This formal method 

 may be more effec- 

 tive in evaluating 

 certain products. 



which lessens their effectiveness as 

 objective taste testers. 



Food technologists have devised 

 methods of overcoming these prej- 

 udices. For example, coding the 

 samples so that the testers cannot 

 identify them; presenting the 

 samples in random order; having 

 the tasters jiresent their results in 

 writing without consultation with 

 other judges; and presenting the 

 samples in containers that are alike 

 and free of taste and odor. 



Data Must Be Significant 



After obtaining the data from 

 the various tests, researchers must 

 analyze them to learn whether or 

 not the taste differences recorded 

 are significant. For example, sup- 

 jjose we hatl presented a series of 

 judges with two samples of tomato 

 juice, each with a different amoimt 

 of salt added; the samples are coded 

 and the judges are asked to select 

 the samjjlc they prefer. Suj^pose 

 there are ten jutlges and each judge 

 tastes each sample foiu' times. \V^e 

 now have 40 judgments. Suppose 

 12 of these indicate a preference 

 for sample A, and 28 for sample B. 

 Our statistical analysis indicates 

 that such a distribution of prefer- 

 ence woidd occin* by chance alone 

 about once in 20 times. This 

 probability is generally taken as 

 the dividing line, and we can say 



that there is a statistically signifi- 

 cant jjreference for sample B. 



Results Not Always Useful 



Now suppose that our results 

 showed 16 for samjjle A and 24 for 

 sample B. This distribution would 

 occur by chance alone about one 

 out of four times. Since such a 

 result coidd occur so frequently by 

 chance alone, in this case, it woidd 

 be stated that no statistically signifi- 

 cant preference was observed. 



Panels Differ 



There are two types of panels to 

 measure the two taste-testing reac- 

 tions. To determine the difference 

 among samples, an "expert" panel 

 is used. This group is usually 

 made up of less than ten people, 

 and is concerned with determining 

 small differences in flavor. 



To determine preference, a much 

 larger "consiuner" panel is used. 

 As in choosing a grouj) for an elec- 

 tion straw vote, the testers must be 

 sure that the people on this panel 

 are truly representative of such 

 factors as family size, economic 

 backgroimd, and occupation. 



Before undertaking such a proj- 

 ect, previous study with the smaller 

 panel is carried out to determine 

 whether differences exist. If such 

 differences cannot be detected by 

 the smaller panel, preference tests 

 aie of little value. 



