Step-wise Regression:
Step-wise regression is one of several computer-based iterative variable-selection procedures. Variables are added one-by-one based on their contribution to R-squared, but first, at each step we determine whether any of the variables (already included in the model) can be removed. If none of the variables can be removed, we determine whether a non-yet-included variable can be added. A variable can be added to the model at a step, removed at a following step, etc.
See Variable-Selection Procedures for criteria of "can be removed" and "can be added."