The identification of environments suitable for selection should enable plant breeders to test reliably a larger number of genotypes with given resources. This research was undertaken to evaluate discrimination ability (DA) and prediction ability (PA) of eight environments involving two locations. Eckartsweier (Eck) 1 and Voelkenrode (Voe); 2 yer. 1984 (84) and 1985 (85); and two harvests, silage harvest (I) and later (II). It was based on dry matter yield of plant (PDMY), ear (EDMY) and stover (SDMY) and on in vitro digestible organic matter (IVDOM), neutral detergent fibre (NDF), acid detergent fibre (ADF) and acid detergent lignin (ADL) of stover of 12 inbred lines and their 66 diallel crosses in maize (Zea mays L.). Linear regression coefficient of the performance of genotypes in an environment on that averaged across all environments and its degree of determination were used as measures of the DA and PA of environments, respectively. In hybrids, the DA of environments differed significantly for all traits except SDMY and ADL. Environment Eck851 showed better PA (≥0.63) than other environments for PDMY, EDMY, IVDOM, NDF, and ADF. Among the 12 two-environment combinations. EckI was a better predictor for PDMY, EDMY, IVDOM and NDF (PA≥0.80). The more productive environments showed better DA and PA for PDMY and FDMY than less productive ones. For IVDOM and NDF the first harvest provided better discrimination and prediction than the second harvest. In inbred lines the DA of environments differed significantly for EDMY only. In the present study, PA and DA seemed to be functions of the environments, PA was improved by combining two environments, and some environments seemed to be suitable for preliminary selection of genotypes for such diverse traits as yield and digestibility.