The sum-peak counting method was applied for 125I absolute activity measurement. Count rate-dependent corrections for random summing and dead time effects were experimentally determined at different count rates and in several experimental conditions. Two NaI(Tl) crystals of different volumes were used. They were connected to pulse recording systems with different characteristics, with or without pile-up rejection live-time corrector circuits (PUR/LTC). A theoretical model was proposed to explain the different behaviours found experimentally. It is shown that count rate-dependent effects are drastically reduced by the use of PUR/LTC.