When I first met Seth he was an assistant professor of psychology at UC Berkeley. At the time he wrote all his code in APL, and couldn't figure out why anyone would use anything else. More recently he used R. To everyone calling him "not a hacker," he was hacking on code before a lot of you were born, though his code was mostly for his own purposes.
I'm a decent R user, have done many self-experiments myself, and read & commented on Roberts's blog for years; I have no idea about APL, but I think you overestimate how much hacking on R code Roberts was able to do - he had a lot of trouble programming the reaction-time code for his Buttermind experiments and I don't recall him ever reporting more complex analyses than a linear model.
