186 



FISHERY BULLETIN OF THE FISH AND WILDLIFE SERVICE 



origin was known, only the catches from vessels 

 that fished a single area were sampled. A sample 

 of about 100 fish was selected for measuring in as 

 nearly random a manner as working conditions 

 would permit. The standard practice at the port 

 of New Bedford was to pack the fish in 125-pound 

 boxes as they were being unloaded. The boxes 

 were accessible to the measurers before being iced 

 and closed, and it was convenient to measure 20 or 

 25 fish from each box. A sample of 100 fish was 

 obtained from 4 or 5 boxes taken one at a time 

 from the unloading line as needed. Usually from 

 1,000 to 2,000 pounds of fish were unloaded be- 

 tween the boxes sampled. The fish were taken 

 from one end of each box from the top to the bot- 

 tom, with a special effort to avoid any selection. 

 In view of the difficulties of obtaining an accur- 

 ately representative sample (Hayne 1951), a 

 slight bias may have favored the large fish; but 

 the same technique was followed throughout the 

 investigations, and the bias, if any, should not 

 affect the interpretation of trends in fish lengths. 



Measurements were of the total length of the 

 fish — from the tip of the lower jaw (with the 

 mouth closed) to the end of the caudal fin. 

 Almost all measurements were recorded on a 

 measuring board slotted to receive an aluminum 

 strip. The measurement was taken by pricking 

 a hole in the strip, which was marked off in two 

 parts to keep separate the records of the lengths 

 of males and females. This method of measuring 

 is very satisfactory, provides a rapid field method 

 suitable for use when fingers are too wet or too 

 cold to write, and is free of "digit bias," which 

 has troubled other investigators who have meas- 

 ured large numbers of fish (Sette 1941). Later 

 in the laboratory, the lengths were tallied to the 

 half centimeter by superimposing a graduated 

 celluloid strip over the marked aluminum strip. 



It became apparent quite early in the study that 

 the sexes differed in size composition, and com- 

 mencing in October 1942 most measurements were 

 kept separate by sex although the total sample 

 was obtained in as random a manner as possible 

 so that the number of each sex measured woidd 

 be representative of its numbers in the landings. 

 After opening a few fish to determine the condi- 

 tion of the sex organs, it was discovered that the 

 yellowtail could lie sexed easily and accurately by 

 holding the white side to the light and looking 



through the fish. In this way, the outline of the 

 ovary extending posteriorly from the mass of 

 viscera can readily be seen even in immature 

 females. 



The program of sampling was planned to ob- 

 tain a sample from nearly every vessel landing 

 that had fished but a single area. It was expected 

 that this would supply representative samples of 

 the entire landings, but pressure of other duties 

 and changes in field personnel made it impossible 

 to maintain the program at the same level at all 

 times. Some gaps also occurred because of the 

 fishermen's habit of working in two or more areas 

 when fish are scarce. This was particularly true 

 in the yellowtail fishery, and many months when 

 the landings from an area were low it was not 

 possible to obtain a sample because the few land- 

 ings made were always mixtures of fish from sev- 

 eral areas. This tendency led to some under- 

 sampling of areas poorly represented in the land- 

 ings at New Bedford. Furthermore, the catches 

 from the Block Island and Long Island areas, 

 which are fished mostly by Bhode Island and New 

 York fishermen, were landed to a large extent in 

 ports not covered by our sampling. 



To obtain the best representation of the length 

 composition of the yellowtail for the period Octo- 

 ber 1942 through December 1947, it would be best 

 to weight the length frequencies by the quantities 

 landed. This would be difficult, however, because 

 of lack of data in numerous quarters and from 

 some statistical areas (table 8). Therefore, we 

 have tested the representativeness of our un- 

 weighted data in two ways: First, by comparing 

 the distribution of measurements with the distri- 

 bution of catch according to area and time, and 

 second, by computing the effect of the maldistribu- 

 tion of the catch on the average length. 



A comparison of the distribution of measure- 

 ments with the distribution of catch shows that 

 the discrepancies were not serious. When con- 

 sidered according to area, it is apparent that the 

 areas off No Mans Land and Nantucket Shoals, 

 which provided the bulk of the catch, were some- 

 what oversampled and the areas off Block Island 

 and Long Island were somewhat undersampled 

 (table 9). The distribution according to years 

 showed similar discrepancies, with 1943, 1944, and 

 1947 being undersampled and 1945 and 1946 being 

 oversampled. However, the distribution accord- 



