Beyond Protocols and Indices in Seed Germination Studies: Frequentist vs. Probabilistic Errors in Germinability Measures

Seed germination indices are commonly used to quantify either the frequency of germination within a tested seed sample or the temporal dynamics of germination events. Among these metrics, final germination percentage, or germinability, is conventionally calculated as the proportion of germinated seeds relative to the total number of seeds subjected to a trial. Although operationally simple, this formulation treats the entire tested seed pool as the biological reference population and may therefore become prone to interpretational error when non-viable seeds form a substantial component of the sample. Here, we examine the consequences of denominator choice in germinability estimation by contrasting the conventional unconditional formulation with a viability-conditioned probabilistic formulation. Using raw data from Cistus salviifolius, a characteristic Mediterranean obligate seeder, we first quantified variation in the proportion of non-viable seeds through a proportion-based meta-analysis. We then applied simulations based on triangular probability distributions, complemented by Beta-PERT sensitivity analyses, to compare the distributional behaviour of conventional and viability-conditioned germinability estimates and to identify the probability ranges in which both formulations converge or diverge. The results indicate that non-viable seed fractions vary substantially among datasets, spanning 12.8% to 55.0% across the six datasets examined. Simulated comparisons further suggest that the divergence between formulations is not uniformly distributed across the germinability spectrum but is most pronounced across intermediate germination ranges. These findings suggest that technical improvements in germination protocols alone may not fully resolve the conceptual limitations of conventional germinability estimates unless viability structure is explicitly incorporated into both experimental design and statistical interpretation.