A practical training program for fluid milk defect judging should focus on initial training of panelists

The sensory quality of fluid milk is of great importance to processors and consumers. Defects in the expected odor, flavor, or body of the product can affect consumer attitudes toward the product and, ultimately, willingness to purchase the product. Although many methods of sensory evaluation have been developed, defect judging is one particular method that has been used for decades in the dairy industry for evaluating fluid milk. Defect judging is a technique whereby panelists are trained to recognize and rate a standard set of fluid milk defects that originate from various sources (e.g., microbial spoilage). This technique is primarily used in processing facilities where identification of sensory defects can alert personnel to potential quality control issues in raw material quality, processing, or good manufacturing practices. In 2014-2016, a preliminary study of defective milk judging screening and training was conducted by the Milk Quality Improvement Program at Cornell University (Ithaca, NY). The study, which included 37 staff and students from the Cornell community, used prescreenings for common odors and basic tastes, followed by uniform training to select, initially train, and retrain defect judges of unflavored high temperature, short time fluid milk. Significant improvements were seen in correct identification of defect attributes following initial training for all defect attributes, with the exception of fruity/fermented. However, following retraining, significant improvements were observed in only 2 defect attributes: cooked and milk carton. These results demonstrate that initial training is important for panelists to correctly identify fluid milk defect attributes, but that subsequent retraining should be tailored toward specific attributes. This study provides a resource for dairy industry stakeholders to use to develop relevant and efficient training methods for fluid milk defect judging panels.