Stalled fertility declines have been identified in several regions across the developing world, but the current conceptualization of a stalled fertility decline is poorly theorized and does not lend itself to objective measurement. We propose a more rigorous and statistically testable definition of stalled fertility decline that can be applied to time-series data. We then illustrate the utility of our definition through its application to data from rural South Africa for the period 1990-2005 collected from a demographic surveillance site. Application of the approach suggests that fertility decline has indeed stalled in rural KwaZulu-Natal, at about three children per woman. The stall, some 20 percent above the replacement fertility level, does not appear to be associated with a rise in wanted fertility or attenuated access to contraceptive methods. This identification of a stalled fertility decline provides the first evidence of such a stall in southern Africa, the region with the lowest fertility levels in sub-Saharan Africa.