Bounded rationality questions backward induction, which however, does not exclude such reasoning when anticipation is easy. In our stochastic (alternating offer) bargaining experiment, there is a certain first-period pie and a known finite deadline. What is uncertain (except for the final period) is whether there is a further period. Whereas backward induction requires information about all later pie sizes and probabilities, forward reasoning is expected to consider only the immediate prospects. Rather than relying only on decision data, we try to assess the cognitive approach such as forward reasoning of backward induction by control of information retrieval. We find that participants who begin with the shortest games before playing possibly longer games, initially resort to backward induction before switching to forward-looking behavior.