Later-age forest fertilizer experiments can be quite difficult to analyze appropriately and in ways that extract all the information inherent in the collected data. Observed responses are likely to be partially confounded with variations in pretreatment stand development, most of which can be removed through analyses of covariance using quanta of initial growing stock as a covariate. Pretreatment growth rate may represent an even more discerning covariate. Rather than use covariance some authors have chosen, instead, to use arithmetical procedures to adjust treatment responses. Reanalysis of a Canadian experiment of this latter kind suggests that such methodology may be less than ideal, and should not be preferred to covariance analysis. A general and systematic procedure for examining forest nutrition experiments is proposed for those involving t treatments, and also those where the t treatments represent n factors at p levels in factorial combination. An example of adopting the recommendation methodology for the first type is given using a completely randomized experiment in naturally regenerated radiata pine in New Zealand with five replications of four treatments. Adoption of the suggested procedures in conjunction with two covariates provides a useful insight into the data, and appreciably increases precision. The system is sequential in structure, necessarily inducing some risk of erroneous hypothesis testing. Such a danger is usually minimal, however, and the suggested system, it is claimed, represents a useful method for isolating treatment and growth effects in forest fertilizer trials. For. Sci. 34(3):769-780.