The aim of this study was to examine different methods for determining growing degree-day (GDD) threshold temperatures for two phenological stages (full bloom and harvest) and select the optimal thresholds for a greater number of apricot (Prunus armeniaca L.) cultivars grown in the Belgrade region. A 10-year data series were used to conduct the study. Several commonly used methods to determine the threshold temperatures from field observation were evaluated: (1) the least standard deviation in GDD; (2) the least standard deviation in days; (3) the least coefficient of variation in GDD; (4) regression coefficient; (5) the least standard deviation in days with a mean temperature above the threshold; (6) the least coefficient of variation in days with a mean temperature above the threshold; and (7) the smallest root mean square error between the observed and predicted number of days. In addition, two methods for calculating daily GDD, and two methods for calculating daily mean air temperatures were tested to emphasize the differences that can arise by different interpretations of basic GDD equation. The best agreement with observations was attained by method (7). The lower threshold temperature obtained by this method differed among cultivars from -5.6 to -1.7 degrees C for full bloom, and from -0.5 to 6.6 degrees C for harvest. However, the "Null" method (lower threshold set to 0 degrees C) and "Fixed Value" method (lower threshold set to -2 degrees C for full bloom and to 3 degrees C for harvest) gave very good results. The limitations of the widely used method (1) and methods (5) and (6), which generally performed worst, are discussed in the paper.