Random-walk and diffusion models for two-choice comparison of paired successive or simultaneous stimuli focus on response time (RT), modeled as the time needed to reach one or the other barrier, and its relation to the response probabilities. Logit P1 = ln[P1/(1-P1)], where P1 is the probability of responding ”first greater,” can be seen as a measure of subjective stimulus difference, d. Signed response speed (SRS), ±1/RT with the sign of the response, yields another d measure. The two measures are highly correlated and, importantly, the intercept in the regression of logit P1 on mean SRS estimates the asymmetry of the starting point relative to the barriers, that is, the bias. New analyses of data from Patching, Englund, and Hellström (2011) show that this bias helps explain the variability of the time-and space order errors. Possible connections of the bias with the parameters in Hellström’s (2003) sensation-weighting (SW) model are explored.