Understanding the mechanisms that maintain the coexistence of plant species is critical to addressing the global biodiversity crisis. Increasing attention has been paid to interactions between plants and soil microbes (plant-soil feedback, PSF), which can promote plant coexistence by increasing stabilizing effects, but also hinder it by generating competitive fitness differences. However, the predictive power of the PSF has been questioned in recent studies because estimates of microbially mediated coexistence have correlated poorly with the outcomes of plant interactions observed in the field. This discrepancy may be due to the approaches typically used in PSF research, such as measuring PSF effects on a single vital rate or using soil conditioned for a short time period and without considering abiotic contexts. Here, I examined the effects of soil inoculum with different training histories and training environments (with and without added nutrients) on germination, seedling survival, and biomass of four grassland species. I then examined whether predictions of microbially mediated coexistence of four species pairs were sensitive to the vital rate identity, conditioning history, and soil training environment. I found that conspecific inoculum trained for longer had increasingly positive and negative effects on germination and biomass, respectively, although the effects of inoculum history varied across species and training environments. Estimates of microbially mediated outcomes were directly related to the vital rate used: when based on biomass and seedling survival, all four pairs were predicted to coexist, but only two pairs could do so when based on germination due to much reduced or even negative stabilization. Although coexistence predictions were not significantly related to conditioning history (including the effects of both variable conditioning durations and combinations of conditioning species) or nutrient treatments, both factors had a significant effect on stabilization. These results suggest that predictions of microbially mediated coexistence may be biased when based on a single vital rate, such as plant growth. To obtain more realistic and accurate outcome estimates, PSF effects should be integrated across different life stages, considering the temporal and abiotic contexts of these effects specific to a focal study system.