The paper demonstrates a method for identifying and measuring the preferences and prejudices of wine critics. Employing expert rating scores from four wine critics assessing Australian premium wines, the analysis further illustrates the significant variability which exists among expert quality scores. A mean pairwise correlation of 0.435, an intraclass correlation of 0.426 and an alpha reliability of 0.748 are identified for 258 commonly assessed wines. These measures of rater consensus are lower than those identified from other wine studies and in other disciplines. Regression models are developed to explain the difference between each specific expert score and the average score of the other experts. The models explain the specific preferences of wine critics employing information which is readily available to consumers. For different experts, the cellaring potential of a wine, its vintage, the source region of the grapes and the variety/style of the wine are found to explain wine preferences to different degrees. The presented information may usefully be employed by consumers when confronted with conflicting wine scores. To aid with purchasing wine products consumers may wish to align their preferences with the identified preferences of specific wine critics.