Survival in larval rearing experiments is difficult to estimate due to accidental losses and periodic sampling. The number of sampled fish can be a large proportion of the stocked ones, making it difficult to calculate the overall survival rate and mortality coefficient as this is based on the initial number. Here, a new method of calculating survival is proposed using the mortality coefficient. When the initial stocking density and sampled and final numbers are known, and assuming that mortality coefficient is constant, the final number of fishes can be represented by the formula N t = e -mt (N 0 -RN Sn e mdn ), where t is rearing period (days), N 0 indicates initial number, N t indicates the survival number at t days of rearing, m is the natural mortality coefficient, N Sn is the sampled number in the nth sampling, and dn is the rearing period until removal of the nth sample. The provisional mortality coefficient is calculated from initial and final stocking numbers. Then values for the natural mortality coefficient are substituted into the formula with successive approximation. The coefficient, which most closely approximates the actual survival, is determined as the best fit natural mortality coefficient. Examples of larval experiments are provided to demonstrate the method and show that survival is often underestimated using traditional methods.