I’ll just throw out there that some (many?) people model reaction times using an ex-Gaussian distribution, which is a sum of a normal distribution (to get the basic, unimodality of the RT distribution) and an exponential distribution (to get the long right tail). People like it, in part, because it allows the location of the normal component to be estimated while isolating it from the exponential tail component (perhaps because such isolation makes ). Examples include this paper by our own @twiecki.
2 Likes