Abstract: Thirteen mathematical functions were compared to describe the lactation curve, using data taken daily, semimonthly, monthly and bimonthly intervals. The objectives of this study were to compare the most widely used functions for their ability to describe first lactation curves of Holstein cows in a herd located in the southeast region of Turkey and to examine their sensitivity to data reduction. Cows reared outside throughout all seasons and milked 3 times daily. The number of observations per lactation varied from 6-300. Criterias used to compare the models were squared multiple correlation coefficient (R²), correlation coefficient (r) between the observed and estimated test day yields, Durbin-Watson (DW) statistic and mean squared prediction error (MSPE). The performance of models showed great differences in the comparison criteria across calving-seasons (S1-December to February; S2-March to May; S3-June to August; S4-September to November) and sampling groups (TD-0 included all daily records from 5-305th day after calving; TD-15 included records with 15 day interval from 5-305th day after calving; TD-30 included records with 30 day interval from 5-305th day after calving; and TD-60 included records with 60 day interval from 5-305th day after calving). The most appropriate model to describe lactation curve was Glasbey (GLAS) when all daily observations in lactation are used and regardless of that the observations are taken semimonthly, monthly, or bimonthly interval.