In this study we consider the problem of estimating, for management pu
rposes, a minimum biomass reference level at which recruitment to a fi
sh stock is seriously reduced. We take an empirical-comparative approa
ch to the problem by examining observations on a wide range of fish st
ocks. Eight methods for estimating spawning stock biomass thresholds f
or recruitment overfishing are investigated. Their behaviour is tested
using stock and recruitment data for 72 finfish populations, each wit
h at least 20 years of data. We considered three classes of thresholds
defined by: (1) the stock size corresponding to 50% of the maximum pr
edicted average recruitment; (2) the minimum stock size that would pro
duce a good year class when environmental conditions are favourable; a
nd (3) the stock size corresponding to 20% of various estimates of vir
gin stock size. The estimators of the first type are generally prefera
ble because they are easily understood, relatively robust if only data
at low stock sizes are available, and almost always result in higher
levels of recruitment above the threshold.