A procedure for assessing birth spacing goals, an important component of fertility preferences, is proposed and applied to 1993 Costa Rican data. Based on a reverse or backward survival analysis, preferred birth intervals are estimated to range between 3·5 and 4·5 years (1·5 years for the interval union to first birth). These intervals are 2 or 3 years shorter than crude estimates from data on open or last closed intervals, which are upwardly biased by selection and left censoring effects. To achieve these spacing preferences, a cohort must spend about two-thirds of the time using contraception (one-third in the interval union to first birth). An inverse association between desired family size and desired birth interval is evident only in parity-specific analyses.