid black line represents the expected time-course of activation within the controlled channel, the dotted green and red lines are the respective activations within the automatic channel, and the solid green and red lines are the overall expected activations from superimposing both channels (green: congruent, red: incongruent). For RDMC, the dashed and the dotted lines are expected activations within the controlled and the automatic channel, respectively. The solid green and red lines are the overall expected activations from superimposing both channels (green: congruent, red: incongruent). The figures were drawn with the following parameter values, DMC: $A=20$, $\tau =80$, $a=2$, ${\mu }_c=0.5$; RDMC: $A_0=0.8$, $k_c=10$, $k_i=30$, $d_c=0.6$, $d_a=0.4$ (see Table 3 and Figure 7 in Lee & Sewell, 2024). DMC is presented in milliseconds and RDMC in seconds for consistency with the original publications

Fig. 2

DMC (upper row) and RDMC (lower row) predictions for the Simon task data. Note. In all panels, model predictions are shown as lines, and observed values are shown as dots. Specific values were obtained by averaging across participants. The leftmost panels (i.e., A and D) show predicted and observed quantiles. The middle panels (i.e., B and E) show delta functions which can be derived by plotting the difference between quantiles (to the same probability) of incompatible and compatible trials against their mean. The rightmost panels (i.e., C and F) show CAFs, derived by first binning RTs and then calculating the proportion of correct responses per bin

Fig. 3

DMC (upper row) and RDMC (lower row) predictions for the flanker task data. Note. Figure legend is identical to Fig. 2