When analyzing subjective expectations, researchers commonly apply (non-)parametric approximations to point identify beliefs. We propose a new take on this type of data that does not impose a functional form on expectations. Using the widely researched example of subjective survival expectations, we construct bounds for subjective survival curves. These bounds allow us to partially identify subjective life expectancy. We show that the informativeness of the bounds depends on our willingness to interpolate beliefs between data points. If we do not smooth between the elicited points on thesurvival functions, the bounds are too wide for useful inference. However, if we do interpolate and allow for a limited amount of rounding, the resulting bounds are narrow enough to show variation in life expectancy with age and self-reported health, the strongest predictors in point identi ed models. Finally, we match the subjective data to life tables. While analysis that point identifies life expectancy, either parametrically or non-parametrically, rejects consistency of expectations with actuarial forecasts for women, the bounds show that allowing for rounding renders the subjective dataconsistent with forecasts on average.