This part of the document describes in more detail the choice of species and the way in which the exercise was carried out.
Choice of species
The original aim was to consider all species whose most recent world catch exceeded 50,000 metric tons (MT) per year together with a few specially important species of a lower catch level. By omitting species caught mainly for fishmeal (Gulf and Atlantic menhaden, sandeels, Norway pout, for example), and many of the groups of species designated in FAO catch statistics as 'X nei (‘nei’ means ‘not elsewhere included’) it was hoped that about 120 species or groups of species could be covered. It was thought possible that no data would be available on a number of species. In practice, it has been found that, with relatively few exceptions, sufficient data emerged from a literature search to enable most species and groups to be dealt with, with the obvious and necessary exception of broad categories such as ‘marine fish nei’ The collection and assessment of literature data began with the most heavily caught species and proceeded down the catch tonnage list. The assessment of the data and selection of preferred yield and composition values has been completed for about 130 species or groups down to a catch level of 100,000 MT per year, though the initial data collection has covered most species down to 50,000 MT per year.
It proved possible, and in some cases necessary, to look at related species together (hakes are a good example); these grouped species sometimes included species caught at less than 100,000 MT per year.
Collection of the data
As far as possible only original source data have been collected. In a very few cases, where the original literature was inaccessible for some reason, secondary sources such as abstracts and compilations have been admitted. References to the original literature were collected from various sources:
a full computer-based search of Food Science and Technology Abstracts from 1969
a card index and a computer-based file held in Torry Research Station
a library search of books, including previous compilations of data
personal files
cross references from known sources
random discovery
In reference to the last source, it was somewhat disconcerting, when looking up a known paper in a journal, to come across hitherto unknown papers of some value. This not uncommon experience does raise concern about the efficiency of abstracting services and of the means of searching abstracts.
No claim is made for comprehensiveness in the literature search carried out: additional sources continued to become known right up to the end of this exercise. Nevertheless, the amount of data available overall was higher than expected, but not consistently so.
Some species with a high catch level, Black sea sprat for example, turned up almost no data, while species or groups much further down the catch tonnage table proved to have extensively studied. Thus Dentex, seabreams, etc, n.e.i. (Sparidae) provided more data than could conveniently be assimilated, partly perhaps because it is a large family with many important genera, and partly perhaps because many species are highly esteemed edible fish and thus attract the attention of laboratory workers.
The failure of this survey and others to search the literature comprehensively is confirmed by the lack of really good agreement between selected values in this report and those in other compilations. It is recognised that Russian and Japanese literature is inadequately represented in this exercise and this almost certainly applies also to literature in other languages.
No attempt has made to standardize the format of the bibliographic references cited throughout this document. Rather, they are cited in accordance with the style used by the systems of service through which references to the papers found through random discovery, that is to papers of interest not identified through any system or service, but only through examination of the same source documents. Despite the efforts made towards standardization and comprehensive coverage in these systems, it is obvious that individually they fall far short of the needs for study of this kind.
Quality of collected data
Much of the data collected is of rather low quality, not because the actual measurements of yield and analytical results are themselves inaccurate, though a few clearly are, but because the authors have not given sufficient information to enable their results to be interpreted, or because they have not appreciated the possible variation in the quantities they have measured.
As far as adequacy of information is concerned, yield data in particular are universally ill-defined. The term ‘fillet’ is used without making it clear whether the skin is still present and how far the fillet has been trimmed to remove belly-flaps and small bones such pin-bones. The term ‘edible portion’ and similar terms are widely used but are ill defined or, more commonly, not defined. Even within a single country, where the handling of fish should be consistent, marked differences are apparent: for example, in two extensive surveys of philippine fish carried out around differ markedly, measurements of the edible portion of (for a species not included in the present survey).
As far as composition is concerned, it is not always clear whether the material analysed is muscle tissue only, or includes skin or even bone. A judgement has often to be made on whether to accept the data or not.
Many authors, as far back as last century, have expressed their awareness of the possible seasonal variation of composition (and, less often expressed, of yield), but few have attempted to fully investigate the problem, probably because to do so would usually be impracticably expensive in cost and time. surveys of variation in composition throughout the year, or a major part of the year, are not rare. However most such surveys ignore the possibility of variation with location of catching, with the size of the fish, with the sex and with post-catching treatment. An ideal survey would analyse several fish of at least two sizes in the commercial range, of both sexes, at weekly intervals throughout the year or the period of availability, from all exploited grounds.
The yield, by a fully-described procedure, and the composition of the flesh would be determined as soon as possible after catching (though preferably after resolution of rigor mortis). The final calculated mean values of yield and composition should be weighted to correspond with the commercial catch distribution in time and place. The possibility of year to year variation should be investigated. Such and ideal survey has not, and probably never will be carried out.
Much of the available data comes from small numbers of fish: analyses of a single fish are not unknown even in recent work. It is often the case that the number of fish examined is not stated. Where the value in the literature is known to be a mean of a batch or of a number of batches, this information is usually given in the monographs in Part 3 of this report.
Where a species is caught over a wide area, the place of capture is often included in the details in the monographs. No actual evidence of a geographical effect on yield or composition has been discerned, except in the single case of Atlantic herring. The apparent absence of such and effect, generally, may be because of inadequacies in the data.
Use of collected data
a) listing of data
All the relevant, or apparently relevant, data collected have been listed in the monographs. Relevant data are those that relate to the particular species or group being studied, or to a closely related species or group. The extent to which related species can be called in support is discussed said that in some cases there was little choice in the matter, as the amount of data on the studied item was inadequate.
The collected data are listed in the monographs, separately for yield and composition: within each monograph the order of species usually follows that used in the ‘Systematic list of aquatic organisms’ printed in the FAO Yearbook of Fishery Statistics- Catches and Landings. In many cases the scientific names, or common names, of the species given in the source literature are not the current ones. Where clear identity of scientific names can be established, this is indicated in the monographs. In general, when a scientific name is not given in the literature, the data have not been listed: in a few cases accurate identification of common names was possible and the data were accepted. Within each species or genus, the data are not in any significant order. Similarly, the source references are listed in the order in which they were used (with a few exceptions) and so the order is of no significance; it is neither alphabetical nor chronological. Three or fewer authors are named in full: more than three are named as ‘First author et al.’.
The list of sources is separate for each monograph. Many references occur in two or more monographs: in the present form each monograph is complete in itself.
b) means and ranges in source data
The data collected can take several forms:
a single figure, which may relate to a single fish, or may be the mean of a known or an unknown number of fish;
separate figures, for individual fish or different batches;
the range of values (i.e. maximum and minimum only) for a number of fish or a number of batches;
the range, together with the mean values, again of a number of fish or a number of batches;
the mean and the standard deviation of the values for a number of fish.
As far as it can be discerned from the description of the work given by the original source, the data have been listed in the monographs in one of the above forms. Often it is not possible to know whether the figure quoted is for one fish or more. Standard deviations are very rarely available.
The data are normally listed with the number of significant figures given in the original source: in one or two cases unrealistic figures have been rounded to two decimal places.
c) assessment of data
As far as possible the selected, preferred, values of the yield and composition are based on comprehensive sets of data. Surveys covering all or a major part of the year, especially if they cover more than one fishing ground, are very desirable. Two concordant surveys are a rare luxury. Where surveys are available, additional data on smaller numbers of samples are generally discounted unless they appear to contribute additional useful information. Where surveys are not available, a judgement has always has to be made in deciding which results are to be incorporated in the calculation of the selected, preferred, value. In cases of doubt, results have been included rather than excluded.
d) calculation of selected values
Where two or more good surveys have been found, the selected values are the means of the these results. Where only one survey is available, its results are often taken as the selected values, provided they are consistent with any remaining, more fragmentary, data. If surveys are felt to be not properly representative, means of the survey and other data are calculated.
In calculating overall mean values, the separate collected mean values or single values are used. When only a range is given in the original, the mean of the range is taken into any overall mean.
Means are always simple arithmetic means. It could be argued that the mean ought to be weighted in some way towards the more comprehensive or more reliable results. Unfortunately the basis for such weighting does not exist.
For instance, suppose that two sets of results are available, the mean of 100 fish caught in December and the means of 10 fish caught equally in December and June. One could weight the result 10 to 1 in favour of the first set, because 10 times as many samples were studied. Alternatively, they could be weighted 2 to 1 in favour of the second set because fish were caught on two occasions instead of one. The latter is probably preferable in this particular example, but if only 2 fish had been examined in the second set of results, one in December and one in June, weighting in favour of the 100 result mean would appear more sensible. In any actual situation no weighting procedure can properly be justified, apart from the weighting implicit in the preference given to major surveys.
e) energy values
The energy values have been calculated using the following factors:
| Protein | 4.27 kcal/g |
| Fat | 9.02 kcal/g |
| Glycogen | 4.11 kcal/g |
Conversion to kilojoules was by multiplying by 4.184.
There is no universal agreement on the factors to be used, despite long recognition of the desirability of having agreed factors. The factors above are USA values (ref. 1) for meat and fish, and are used in other recent FAO studies of food composition. In some countries average values for all foods are preferred. In the UK (ref. 2) the values used are:
| Protein | 4 kcal/g |
| Fat | 9 kcal/g |
| Carbohydrate (as monosaccharide) | 3.75 kcal/g |
For a few random species of fish, these factors give energy values a few percent lower than those in Table 1, but the differences are unlikely to be important in practice.
Are related species similar?
In some cases, sufficient data are available on a particular species to enable mean values to be selected. In many cases, however, additional data must be found. There is a presumption that species that are taxonomically related will be biologically more similar to each other than to other more distant species and so should be similar in yield and composition. It has long been clear that species in the family Gadidae are very similar in composition, and, in particular, have a very low fat content. Clupeidae are known to have moderate to high, but seasonally varying, fat contents. It is not known, however, to what extent taxonomic propinquity can always predict compositional or yield similarity.
In the present exercise additional data, when required, has been sought initially in the same genus as the species under study, and, with less confidence, in other genera of the same family. When two or more species of the same genus have come within the scope of the present exercise they have generally been considered together. Thus all Merluccius species are dealt with in the same monograph and no significant differences between species were found. Similar results with Sardinops, Sardinella, Engraulis, Oncorhynchus and other genera give some confidence in the belief that closely related species will be similar in yield and composition, though there are exceptions. When one considers the similarity between genera, the situation becomes more difficult. In cases where fish of a different genus have had to be included in the data collected for a particular species or group, in the absence of adequate data within its own genus, it has usually been found that the variability of the results masks any differences between genera. When related genera have been studied separately, in different monographs, differences between genera may become apparent, as the amount of data is greater. All Gadidae studied did prove to be similar, but this was not so far Clupeidae, Scombridae or Carangidae, three extensively studied families. For example, in Clupeidae, the selected fat contents of Clupea harengus and C.pallasi are 13.8% and 11.6%, while the value selected for Sardinella species is 4.5%. Among Scombridae, the fat content varies from 3.2% for Rastrelliger species to 11.4% for Scomber scombrus, while protein is 18.7% in Scomber scombrus and 24.2% in Katsuwonis pelamis.
Less marked differences are seen in Carangidae. Some of the variation may arise from the incomplete nature of the data, but it seems likely that different genera within a family may often, but not always, be somewhat different in composition. Because of the greater variability of the data, conclusions about yields are less certain.
The conclusions, tentative at the moment, is that while species in the same genus are likely to be quite similar, species or groups in different genera may differ significantly. Care must therefore be exercised when trying to apply data for one species to any other than a closely-related species.
Appropriate yield figures
Two yield figures have been derived for most finfish, of skinless fillets and of total edible flesh. In many developed countries fish is sold to the consumer mainly in the from of fillets, often skinless, usually closely traimmed, or products made from blocks of skinless fillets. The literature, it has already been said, seldom given an indication of how the fillet has been prepared and trimmed. Small differences in filleting traditions in different countries will mean that an average fillet yield based on results for several countries will hide such differences. In theory one might try to chose a fillet yield appropriate to a particular country from results obtained in that country, but in practice the variability and/or inadequacy of the data would make such an attempt difficult to justify.
Fillet yields are sometimes known, and sometimes suspected, to include skin. Where necessary the presence of skin has been allowed for by subtracting 5 % of the fillet yield. This is, of course, a rough approximation since the thickness of fish skin can vary greatly between species. It was not felt to be realistic to try to estimate the proportion of skin in individual species: some data are available, especially in Russian literature, but only for a small number of species.
Some fish skin is eaten and will make some nutritional contribution. The high connective tissue content of skin means that its quality as protein is low.
This may be a partial justification for excluding it from fillet (and flesh) yield. In any case, no significant information appears to be available on the extent to which fish skin is eaten in different countries.
In many countries, especially developing countries, it is much more usual for fish to be supplied to the consumer either whole, usually gutted and often headed, or in a dressed or split form, perhaps smoked or dried. In these total edible flesh values selected for each species are intended to give a measure of such consumption. Again, as for the fillet yield, the variability of the variable data is high. Part of the variability will result from the different methods used to recover the flesh; the exact procedure is seldom indicated in the original sources. In a very few cases the flesh yield is from trials with deboning machines. It is claimed that these machines can recover up to 50% more edible flesh than by filleting, but not all the data support this. It is not impossible to find a quoted yield by deboning that is lower than the fillet yield. One possible explanation of this anomaly is that the machines have been used on a batch basis in the laboratory rather than on a continuous commercial basis.
The ratio of the selected values of total flesh yield to those for fillet yield is often surprisingly low. One must suspect that some of the total flesh yields in the literature are actually fillet yields (possibly including belly flaps), but unless this is certain the data have been accepted at face value.
Final Considerations
This exercise has not attempted to address questions such as loss of nutritive value after catching (much of the data refers to recently caught fish) by leaching and spoilage, losses by infestation and waste at different stages. It might be suggested that a detailed questionnaire to appropriate authorities in each country might supply some data on these questions and also, perhaps, on the forms (fillet, whole etc) consumed and to what extent skin is eaten.
Implicit in many comments in this part of the document is the hope that future studies of yield and composition will be more extensive and planned to answer the need for the sort of data that has proved difficult to obtain and that the description of the work and the results are more detailed.
References
Merrill, A.L. and Watt, B.K. Energy value of foods - basis and derivation. Agriculture Handbook No. 74 (US Dept of Agric. 1973)
Paul, A.A. and Southgate, D.A.T. McCance and Widdowson's The Composition of Foods. 4th edition (HMSO, London, 1978)
