Hierarchical models in Bayesian analysis often face uncertainty in choosing priors for hyperparameters. Using improper priors from non-hierarchical models can lead to problems. A study found that hyperpriors on the boundary of admissibility are good choices for objective priors in normal hierarchical models. These hyperpriors are diffuse enough to be sensible without causing issues. The study considered various priors but did not reach a specific conclusion. A new objective prior has been proposed for all normal hierarchical models based on admissibility, ease of use, and performance.