Model tests at ballast and design draught are used toconvert the sea trial results from the ballast trial draughtto the contractual design draught. Correlation allowancesin model test results and their effect on the trialperformance prediction are of major importance. Alarge spreading in the values of the correlation allowancesfor design draughts have been found for merchantvessels tested at different model test institutes, but atballast trial draught the spreading is much less. Can ithappen that some institutes on purpose select favourablecorrelations allowances? Or should we accept a largespreading in correlation allowances and have theseindeed been confirmed by sea trials at design draught?Some statistical analysis will be presented, as well as adiscussion on the accuracy of model and full scale tests.