The herbage yield responses to phosphorus (P) fertilization show large variation and, unlike to cereals, initial soil test P (STP) does not always predict such variation. By using meta-analysis, we reviewed quantitatively 38 Finnish field experiments related to the effects of P fertilization on herbage dry matter (DM) yields in the 1960s – 2000s. The current meta-analysis has several advantages over other type of research aiming to summarize the effect of P fertilization on herbage production. It involves large number of studies, diversity of soils and cultivation zones, and it enables to highlight the sources of variation in responses across the studies. Along with clay and coarse-textured mineral soils, we observed 20 studies on organic soils, whereas previous European and Canadian studies on the same topic involved mostly mineral soils. Moreover, we developed yield response models for different soil textures and initial STP levels. The summarized effect of P fertilization (mean 50 kg P ha-1) on DM yields was 13% over the control (N and K fertilization) with large variation across the studies from –10% to 70%. The major sources of variation were soil texture, soil acidity (for organic soils), initial STP and the yield level in control. For example, the yield response was three times larger on organic soils than on clay or coarse-textured mineral soils. Moreover, the yield response was double on slightly acidic organic soils (29%, n = 8) compared to that on moderately acidic ones (15%, n = 11). Thus, for acidic organic soils, even with low STP, P fertilization may be practically useless until the soils are first limed. Meta-regression proved that generally the responses reduced with increasing initial STP across the studies and reached zero at high STP (PAc, 21 mg l-1). However, in some studies with high control yields, no responses were observed even at low STP, since the responses decreased linearly with increasing control yields. Finally, the responses did not statistically differ between the cultivation zones, cut frequency, decades, or between the studies having different duration of experiments. Variation in herbage yields that is often observed in P fertilizer experiments is only partly related to increasing P rates, but also other factors should be taken into account. Further studies should address to the role of, e.g., soil structure in yield development, as it is evident that additional relevant explanatory variables would be needed to adequately explain the variation in herbage yields.