Applicability was evaluated of second-, third- and fourth-order Legendre polynomials (LPs) as lactation curve models for prediction the 305-day lactation yields. First lactation milk, fat and protein yields were calculated using the standard lactation curve method (LCM) and data collected during lactations truncated at 60, 100, 200 and 305 days-in-milk. All results were compared with the official lactation yields of cows. Data consisted of 5,289,576 test-day yields from 668,964 lactations of Black-and-White heifers. Standard lactation curves were modelled by LPs of the second (L2), third (L3) and fourth (L4) orders within 64 classes of genetic group by age at calving by season of calving.LPs of the highest order (L4) were the best-fitted lactation curve models, followed by L2 and L3 polynomials, showing that LPs of even orders are more suitable for fitting lactation curves. It is concluded that 305-day yields of heifers can be predicted with sufficient accuracy when the lactation curve parametres are derived using tests from the first 200 days-in-milk.