shellfish, Tilefish, ocean perch, "little tuna," and California sardine were 

 subjects of less extensive special studies. These 6 species account for 93 

 of the l6l dishes included. Fresh-water fish were represented by only 1 

 species. Thus, the number of analyses reported for the various species has 

 no relation to the commercial importance of the species. Actually, selection 

 of species to be included makes very little difference in maqy of the recipes, 

 especially for a lean species or for a dish in which the fish is a relatively 

 minor ingredient in the recipe. In such cases very nearly the same data on 

 composition are obtained even though a number of different species of fish 

 are used. 



Analyses were limited to the determination of the usual proximate com- 

 ponents: moisture, protein, fat, ash, and carbohydrate. The last was deter- 

 mined by difference. Inclusion of analyses for vitamins and mineral elements 

 would have required limitation of the samples studied to a very small fraction 

 of the samples included in the original series, and it was felt that greater 

 value would be derived from proximate composition of a much larger and more 

 representative group of samples, 



METHODS OF SAMPLING AND ANALYSIS 



Sampling — The method of sampling was influenced by the source of samples. 

 As test-kitchen dishes, their primary purpose was for use in palatability tests 

 by taste pajiels. They were judged for appearance, flavor, and texture. Ideal 

 sampling for chemical tests would require grinding and mixing the entire dish 

 before removal of sample aliquots. In the present investigation, the dish as 

 served to the taste panel was necessarily also the source of samples for cinal- 

 ysis. Consequently, from 10 to 20 percent or usually about 15 percent of the 

 dish was available for testing. Roughly one serving portion was used in most 

 instances. For such dishes as loaves, cakes, fritters, and most casseiroles 

 and salads, the ingredients were reasonably well mixed and this condition did 

 not impose undue sampling error. For such dishes as soups, chowders, stews, 

 stuffed fish, and curries or Creoles, vrtiere a fish mixture is served in a rice 

 idng, removal of a representative sample was difficult, and a greater range in 

 analysis of duplicate samples was observed for dishes of this type. 



The variability introduced in this way was smaller than would be expected, 

 judged by the few dishes of this type for which 4 to 6 samples were available. 

 Values that appeared to have appreciable sampling bias have been omitted frcan 

 the averages. All values recorded in the tables have been rounded to the 

 nearest one- tenth of a gram. 



Grinding — A small hand or power grinder was used for most samples, and the 

 ground material was passed through 2 or more times to improve mixing. The 

 consistency of some samples such as chowders, bisques, and stews permitted the 

 use of a blender for disintegration. 



Moisture — From 5 to 15 grams of the ground and mixed samples were weighed 

 in aluminum drying cans and dried overnight, about 20 hours, in an air oven 

 at 105° C, The loss in weight was reported as moisture. Triplicate analyses 

 normally checked within 0,5 percent, 



-3- 



