peer reviewed ; A particularity of the Luxembourg milk recording is the use of different schemes. Two principal schemes are applied: the scheme “S” applied on 69.1% of the total herds (n=712) and consisting in one proportionate sample of all daily milkings, and the scheme “T” (21.6% of the total herds) which consists in one sample of only one milking (morning or evening milking) (and alternating milking time from month to month). The problematic is that application of different schemes could influence the milk components (protein and fat yield) and the milk fat components (saturated and unsaturated groups of fatty acids) genetic parameters estimation and to prevent all comparisons between dairy population under different milk recording schemes. A total of 47,613 and 44,833 test-day records were obtained, respectively for schemes “S” and “T” from Holstein cows in first lactation in Luxembourg dairy herds. The used model included as fixed effects: herd x date of test, class of age, and month x year. Random effects were permanent environmental, additive genetics, and residual effects. The main objective of this work is to study the effect the choice of milk recording schemes (“S” or “T” schemes) on milk yield and milk components genetic parameters. A solution could be to add a fixed effect taking in account the milking time. The second objective is to study the effect of milking time (morning or evening) on genetic parameters estimated in the case of scheme “T”. According to the results, genetic parameters were statistically different between the schemes “S” and “T” for milk yield (P value < 0.0001). Further, the classifications of bulls according to their breeding values were very different when values were estimated on basis of scheme “S” or “T” (Spearman correlation value of 0.51 for milk yield for example). In conclusion, using several milk recording schemes do not allow any comparison of genetic parameters between dairy cattle’s.