Assessing the plant nitrogen (N) status is one of the major challenges for improving N fertilization management and reducing its environmental footprint on forage-based agricultural systems. We re-analyzed a dataset of biomass (W) and plant N concentration (%N) of tall fescue (Festuca arundinacea Schreb.) using a Bayesian statistical model for estimating the coefficients of the critical N (%Nc) dilution curve (Nc = A1 W − A2) with two objectives in mind: i) to revise the reference %Nc dilution curve established for tall fescue by Lemaire and Salette (1984), and ii) to analyze how the determination of %Nc curves is affected by the structure of the dataset (number of sampling dates and N rates) along with the range of shoot %N and W achieved. Our analysis suggests that the original tall fescue %Nc curve of Lemaire and Salette (1984) was overestimated. A critical %N curve for tall fescue was obtained using the Bayesian method across 14 unique genotype × environment × management (G × E × M) conditions (Nc = 3.93 W -0.42). We show that the high uncertainty associated with the %Nc curve could be reduced by increasing the number of experiments. When a single dataset was used, 95 % credibility intervals (95 % CI) were [2.52, 5.66] for A1 and [0.06, 0.63] for A2. However, 95 % CI were reduced up to a 73 % when the critical N dilution curve was based on 14 studies. We also show that a minimum of five studies (+100 data points for our study) are needed to avoid large biases and uncertainty levels in coefficient estimations with the Bayesian method. However, these studies must be carefully designed. The reliability of critical %N curves is greatly reduced when they are estimated with datasets comprising only a few data under non-limiting N conditions or data covering only very low or high W values. Our results suggest that more reliable critical N dilution curves for species can be developed by grouping numerous datasets covering a broad range of G × E × M conditions