Economic and behavioral theories arrive at different conclusions about the effect of being allowed to borrow from one’s defined-contribution (DC) retirement plan on people’s contributions to DC plans. Traditional life-cycle models unambiguously suggest that the borrowing option makes people better off than not being able to borrow. Households consequently contribute more to their DC plans than they would absent the borrowing option. Previous research finds that the ability to borrow from a DC plan increases contemporaneous contributions, consistent with traditional models. Behavioral finance, in contrast, suggests that some workers may operate with nonlinear time discounting. They plan on saving more in the future but change their mind and save less than initially planned as time passes. These workers may enjoy higher lifetime utility if they have no loan option because DC plans serve as commitment devices for retirement saving. The money cannot be used prior to retirement. Absent this commitment device, contributions may be lower for some households than would be the case without a DC loan option. We study DC plan contributions for households with heterogeneous preferences about discounting. We separate households into those that demonstrate inconsistent (or paradoxical) borrowing behavior, which may reflect nonlinear time discounting, and those with more consistent borrowing behavior. We find that a DC loan option raises current savings, but does so more for households with consistent borrowing behavior than for those with inconsistent borrowing behavior.