preferred in 5 instances. On two occasions the judges rated the con- 

 trol and the alkali process sample equally superior to that contain- 

 ing the boiling process oil and on one occasion the control and the 

 sample vrlth boiling process oil added were preferred equally over the 

 sample to which the alkali process oil had been added. 



One judge gave samples containing the boiling process oil the low- 

 est average flavor rating and a second judge rated this pack lowest for 

 four out of the five species over the entire period of the examinations. 

 This was one of the few consistent trends in the observations of the 

 tasting panel. Other judges thought they could distinguish particular 

 samples on the basis of oil preparation method, but their average fla- 

 vor ratings did not consistently agree with their preconceived opinions. 



In order to check on the number of points required to indicate 

 significant differences in flavor scores for the several individuals, 

 a comparison test was conducted, A single can of red salmon with alka- 

 li process oil added was opened and the entire contents were blended 

 thoroughly, V/hen the blended material was as nearly homogeneous in ap- 

 pearance and composition as possible, it was divided into three equal 

 portions. The judging panel manbers were asked to rate the samples, 

 marked A, B and C, just as in all other instances. They were told only 

 that the samples were red salmon. Of the nine judges, three found no 

 difference, two had a flavor rating spread of one point between the 

 replicate samples, two judges had a two point spread, and two fovind as 

 much as three points difference. 



If the flavor ratings for the eight monthly examinations are eval- 

 uated in the light of this diversity of organoleptic sensitivity in the 

 judging panel, it becomes evident that a difference in average prefer- 

 ence score of approximately 1,3 points should be found between samples 

 in any given examination to indicate perceptible flavor differences. 



In organoleptic tests it is difficiilt to judge flavor without a 

 conscious or unconscious bias from texture and color when flavor is the 

 variable being evaluated. The texture of the salmon was materially dif- 

 ferent in the lots to which the oil was added. From comments of the 

 tasting panel, it was evident that these differences in fat content were 

 apparent to than. Each judge was asked whether he preferred lean, medium 

 fat, or fat fish. The stated preference was then used in studying the 

 flavor ratings for the fish samples. Two of the judges who had a decided 

 preference for lean fish chose the control samples as best for 60 percent 

 of the samples tried. On the other hand, the persons signifying a 

 preference for medium fat and fat fish rated the control samples best in 

 approximately 60 percent of the tests. This would seem to mini-nize any 

 bias in flavor scores attributed solely to differences in fat content. 



% 



