I neither "believe" nor "not believe" in ESP. I maintain philosophic doubt about all phenomena. (See Practical Dharma, Lojong 2.) While I eagerly await the publication of the full article, in the Journal of Personality and Social Psychology, Bem (in press), any statistical analysis that purports to support a very controversial or new idea, should present especially strong evidence.
B.F. Skinner got into trouble when he said the purpose of a science of behavior is to "predict and control" (from Beyond Freedom and Dignity). Critics assumed he meant the purpose of behavioral science is to predict and then control people. "Predict and control" is a hallmark of behavior analysis and relates to scientific knowledge. (The relationship between actual controlling variables and the organization of society is too complex to get sidetracked into here.)
The question of Psi as causal in behavior is the type of problem behavior analysts address with the demand that adequate proof must be demonstrated in repeatable functional relationships.
From my perspective, who cares if there is a statistical relationship due to factors we can not explain or control. What does this tell us except there is unexplained variance in the data (that begs to be explained).
Until otherwise demonstrated, I am (temporarily) operating under the assumption that the results reported in the upcoming Bem article are due to an unobserved and unexplained concurrent factor that explains both the "current" and "future" responses. Future research should attempt to find this independent variable and then demonstrate it can be controlled to reliably produce the dependent variable. (Maybe this independent variable will be Psi or maybe it will be a subtle expectation or demand characteristic, which has a long research history in the Jounal of Personality and Social Psychology - and IMJ is why the Journal published this.)