We study lower and upper bounds on the parameters for stochastic state vector reduction, focusing on the mass-proportional continuous spontaneous localization (CSL) model. We show that the assumption that the state vector is reduced when a latent image is formed, in photography or etched track detection, requires a CSL reduction rate parameter λ that is larger than conventionally assumed by a factor of roughly 2 × 10 9±2 , for a correlation length r C of 10 −5 cm. We reanalyze existing upper bounds on the reduction rate and conclude that all are compatible with such an increase in λ. The best bound that we have obtained comes from a consideration of heating of the intergalactic medium (IGM), which shows that λ can be at most 3 × 10 9±1 times as large as the standard CSL value, again for r C = 10 −5 cm.(For both the lower and upper bounds, quoted errors are not statistical errors, but rather estimates based on modeling uncertainties.) We discuss modifications in our analysis corresponding to a larger value of r C . With a substantially enlarged rate parameter, CSL effects may be within range of experimental detection (or refutation) with current technologies.2