Figure 3.

Expected behavior of Model 1, the zero-drift diffusion model (drift=0) with threshold adaptation (θ_i = ξ/R_i). A) Surface shows expected proportion of 1-responses as a function jointly of reward rate for 1-responses (R₁) and reward rate for 2-responses (R₂); radial lines in the R₁,R₂ plane show contours of constant choice probability. B) Expected response time as a function of R₁ and R₂; this 3D plot is rotated relative to panels A and C to make the shape of the surface more easily discernible; since its height goes to infinity, it is also truncated to 5 sec. C) Expected 1-response rate (B₁) as a function of R₁,R₂ (notice the collapse to 0 along both the R₁ and R₂ axes). D) Scatterplot of random R₁,R₂ pairs, representing blocks of 100 responses in which those reward rates were obtained. E) Scatterplot showing expected choice proportions plotted vs. relative reward rate in those blocks of responses; the best-fitting line (solid) is superimposed, and its equation is displayed. Matching predicts a slope of 1 and intercept of 0 (dashed line). F) Choice ratio B₁/B₂ plotted vs. the reward ratio R₁/R₂ on a log-log scale, which is frequently used for highlighting generalized matching behavior. Predicted behavior is plotted as a dashed line; best-fitting line is solid.