The simple form would be to get a group of people, have them state a major (minor?) life goal, assign them randomly to two groups, one of which does affirmations, and one of which does not. Time period and measure of success would depend on what exactly your theory was, but say 1 year and count all quantifiable steps in the target direction through a self-reporting format, with experimenter followup. (Alternatively, if you have funds for it, keep the experiment running until a given percentage reach their stated goal, and compare success rates; however, multiyear studies are much more expensive and difficult -- and have more participant dropouts).