Quantifying predator–prey body size relationships is key to understanding food webs. Food web models often assume that all predator individuals prefer the same relative body size of prey, using a single constant called preferred predator–prey mass ratio (preferred PPMR). In contrast, empirical studies have shown that relative prey body size in diet varies with individual predator size. We point out that this apparent inconsistency arises because PPMR has been measured only through dietary data (realized PPMR) without considering environmental prey availability, suggesting that preferred PPMR may be size-independent. Here, we revisit the assumption of size-invariant preferred PPMR in food web models. We compare two measures of PPMR calculated from prey compositions in predator diet and environmental prey composition (realized PPMR vs. environmental PPMR), respectively, and consider their deviations as a proxy of individual variations in relative prey size preference (preferred PPMR). We apply this idea to long-term dietary data of an omnivorous predatory fish species collected from Lake Biwa over four decades. Our results showed that the preferred PPMR is size-independent when the foraging mode of the predator is considered while the realized PPMR is size-dependent regardless of the foraging mode. We suggest that the apparent inconsistency between theoretical assumption and empirical observation of PPMR is due to the conceptual confusion. Further, in contrast to the previous arguments based on realized PPMR, we provide the first evidence of size-invariant preferred PPMR.