Irrigation is a key factor for improving raspberry fruit production, especially under summer droughts characteristic of lowlands. However, when the available irrigation water is insufficient to meet crop water use, a kind of regulated deficit irrigation has to be applied. In such cases it is important to determine the limit of the water application reduction, during the whole vegetation period or in specific phenological phases, to provide reasonable levels of the yield, respectively of the net incomes from irrigation. For that purpose, there are empirical parametric equations (models) developed in order to describe the relationship between the yield and the water application rate. These equations, however, must be calibrated for each crop and towards the site specific soil and climate conditions. In the present paper, three independent models of the ‘water-yield’ relationship are calibrated and compared. The experimental work was carried out during an eight-year period (third to tenth vegetation) in a 400 m2 raspberry plantation of the primocane-fruiting Lyulin cultivar. Seven irrigation treatments were studied in four replications. During the main phenophases –intensive growth (F1); blossom (F2); and fruiting (F3) – the water was applied in amounts recovering 100%, 75% and 50% of ETC respectively. Fertilizers were applied through the irrigation system, the fertilization dose being the same in all treatments. According to the obtained results, the relationship between the yield and the water application rate was approximated with high reliability by all studied models, but the equation parameters were different each year. The averaged over all experimental years’data show that the raspberry susceptibility to reduction of the water application rates was comparable in the phases of blossom and fruiting. The raspberry susceptibility was significantly lower during the phase of intensive growth, probably because of the larger rainfall amounts during that period. It can be concluded that when the examined models were calibrated by years they approximated the ‘water-yield’ relationship with high reliability (R=0.71÷0.98). The models were more universal when calibrated using averaged over all experimental years’ data, but in that case their precision was relatively lower.