Risk assessment of infant using a realistic persistent organic pollutant (POP) exposure through breast milk is essential to devise future regulation of POPs. However, recent investigations have demonstrated that POP levels in breast milk collected from the same mother showed a wide range of variation daily and monthly. To estimate the appropriate sample size of breast milk from the same mother to obtain reliable POP concentrations, breast milk samples were collected from five mothers living in Japan from 2006 to 2012. Milk samples from each mother were collected 3 to 6 times a day through 3 to 7 days consecutively. Food samples as the duplicated method were collected from two mothers during the period of breast milk sample collection. Those were employed for POP (PCBs, DDTs, chlordanes, and HCB) analysis. PCB concentrations detected in breast milk samples showed a wide range of variation which was maximum 63 and 60% of relative standard deviation (RSD) in lipid and wet weight basis, respectively. The time course trend of those variations among the mothers did not show any typical pattern. A larger amount of PCB intake through food seemed to affect 10 h after those concentrations in breast milk in lipid weight basis. Intraclass correlation coefficient (ICC) analyses indicated that the appropriate sample size for good reproducibility of POP concentrations in breast milk required at least two samples for lipid and wet weight basis.