FISHERY BULLETIN: VOL. 71, NO. 3 



species of fish, including results obtained by 

 the NMFS, and those made available by the 

 FDA and the Canadian Government Inspection 

 Service. The data were all obtained using 

 variations of the following procedures: Arm- 

 strong and Uthe (1971); Uthe (1971) and 

 Uthe, Armstrong, and Stainton (1970). Diffi- 

 culty was experienced in cross referencing in 

 some cases since the Market Facts survey used 

 common names for fish which did not always 

 represent unique species. For example, the 

 survey lists mackerel. This could include 

 Spanish mackerel, Scomberomor'us maculatus; 

 king mackerel, Scomberomorus cavalla; or 

 Atlantic mackerel. Scomber scombrus. It would 

 not refer to jack mackerel, Trachurus sym- 

 metncus, which would appear under canned 

 mackerel. In this case, average mercury levels 

 were calculated for mackerel by weighting levels 

 for each of the three species referred to above 

 according to the 1969 U.S. landings (Riley, 

 1970). 



In the case of mixed categories, levels were 

 calculated for each, making assumptions on 

 their frequency of use. Thus in the case of 

 fish sticks, it was assumed that these con- 

 sisted of 40% cod, 40% haddock, and 20% sole. 

 The risk of introducing serious errors by this 

 somewhat arbitrary procedure was not con- 

 sidered high since all the species likely to be 

 involved in these grouped categories have rela- 

 tively low mercury levels (0.1 ppm) and do not 

 differ greatly from each other. The case of 

 light meat tuna presents a particular problem 

 since the consumption is relatively high and 

 the levels both within and between species 

 vary considerably. 



In other cases also it is difficult to be sure 

 of the extent to which the analytical data for 

 particular kinds of fish is representative of 

 mercury levels in the corresponding commercial 

 supply since methylmercury levels in fish may 

 vary substantially with biological and environ- 

 mental factors and since there are practical 

 limits to the number and range of samples 

 which may be obtained and analyzed and to 

 their application to the market supply. More- 

 over, sampling for regulatory purposes may 

 be concentrated on those segments of the catch 

 whose size and catch location may give 



reason to believe they are most likely to exceed 

 the guideline, e.g., large fish, so that the results 

 obtained in the course of regulatory activities 

 may be unrepresentative of the whole catch. 

 For this study the sources of all available data 

 for each species were reviewed and, to the 

 extent possible, data believed to be unrepresenta- 

 tive were eliminated. The mercury data used 

 also reflect that in the fishery resource, not 

 that in the market place, since FDA controls 

 have eliminated much, if not all, of the high 

 level fish. The figures used cannot therefore 

 be regarded as indicators of levels found in 

 the present supply, but rather that which 

 would exist under various guideline conditions. 



To assess the effects of possible errors in 

 the mercury levels used, the sensitivity of the 

 estimates produced by the program to varia- 

 tions in input levels of each species in the 

 program was tested. Preliminary runs were 

 made in which the input mercury levels were 

 increased for each kind of fish in turn. In the 

 case of kinds of fish having low levels initially, 

 the level was increased by 100% . Where the 

 initial levels were over 0.20 ppm, the levels 

 were increased by 25-70% with actual increases 

 no less than 0.10 ppm. Sensitivity was judged 

 by differences from the standard run in the 

 level of intake not exceeded by 99% of the 

 sample (99% level), by changes in the intakes 

 of the highest family. The preliminary 

 runs showed that increases in 40 of the kinds 

 of fish in the survey had no measurable effect 

 on the higher intake levels. This means that 

 relatively large errors in the mercury input 

 data for these kinds of fish data would not be 

 critical to the reliability of the end result. Of 

 the remainder, only variations in the input 

 data of white meat tuna, light meat tuna, and 

 swordfish showed marked effects on the end 

 results. These were subjected to further sensi- 

 tivity testing using revised data as were also 

 mackerel, red snapper and the "all other fish" 

 category since mercury data available for these 

 was limited. The results of these tests, shown 

 in Table 2, confirms that only the input levels 

 of light meat tuna, white meat tuna, and 

 swordfish markedly affect the end results. 



Finally, in order to estimate the effects of 

 guidelines on the mercury consumption pattern, 



620 



