We show how a simple reparameterization can reduce the number of parameters that need to be estimated by numerical maximum likelihood in site occupancy models. Three examples are provided.