Modern specimens (N = 165) of albacore, Thunnus alalunga, were collected and 18 measurements made on cranial bones and an element of the caudal peduncle. Allometric equations were calculated to facilitate the reconstruction of live fork length from archaeological bones. Modern fisheries trawl data of albacore, bigeye tuna (Thunnus obesus), yellowfin tuna (Thunnus albacares), and skipjack tuna (Katsuwonus pelamis) provided data for equations linking live fork length to body weight for each species. Examination of power curve fits between bone dimension and fork length for 12 unrelated species showed that most conform to a single growth curve, with specialised small-mouthed browsing species as outliers. The curves of three species within one family were effectively co-linear. Since the cranial anatomy of different species of scombrids is notoriously difficult to identify to species, it was hypothesised that the allometric curve for albacore might therefore be suitable for estimating fork length from archaeological bones of the three other species. Scombrid bones from four archaeological sites (Te Ana Pua, Fa’ahia, Hane and Motupore) were measured and live fork lengths estimated. Size-frequency histograms showed ancient catches of the first three were dominated by skipjack, with very few yellowfin and albacore. By contrast, at Motupore, most fish were a larger species, probably yellowfin. Total live body weight was estimated and found to be similar at each site (265 to 374 kg). When meat weights were calculated and nutrient composition evaluated, the overall MNR was found to be 60:38:2. The slightly elevated value of carbohydrate is due to glycogen reserves in tuna white muscle.

Pacific Archaeology, Archaeozoology, Allometry, Scombridae, Tuna, Palaeoeconomics, Macronutrients

When Charles Nordhoff published the first part of his important paper on the “Off-Shore Fishing of the Society Islands” in 1930 in the Journal of the Polynesian Society, the first fish he chose to highlight was albacore (ALB, Fig. 1), and he devoted a full third of the article to it (Nordhoff 1930). Nordhoff lived in the Society Islands for eight years and such was the importance of offshore fishing to the islanders, he said that if he had lived for twice that time, he would have twice as much to say on the topic. Another indication of the importance of albacore was the many names given to it by Polynesians, and, as Nordhoff points out, these are used to distinguish different sizes, appearance at different seasons, the variable strength of different individuals when caught, especially fierce hungry individuals that seize baited lines, and wary ones that refuse to take bait. Having such a large number of words referring to some single feature of the environment is usually taken to indicate the special importance of it to the group of people whose language it is. An oft-quoted example of this, is that among the Eskimo-Aleut languages there are a large number of words that are used to refer to snow (Krupnik and Müller-Wille 2010). Similarly, the Sami people have 175–180 words related to snow and ice (Magga 2006: 34). Incredibly, the Sami have around 1,000 words for reindeer (ibid.: 31). The generic name for albacore in the Society Islands is aahi, which, Nordhoff explains, should be written as ‛a‛ahi to signify glottal stops. The name aahi, and cognates of it (kafi, kahi, ‛ahi, kaakahi, kakahi, kasi, gahi-di-awa, a‛ai, ‛asi‛asi, and kakai, are found in many other Polynesian islands (Biggs’s comparative Polynesian Lexicon Project (POLLEX, Biggs and Clark 1996), though not all refer specifically to albacore. Among the New Zealand indigenous language Māori, no name has been recorded for albacore, and the cognate kaa/kahi refers to some kind of whale or porpoise (ibid.). To this date, not one bone of albacore has been found in New Zealand archaeological sites and, as will be shown below, the bones are not commonly identified in archaeological sites in the tropical Pacific either; this may partly be due to mis-identification.

Figure 1. 

Albacore tuna, Thunnus alalunga. Image source: Cuvier and Valenciennes 1828–1849: volume 8: 123, and plate 215.

The present research was inspired by Nordhoff’s publication about albacore fishing in the Society Islands, starting with measurements of cranial bones of modern specimens of albacore, and the development of allometric equations between bone sizes and live length and weight. The primary purpose of this is to reconstruct the size-frequency of the original catches by pre-European communities. This can be informative on ancient catching methods, prey choice targeting, fishing zones, site occupation seasonality, and other useful information about human behaviour. The estimates of live body weight (BWT) from individual bones are then used to estimate the original meat weight (MTWT), which in turn contributes basic information of macronutrients in ancient human diet and nutrition (Smith 2011).

The taxonomic name for albacore¹ has changed many times over the years and there are at least 22 synonyms outlined by https://indiabiodiversity.org/species/show/233712. In 1831, Cuvier and Valenciennes listed the fish as Thunnus alalunga (Cuvier and Valenciennes 1828–1841, Vol 8: 120), and why this morphed to present-day alalunga is not clear.

Albacore is a member of a family of fish known as Scombridae, with c. 56 species. Historic sources on the taxonomy of this family are Collette (2003), Bannikov (2007), Collette and Chao (1975), Monsch (2012) and Collette and Graves (2019). Unfortunately, considerable difficulty has been encountered distinguishing one species from another from skeletal elements in archaeological sites. For this reason, when calculating minimum numbers of individuals (MNI) from an archaeological site, results are sometimes reported as Thunnidae/Katsuwonidae (McMurtry (1986): Table 1; Davidson et al. (1998): 145), or simply as Scombridae. The development of several methods of both genetic and peptide profiling now permit species to be identified from collagen in ancient fish bone (described below), but the success rate is variable, and the expense means only a few bones can be processed. Although this research is valuable, it is not amenable to mass processing of archaeological bones to generate size-frequency diagrams, which is the essential first step towards the palaeoeconomics of fish in human diet. In the following, I first describe some basic features of the four most common members of the Scombridae family in the tropical Pacific and then give special attention to albacore and develop allometric equations for this species. I then give attention to whether these equations might be applied to scombrid species other than albacore. Finally, I apply the equations to four archaeological sites across the Pacific.

Although this paper is primarily concerned with one species in the Scombrid family, it is necessary to provide a thumbnail sketch of several others as they become relevant later. In the Pacific islands there are four species of tuna that are commercially significant, and most likely to have been the species caught in pre-European times and been present in archaeological sites. These are bigeye tuna (Thunnus obesus), yellowfin tuna (Thunnus albacares), albacore tuna, and skipjack tuna (Katsuwonus pelamis). The New Zealand Ministry of Fisheries refers to these species with acronyms, BET, YFN, ALB, and SKJ respectively (Fig. 2). A fifth species, bluefin tuna (Thunnus orientalis), has been recovered in at least 13 archaeological sites along the Pacific northwest coast of North America over a 5,000-year period (Crockford 1997: 13). This species has not so far not been identified in archaeological sites in island Oceania, and is not considered further here.

Figure 2. 

Four species of tuna discussed in this paper. Source of images: NOAA stock assessment reports. www.fisheries.noaa.gov/species/.

Scombrid proximate analysis

Of special importance to ancient diet is the fat content of available foods (Speth 2025; Leach and Davidson 2023; Leach 2024). Two detailed proximate analyses of albacore and two other scrombrid species have been carried out by Vlieg (1988), Vlieg and Murray (1988), and Vlieg and Body (1988), and their data is summarised in Suppl. material 1: tables S2, S7, and some features graphed in Fig. 3.

Figure 3. 

Results of proximate analysis by Vlieg and Murray (1988), and Vlieg, Habib and Clement (1983) of 58 species of fish in New Zealand waters (whole fish values). Left: Percent energy from protein and fat (whole fish). Right: Relative abundance of unsaturated fat and poly unsaturated fatty acids (PUFA). ALT = slender tuna, Allothunnus fallai.

The proximate analysis shows that whole albacore has a macronutrient ratio of energy from protein fat and carbohydrate (MNR) of 55:45:1. An unusual amount of carbohydrate is mainly from glycogen in muscle and liver (Barrett and Conner 1964: tables 5–7, 12–14), and along with interstitial lipids is thought to mobilize energy for bouts of fast swimming (Hulbert et al. 1979). Tuna are in the mid-range for energy from fat (Fig. 3 left). This is somewhat surprising, since tuna generally have a reputation of being oily fish. Albacore is in the group with high PUFA content (Fig. 3 right). The distribution of nutrients is far from uniform throughout the compartments of this species along with many other fishes, and this explains why certain parts are especially favoured and considered high status food among Polynesians (Leach 2006: 245; Leach et al. 2025: 17–19). For example, for albacore, only 5% by weight of oil is found in the fillet compared with 26% in the skin and 14% in the head. When converted into caloric energy, these differences are even more striking (Fig. 4). By far the most energy is in the skin, followed by the head, frame, fillet and viscera. Ethnographic records show the viscera from some fish species to have been a delicacy among some Māori communities (Buck 1926: 620).

Figure 4. 

Distribution of caloric energy in different compartments of the albacore. Protein is red, fat is yellow and carbohydrate is green. The skin and head are the richest sources of food energy.

The research by Vlieg and Murray included two other scombrid species, ALT and SKJ. As Fig. 3 shows, 73% of the food energy from the slender tuna derives from fat, while skipjack is similar to albacore. Another important finding of their research was that the mean oil content of both albacore and skipjack increased with decreasing SSWT, R = 0.96 (Vlieg and Murray 1988: 493; Vlieg et al. 1983: 247, fig. 2). The range is surprising, from 1.0 to 15% wet weight in white muscle across SSWT of 17 to 28 °C.

Scombrid isotopes and human diet

When reconstructing ancient human diet in the Pacific region from isotope values in human bone collagen (Leach et al. 1989; Leach et al. 1996b; Leach et al. 1997b; Leach et al. 1998; Leach et al. 2003; Kinaston et al. 2013; Leach et al. 2016; Kinaston et al. 2024), it is essential to have an extensive database of isotope values from organisms which form part of the diet of humans. In the Pacific region, 386 values for three isotopes (δ¹³C, δ¹⁵N and δ³⁴S) have been collated for plants, and both marine and terrestrial animals (Leach et al. 2003: 78–82), and an additional 141 values for two isotopes (δ¹³C and δ¹⁵N) by Kinaston et al. (2013: 5–6) and Kinaston et al. 2015: 35–36). Only two analyses of Pacific scombrids are included in these databases: all three isotopes are reported for the jack mackerel Trachurus declivis (Leach et al. 2003: 80), and one isotope (δ³⁴S) for the blue marlin Makaira mazara (ibid: 85). Fortunately, there is now archaeological data available to fill this gap, and this is included in a thumbnail sketch of marine fishes in this region (Fig. 5).

Figure 5. 

¹³C and ¹⁵N isotope values from fish bone collagen: blue = inshore species, red = oceanic species: Thunnidae spp. and Scombridae ?sp. from archaeological sites. Data source: Leach et al. (2003: 83, appendix 2); and Vika and Theodoropoulou (2012: 5, Table 2).

These isotope values reveal a considerable difference between inshore and offshore species of fish, and this needs to be taken into account when reconstructing human diet from human collagen in cases where archaeological sites contain significant quantities of scrombrids.

A sample of 165 specimens of albacore was obtained with assistance from the New Zealand Ministry of Agriculture and Fisheries from a research trawl dated 20 January 1990 (Kendrick and Bentley 2010). The fork lengths and weight were recorded on deck. The fork length was recorded to 5 mm precision, and the whole ungutted fish weighed with 100 g precision. Each head was removed for later processing and bagged and labelled with a catalogue number. These were given to the Archaeozoology Laboratory, at the Museum of New Zealand Te Papa Tongarewa. Each was given a museum catalogue number as it was processed in the laboratory. The bones used for measurement are five paired cranial bones: the dentary, articular, quadrate, premaxilla, and maxilla, and a special bone, the hypural in the caudal peduncle, which is diagnostic of fishes belonging to the families Scombridae (including Acanthocybium spp.), Xiphiidae and Istiophoridae (dicussed below). The five bones listed have been used for many years to quantify ancient fish catches from archaeological sites in the Pacific and New Zealand (Leach and Davidson 1977; Leach and Ward 1981; Leach 1986a; Leach and Boocock 1993). The heads (Fig. 6) were simmered in water until the five paired bones could be removed, cleaned of any connective tissues with tap water, and kept in open labelled petri dishes so they could dry naturally. When the bones were fully dry, the lids were placed on the dishes and stored for measurement.

Figure 6. 

Cranial preparation albacore tuna, Thunnus alalunga. Image source: https://www.fda.gov/ucm/groups/fdagov-public/documents/image/UCM059409.jpg.

The five bones mentioned do not always survive intact in archaeological sites. Hence it is desirable to include measurements which can be applied to incomplete bones. For this reason, more than one measurement was made on any one bone. Whenever possible on archaeological specimens the largest dimension is best to choose, as this yields the most reliable estimate of the original fish size. Thus, there is a series of measurements appropriate to whole bones and another series appropriate to various kinds of fragment. The dimensions chosen are illustrated in Fig. 7. These closely parallel those employed by archaeozoologists on other species (Rosello-Izquierdo 1986:35, Libois and Libois 1988; Sternberg 1992; Wheeler and Jones 1989:139 ff.).

Figure 7. 

Cranial elements of albacore used for measurements. The right bones are illustrated, but left bones were also measured. Measurements are taken between landmarks which are indicated with hollow dots and letters. The method of measurement is described below. When coding measurements, the three characters in the illustration are used, eg: RA2 = Right articular measurement #2 (see Suppl. material 1).

The anatomical landmarks used in this study are indicated on Fig. 7. Each measurement has a unique computer code with three characters. Thus, RD1 refers to the Right Dentary and the first measurement made on that bone. In cases where the terminology ‘maximum length’ or ‘maximum height’ is used, this implies that the measuring callipers are rotated about the nominated landmarks until a maximum value is obtained. The definition of each measurement is provided in Suppl. material 1. The number of these bones identified for any one species is generally considerably lower than for other bones in the anatomy of a fish. Moreover, in particularly large assemblages, the quadrate is sometimes excluded from the analysis, because of difficulties distinguishing between some species. The quadrate is quite robust and an adequate sample of measurements can be taken on whole bones. Two measurements are indicated for both dentary and articular, four on the hypural, and one for premaxilla, maxilla and quadrate.

The purpose of the three-character code is to permit simple coding of measurements on plastic bags which contain identified fish bones from archaeological sites. These are later entered into a database according to the original archaeological provenance. The appropriate equation for estimating live fork length and weight is selected using these three character codes. Mitutoyo digital callipers model 500–322 were used for linear measurements, which are recorded to ~ 0.01 mm precision, and a Sartorius model BA310S balance was used for weight measurements with a precision of ~ 0.001 g.

Even with the benefit of an annotated illustration (Fig. 7) and formal definitions (Suppl. material 1), it is not a simple matter to make the correct measurements on archaeological bones. It is desirable to learn the correct methods by re-measuring bones in the modern comparative collection and cross-checking measurements against those taken earlier. Subtle differences in the orientation of callipers, even when placed on the correct landmarks, can cause substantial percentage errors.

When completed, the database of modern measurements costs for albacore consists of 165 rows of specimens. and 22 columns of data for each: Specimen number, catalogue number, fork length, live ungutted weight, seven left, seven right measurements, and four on hypurals.

Distinguishing between live specimens in the genus Thunnus is not always simple (Viñas and Tudela 2009: 1), let alone from fragments of bone found in archaeological sites. Meat from tuna that is traded can be very difficult to ascribe to species. For these reasons considerable research has been undertaken to develop reliable genetic profiling (Weist et al. 2024; Davies et al. 2011; Viñas and Tudela 2009). This genetic research suggests that north and south Pacific albacore are separate stocks (Vaux et al. 2021). Both trace element research (Davies et al. 2011), and peptide profiling (Andrews et al. 2022; Winter 2023) are also proving a useful means of differentiating different species.

In spite of having access to considerable modern collections of scombrids at the Museum of New Zealand, the present author has always experienced difficulty identifying archaeological bones of scombrids to species level. The problem is compounded by the fact that archaeological bones are typically fragments, and distinctive features are not always present. It is usually possible to separate most scombrid bones from other species in an archaeological collection; however, the number that can reliably be ascribed to species is, at best, much smaller. This greatly reduces the prospect of reconstructing a reliable size-frequency distribution of any one species. As mentioned in the introduction, this is the necessary first step in building a quantitative picture of fish in ancient human diet. It has been noted by more than one author that scombrid vertebrae are readily distinguished from other families of fish, and some very useful research has been done following Jean Desse’s leadership in Europe (Lambrides and Weisler 2013). However, they note that when comparing recovery from two sieve sizes that when they used 6.4 mm sieves, 37% of vertebrae could not be identified, and for 3.2 mm sieves this rose to 60% (Lambrides and Weisler 2015: 7).

In spite of well documented difficulties, some researchers have reported success in distinguishing species of tuna using archaeological bones. For example, Lambrides and Weisler note that “species level identifications of Katsuwonus pelamis were achieved using a few cranial elements (ie., basypterygia, dentaries, scapulae, and quadrates), but caudal vertebrae were found to be the most useful, particularly C21–C25” (Lambrides and Weisler 2017: 10). They also note that the species identifications of skipjack were done on vertebrae (ibid.: 12), unfortunately not providing guidelines of how these could be distinguished from other scombrids. They reach a fair conclusion that “there is a need to complete species-level identifications of archaeological tuna remains” (Lambrides and Weisler 2017: 26). In some cases bone size might be a helpful guide to which species is present, assuming that someone has collated the necessary comparative data. Another option is to guess the most likely species, from their modern distribution in the Pacific and the location of the archaeological site in question. There are obvious hazards in such protocols, not to mention the possibility of distributional changes of species over the last 3,000 years, associated with changes in SSWT and the location of gyres and currents. Although bone identification is not an exact science, the best approach is, whenever possible, to base identifications on features of anatomy that have been proven to be distinctive from all other possible species.

The present author is not the only one who finds difficulty distinguishing archaeological bones of this family of fishes. Buckley, Pinsonneault and Brassey et al. point out that scombrid bones can [only] be used to identify to family level (Buckley et al. 2021: 136), and this was the reason for exploring the use of biochemical markers. Such techniques offer exciting new possibilities, but are unlikely to be extended to thousands of bones that archaeologists are normally confronted with.

Historical publications on scombrid comparative osteology were canvassed for possible guidance on distinguishing species from bone fragments. There are exceptional illustrations of fish cranial and infra-cranial anatomy in early historical literature, and some of these are helpful when there are substantial portions of the anatomy present in archaeological sites. Examples that included scombrids are Allis (1903) and Sylvia (1955). These provide excellent illustrations of individual elements with notes on distinctive details, such as the elements of the caudal peduncle. Collette and Chao (1975: 579) provide detailed illustrations of dentaries and articulars of Sarda spp. for comparison. Kishinouye (1923), Nakamura (1965), and Gibbs and Collette (1967: 78, 81) provide quality comparative drawings of the caudal peduncle and other vertebrae of scombrids. Godsil and Byers provide especially useful comparative drawings for five species of tuna, illustrating the hyomandibular, metapterygoid, opercle, quadrate, subopercle, interpercle, epihyal, ceratohyal, and cleithrum (Godsil and Byers 1944: 120–122). The differences between species are subtle, and hand specimen comparative material is needed for identifications to be made. There is also a species by species description of Pacific fish cranial elements for archaeologists by Dye and Longenecke (2004), unfortunately, only including one species of scombrid, skipjack tuna.

In spite of all the high-quality descriptive anatomy that has been published over the years, only a small amount is aimed at distinguishing between each of the species of tuna, currently thought to be as many as56. The present author has access to multiple specimens of three species (albacore, yellowtail and skipjack, and access to other boxed specimens) but has great difficulty distinguishing between these with consistent reliability using five of the cranial elements and one special bone (dentary, articular, maxilla, premaxilla, quadrate, and hypural). The only two that give much promise are the dentary and hypural, and these are illustrated for three species in Fig. 10. Note, the accessory hemal spine (parhypural) on F is loosely attached on the specimen illustrated. The connective tissue holding it in place degrades and the spine detaches in archaeological specimens. It is loosely attached in the specimen illustrated, and easily removed. The parhypural spine does have distinctive anatomy and is therefore quantifiable, but would be difficult to distinguish among a mass of other fish spines. The hypural element itself is very similar from one species to another among those identified in archaeological deposits belonging to Scombridae, Xiphiidae and Istiophoridae (see for example illustrations by Fierstine 1968: 15, figs 12–13; Izumi et al. 1972: 30; Izumi 1983: 273–274; Potthoff 1975, fig. 23; Potthoff and Kelly 1981: 172–175; Leach et al. 1987; Monsch 2000: fig. 7.43). Moreover, there are so many species in the family Scombridae alone that it is doubtful that correct species identification could be made from this bone based only on anatomy. In this respect, of particular importance to Pacific archaeologists is that the hypural of Acanthocybium solandri is not easy to distinguish from species of tuna. If bone size was used to aid identification, there is so much overlap from one species to another that even distinguishing between Thunnus spp. and Euthynnus spp. is difficult. See for example, illustrations of the latter by Marrast and Bearez (2019: 190, fig. 1). The caudal peduncle of species in these three families of fish is highly specialised for active and sustained swimming, inspiring one biologist to remark “The acme of this type of specialisation must surely be in the extraordinarily beautiful tail-end of the bonito (Katuswonis)” (Ford 1937: 8). Finally, the two preurial bones (G), resembling a flagpole cleat, are also so similar from one scombrid species to another that identification better than Scombridae or Thunnidae/Katsuwonidae is, in my opinion, doubtful.

Figure 10. 

Comparison between three species of scombids. A, D. Albacore tuna; Thunnus alalunga; B, E. Yellowfin tuna (Thunnus albacares); C, F, G. Skipjack tuna, Katsuwonus pelamis. Bones D and E are inverted.

In the case of the dentaries illustrated in Fig. 10 there is some room for optimism. The red arrow points to a ridge along the inner surface of the dentary, and this is arguably distinguishable between at least the three species illustrated. For specimen A the bony ridge marked with the red arrow terminates at the margin of the intermandibularis anterior muscle; for B; it extends beyond this margin; and for C it terminates before the margin. The cavity under this ridge is associated with the intermandibularis anterior muscle. This is a small muscle whose fibres run between the anterior portion of the two halves of the dentary. Finally, the blue arrow points to the ventral, or postsymphyseal notch, and the shape of this is a consequence of how high and robust the anterior wall of the dental symphysis is. The more robust the wall, the more pronounced the notch. The notch is deeper for C than the other two species, and the chin is wider for B, and narrower for both A and C. These differences are subtle, and have not been checked against many other species of scombrid, so whether this optimism survives comparison with a larger range of species and across individual variation remains to be seen.

Since there are difficulties distinguishing between scombrid species from their bones, is it possible that the allometric equations for one species could be used for another? When this prospect was first mooted with colleagues the initial response was “surely not possible”, in which case it should be a simple task to reject the null hypothesis with elementary statistical comparison.

Cranial bone measurements and fish length estimate

Since the best-fit relationship between any one cranial measure and the live length of the fish has been shown to be a power curve, there are several ways that this hypothesis could be tested. For this test we chose the largest, and therefore most reliable, measure – the length of the dentary (LD1), and the first test was a simple visual test of the relationship between LD1 and the live fork length (FL) for 12 species illustrated in Fig. 11. These data derive from earlier allometric studies of these species, all fully published, with the exception of albacore (this paper).

Figure 11. 

The power curve fit between LD1 and Fork Length (FL) for 12 species of fish (Nemadactylus macropterus, Odax pullus, Thyrsites atun, Parapercis colias, Pagrus auratus, Arripis trutta, Pseudophycis bachus, Notolabrus celidotus, Notolabrus fucicola, Pseudolabrus miles, Thunnus alalunga, Anguilla australis). Data derived from earlier published research.

When the data for these 12 species was plotted it was surprising to see that a number of species appeared to more or less follow the same curve as albacore, while others clearly departed from it. In the case of the four main outlier species, the fork length grows much faster than the dentary, so that, in relative terms, the mouth parts end up being smaller at maturity than the other species. There are good biological reasons for such growth patterns. For example, it has been shown that for the same body length, omnivorous fishes tend to have smaller mouth areas than carnivorous ones (Karachle and Stergiou 2012: 79–80). Although this topic is beyond the scope of the current paper, I observe in passing that the four outliers in Fig. 11 all browse on smaller organisms with variable targets. One species, of the Odacidae family, is a vegetarian.

In this foregoing simple experiment, the expectation was that the visual test would reject the null hypothesis outright, requiring no statistical test for confirmation. The fact that six quite different species plotted along one similar growth trajectory was therefore unexpected, and closer attention is called for (Fig. 12).

Figure 12. 

Relative growth rate of dentary compared with body length for six species: cyan dots = barracouta (BAR), Thyrsites atun; blue dots = blue cod (BLU), Parapercis colias; yellow dots = snapper (SNA), Pagrus auratus; green dots = kahawai (KAH), Arripis trutta; red dots = red cod (RED), Pseudophycis bachus; and red dots also = albacore (ALB). These species belong to quite different families, yet follow a surprisingly similar relative growth curve. There are good grounds for expecting departure from co-linearity in the case of very large specimens of any one species. This is because the same percent variation in condition factor results in increasingly greater absolute variation from small to large (see Fig. 18).

A simple test for co-linearity is to examine the parameters for exponent and constant. These values are given in Suppl. material 1: table SS3, and illustrated in Fig. 13. On this test, two species are within two standard errors of the parameters of albacore, and could therefore claim to be co-linear. Finding even one other species was a surprise.

Figure 13. 

The power curve parameters of albacore LD1 against FL compared with 11 other species of fish. The green lines mark off confidence boundaries for albacore at 2SE.

From a practical point of view, such a test for co-linearity is not necessarily the most important issue when it comes to reconstructing the nutritional value which each species contributes to ancient diet. What matters most is the weight of meat (MTWT) which the bones represent, and this is estimated from the reconstructed body length from the bones and BWT. So, the most important consideration is what the practical significance is of these slightly different best fit curves. One way to evaluate this is to consider what happens if the wrong equation is used to estimate fork length from bone dimensions. For example, instead of using the equation appropriate to albacore (Suppl. material 1: table SS3. Fork Length = 2.9943*LD1^0.8404), what happens if we use the equation for, say, blue cod? That is, what practical difference does it make to the estimated fork lengths? This was tested for the entire range of LD1 measurements of albacore (ALB), using the equations of the other 11 species in Suppl. material 1: table SS3. We were surprised to find that the results for three species were within one SE of the estimate boundary for albacore. These are illustrated in Fig. 14.

Figure 14. 

The effect of using incorrect equations for reconstructing fork length across the full-size range of LD1 cranial measurement on albacore bones. The Y axis shows the error that is introduced as a result (mm difference in estimated fork length). This shows that equations for quite different species could be used for albacore without significant error. Further support for this is a recent allometric study of 13 species of scombrids which provides supporting evidence of co-linearity between live length and bone dimensions (Elliott Smith et al. 2026: fig. 2).

Fig. 14 shows that of the 11 species for which equations have been established, in the case of three, there is no practical disadvantage in using the wrong equation. For BLU, SPO, and BAR, the estimated fork lengths are well within ± 1 standard error of the estimate (39.8 mm). A further species is just outside this range, and other species are within 2SEs.

From the foregoing there are grounds for optimism that the equations derived for albacore may be useful for estimating live fork length of other species of scombrid. If so, it will help to overcome the problem of the cranial bones of Pacific scombrids being so similar. This is not the first time that such an issue has emerged in archaeology, and it is instructive to outline how it has been dealt with earlier. The example concerns the allometric equations for three species of labrid (Fig. 15), studied by Leach and Davidson (2001). To my knowledge this is the only comparable data available that can be compared with this present study of albacore. Specimens of these three species are prevalent in inshore environments in New Zealand and common in archaeological sites, but as with scombrids, the cranial anatomy is very similar, creating difficulties identifying the correct species from archaeological bones. In this case, there were sizeable modern collections of all three species, so all regression equations were able to be established. The potential effects of choosing one equation over another was carefully examined and it was concluded that there was little practical difference when estimating fork length from archaeological bones. The power curve relationship between LD1 and fork length for the three species is illustrated in Fig. 16. In this case, a regression equation based on all three species combined was chosen for estimating fork length from archaeological bones. Although the size-frequency distribution of modern adult specimens of this species overlap, they are largely distinctive. This facilitated the decomposition of the archaeological measurements into the original three species using MIX software developed by Peter Macdonald at McMaster University for this very purpose. This was based on earlier research (Macdonald and Pitcher 1979; Schnute and Fournier 1980; Everitt and Hand 1981; Titterington et al. 1985; Macdonald 1987; McLachlan and Basford 1988). In the case of labrids, this approach proved to be very successful in coping with archaeological bones. Could this also be possible with tuna species? This depends, in part, on how distinctive the size-frequency curves are for the different species.

Figure 15. 

Three species belonging to the family Labridae which have similar cranial anatomy. A. Pseudolabrus miles, B. Notolabrus celidotus, C. Pseudolabrus fucicola. Photos courtesy Museum of New Zealand, Te Papa Tongarewa. Original artwork by Frank Edward Clarke.

Figure 16. 

The best fit power curve for three species of labrids.

Conclusion on length estimate

Regardless of which species of tuna is present in an archaeological site, live fish length could be estimated with reasonable precision from cranial bone measurements using one equation for other species of scombrid. Although comparative analysis of modern data of these four tuna species has shown overlapping size distributions, it should be possible to identify the relative abundance of each tuna species that is present in an archaeological assemblage, by examining the size frequency distribution of fork length and comparing this with those in Fig. 17. Alternatively, a case could be made for using Peter Macdonald’s MIX algorithm, as was done with labrids.

Figure 17. 

Size-frequency distributions of four common species of scombrids in the Pacific. Source of data: New Zealand Ministry of Fisheries.

Estimate of body weight from length

From the foregoing, assuming that acceptable estimates of live length have been obtained from an archaeological site, the next task to consider is whether there is a means to estimate the live body weight (BWT) of each specimen from which meat weight (MTWT) can be calculated. Fortunately, there is a database of fork length coupled with live weight available from the New Zealand Ministry of Fisheries³, from which data for the four species was extracted and examined. The results are shown in Fig. 18.

Figure 18. 

Scatterplot of live weight against fork length for four species of tuna, together with the best fit power curve of the relationship between the two. Data source: courtesy New Zealand Ministry of Fisheries. The parameters of each curve are provided in Table 3.

The important question now is — can one curve be used for all species? Once again, this depends on what is the practical difference between the curves in Fig. 18. The standard errors of the estimates of the two parameters of each curve were examined to see whether any are co-linear. The results are provided in Suppl. material 1: table SS5 and illustrated in Fig. 19.

Figure 19. 

Left upper: Standard errors of the estimates of the exponent and constant values of the regression line between ln(FL) and ln(BWT) for four species of scombrids. Right upper: plot of the four power curves of fork length (mm) and live body weight (g). The black line is BET, blue is YFN, red is ALB, and SKJ is cyan. Below left: The percent error in estimating weight between the two curves that are widest apart (ALB and SKJ).

This reveals that the curves for ALB and BET are effectively co-linear, as are YFN and SKJ, although the two pairs are significantly different. Once again, the important question is what the practical difference is between these four curves for estimating MTWG. This is easily examined by considering the difference along the Y axis between the curves that are widest apart: ALB and SKJ. This is shown in Fig. 19, bottom left. For example, a fork length estimated from an archaeological bone of 300 mm would produce an estimated weight of 426 g using the SKJ curve, and 618 g using the ALB curve, that is a difference of 192 g, which is 31% of the ALB value. There is zero difference at a fork length of 1068 mm. This shows that errors might be acceptable when fish are greater than c. 600 mm fork length, but would not be acceptable when FL < 600 mm.

It has been noted that tuna are relatively rare in archaeological sites in the Pacific (Ono 2010: 299). Amesbury and Hunter-Anderson have produced a summary of the presence of pelagic species in Guam and the Mariana Islands (Amesbury and Hunter-Anderson 2008). This shows that several species of tuna, marlin and dolphinfish were regularly caught by pre-European people in this region of the Pacific. Lambrides and Weisler produced valuable information on a wide range of fish species from excavations on Ebon atoll in the Marshall Islands dating from the last 2000 years. They found the caudal vertebrae C22–C25 to be the most valuable element for identification and quantification of scombrids. More generally, these authors have made significant inroads in archaeological fish quantification using vertebrae (Lambrides and Weisler 2013, 2015). The special value of the hypural bone in the identification of scombrids, and its value for estimating life size has long been recognised in archaeological studies, but it has been noted that identification to species is not so easy (Davidson and Leach 1988: 340–341; Leach et al. 1988: 39; Leach et al. 1990: 22). Ono and Intoh describe the presence of scrombrids (100 MNI and NISP 1,041) from a total MNI for all fish species of 281 (NISP 2,320) from their excavations on Fais, ranging in age from AD 200–1400 (Ono and Intoh 2011: 262; see also Leach et al. 1994). Ono and Intoph described vertebrae and caudal peduncles as being similar to skipjack and yellowfin (2011.: 267). Most of the tuna identifications were from vertebrae (NISP = 951), caudal peduncle (76), and only 14 cranial bones (ibid.: 268).

Buckley et al. (2021) describe collagen peptide mass fingerprinting (also known as ZooMS) on 135 scombrid bones (11 hypurals) from an excavation of the Hanamiai archaeological site, on the leeward side of Tahuata island, Southern Marquesas, dating to c. AD 1250–1900. Seventy one of 73 hipurals were identified as skipjack, one as Thunnus sp., and one as unknown sp. (Buckley et al. 2021: 5, Table 4). Thirty-eight bones contained insufficient collagen for analysis (28%). Other researchers report a failure rate for ZooMS of 50% on archaeological fish bones (Dierickx et al. 2023). At Hanamiai there was a NISP of 100 scombrids.

Several studies have been undertaken of the tuna recovered during the excavations of a site on Motupore island, Bootleg Bay, Papua New Guinea. McMurtry’s analysis by fish family included an MNI of 57 Thunnidae/Katsuwonidae in a total of 442 (McMurtry 1986: Table 1). Of special interest in this analysis was an MNI of four Istiophoridae/Xiphiidae, which is unusual in Pacific assemblages. Bones of these large predators have also been recovered in excavations in the Marianas (Leach et al. 1988: 35, Table 4). In the detailed study of the Motupore excavations by Allen he notes that a species of scombrid, known as the striped mackerel, Rastrelliger kanagurta, spawns in Dagulata estuary within Bootless Bay in June and July (Allen 2017: 506). He cites Pulsford’s ethnographic description of yellowfin tuna fishing in Bootless Bay. Pulsford did not provide the taxonomic identity of the species that his observations relate to, referring only to it by the indigenous name of kidukidu (Pulsford 1975: 107). In the southern coast of Papua New Guinea this name is used for both dogtooth tuna, Gymnosadia unicolor and frigate mackerel, Auxis thazard (Coyne 2008: 16). Skipjack tuna is also present in this vicinity and caught by indigenous people, and yellowfin tuna migrate into Bootless Bay from May to October (Saulei et al. 2021: 27). Careful examination of the two photos published by Pulsford confirms that the species caught on that occasion was Thunnus tonggol (longtail tuna)⁴. Unfortunately, it must be concluded that the species(s) of tuna present in the Motupore excavations is, at present, not known with certainty. The original identification by McMurtry as Thunnidae/Katsuwonidae still stands.

There is information in the data collated about the modern tuna fishery in the vicinity of Papua New Guinea that has potential interest to archaeologists. Kumoru and Koren provide catch data by both seine and longline methods between 2001 and 2005 for skipjack, yellowfin and albacore (Kumoru and Koren 2006: 9–11). This is summarised here in Suppl. material 1: table SS5. This reveals that almost no albacore are caught by seine (0.4%), but 56% of the longline catch is albacore. The identity of tuna caught by indigenous people in Bootless Bay is also raised later.

Fraser has reviewed the presence of tuna in archaeological sites throughout the tropical Pacific (Fraser 1998, 2001). She provided a thorough review of tuna from 21 archaeological sites in the Pacific, based on collections in the Archaeozoology Laboratory at the Museum of New Zealand (Fraser 1998, 2001). These covered a broad geographical range, from the Marianas and Hawaii to Papua New Guinea, New Caledonia, Cook Islands and the Marquesas in the east. A few more recent excavations have been added to the database, and the MNI and NISP values from a selection of 37 archaeological sites are provided in Suppl. material 1: table S9, together with cross references to the source of information in each case. Unfortunately, some paired MNI and NISP are not available, reducing the number of paired values to 31 sites. The percent values by MNI and NSIP are plotted in Fig. 20.

Figure 20. 

The relative abundance of Scombridae in 31 of 37 archaeological sites from the tropical Pacific, by both MNI (x axis) and NISP (y axis), together with confidence ranges expressed as equi-probability ellipses (Snedecor and Cochran 1967: 210–211; Leach and de Souza 1979: 32). A few site numbers are cross-referenced in Suppl. material 1: table S9.

This depiction shows that tuna species generally occur in low abundance (<10%) in 24 of the 37 sites, where they are present at all. The most notable exceptions are marked with a number in Fig. 20. These are as follows: #1 and #2: the two independent studies of the collections from Fa’ahia on Huahine in the Society Islands, #3 is the site on Ebon in the Marshall islands, #6 the site on Fais in the Caroline Islands, #11 the site on Atafu in the Tokelau Islands, #13 the site at Hane on Ua Huka in the Marquesas Islands, and #36 is Te Ana Pua on Ua Pou in the Marquesas Islands. In the 1997 publication about Te Ana Pua it was noted that although tuna bones could not be identified to species, one bone stood out as being especially suitable for fish size to be estimated — a robust element of the caudal peduncle known as the hypural bone, and present in several species of tuna, wahoo, marlin and swordfish. The hypurals from the site were measured and a size-frequency distribution suggested that more than one species might be present at Te Ana Pua, “such as Katsuwonus pelamis (skipjack), Sarda sp. (bonito), or perhaps Rastrelliger sp. (mackerel)” (Leach et al. 1997a: 57, figs 4, 5). At the time, these were merely guesses; however, we now have a method for estimating life fish size from bones of tuna, and this is now attempted using some archaeological collections.

Albacore tuna is one of a number of species in the Scombridae family, four of which are of commercial significance in the topical Pacific region (bigeye, yellowfin, albacore and skipjack). While bones of tuna are found in archaeological sites they are not common, except in a few cases. Both cranial and infra-cranial bones of tuna are very distinctive from other families of fish, but in spite of considerable differences in size within the family, the bones are not easy to identify to species from anatomical features. Genetic and peptide fingerprinting can differentiate tuna species and is a valuable tool, but unlikely to apply to thousands of bones typically found in archaeological sites. Archaeologists interested in palaeoeconomics and nutritional aspects of ancient diet rely on estimates of meat weight of different species in archaeological sites. In the case of tuna, the difficulties in identifying bones to species level, degrade the calculation of MNI and/or NISP, and frustrate the ability to estimate live fork length and body weight from bone measurements. While examining the allometric relationship between bone size and live dimensions it became clear that the growth of the viscerocranium and postcranial compartments proceeds at different rates through life from one species to another. For example, the postcranial compartment of eels grows much faster than the mouth parts when compared with other species, resulting in a long body and small jaws. Although this may seem obvious, what is not so obvious is the unanticipated finding that species from widely different families followed almost identical allometric trajectories. Analysis of three species in the Labridae family revealed that the relative growth curves of cranial bones and live length were interchangeable. It was therefore suggested that there were reasonable grounds to expect members of the Scombridae family to follow closely similar growth curves, effectively co-linear, and that the allometric equations of albacore should be an acceptable proxy for other species in the family. Although the relationship between cranial bone size and body length is approximately linear (exponent = 1.0), the best fit is a power curve with an exponent between 0.75 and 0.95. Equations were therefore proposed for 18 cranial bone measurements and fork length from research on modern albacore which could be applied to bones of the three other species of tuna being considered. In the case of estimates of live body weight from bone measurement, the relationship is closer to cubic, and even small differences in the curves from one species to another would amplify errors if an equation for one species was applied to bones of another. Separate power curve fits between fork length and body weight were therefore established using measurements of large numbers of specimens of the four species of tuna from modern research trawls.

Cranial bones identified as tuna from four archaeo­logical sites in the Pacific (Motupore, Te Ana Pua, Fa’ahia, and Hane) were measured and fork length estimated using the equations derived from the albacore study. Size-frequency diagrams were constructed for the archaeological samples, and these were compared with similar diagrams of modern trawl catches of the four species of tuna. The archaeological fork lengths were then split up into groups that most closely matched those of the modern species. At Te Ana Pua, Fa’ahia and Hane, the dominant species was skipjack, with a few albacore at Te Ana Pua. At Motupore most of the catch is judged to have been yellowfin, and a few skipjack.

Live body weight was then estimated for each, using equations of the appropriate species. The mean body weight is then multiplied by the MNI at each site to yield the total BWT of tuna. The values for the four sites are quite similar, ranging from 265 to 374 kg. When MTWT values were calculated and nutrients from each body compartment evaluated, the overall MNR was found to be 60:38:2. The slightly elevated value of carbohydrate is due to glycogen reserves in tuna white muscle. This shows that tuna was an especially rich source of protein to the diet of pre-European Pacific islanders. Although the main objective of this paper has been to help build the infrastructure for economic interpretations in archaeology, judging from the comments made by Nordhoff in the introduction, the attitude of Polynesians when fishing for tuna was probably less about food, and more about drama on the high seas, pitting one’s wits against a formidable marine adversary. In this respect, the cousins of Polynesians in the vicinity of the Mariana islands are especially notable for their battles with marlin and swordfish from canoes – but that requires another publication.

The research reported in this paper is not without its limitations. Above all, further research is needed on modern specimens of the remaining three species of tuna to shed more light on the relative growth rates between bone size and body length.

The Ministry of Agriculture and Fisheries kindly provided the specimens of albacore for this project from a research trawl dated 20 January 1990, as well as much needed historical data on fork length and landed weight for four species tuna from their database. Associate Professor Alessio Datovo, Curator of Fishes at the Museum of Zoology, University of Sao Paulo, Brazil provided access to his exceptional illustrations of scombrid soft tissue and osteology, and helpful advice on anatomical issues; Garry Law provided useful discussion and advice on a number of parametric statistical issues; and Janet Davidson read drafts of this manuscript for editorial blunders. Martin Lewis and Amy Phillips, reference librarians at the Museum of New Zealand, gave much needed help during literature research and provided many documents by interloan. I am grateful for the generous assistance from all these people.

The research on tuna presented in this manuscript was part of the Bridge and Barrier Research Project. I thank the Foundation for Research Science and Technology for their financial support of this project (Contract #MNZ801), and the Museum of New Zealand, Te Papa Tongarewa, for giving institutional support.

Finally, I would also like to give my special thanks to Ross Whymouth for providing a modern computer system that was capable of running heritage software in a virtual environment, which facilitated the completion of this project. All numerical processing in this paper was accomplished with code written by the author in Turbo Pascal version 5, and all graphics was produced with the Graphics Language Editor GLE version 5.5 in a Dos environment, courtesy of Chris Pugmire, formerly of Physics and Engineering, DSIR.

This paper has benefitted from helpful comments from four referees, for which I offer my thanks. I am also grateful for advice from, and careful attention to detail by, the Editor Phil Sirvid.