For the last 15 years I've been employed by a software company that markets a matrix programming language (GAUSS), and part of my responsibility is to provide support. But it turns out that I spend a great deal of my time teaching people about computation on computers. I find social science researchers woefully undertrained in numerical issues. My presentation will give a general picture of what needs to be known about numerical methods on computers starting with floating point numbers in a 32 bit operating system. I have 36 years of experience solving statistical problems on computers and I would like to share that experience with the members of CSSS.