Background and aims: Germination as a function of soil water potential (h) is modelled using polyethylene glycol (PEG). But, PEG would not consider soil properties. Our objective is to show the limitations of PEG to model germination in real soils. Using a new device, the tension germinator (TG), we show the interaction between soil type, h and seed characteristics on seed germination.TG (A Mariotte reservoir that supplies water at constant h to a porous substrate on which seeds are deposited). Barley (Hordeum vulgare L.) and vetch (Vicia sativa L.) seeds were placed on TG with loam (TG-loam) and sand (TG-sand), and h of 0, -0.002, -0.006 MPa. Then, the imbibition curves were monitored. PEG experiments (0 to -2.5 MPa) were performed to estimate the critical h, hPEG, or h from which the imbibition curve decreases compared with that at 0 MPa. PEG curves for 0 > h > -0.01 MPa were compared with TG.PEG and TG curves were not different at 0 MPa. hPEG for barley and vetch was within [0, -0.01] MPa. While no differences were observed between PEG at [0, -0.01] MPa and TG-loam curves, TG-sand curves at -0.002 and -0.006 MPa were different to those of PEG. Conversely to PEG at -0.01 MPa, no imbibition was observed in TG-sand (-0.006 MPa). A negligible influence of h in TG-loam was observed.PEG is not adequate to describe seed germination in soil. But the TG allows monitoring seed germination in real soils and controlled h.Methods: Germination as a function of soil water potential (h) is modelled using polyethylene glycol (PEG). But, PEG would not consider soil properties. Our objective is to show the limitations of PEG to model germination in real soils. Using a new device, the tension germinator (TG), we show the interaction between soil type, h and seed characteristics on seed germination.TG (A Mariotte reservoir that supplies water at constant h to a porous substrate on which seeds are deposited). Barley (Hordeum vulgare L.) and vetch (Vicia sativa L.) seeds were placed on TG with loam (TG-loam) and sand (TG-sand), and h of 0, -0.002, -0.006 MPa. Then, the imbibition curves were monitored. PEG experiments (0 to -2.5 MPa) were performed to estimate the critical h, hPEG, or h from which the imbibition curve decreases compared with that at 0 MPa. PEG curves for 0 > h > -0.01 MPa were compared with TG.PEG and TG curves were not different at 0 MPa. hPEG for barley and vetch was within [0, -0.01] MPa. While no differences were observed between PEG at [0, -0.01] MPa and TG-loam curves, TG-sand curves at -0.002 and -0.006 MPa were different to those of PEG. Conversely to PEG at -0.01 MPa, no imbibition was observed in TG-sand (-0.006 MPa). A negligible influence of h in TG-loam was observed.PEG is not adequate to describe seed germination in soil. But the TG allows monitoring seed germination in real soils and controlled h.Results: Germination as a function of soil water potential (h) is modelled using polyethylene glycol (PEG). But, PEG would not consider soil properties. Our objective is to show the limitations of PEG to model germination in real soils. Using a new device, the tension germinator (TG), we show the interaction between soil type, h and seed characteristics on seed germination.TG (A Mariotte reservoir that supplies water at constant h to a porous substrate on which seeds are deposited). Barley (Hordeum vulgare L.) and vetch (Vicia sativa L.) seeds were placed on TG with loam (TG-loam) and sand (TG-sand), and h of 0, -0.002, -0.006 MPa. Then, the imbibition curves were monitored. PEG experiments (0 to -2.5 MPa) were performed to estimate the critical h, hPEG, or h from which the imbibition curve decreases compared with that at 0 MPa. PEG curves for 0 > h > -0.01 MPa were compared with TG.PEG and TG curves were not different at 0 MPa. hPEG for barley and vetch was within [0, -0.01] MPa. While no differences were observed between PEG at [0, -0.01] MPa and TG-loam curves, TG-sand curves at -0.002 and -0.006 MPa were different to those of PEG. Conversely to PEG at -0.01 MPa, no imbibition was observed in TG-sand (-0.006 MPa). A negligible influence of h in TG-loam was observed.PEG is not adequate to describe seed germination in soil. But the TG allows monitoring seed germination in real soils and controlled h.Conclusions: Germination as a function of soil water potential (h) is modelled using polyethylene glycol (PEG). But, PEG would not consider soil properties. Our objective is to show the limitations of PEG to model germination in real soils. Using a new device, the tension germinator (TG), we show the interaction between soil type, h and seed characteristics on seed germination.TG (A Mariotte reservoir that supplies water at constant h to a porous substrate on which seeds are deposited). Barley (Hordeum vulgare L.) and vetch (Vicia sativa L.) seeds were placed on TG with loam (TG-loam) and sand (TG-sand), and h of 0, -0.002, -0.006 MPa. Then, the imbibition curves were monitored. PEG experiments (0 to -2.5 MPa) were performed to estimate the critical h, hPEG, or h from which the imbibition curve decreases compared with that at 0 MPa. PEG curves for 0 > h > -0.01 MPa were compared with TG.PEG and TG curves were not different at 0 MPa. hPEG for barley and vetch was within [0, -0.01] MPa. While no differences were observed between PEG at [0, -0.01] MPa and TG-loam curves, TG-sand curves at -0.002 and -0.006 MPa were different to those of PEG. Conversely to PEG at -0.01 MPa, no imbibition was observed in TG-sand (-0.006 MPa). A negligible influence of h in TG-loam was observed.PEG is not adequate to describe seed germination in soil. But the TG allows monitoring seed germination in real soils and controlled h. [ABSTRACT FROM AUTHOR]