The varied statistical procedures used for model validation as well as fundamental and practical objections formulated by many authors to these procedures have led modelers to use non-inferential methods for empirical validation of models. The procedure presented by <A HREF="#freese">Freese (1960)</A> to validate forestry models and the modification by <A HREF="#rennie">Rennie and Wiant (1978)</A> are perfectly applicable to other areas. These procedures and the proposed extensions allow covering a scenario with data coming from models that may or may not have biased estimates, where the maximum admissible error is expressed either or both in terms of the same units or as a percentage of the real value. The illustration of the methodologies used data from a perennial grassland growth system.