Policy-makers from both sides of the political aisle are actively seeking ways to encourage retirement saving among middle- and low-income households. A growing body of evidence suggests that making it easy for these families to save, and presenting them with clear and effective financial incentives to do so, succeeds in generating significantly higher contributions to retirement accounts. One of the most auspicious ways to make it easier for households to save, for retirement and other purposes, is by allowing them to directly deposit part of their income tax refund into a savings vehicle. This policy brief examines ways of encouraging households to save at one of their most ?savable? moments: when they learn they will receive a substantial federal tax refund. Currently, the Internal Revenue Service (IRS) permits refund recipients to identify a single destination for refunds. Many have called for the IRS to allow refund recipients to split their refunds, to facilitate saving in the same way that most payroll systems let employees ?split? their paychecks and deposit part in savings products. The splitting proposal has been adopted by the Bush Administration and has been in the budget for the last two years. Twelve Members of Congress wrote to IRS Commissioner Mark W. Everson on January 31, 2005, requesting that the IRS adopt technical changes that would enable split refunds. On March 25, 2005, Commissioner Everson replied that the IRS was ?working toward making this program available as quickly as possible? and set a deadline of the 2007 filing season. This policy brief addresses some practical issues with this idea, which are also partially addressed in an upcoming Retirement Security Project policy brief by Mark Iwry. In particular, we advance the idea of establishing a refund-based savings option available to all at tax time?U.S. savings bonds, purchased out of a filer?s refund. The policy brief describes results from an experiment we conducted that underscores the potential benefits from this split refund proposal. The experiment was not focused specifically on retirement savings, but nonetheless provides insight into the process of saving for other purposes?as well as the steps that would be necessary to encourage households to set aside part of their income tax refund specifically for retirement purposes. The core of the experiment involved offering refund recipients the ability to ?split? or ?bifurcate? their refunds, i.e., to direct refunds to multiple destinations. One of these destinations was a simple bank savings vehicle. The results were quite promising in some ways. A meaningful fraction (about 20 percent) of all refund recipients sought to save part of their refunds using our program. Nearly three-quarters of these were ?greenfield? savers, who reported no prior savings. Participants reported high levels of satisfaction and appeared to have spent their money on their predetermined goals. At the same time, however, the vast majority of those making contributions appear to have withdrawn the funds relatively quickly. Although these withdrawals may have fulfilled various goals that households had earlier identified, they nonetheless mean that little of the deposited funds were retained as long-term savings. To the extent such retention is a goal for policy-makers, other steps may be required to make these assets more ?sticky,? possibly including framing them as retirement savings.