A species of fish, Odax pullus, predominantly vegetarian, common in weedy marine zones in New Zealand, is the subject of this paper. The species is primarily caught by setting gill nets today, and seldom takes a baited hook. It has variable abundance in archaeological sites, more commonly in higher latitudes. It is very rich in iodine, and is useful in preventing goitre, since the element is low in New Zealand soils. It has very low levels of unsaturated fatty acids and low levels of omega 3 lipids. Modern specimens were collected and cranial bones extracted and measured to establish equations to back-calculate fork length (FL) and ungutted weight (BWT). Power curves were fitted, and constants provided to enable live characteristics to be estimated from archaeological bones. Bones from nine archaeological sites were studied and FL and BWT estimated. Length-frequency distributions were mostly non-normal, suggesting more than one method of capture, possibly set nets or hoop nets with different mesh size. The modern practice of only eating the fillet harvests only 34% of nutrients, with a MNR of 89:10:1. Ethnohistorical observations suggest more than the fillet was formerly eaten, especially fat-rich components. We estimate a MNR of 78:21:1 was more likely if some of the head, viscera and skin was eaten. In southern latitudes, where carbohydrate-rich foods were in shorter supply, harvesting these fat-rich parts was a sensible survival strategy. The archaeological catches show 20–45% were fecund individuals, below modern legal size. This is consistent with other marine species from archaeological sites.

Allometry, Aotearoa, Archaeozoology, Ichthyoarchaeology, Macronutrients, Odacidae

Quantitative reconstructions of subsistence economics of ancient societies depend on reliable information of the weight of each ingredient of food. For each animal species in the food web, metrical analysis of bone size in archaeological sites can be used to estimate the live ungutted body weight (BWT). From background knowledge of protein, fat and carbohydrate values of each species, the usable meat weight is then used to calculate the macronutrient ratio (MNR) and caloric energy for each ingredient in a human diet. Thus, animal bone size is part of the primary data that is used for reconstructing palaeoeconomics. In addition, analysis of bone size can be used to reconstruct the live fork length (FL) and size-frequency diagrams of each species. These distribution diagrams and dispersion statistics help to identify changes through time that may have a number of causes, such as climate change, and the impact of humans on environmental resources. Over the last 30 years a series of specialised publications have appeared, focusing on individual species of New Zealand fish that were important in pre-European times: species in the Labridae family (Leach and Anderson 1979a; Leach et al. 1997a; Leach et al. 1999; Leach and Davidson 2001); Sparidae (Leach and Boocock 1995); Arripidae (Leach et al. 1996a); Gempylidae (Leach et al. 1996b); Pinguipedidae (Leach et al. 1997b); Moridae (Leach et al. 2001); and Latridae (Leach et al. n.d.). The primary purpose of each of these studies has been to provide reliable allometric equations for archaeologists to enable live fish FL and BWT to be estimated from archaeological bone assemblages. This current study is of a species in the sub family Hypsigenyini of Labridae (formerly Odacidae family¹), a fish variously known as greenbone or butterfish, and mararī (Odax pullus). A second species, Odax cyanoallix, is mainly found in the sub-tropical waters of the Three Kings Islands to the north of New Zealand, straggling to Northland in the north of New Zealand.

Figure 1. 

The New Zealand greenbone Odax pullus. Art work from Doogue and Moreland (1966: 256, courtesy of artist Eric W. Heath).

The reason for the name greenbone is that it has a tendency for the bones to turn to a green colour fairly quickly after the fish dies, and for this reason the fish was unpopular in the 1930s in New Zealand, consumers thinking that the fish was rotten. However, as Graham, the director of the Marine Biological Laboratory at Portobello pointed out, the colour was due to precipitation of an iodine salt, as the flesh of this fish is very rich in iodine, something which makes it very useful for prevention of goitre (Graham 1956: 262). Paul commented regarding this point that the origin of this information has not been found (Paul 1997: 3); however, there has been considerable research on this issue, since New Zealand soils are naturally low in iodine, and endemic goitre was noted in the 19th century. From 1924 onwards, salt was iodised to alleviate this. Goitre would have been endemic amongst pre-European Māori without a means to augment iodine in their diet. Consumption of greenbone and mussel shellfish would have helped, along with seaweeds such as karengo (Porphyra columbina), which is rich in iodine (Cambie and Fergusson 2003: 115). Hercus and Aitken carried out pioneering research into the matter in 1932, examining iodine levels in seaweeds and species of fish and shellfish obtained from the Portobello Laboratory, and confirmed the high levels of iodine in New Zealand greenbone and shellfish (Hercus and Aitken 1933: 65–66, Suppl. material 1: table S1).

The Māori name for this fish is mararī, or rarī. This name is common in many parts of the Pacific, often, but not always, applied to Labridae. The word has many cognates and has been reconstructed to Proto-Polynesian *m(a,o)lali by Hooper (1994: 217–218). The word maa-lali and its cognates (Māori mārari) refer to things that are smooth, wet or slippery. This is an appropriate description for the greenbone fish, which has very tiny scales and no sharp spines and is very slippery when wet. Peter Buck also recorded this species with the name koeaea (Hiroa 1926: 646), a name more often associated with species of New Zealand whitebait (Galaxias spp.); however, Williams (1971: 122) certainly attributes the name kōeaea to greenbone, and confusingly also to marblefish (Aplodactylus arctidens). Considering the abundance of greenbone in areas where marblefish thrives, surprisingly little is recorded in early historic ethnographic literature about Māori use of this species of fish. The most useful observation was made by Knox in 1870 when describing greenbone fishing around Mana Island near Wellington; they are “fished by means of a net in the form of a bag with a hoop round the mouth, and secured by means of a rope to a branch of the kelp... the net is set amongst the kelp, where the rise and fall of the tide produces a kind of free run, which the fishermen avail themselves of in setting their net, and upon returning they find it full of the fish of all sizes” (Knox 1870: 131). An example of such a hoop net was illustrated by Sydney Parkinson during Captain Cook’s visit to the Marlborough Sounds in 1769 (Cook 1968: fig. 41). In 1923 Peter Buck, along with Elsdon Best and two other colleagues, made what they describe as an ‘ethnological expedition’ to the coastal areas of Gisborne and Bay of Plenty to Te Kaha to record the fishing activities of Māori communities in the area (Hiroa 1926). The publication that resulted is a mine of useful information about netting activities among the rocky reefs of the area. Great emphasis is given to marblefish (kehe) which browses on algae in the same seaweed rich habitats as greenbone, but almost no comment is made about the latter. Both species are almost entirely caught by netting. Buck provided photos of several sizes of hoop nets, some with very fine mesh.

Greenbone are found throughout New Zealand, around kelp-rich rocky shore habitats, but are not so common in the warmer waters of the far north. It is primarily a vegetarian fish, feeding almost exclusively on Ecklonia radiata (golden kelp) and Carpophyllum spp. (flapjack algae) (Clements and Coat 1993: 213, also Trip et al. 2016: 23). Their teeth are fused into a sharp beak-like jaw, similar to the tropical parrotfishes (tribe Scarini in the Labridae), and they bite pieces from the kelp, leaving a neat oval-shaped hole in the frond. They have three sharp-edged pharyngeal bones in the throat (Fig. 5 lower), and these cut the piece of kelp into smaller pieces before swallowing. Stomach contents also contain small crustaceans and shellfish. Greenbone are a beautiful deep blackish green or blue colour and thick in cross-section, providing good-sized fillets for European consumption. Modern supermarket specimens average 30–50 cm length, but can reach 70 cm or more. According to Trip et al. (2011: 186 and fig. 3). greenbone become sexually mature by the age of 1.1 to 1.5 years (FL c. 220–260 mm) and change sex at 2.8 to 3.5 years (FL c. 360–390 mm). Thompson (1981: 213–215; Fisheries New Zealand 2024: 262) provides some useful information about the species in the Leigh Marine Reserve north of Auckland. She notes that they are diurnal, something Graham also commented on (Graham 1956: 264), and that they are most abundant in waters less than 13 m deep. In the Leigh Reserve they reach densities of 14 fish per hectare, and are also present in deeper waters where Ecklonia forests are found, but at densities of about 2.5 fish per hectare. In theory, the vegetarian habits of greenbone preclude the use of hooks, but they are known to take a tiny hook baited with a shrimp and even a fragment of kelp will occasionally capture them. However, this is a species which in pre-European times would have been taken almost exclusively by netting or spearing. Graham describes them as having a range of colours, depending on environment and amount of light present. He also notes their habit of apparently sleeping in a vertical position in an aquarium, instantly returning to a swimming position when lights are turned on (Graham 1956: 264). These comments and those of Thompson are somewhat at variance with the use of set nets by modern fishermen, whose nets are often set in the evening and collected in the morning after a tidal change. Personal experience of the authors is that night netting is very productive, and when seen with a torch during night diving, these fish seem quite active.

The most useful information about the food value of greenbone comes from the detailed study of New Zealand fishes by Vlieg (1988). He carried out proximate analyses of 62 different species of New Zealand fishes. In most cases, 6 specimens of each were studied. Each specimen was dissected into five sections (head, viscera, frame, skin, and flesh), and each portion was weighed and tabulated by proportion of the live BWT. This information for greenbone is summarised in Suppl. material 1: table S2.

The macronutrient ratios of the more common species of New Zealand fishes are plotted out in Fig. 2. Carbohydrate values range from 0.15 to 0.25% by BWT (Vlieg 1988: 5), corresponding to a macronutrient ratio (MNR) of c.1%, and in many cases can be ignored. The MNR of greenbone is 67:32:1 (from total nutrient values). This species has low fat reserves, something noted by early researchers of New Zealand fishes (Hector 1872: 115). On the right of Fig. 2, the total amount of unsaturated fatty acids is plotted against the n-3 polyunsaturated fatty acids (PUFA) for the same selection of New Zealand fish species (extracted from data in Vlieg and Body 1988). The PUFA values in the graph combine two omega-3 lipids: EPA (eicosapentaenoic 20:5) and DHA (docosahexaenoic 22:6). Greenbone has low levels of omega 3 (9.6% of available lipids w/w). It also has very low levels of unsaturated fatty acids compared with other fish species (43.7% of all available lipids w/w).

Figure 2. 

Macronutrient ratios (MNR) and lipid values of 58 common New Zealand fish species.

One more characteristic of greenbone that should be briefly mentioned is the isotope signature of its flesh. This is important in reconstructions of ancient human diets from the isotope signature in collagen from human bone. The values determined for specimen AA569 in the Archaeozoology Laboratory are: δ¹³C = -14.9‰, δ¹⁵N = 9.2, and δ³⁴S = 17.2‰ (Leach 2003: 48–58). This is the lightest δ¹³C value of any fish species studied (ibid.: fig. 6), and presumably reflects the origin of the carbon from a rocky substrate, where the plants the fish consume get their carbon from, rather than from seawater. Any humans relying on this species for a substantial amount of protein in their diet would have a bone signature that would appear to be terrestrial rather than marine. The δ¹⁵N is also very low (ibid.: fig. 8), and the same warning applies. However, the δ³⁴S signature of greenbone is typical of all other marine species studied (ibid.: fig. 10), suggesting that seawater is the main source of sulphur in its diet.

A sample of 306 specimens of greenbone was obtained from a number of places around the northern coastline of Cook Strait New Zealand between 1991 and 2001 (Fig. 3).

Figure 3. 

Three locations along the northern coast of Cook Strait where greenbone were netted for the comparative osteological collection in the present study. The most northern location is in the vicinity of Waikanae. The next cluster of red dots is around Wellington Harbour, Orongorongo, Turakirae and Ocean Beach. The southernmost two red dots are at Cape Palliser.

Personal catches were made in Palliser Bay, heads were available from commercial fishers at Waikanae, and the Ministry of Fisheries (MoF) carried out research on set net selectivity between Wellington and West Palliser (Dunn and Paul 2000: fig. 2). All specimens were taken with gill nets, as discussed later. The standard fork length (FL, Fig. 4) was taken to 1 mm precision, and the whole ungutted fish weighed to 1 g precision (BWT).

Figure 4. 

Definition of three different ways of measuring live fish length. Image source Paul (2000: 110). Courtesy of Larry Paul.

Each head was removed for later processing, and bagged and labelled with its catalogue number. The bones used for measurement are five paired cranial bones: the dentary, articular, quadrate, premaxilla, maxilla, and upper and lower pharyngeal bones. These bones have been used for many years to quantify prehistoric fish catches from archaeological sites in the Pacific and New Zealand (Leach and Davidson 1977; Leach and Ward 1981; Leach 1986; Leach and Boocock 1993). The bagged heads were taken to the Archaeozoology laboratory at the Museum of New Zealand and each was simmered until the five paired bones and otoliths could be removed, cleaned of any connective tissues with tap water, and kept in open labelled Petri dishes to dry naturally. When the bones were fully dry, the lids were placed on the dishes and stored for measurement.

The bones mentioned above do not always survive intact in archaeological sites; so it is desirable to include measurements that 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 the 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 forms of bone fragment. Each measurement has a three character code, eg: RP2 = Right premaxilla second measurement. The landmarks chosen for measurement are illustrated for the right bones in Fig. 5.

Figure 5. 

Cranial elements of greenbone used for measurements (image source Leach 1997). The right bones are illustrated, but left bones were also measured. Measurements are taken between landmarks which are indicated with solid dots. The method of measurement is described in the text, and each dimension is defined in Table 1. Each measurement has a three character code, eg: RP2 = Right premaxilla second measurement.

These measurements 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.; Morales and Rosenlund 1979). In cases where the terminology ‘maximum length’ or ‘maximum height’ is used, this implies that the measuring calipers are rotated about the nominated landmarks until a maximum value is obtained. Bone measurements were taken with a precision of 0.01 mm, and experiments on repeatability showed accuracy to 0.1 mm. The definition of each measurement is provided in Table 1.

The three character code permits simple coding of measurements on plastic bags containing identified fish bones from archaeological sites. These are entered into a database according to the original archaeological provenance. The appropriate equation for estimating live FL and BWT is selected using these three character codes. Mitutoyo digital calipers 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.

In our experience, even with the benefit of an annotated illustration (Fig. 5) and formal definitions (Table 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 calipers, even when placed on the correct landmarks, can cause substantial percentage errors.

When completed, the modern database for greenbone consists of 342 rows of data for 306 specimens. and 32 columns of data for each: Specimen number, catalogue number, FL, BWT, and 14 left and 14 right measurements.

One objective of this study was to establish reliable regression relationships between bone dimension and FL and BWT, which can be used for estimating live size from archaeological bones. Two different approaches have been suggested when estimating BWT. A One-Step Model is to find the relationship between bone dimension and BWT. A Two-Step Model involves first, establishing a relationship between bone dimension and live FL, and second, the relationship between FL and BWT. Previous research of this kind, referred to earlier, has concluded that the two-step method is preferred. It is a formidable task extracting the necessary bones and measuring 30 different variables on each fish specimen, so the comparative material for the osteological analysis is limited. By contrast, in modern fisheries research, FL and BWT are often on thousands of specimens, and such data can generate far more reliable equations linking the two. Some researchers measure the gutted weight of the fish (GWT) to avoid variation due to stomach content.

The relationship between live FL and BWT

Several types of curve fitting procedures have been explored when trying to find an acceptable link between biometric variables such as length and weight. The general equations for estimating Y from X are as follows (A = constant, B = slope):

Linear fit Y = A + B * X

Exponential fit Y = A * exp(B * X)

Logarithmic fit Y = A + B * ln(X)

Power Curve fit Y = A * X**B

Cubic fit Y = A + B * X**3

Each of these models has been used in studies of fish bone dimension and live characteristics in earlier research cited above. In the case of bone to live FL, the power curve is preferred. On the relationship between length and weight, one view is that a cubic function has an arguably sound theoretical basis. This is traceable back as far as Galileo (Frose 2006: 241). However, biometric forms do not really conform to a box with rectilinear sides. As Ritchie (1969) points out, a cubic relationship assumes that growth is isometric – that is, that the growth of any one part of an organism reflects the growth rate of the whole body. However, many studies of fish have shown that the growth rate is allometric rather than isometric, and that a logarithmic model may be more appropriate (Ritchie 1969: 105). In the case of greenbone, some researchers who have studied the growth parameters of this species have provided three best fit curves – one for juveniles, and one each for males and females, all being linear equations. However, given that this species is a protogynous hermaphrodite and changes sex from female to male at about 350–400 mm (Trip et al. 2011: 176), it is a moot point as to the usefulness of distinguishing male from female when finding best fit allometric equations. This is especially important with archaeological specimens where the sex of bones is not known. At worst, the effect of not being able to determine sex from archaeological specimens is to widen confidence boundaries (increase standard error of estimates).

The results of earlier attempts to link FL with BWT of greenbone are summarised in Suppl. material 1: table S3, and plotted in Fig. 6. In the case of Crabb’s1993 study, no equation was provided; however, his data points are plotted to indicate that not all measurements conform to a simple growth curve. The circumstances of the measurements in each of these studies was not uniform. Some were taken on board a fishing vessel and may perhaps be less reliable. In the present study, measurements were taken in a laboratory with great care. By far the largest collection of greenbone (902) that has been studied to date was collected over a long period by spearing (Trip et al. 2014: 868) and included very small and very large specimens. FL measurements were made to 1 mm and BWT to 1 g (small individuals to 0.1 g). FL and BWT data from this research was made available for this present project by K. Clements (2024 pers. comm.). This data was also useful for establishing a relationship between gutted weight (GWT) and ungutted weight (BWT), shown in the bottom of Fig. 6. The line of best fit in this case was linear: BWT = 1.186 × GWT +2.89 ± 48.8 (units g). We also used this data (N = 719) to establish a relationship between FL and BWT (Suppl. material 1: table S3).

Figure 6. 

A–E: Four attempts at modelling FL and BWT of greenbone (A to E), and the relationship between GWT and BWT (F). A. From Ritchie (1969) and his suggested linear best fit lines for male and female specimens. B, C. Red dots from Paul et al. 2000 and their suggested logarithmic line of best fit. Green dots superimposed are from Crabb’s study (1993), no line fitted. D. Yellow dots this present analysis. E. Yellow dots from Trip et al. (2014), see text. F: red dots linear relationship between GWT and BWT, see text.

The curve established using Trip’s data (Clements 2024 pers. comm.) is the most reliable for use in the two-step model for archaeological bones, mentioned earlier, and these parameters are used for BWT estimates in this current paper, and appear in the top of Table 2. However, we have presented contrasting results from other studies above because we will later use these to trace the implications of the different equations in Suppl. material 1: table S3 on archaeological research. It may be noted in passing that these growth parameters have a role in modelling maximum sustainable yield (MSY) when determining commercial fishing quotas in New Zealand under the quota management system (QMS).

Table 2.

Download as

CSV

XLSX

Power curve constants for estimating live FL from bone dimensions of greenbone. The first equation is for estimating BWT from FL. The others are for estimating FL from bone dimensions. Units are g and mm.

Measurement	Constant	Power	Standard Error	R
FOR–WGT	7.3989x10-7	3.4958	49.6	0.99
LD1	30.70	0.8695	16.6	0.95
LD2	51.34	0.8445	22.0	0.92
LA1	47.91	0.8569	16.6	0.95
LA2	39.65	0.8781	19.2	0.94
LQ1	57.50	0.7423	17.7	0.95
LQ2	121.52	0.7610	20.8	0.93
LP1	37.06	0.8229	14.7	0.96
LP2	13.23	1.1134	26.7	0.88
LP3	43.53	0.9607	21.3	0.92
LM1	50.33	0.7770	16.3	0.96
LM2	120.89	0.7026	29.2	0.85
LC1	67.25	0.8408	19.4	0.94
IC1	28.93	0.8299	16.6	0.95
IC2	73.83	0.7066	18.3	0.94
RD1	30.17	0.8764	16.4	0.96
RD2	51.17	0.8447	21.8	0.92
RA1	46.89	0.8649	16.7	0.95
RA2	39.08	0.8833	19.4	0.94
RQ1	56.85	0.7482	17.9	0.95
RQ2	119.99	0.7681	20.6	0.93
RP1	35.60	0.8359	14.3	0.97
RP2	12.83	1.1225	26.4	0.88
RP3	43.02	0.9637	21.6	0.92
RM1	50.75	0.7737	17.6	0.95
RM2	117.69	0.7185	29.6	0.85
RC1	66.24	0.8488	19.4	0.94

The relationship between bone dimension and live fish FL

The various bone dimensions and method of measurement have already been described, and further details are provided in Suppl. material 1. Earlier studies of fish biometrics have shown the power curve fit to provide the best results for estimating live fish FL from bone dimensions (Leach et al. 2001; Leach et al. n.d.), and this model is followed here. When the analysis was carried out on each pair of bone dimension and FL, a series of constants was obtained for each power curve fit. These parameters are given in Table 2, together with the correlation coefficient R, and the standard error of the estimate. The curves of best fit for each measurement and the associated bone measurements are illustrated in Figs 7, 8.

Figure 7. 

These graphs show the individual bone measurement for each greenbone specimen plotted against the FL, together with the power curve fit (bold line). The dashed lines show the 68% confidence margins.

Figure 8. 

These graphs show the individual bone measurement for each greenbone specimen plotted against the FL, together with the power curve fit (bold line). The dashed lines show the 68% confidence margins.

A concrete check of the method of estimating FL and BWT was carried out using the present comparative collection of 306 specimens, and the results are provided in Suppl. material 1: table S4 for the smallest specimen, one in the mid-range, and one of the largest individuals. The FL was estimated from the RP1 measurement, and BWT was estimated from the estimated FL. The power curve equation Y = A*X**B was used for both estimates, using the relevant constants in Table 2. Thus, the two equations are: FL = 35.60*RP1^0.8359, and BWT = 7.3989E^-7*FL ^3.4958. An alternative computational way of writing these two equations would be: FL = 35.60*exp(0.8359*(ln(RP1))), and BWT = 7.3989E^-7*exp(3.4958*(ln(FL))). The standard error of the estimates for RP1 to FL is ± 14.3 mm (21^st row in Table 2) . The errors estimating FL ranged from -29 to +7 mm in this simple test. The standard error for estimating BWT is ± 49.6 g (1^st row in Table 2). The errors in the three test specimens ranged from -71 g for the largest specimen to +8 g. Although these errors are small in this simple test, we should expect some rogue values when estimating BWT in any species of fish because there is far greater variation of condition among large individuals than smaller ones. The mean residuals across all 306 specimens in the comparative collection were: FL 0.2 mm, and BWT -56.0 g.

A survey of 126 archaeological sites in New Zealand with a total Minimum Number of Individuals (MNI) of 40,433 for all species of fish revealed that greenbone is represented with a total MNI of 2,967 (Leach 2006: 334). That is the sixth most abundant of 36 families of fish represented in these sites. The top six families are represented by barracouta (Thyrsites atun Gempylidae), blue cod (Parapercis colias Pinguipedidae), snapper (Chrysophrys auratus Sparidae)², labrids (Labridae), red cod (Pseudophycis bachus (Moridae), and then greenbone.

The distribution of greenbone in New Zealand archaeological sites is shown in Fig. 9. This indicates that the species is present in catches from the far north to the far south, but far more common around Cook Strait, Stewart Island (Foveaux Strait), and in the Chatham Islands.

Figure 9. 

Relative abundance of greenbone in 126 archaeological sites around New Zealand and Chatham Islands. Plain white circles are sites where the species is absent. The size of the black circles indicates the relative abundance of the species in any one site (logarithmic scale). The locations of nine archaeological sites discussed in the paper are indicated with arrows.

Nine archaeological collections which had bones from greenbone in the Archaeozoology Laboratory at the Museum of New Zealand were chosen for detailed analysis. Their locations are marked on Fig. 9. The methods used for establishing MNI and relative abundance have been outlined earlier, and the results are extracted from Leach (2006: 331–333) and presented in Suppl. material 1: table S5 and illustrated in Fig. 10. The presence of greenbone in these sites varies considerably. This not only reflects the latitudinal range of the species, but also local environmental conditions. Greenbone thrives in habitats rich in fields of seaweed. A thumbnail sketch of each of these archaeological sites is given below.

Figure 10. 

Relative abundance (percent of total MNI) of different fish in the nine archaeological sites studied.

The influence of sieve size on recovery

In recent years much has been written about potential bias in species composition and their relative abundance during analysis of archaeological sites due to questionable recovery methods (Campbell 2016; Campbell and Nims 2019; Nims et al. 2020). The main culprit cited is either a failure to use sieves or using a mesh size which is considered too large. A seminal study by Seersholm et al. (2018) highlighted this issue when DNA was recovered from bulk samples of non-diagnostic bone fragments resulting in species being identified that are not normally seen in tabulations from archaeological sites. The question therefore arises whether the reconstructed size-frequency curves provided here for greenbone have been influenced by poor recovery methods. With the exception of the Kokohuia site, all archaeological sites described in this paper were excavated by students of the senior author using recovery methods defined in 1969 during the Palliser Bay expedition (Leach and Leach 1972). Biological specimens above ¼ inch (6.35 mm) were retained, and select samples above 1/8 inch (3.2 mm). This recovery protocol was used for the Black Rocks samples described in this present paper (Anderson 1973: 59). The same was followed for the Mana excavations (Horwood pers. comm. 2024). In the case of Kokohuia, 4 mm sieves were used (Leach et al. (1997c: 102)). In the Chatham Islands, everything was passed through a 1/12 inch (2.1 mm) sieve to remove fine sediments, and the remainder residue then sieved through 4 mm (McIlwraith 1976: fig. 9; Sutton 1979b: 338). The bones measured from the Waihora site permit us to estimate the size of any greenbone that may have been missed with a sieve mesh of 6.35 mm. Of the 4,464 bones measured, 106 bones were recovered that were smaller than this sieve size, ranging from 3.47 to 6.36 mm. These had estimated FL of 288–318 mm. This analysis was repeated with the measurements of only the smallest bone, the superior pharyngeal, producing the same result. We can therefore be confident that any greenbone in the sites that were not recovered would be < c.318 mm FL. This cut off value is marked on Fig. 11 with a dashed red line. In all seven cases shown in Fig. 11, the red line is well to the left of the downward slope of the size frequency curve. This indicates that any fish that potentially may have been missed by the sieves would be small specimens that would be able to swim through the net mesh size being used to catch this species. This conclusion is reinforced by Clements’ data, which deliberately included many small individuals that could be speared. Of the 737 specimens the total BWT of specimens below 318 mm FL was only 7.8% of the total BWT. We conclude from this analysis that the main benefit of using sieves < ¼ inch would be to recover evidence of very small species of fish such as anchovies. Such information plays a major role in palaeoecology and environmental science, and sheds light on mass capture of small species such as New Zealand whitebait (Galaxias spp.). For the most part, species recovered from < ¼ inch mesh sieves has minor significance compared with the total MTWT calculated from larger species.

Figure 11. 

Size-frequency distributions of greenbone FL reconstructions from cranial bone dimensions, ordered by decreasing mean FL from top down. The curves for Mana Island 1 and Kokohuia are not shown, as only seven bones were able to be measured from these two sites. The heavy black dashed line is the minimum legal size for greenbone in New Zealand. The heavy red dashed line is the maximum fish size that could be missed using a 6.3 mm sieve (see text).

FL and BWT reconstructions of greenbone in the nine sites

The cranial bones of greenbone from these nine archaeological sites were measured as described earlier. A summary of the bones able to be measured compared with the Number of Identified Specimens (NISP) and MNI in the sites is provided in Suppl. material 1: table S6. This tabulation shows that the most common measurable bone is the inferior pharyngeal, partly because of its durability. This is followed by the dentary and premaxilla. The left/right ratio of bones measured is 1.03:1, compared to 1.02:1 of the NISP ratio for all sided greenbone cranial anatomy, suggesting negligible bias. The total number of measurements was 5,514, which is 41% of the bones identified as greenbone. The greenbone NISP/MNI ratio is 4.1:1, which is close to the ratio for all taxa identified, 4.3:1, again showing negligible bias.

Once all the bones were measured, the FL of each specimen were estimated using the constants provided in Table 2. In a second step, the live BWT of each was then estimated from the FL, using the constants at the head of Table 2. In cases where more than one measurement was possible, the largest dimension was recorded. When the estimates of original fish size have been completed, size frequency diagrams can be constructed, representing the original catches. This is presented in Fig. 11, and the dispersion statistics in Fig. 12. This diagram permits easy identification of statistically significant changes between any one assemblage and another. For example, the greenbone from BR4 are significantly smaller than those from BR2 and BR3, as their standard errors do not overlap at 95%. The size of greenbone at CHA and WAIH are not significantly different, but both are significantly smaller than those from CHB. Any departure from normal distributions is also shown in Fig. 12. Departure from normal is when g1<>0 and g2<>3, but in both cases the values of w1 and w2 in Suppl. material 1: table S7 need consulting. These are normal deviates with zero mean, so a value ≥ 1.95 is significant p = .05 and ≥ 2.57 is significant at p = .01 (Rao 1952: §5f3.1).

Figure 12. 

Dispersion statistics of the assemblages of the reconstructed greenbone FL, with 68% and 95% standard error ranges. BR2 is not shown in the right hand plot as the skewness value is off scale at 1.47.

Examination of the values for skewness shows that only the greenbone catches from BR3, Kokohuia and the two Mana Island sites are normally distributed. With the exception of BR2, the others display modest positive skewness. One very large fish was caught by the people at BR2, with estimated FL of 685 mm, and this is the reason for a large value g1 (offscale in Fig. 12). Three kurtosis values are significant: BR2 and Waihora are leptokurtic, and CHB is platykurtic. There are hints of multi-nodality in some of these size-frequency plots, which may suggest more than one harvesting method, such as nets with different sized mesh.

There are only sparse ethnographic records of Māori fishing for greenbone, but some form of hoop net set in weedy areas has been recorded. Other methods cannot be ruled out and set nets are also a possibility. During the gathering of modern greenbone for the comparative osteological collection, we were fortunate to acquire specimens from a separate research project of set net selectivity. The purpose of that project was to assess the suitability of current mesh size limits used in the commercial and recreational fishery. The legal minimum mesh size of 108 mm was chosen for selecting a minimum legal FL of 350 mm (Dunn and Paul 2000: 3). They concluded that the current 108 mm mesh size had a probability of catching sub-legal sized fish of less than 0.5% (ibid.). This follows an earlier similar project but with different objectives by Hickford et al. (1997). They note that fish with smooth fusiform shapes such as greenbone are selected predominantly by size in gill-nets. If their maximum girth is less than the perimeter of the mesh they can swim through the net. However, if the opercular girth is greater than the mesh perimeter they are caught. In the case of greenbone, their sinuous swimming motion and weak pectorals do not allow them to swim backwards out of a gill net or to stop quickly. In both studies, set nets were constructed with panels joined together, each panel having a different mesh size. The same pattern of increasing mean FL with larger mesh size was shown in both studies. The catch size frequency by mesh size for both studies is illustrated in Fig. 13.

Figure 13. 

Distribution of greenbone FL from Waihora, together with the distribution of the FL of fish captured with different mesh sized set nets. The red lines were calculated from the original catch data used by Dunn and Paul (2000), and were provided courtesy of Larry Paul (pers. comm. 2000). The blue lines were calculated from data in Hickford and Shiel (1997: 252: table 1). In three cases, very low values of N gave SD which are considered unreliable and were replaced with the mean of the four red curve SDs of 37.5 mm. The minimum legal size for greenbone is 350 mm, shown as the vertical dashed line. The minimum legal mesh size for greenbone set nets is 108 mm.

Dunn and Paul (2001) used their catch data to model gill net selectivity, and produced a series of useful probability curves of FL for each mesh size (ibid.: 17, fig. 5), which were then used to justify maintaining the current legal minimum mesh size at 108 mm. Ultimately, the veracity of this requirement rests on the biological theory relating to recruitment and replenishment of the stock for this species. Current estimates are that this species is sexually mature at a mean age of 1.2 years (mean FL = 241 mm) in northern waters, and 3.9 years and 252 mm in southern waters (Trip et al. 2014: 871), and that there is an exponential relationship of increasing abundance with higher latitude (ibid.: 872, fig. 5). It has also been shown that while females spawn between 250 and 475 mm FL, males only spawn over 350 mm (Trip et al. 2011: 184, fig. 5). There is little doubt that having a minimum mesh size limit for gill nets enables sexually mature female fish to swim through a gill net and thereby assist with maintaining a reproducing population of greenbone. It is not quite so clear whether the current choice for legal limit of 108 mm mesh is fully supported by the science, and that a smaller mesh size may achieve the same objective. Examination of Fig. 11 certainly shows that pre-European Māori were taking many fish which, by today’s standards, were undersized, and certainly among those individuals that were reproducing. This raises an important issue concerning the care and management of marine ecosystems, and what are suitable conservation measures to protect fecund members of a population. The archaeological dimension of this and the ancient New Zealand fishery has been discussed in detail elsewhere (Leach 2006: 223–224; 274–304), where it was suggested that the ‘take everything, regardless of size’, may not be as damaging to coastal ecology as is widely believed.

From the foregoing, the question arises – is it possible to estimate the mesh size of nets used in the pre-European period from the reconstructed FL in each archaeological site considered here? Studies of entanglement rates in set nets have shown that greenbone are almost always gilled or wedged in nets, and lack of entanglement is attributed to their soft fin rays and small scales (Hickford and Schiel 1996: 673). This will doubtless be the main reason why capture using set nets of different mesh sizes has been shown to be so selective by fish size. That is, for any one mesh size, small fish swim through the net and larger ones evade capture in the gill. This is proven by the selectivity study, illustrated in Fig. 13. In short, gill nets for greenbone are size selective. It will be recalled that Fig. 11 shows the size-frequency distribution of pre-European Māori greenbone catches in Palliser Bay, Mana Island, and Chatham Islands. Even cursory examination of these distributions reveals two major points: 1: a high degree of fish, which, by modern management standards, are undersized; and 2: The broad range of FL in each case could not possibly have resulted from one mesh sized gill net. Moreover, for the largest catch (from Waihora) there are no obvious signs of multi-nodal elements in the distribution, which might be decomposed into those appropriate to multiple mesh sizes (Macdonald and Pitcher 1979). A case could be made for multi-nodality in the catches at CHA, CHB, and MANA2 (see Fig. 11). In addition to gill nets, there is an alternative which would account for the broad ranging size-frequency distributions observed. That is, the use of drag nets with small mesh. Such nets, pulled through weedy areas, capture everything that cannot escape. Peter Buck, during his ethnographic research in the Bay of Plenty, observed several variants of these, including large scoop nets for use in weedy areas (Hiroa 1926: plate 105). The description by Knox of greenbone fishing off Mana Island, cited earlier, certainly showed that setting hoop nets across a tidal range was effective.

It will be recalled that we earlier drew attention to the fact that several methods have been proposed to estimate live BWT from FL (Suppl. material 1: table S3). Three of these suggestions were tested in two ways; firstly, the ability of each to estimate the correct BTW values recorded in the comparative collection of 306 fish, using the parameters established from our comparative collection. In this case, the mean estimated BWT was 1180.50, compared to the correct BWT of 1188.78, a difference of -8.3 g. Secondly, when using the parameters suggested by Paul et al. (2000: 9). The mean estimated BWT was 1235.98 g, a difference of 47.2 g from the correct value (N = 212). Finally, when using Clements’ data, the mean estimated BWT was 1248.03, a difference of 59.2 g from the correct value (N = 719). The veracity of the three equations was also tested with the large archaeological sample from the WAIH site in the Chatham Islands (N = 4,464). Again, firstly using the equation derived from our comparative collection, the mean BWT for the Waihora catch was found to be 773.93 g. This compares with 786.69 g using the Paul et al. (2000: 9) equation, and 768.92 g using the data supplied by Clements. Although the difference in errors of the last two is quite similar, Clements’ data is considered superior on the grounds that there are considerably more both small and large individuals than in the other two sets (see Fig. 6). These parameters were therefore finally chosen, and appear in the first line of Table 2.

The estimates of FL of each specimen in the archaeological sites was converted to BWT using these constants, and the BWT values are summarised for each archaeological site in Suppl. material 1: table S8.

The total BWT of fish represented in each of the sites examined is, by itself, not a very useful measure. As with every excavation, whatever is recovered is only a snapshot of the activities of a group of people. In addition, a size-frequency distribution of the BWT of individual fish caught adds little more than the FL distributions. A more meaningful measure is the mean BWT of the sample of fish taken. This is important information when evaluating the quantitative value of nutrients to the diet of prehistoric people. Smaller fish have less BWT per fish, and as a result contribute fewer nutrients than larger ones.

The mean BWT of greenbone at each site is provided in Suppl. material 1: table S7. It may be noted in passing that the average BWT of greenbone specimens caught in Palliser Bay and Mana Island are considerably greater than are those from the three sites in the Chatham Islands. The reasons for this are unclear. Studies of growth-by-age patterns have revealed a latitudinal cline in New Zealand (Trip et al. 2014: 870), but Chatham Islands were not included in the survey. It is possible that sustained regional harvesting over hundreds of years resulted in reduction of mean fish size. Further archaeological research on a series of dated sites would be needed to evaluate this.

Although Polynesians normally eat considerably more of any one fish than would be common among Europeans today, there is, of course, always waste. Smith has suggested adopting a rule of thumb value of 70% useable meat weight (MTWT) of the mean body BWT for each taxon (Smith 2011: 12). This is far more than modern European utilisation of fish, where MTWT (the food) is only c. 34% of BWT, the rest being subject to complex industrial processing. Although Smith’s suggestion is perfectly reasonable, Vlieg has provided a valuable database of the mean composition of each portion of the anatomy of many New Zealand fish species, and each has a unique macronutrient ratio (MNR), given in Suppl. material 1: table S2. It therefore seems sensible to evaluate the full nutritional benefit of each of these archaeological fish catches, by taking advantage of Vlieg’s research. As has been described elsewhere (Leach and Davidson 2023; Leach et al. n.d.), in some areas of New Zealand where carbohydrate-rich foods were unreliable, the quest to find caloric energy from fat assumed considerable importance. In this respect, lipids are more concentrated in the skin and entrails of fish rather than the meat, and there is abundant ethnographic evidence that early historic Māori favoured these body parts more than the meat, especially the viscera (Leach et al. n.d.: 20). A conservative assessment of the benefit a pre-European person might obtain from the different body parts of greenbone is that the percent usable food would be c. 40% from the head, perhaps as much as 90% from the viscera, from the frame as little as 10%, as much as 80% from the skin, and 100% from the fillet part. Assuming these assessments are reasonable, we can then evaluate the macronutrients from greenbone at each of the nine archaeological sites being considered. The details for this are presented in Suppl. material 1: table S8.

Since the mean fish BWT in each site is different, the MTWT is different, and for this reason the total caloric energy of each macronutrient in each fish is also different. Thus, at Kokohuia, the mean caloric energy per fish is only 335 kcal, compared with 907 kcal at Mana Island 2. This illustrates the benefit of reconstructing the size of each fish specimen from bones rather than working with one average value for a species.

The MNR is easily calculated from the data in Suppl. material 1: table S8. The caloric value per fish from protein, fat and carbohydrate at BR2 are 599.7, 164.6, and 7.0; so the MNR = 78:21:1. As should be expected, the MNR is the same value for all archaeological sites in Suppl. material 1: table S8. The MNR for a modern European eating only the fillet of greenbone would consume food with an MNR of 89:10:1, illustrated in Fig. 14. This is very wasteful of valuable nutrients. On the right is the MNR which is more likely to represent the behaviour of pre-historic people, consuming more than just the fillet. Such an efficient way of mining nutrients from fish results in less demand on finding sources of carbohydrate to achieve a balanced diet.

Figure 15. 

The nine archaeological sites are differentiated by the caloric value of each macronutrient which each fish contributes to the economy (from data in Suppl. material 1: table S8).

At a macro level, the foregoing permits the calculation of the total nutritional value of greenbone to the people at these nine archaeological sites studied. For example, the greenbone MNI at BR4 was 62 (from Suppl. material 1: table S8). The mean BWT of greenbone from the site was 836.95 g (Suppl. material 1: table S7). Thus, the total BWT is 51,891 g. However, not all of this usable. Our assessment is that the usable protein would yield 30,834 kCal, with MNR of 78:1. The macronutrients represented by the greenbone catch at the nine sites studied is illustrated in Fig. 15. This highlights a point that the highest caloric reward for effort is from the largest fish, because by far the greatest caloric value is from fat (9 kcal/g) compared to protein and carbohydrate, which are only 4 kcal/g. Thus, the fish from MANA1, MANA2, BR2 and BR3 are far more nutritious than those from the other five sites. Although it may seem a platitude that small fish contribute less to an economic system than bigger fish, there are implications concerning optimal foraging strategies. Preferential choosing of larger specimens provides greater caloric return for effort. The evidence from the nine sites studied here does not support such a harvesting strategy. Instead, the data favours capture and keep regardless of size.

Figure 14. 

Macronutrient ratios of greenbone showing available nutrients available to humans. On the left, is when humans consume only the fillet. In the centre, is the whole fish value. On the right, is when humans consume 40% of the food in the head, 90% from the viscera, 10% from the frame, 80% from the skin, and 100% from the fillet.

The New Zealand greenbone, Odax pullus, is one of the few vegetarian fishes in New Zealand and is highly regarded as a table fish with firm thick fillets. It is easily caught with gill nets set in weedy areas around the New Zealand coastline. Although the fish is present throughout New Zealand, it is increasingly common at higher latitudes. This is reflected in studies of archaeological sites with greater greenbone abundances in assemblages from southern areas. In the early 20^th century, this fish was not popular among Europeans because of a pronounced green colour when filleted. This was later shown to be due to the presence of iodine in the flesh at very high levels, and was beneficial at warding off goitre, a health problem until iodised salt was introduced. It has also been shown that greenbone has low levels of omega 3 of the available lipids, and very low levels of unsaturated fatty acids compared with other fish species. An unusual feature of greenbone is the isotope values in the flesh. Only δ³⁴S reflects the seawater origin of this fish. Both δ¹³C and δ¹⁵N reflect a terrestrial origin, presumably because the species of kelp eaten derive much of their chemistry from the rock substrate they are attached to. This has implications when studying collagen isotopes from human bone.

Although Māori in the historic era caught greenbone, along with many other species of fish, there is a surprising lack of detailed ethnographic descriptions which specifically relate to this species. A detailed ethnographic study by Peter Buck and Elsdon Best of the netting activities of 20^th century Māori barely mentions greenbone (Hiroa 1926). By contrast, bones from this species are quite common in archaeological sites in both New Zealand and the nearby Chatham Islands. Previously there was no method by which the live size of these archaeological fish specimens can be estimated from the bones they leave behind. Consequently, there is no way to estimate the BWT and other nutritional details which these fish contributed to prehistoric diet and other aspects of palaeo-economics. This present research was aimed at filling this gap by establishing the relationship between bone dimensions and live FL and BWT for this species.

We began the metrical research by collecting 342 specimens of greenbone in the Cook Strait area using gill set nets. Fresh measurements were taken and cranial bones extracted. Specialised pharyngeal bones from this species are especially durable and very useful for estimating live dimensions. Power curve relationships were established between bone size and FL. The BWT was estimated from these FL using a power curve derived from a much larger sample of fish captured and studied by other researchers.

An examination of 126 archaeological sites, where fish remains had earlier been studied, revealed that the greenbone family of fishes was the sixth most abundant, and the bones from nine of these sites were available for study: one from Northland, five from the Cook Strait area, and three from the Chatham Islands. The bones from these were measured and power curve equations used to estimate the original live FL and BWT of each fish. Size-frequency analysis suggested non-normal characteristics in five cases. The largest and most reliable sample had significant +ve skewness and was leptokurtic, hinting at more than one capturing method. It was noted that pre-European Māori and Moriori were capturing many fish, which by modern standards, are well below minimum legal size. We did not explore this subject further as it has been well covered elsewhere. Since this species is primarily caught by gill set nets in historic times, these variable size-frequency curves raise the possibility of estimating the mesh size of prehistoric nets. Greenbone do not have sharp spines and are infrequently caught by tangling, so the size of fish caught is largely determined by the size of the mesh. Contrary to these suggestions, we cannot be sure that gill nets were ever used to catch this species in pre-European times, as spearing, drag nets, and hoop nets are also possible, and several may have been in vogue.

Estimates of the mean BWT for each of the nine archaeological sites displayed considerable variation, with catches in the Cook Strait region having heavier BWT than those in the Chatham Islands. It is common among Polynesians to consume more than just the fillet of fish caught, and the viscera, skin and head are especially nutritious, because lipids are concentrated in these components. We estimated the nutritional value of the usable meat (MTWT) in these archaeological catches, taking this into account. We found that the MNR of greenbone for Polynesians would be c. 78:21:1, compared to the European custom of only eating the fillet, which has a MNR of 89:10:1. Such a behavioural difference is significant when trying to achieve a balanced diet in economic systems where there is a super abundance of protein available, and limited carbohydrate. This study revealed a considerable range in the energy deriving from the three main nutrients by fish size from the nine archaeological sites. The energy from fat per fish at one of the Mana Island sites was nearly three times that at the Kokohuia site in Northland. This shows the benefit of targeting larger fish as an optimal harvesting strategy.

The research on greenbone was part of the “Bridge and Barrier Research Project”. The authors would like to 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. Larry Paul at NIWA kindly contributed fish specimens from his study of gill net selectivity, which helped to swell our comparative collection, and also provided useful statistical data on fish and net mesh sizes; Kendall Clements, Auckland University, generously provided greenbone measurements from one of his research projects; Peter MacDonald, McMaster University, kindly carried out the mathematical analysis required for estimating gill net mesh sizes from archaeological fish catches; Chloe Steer and Damien Hewett, Ministry for Primary Industry, provided important historical fish measurements from the Ministry of Fisheries database; Garry Law, Law Associates Ltd., gave useful advice on mathematical issues; Jim Samson, Archaeozoology Laboratory, Museum of New Zealand, and Greg Walter, Department of Archaeology, University of Auckland, helped a great deal with processing fish specimens; and Andy Dodd, Subsurface Ltd., contributed the map of the lower North Island that appears in this paper. To all these people we offer grateful thanks for their generous assistance. Finally, modern-day science would be greatly limited were it not for the diligence of reference librarians hunting out copies of obscure literature for consultation. In this respect we thank Martin Lewis and Amy Phillips, reference librarians at the Museum of New Zealand. The authors thank reviewers of the manuscript for constructive comments.