Volume 101 1992 > Volume 101, No. 2 > The Papatowai site: new evidence and interpretations, by Atholl Anderson and Ian Smith, p 129-158
THE PAPATOWAI SITE: NEW EVIDENCE AND INTERPRETATIONS
The large Archaic phase site at Papatowai, in the Catlins district of south Otago (Fig. 1), has occupied an important place in New Zealand's prehistory as one of a few localities in which evidence indicative of the earliest settlement period has been uncovered. It was at Papatowai in 1938 that artificially perforated moa eggs, a key type of early Archaic phase material culture, were first identified and examples of two other important types of that assemblage, imitation whale teeth pendants and necklace reels, were first recovered in association with evidence of moa-hunting (Duff 1956: 268-9). Until recently, Papatowai also held the distinction of having been radiocarbon-dated more extensively than any other Archaic site, and the results included some which were widely regarded as among the earliest in New Zealand's prehistory (Anderson and McGovern-Wilson 1990, Anderson: in press).
In addition, Papatowai has played a central role in the description and explanation of economic change within the early era. Analysis of faunal remains in relation to stratigraphy and chronology at Papatowai disclosed more clearly than at any other site a dramatic decline in hunting of moas and seals and a corresponding rise in the exploitation of small game, including fish and shellfish. Therefore, just as the Palliser Bay project (Leach and Leach 1979) advanced an influential model of economic adaptation and change for the central region of New Zealand, in which marginal horticulture occurred, so the archaeology of the Catlins, and of Papatowai in particular, provided the orthodox model of economic adaptation and change in southern New Zealand (Lockerbie 1959, Hamel 1977, Anderson 1983, Davidson 1984).
In this paper we reconsider the archaeological basis of that model at Papatowai. Three forms of evidence are critical to this reassessment: the stratigraphy, the stratigraphical distribution of faunal and artefactual remains, and evidence concerning the period and length of occupation. We examine critically earlier data and interpretations in the light of our own investigations at Papatowai in 1990.
THE PAPATOWAI SITE
The rich middens at Papatowai were described first more than a century ago by Sir Thomas McKenzie who fossicked among them briefly, finding some “flint” implements among “…immense quantities of shells, fish, bird,- 130
moa, human and dog bones …” (Otago Daily Times, January 30, 1889). Excavations on a small scale were begun by David Teviotdale in 1933 but his main seasons were in 1936 and 1937 (Teviotdale 1937, 1938a, 1938b, and diaries summarised in Hamel 1977:199-228. There is another report, evidently the basis of the 1937 paper, in Buick 1937:121-31). Leslie Lockerbie, who had worked with Teviotdale, undertook his own excavations in 1941 and 1956 (Lockerbie 1950, 1953, 1954, 1959), and Jill Hamel excavated an additional area in 1971 (Hamel 1977:229-53; 1978). Our own excavations occurred during Easter, 1990.
These various investigations covered only a small part of a large site of unknown extent (Fig. 2). The main site area, so far as can be determined, lies on an undulating terrace of low dunes, approximately 120 m by 80 m in extent at the distal or western end of the Tahakopa sand spit. There is no very obvious - 131 arrangement to the dunes, but most seem to run approximately north-east-south-west (Teviotdale 1937:136 sketch map). The low relief is, in part, a consequence of the filling of dune swales up to two metres deep with cultural debris from the period of occupation. To the east, the main site area is bounded by dunes up to 12 m high.
Eroding midden can be traced discontinuously today along the river edge of the site and it is most conspicuous in the 30 metres immediately south of the high dunes. It is in this area that excavations have been concentrated, but disturbed ground is apparent elsewhere on the terrace as well. The only place in which we felt confident that intact deposits would remain was where the old coach road, formed at the beginning of the century, ran down from the high dunes on to the terrace, since it appeared that the pre-European ground surface had been buried in that area by sand pushed down during construction.
There are probably other intact areas as well, although they are difficult to find under the forest litter. There is evidence that the site runs under, or is patchily represented among, the high dunes. Middens are exposed on the forest floor up to 200 m east of the main site area and Teviotdale's diaries record excavation of several outlying areas which contained abundant moa bone and are probably parts of the main site (areas SP and JP, described by Hamel 1977:218-20).
PREVIOUS EVIDENCE AND INTERPRETATIONS
Before describing our excavations, it is pertinent to review observations made by earlier investigators and the interpretations that have been drawn from them. One important matter is whether the Papatowai site had a single stratigraphical structure, perhaps one which is also found at other early sites in the Catlins district, such as Pounawea and Hinahina (Lockerbie 1959).
Earlier Stratigraphical Observations
The stratigraphy of previous excavations has been discussed in detail by Hamel (1977). She notes that Teviotdale observed a stratigraphic dichotomy between an upper shelly layer and a lower black layer rich in bones, though in several places he recorded the existence of an intermediate brown sand layer. Lockerbie reached a similar conclusion in his synthesis of Catlins prehistory (Lockerbie 1959), as Hamel (1977:201) explains:
It incorporated a black layer representing moa hunting with [sic] a varied kit of adzes, many of them massive and skilfully flaked, unbarbed fish hooks, large silcrete blades and ulu-type knives … an intermediate layer with an ashy matrix showing a decline in moa hunting and changes in the proportions of artefact types; and a top- 132 - 133
layer with barbed fish hooks, fewer and smaller flake knives, little evidence for live moas and abundant shells and fish bones.
Hamel also sought to construct a single stratigraphic model for the site. The process involved setting aside the Lockerbie sequence and “… establishing the equivalence of given layers from different sections to form components, and then interleaving the components to form the Papatowai sequence” (Hamel 1977:202). This was, she conceded, an uncertain procedure dependant upon finding common characteristics.
In the event, Hamel relied on the same markers for correlation of layers as those chosen by Teviotdale and Lockerbie: layer colour (especially whether brown, black or neither), layer texture (greasy, ashy), the presence or absence of shell midden, and the relative abundance of moa or mammal bone, or of major artefact forms such as adzes and flake knives. Occasionally, radiocarbon dates were used to support particular conclusions.
The “General Sequence” developed by Hamel (1977:220-3), contained five components. At the bottom was a “Lower shell” component, represented discontinuously, and above it a “Black” component thought to be a single continuous layer. Perhaps of the same age was a discontinuous “Shell and bone” component below a discontinuous but widespread “Working floor”. The uppermost unit, which had a similar distribution to the working floor, was an “upper shell” component. Excepting minor variations, such as the absence of a lower shell component in the Lockerbie sequence, Hamel's model was essentially that envisaged by her predecessors (e.g., Teviotdale 1937:137; Lockerbie 1959:84-5).
However, it is apparent from Hamel's painstaking analysis of recorded stratigraphic sections that none of them discloses a full set of the components of the general sequence and, in fact, that she was unable to fit Lockerbie's 1956 stratigraphy into it at all. Generally only one or two of the layers of the general sequence have been identified by Hamel in each section. The “upper shell” component was missing in parts of both of Lockerbie's excavations. The “black” layer was discontinuous in the 1956 excavation, seems to have been missing between Teviotdale's areas 1 and 2, and was certainly absent in his excavation IIIb. Hamel's own excavation (TT1) also lacked the “black” layer in one of the two squares (Hamel 1977:212). South-west of these excavations were four areas excavated by Teviotdale (IIb, IId, IIe and 3). According to Hamel's (1977:214-7) interpretation of his notes, the “black” layer was missing from all of them. This was also the case at two other Teviotdale areas (SP, JP) which lie south-east of the main concentration of excavations (Fig. 2). In fact, for the 15 recorded excavations, Hamel (1977: Table 4:14), argues that the “upper shell” and “working floor” components - 134 occurred in nine cases, the “black” layer in seven, and the “lower shell” and “shell and bone” components occurred in only three each.
Apart from a certain uneasiness which must be felt in accepting a general sequence that is so seldom represented by actual stratigraphy, there are more specific grounds for concern. Firstly, since the “shell and bone” component occurred only as a single layer, except in one case (the Teviotdale excavation JP) where it was underlain by a “lower shell” component, no stratigraphic relation could be established between it and the “black” layer or the “upper shell” component. The argument for its stratigraphic position rests entirely, therefore, upon a subjective assessment about a similarity between the “black” and “shell and bone” components in quantities of moa and mammal bone (Hamel 1977:219-20). That aside, there is no reason why the “shell and bone” component may not be equivalent to “upper shell”, or any other layer, or none of them.
Secondly, there is considerable variation in the units attributed to the “working floor” component. Generally speaking, this is defined by its stratigraphic position between the “upper shell” and “black” layers, but it is described first by Hamel (1977:204) as the upper surface of the “black” layer (in Teviotdale's area 1) which appeared as a compacted “floor” strewn with numerous bone artefacts. In area 1a the “working floor” is defined by some adzes lying on top of the “black” layer, even though more were found within that layer as well (Teviotdale 1938a:27; Hamel 1977:205), and in Lockerbie's 1941 excavation, remains of fishhook manufacture were mainly found within the “black” layer (Lockerbie 1953:16-17). In most of Teviotdale's excavations, however, artefacts and worked bone were concentrated in the “upper shell” midden, as he states explicitly (Teviotdale 1937:138,139). Elsewhere, the “working floor” was defined by Hamel (1977) as an ashy layer between the “black” and “upper shell” components, or a discoloured sand lying below a shell layer (with a black layer beneath absent in some places), or simply a single ashy-shelly layer with no other components present. Hamel (1977:208) seeks to explain the numerous stratigraphic anomalies thus:
If most artifacts were deposited during the period of the Working Floor component, they could have been trampled into the upper parts of the black layer, and in other areas scuffed up into the base of the single ashy-shelly layer which Teviotdale recognized above the black layer.This comes very close to an admission that the “working floor” is not actually to be regarded as a stratigraphic component at all. Rather, it is a term for any - 135 horizon at which artefacts or evidence of their manufacture appeared most noticeable in the stratigraphic section.
Consequently, since the “lower shell” component was uncommon, since the stratigraphic status of the single ashy-shelly layer (i.e., the “shell and bone” component) in the peripheral excavations is unclear, and since the “working floor” cannot be considered a stratigraphic unit, the most that may be safely asserted of the Papatowai stratigraphy as a whole is that there is often a black layer underlying shell midden in the north-eastern or main area of excavations, but that only a single shell and bone midden layer occurs generally in the surrounding areas. The black layer was particularly rich in ovenstones and seal and moa bones, while the shell and bone midden contained numerous artefacts and worked bone debris from fishhook manufacture. This suggests that big-game cooking was largely confined to the north-east corner under the high dunes, perhaps for shelter from the prevailing north-east wind in summer, and that the remainder of the site consisted of butchering, manufacturing, discard and living areas. Since only one “fireplace” has been recorded, in Teviotdale's(1938a:28) area 2, and no postholes, nothing is known of the domestic structures or settlement pattern.
If this analysis of the stratigraphy does not suggest as lengthy an occupation as that implied by the more complex general sequence, then the same conclusion emerges from analysis of the distribution of artefact types in the site. Lockerbie (1959:84-5), argued that there was evidence of change in adze and fishhook types, and in the size of flake and blade implements between the bottom and top layers of Catlins sites. Hamel sought evidence in support of this model but her results were not encouraging.
She noted (Hamel 1977:212) that, while artefact types changed very little through the layers, there was a change in the relative abundance of types. However, reference to her semiquantitative tabulation of faunal and artefactual types with layers (Hamel 1977:Table 4:13) reveals no apparent trends in types or proportions. In addition, it seems that there was a “… tendency for a restricted area to have the same function in two or three different layers” (Hamel 1977:223), that is, for concentrations of adze deposition or fishhook manufacture to extend through several layers in the same place. This is a clear indication that the layers were deposited at more or less the same time, at least in particular areas.
This leaves only the question of whether differences in the relative - 136 quantities or types of animal remains might support a hypothesis of long occupation, as has often been concluded on the basis of the earlier evidence (e.g., Anderson 1983). It is important to make a clear distinction here between stratigraphic differences in the relative abundance of taxa, and conclusions about changes in subsistence strategies which may be drawn from them. We have no doubt that Teviotdale and Lockerbie were correct in drawing attention to a marked concentration of moa and seal bone in the lower layer of the main site area and to an equally apparent abundance of fish bone, small bird bone and shellfish remains in the upper layer. Likewise, although we would rearrange somewhat the stratigraphic distribution of Hamel's TT1 faunal data (below), we accept the trend evident in them.
As explained above, however, most of the site seems to have consisted of a single cultural unit, Hamel's “shell and bone” layer, in which remains of big game, notably moas, were mixed with dog, small bird and fish bone in a dense shell midden. Unfortunately, there are no quantitative data from these parts of the site.
A simple interpretation of the distribution of faunal remains would be as follows. There was a preferred big-game cooking locality in the north-east or main site area and bones from which the meat had been stripped were often discarded into or beside the ovens. Small game, fish and shellfish were generally cooked contemporaneously elsewhere in the site, probably at many points (Hamel (1977), reported “scoop hearths” in area TT1) and refuse from this activity was dumped indiscriminately along with remains of big-game butchery and consumption. The bulky shell and bone dumps encroached eventually over the main oven area.
The principal objection to this and other propositions which imply occupation during a single phase, and probably for a fairly brief period, has always been the radiocarbon dates.
Until 1990, there had been 15 radiocarbon dates on samples obtained from Papatowai (Hamel 1977:Table 4:17). Seven samples were submitted from the Lockerbie 1956 excavation. Four were of moa bone carbonate (NZ 137-140), and they can be rejected on the ground of probable contamination by isotopic equilibration (Hamel 1978:53), along with one charcoal date (NZ 135) which was a duplicate run of the sample for NZ 134 (and which produced the same result). All the charcoal samples were unidentified. The remaining 10 dates (Table 1), formed a fairly consistent pattern of age with stratigraphy and indicated that occupation had occurred over a period of 300-400 years. However, the potential impact of “old wood” in charcoal samples was not fully appreciated in this series and it is likely to have been very significant - 137 indeed. The older dates from the main ovens may reflect use of larger and older pieces of wood than those for smaller cooking fires elsewhere in the site.
Hamel dated one shell (Paphies australe) sample (NZ 1333) and seven charcoal samples (Table 1). Of the results obtained from the latter, Hamel (1978:53) remarked that “… the chronology of the dates agrees with the stratigraphic order”. But all the samples consisted mainly of totara (Podocarpus totara/hallii), and the oldest date (NZ 4267) was on a 90% totara sample while two (NZ 4271 and 4272), of the three youngest dates were on samples of 55-60% totara. Thus, the broad consistency of dates with layers (Hamel 1977:Table 4:21, Fig. 4:13) admitted two explanations — relative predominance of old wood or relative age at death. Clearly, additional samples in which the old wood problem was minimised were required to resolve the uncertainty.
EXCAVATIONS IN 1990
Our investigations in April 1990 set out to examine areas as close as possible to the main excavations by others. Our purpose was to re-examine stratigraphic profiles in these areas, recover material for radiocarbon dating, and collect further samples for comparative faunal and artefactual analysis.
Hamel's (1977: fig. 4.8) map represents the best estimate of the location of former excavations, 1 In the absence of surveyed plans of these, it cannot be precise, but the earlier excavators used the same coach road as a reference line as well as the sharply defined edge of the high dunes along the north-east edge of the site. Hamel also consulted with Lockerbie on site and, of course, Lockerbie had worked with Teviotdale. In the main site area, then, the margin of error may be only a metre or two at most.
Area THK, a 3 x 1 metre trench, was positioned along the road scarp (Fig. 3) as near as possible to the estimated locations of Teviotdale's Area 1 and Lockerbie's 1956 excavation. Beneath a surface litter, up to 10 cm in depth, six stratigraphic layers were identified (Fig. 4).
Layer 1 was a sterile dark brown humic sand 10-20 cm deep. Layer 2 was a substantial deposit of clean yellow sand, up to 50 cm thick on the uphill side of the excavation but thinning to 10 cm downslope. It was probably pushed downhill during construction of the coach road. Layer 3 was a sparse shell midden which occurred only in square A3. Its greatest depth, of about 25 cm, was in the southern baulk (Fig. 4:B-C), indicating that that midden extended further in that direction. Layer 4 was a yellow-grey sand containing occasional flecks of charcoal, reflecting burning elsewhere on the site. Layer 5 was a dark grey to black greasy sand containing numerous fragmented ovenstones and pieces of charcoal, some of which occurred in dense patches.- 138 - 139
The patchy colouration of this layer, along with its contents, indicate that it was composed largely of oven rake-out material. Layer 6 exhibited the characteristic pink, white and grey colouration of burnt sand. Layer 7 was a sterile yellow sand that extended to the base of the excavation.
Comparison with earlier excavations: Our observations (Fig. 4) are consistent with stratigraphy described by Teviotdale (1937:137-9) on the high-dune margins of his Area 1; that is, the upper shell layer became discontinuous and the black layer contained practically no artefactual or faunal material, only ovenstones and charcoal. Similarly, the upper shell and black layers were separated by a layer of sand (cf. Hamel 1977: Fig. 4:10b, and note difference with Fig. 4:10a and Teviotdale's (1937:137) section drawing which represent stratigraphy elsewhere in Area 1). The stratigraphy of Lockerbie's 1956 excavation does not match ours at first sight mainly because he found two black layers (1959:84 and plate II), but our black layer was partly split in - 140 places and we note in Lockerbie's photograph (1959: plate Ib) that his two black layers coalesce. Given our view (above) that cooking was concentrated in this area of the site, it is to be expected that there would be many local variations in the stratigraphy which reflect the repeated cutting and use of ovens. Lockerbie's sections (1959:84 and plate II) also show shell occurring as discontinuous lenses, as in our experience.
Hamel's TT1 excavation had taken place on the flat ground east of the coach road just below the foot of the high dunes. Examination of the ground surface in this area showed few archaeological remains other than a scatter of shells immediately adjacent to a probable fossicker's pit on a low spur extending westward from the foot of the high dunes (Fig. 3). TT1 was relocated, with Jill Hamel's assistance, in a shallow hollow just north of this spur.
Area PPT was laid out immediately south and east of TT1. Forest litter 5-15 cm in depth was cleared from the area and beneath this five stratigraphic layers were identified, along with two lenses of more restricted distribution (Fig. 5).
Layer 1 was a sterile brown humic sand and root mass. Layer 2 was a light grey sand containing occasional bones, shells and fragmentary ovenstones. It was 25 cm thick in square B4, but only 1-2 cm in square A1. Layer 3 was distinguished from layer 2 by a greater density of shell midden and a slightly darker grey sand matrix. However, it soon became clear that layers 2 and 3 represented a single stratigraphic component, with midden concentrated at its base. We shall call this the layer 3 midden. The density of midden also varied horizontally (Fig. 6a) with the greatest concentration to the east. Layer 4, a brown sand, occurred over the entire area and in most places extended to the base of excavation. It was sterile except for a single moa bone found near the southern end of square A1. Layer 5 was a dark grey charcoal stained sand, entirely lacking in other remains. It occurred only around the southern and eastern edges of TT1 (Fig. 6b), and the stratigraphic profiles (Fig. 5) indicate that it is a lens within layer 4. Two further stratigraphic components were identified along the western edge of squares A1 and A2: Lens A between layers 4 and 5, and Lens B between layers 2 and 4 (Fig. 5c). Both lenses were presumed to be the result of subsurface disturbance.
Five test pits were also excavated in the vicinity of PPT (Fig. 3) with the objective of determining whether any of the same stratigraphic components could be identified over a wider area. Test pits A and B were excavated to a depth of 1 metre at the foot of the scarp immediately west of the coach road. Both showed a brown sand and humus (10-20 cm) overlying sterile yellow- 141
Figure 5. PPT - stratigraphic profiles. a and b. reconstructed from levels taken during excavation; c. drawn from western baulk; d. after Hamel 1977:Fig. 4:12. Location of sections shown in Figure 6.
sand. At 70 to 80 cm below the surface, this graded into a sterile brown sand. To the south of PPT, Test pits C, D and E all revealed a similar stratigraphic profile. This comprised a brown sand and humus matted with roots (8-15 cm) overlying a grey-brown sand containing sparse shell midden (12-15 cm). Beneath this was a sterile light grey sand (10-15 cm) which graded down into brown sand.- 142 - 143
Comparison with TT1: The stratigraphy observed in our excavation differs from that described by Hamel (1977:211-2, 232, Fig. 4:12). She reported three cultural layers: an upper shell midden, a discoloured sand containing faunal remains and artefacts, and a black layer rich in faunal remains. We found two cultural layers: a widespread layer of shell, bone and artefacts (layer 3 midden); and a dark grey sand, stained with charcoal but entirely lacking in faunal and artefactual remains (layer 5).
There can be little doubt that Hamel's upper shell equates with our layer 3 midden. Leaving contents and colour aside for the moment, it is not difficult to equate her black layer with our layer 5. Juxtaposition of the two most closely adjacent section drawings from PPT and TT1 (Fig. 5 c and d) show that this interpretation is stratigraphically justifiable.
The differences in contents and colour between Hamel's observations of the lower layer and our own can be explained by two factors. Firstly, our excavation showed that layer 5 occurred only around the eastern periphery of TT1 (Fig. 6b), and we note that the black layer was absent in the western half of that excavation (Hamel 1977:212). Thus, it was a layer of limited horizontal distribution within which faunal remains appear to have been confined to the central portion which Hamel excavated almost in entirety. Secondly, Hamel's colour terminology suggests that her layer names were not so much descriptive as classificatory. She described this layer as “black” because it was at the base of her stratigraphic profile.
The most significant difference between Hamel's observations and our own concerns her middle or discoloured sand layer, which she assigned to the working floor component of the general sequence (Hamel 1977:232). We found no evidence of that component at all and can only suggest that Hamel's discoloured sand layer is either the base of our layer 3 midden, or material disturbed and mixed in the course of oven construction and use. This may reflect, once again, imposition of the general sequence upon interpretation of a particular stratigraphic profile.
These excavations at THK and PPT provide clear support for our reinterpretation (above) of the stratigraphy recorded during previous investigations. They revealed only two cultural layers, an upper shell midden and a lower charcoal-stained sand. The latter was deep and dark in colour at THK, within the main site area, but beyond, at PPT, it was shallow, lighter in colour and of very limited areal extent. Before considering what this might mean in relation to the length of occupation and economic change at the site, it is necessary to describe further evidence recovered during our excavations.
Faunal remains were identified using comparative material in the Anthro- - 144 pology Department, University of Otago and, in the case of moa bones, with the assistance of Trevor Worthy (personal communication).
THK Layer 3. A small assemblage was recovered from the lens of sparse shell midden at the southern end of area THK. All the excavated contents of this deposit were retained for analysis. Approximately 5 litres was taken as a bulk sample, while the remainder was sieved (2 mm mesh) and all shells, bones and stone artefacts retained. Sieving of samples from all other stratigraphic components in this area failed to yield any faunal or artefactual remains.
The THK midden was dominated by shellfish — 98.4% of the total Minimum Number of Individuals (MNI) identified — mostly pipis, cockles and blue mussels (Table 2). More than half of the bones in the assemblage were from fish, but these represented only a few individuals, mainly barracouta (Table 3). Birds were more numerous, with parakeets and pigeons predominating. Both mammal and moa remains were sparse, the former comprising a dog tooth, one piece of sea-lion rib and two unidentifiable fragments, and the latter four long bone fragments and two tracheal rings.
PPT Layer 3 Midden. A larger assemblage was recovered from the layer 3 midden in Area PPT. Two procedures were used in sampling this deposit. Bulk samples were collected from squares B1, C3 and D2 for detailed analysis of shellfish species composition. All other excavated remains were sieved, with bones and stone artefacts being retained for analysis. The only faunal material encountered elsewhere in this area was a complete right femur of Emeus crassus, discovered a few centimetres below the surface of layer 4 in square A1.
The relative abundance of molluscan species in the bulk samples closely matches that observed in THK, with the three major species in the same order (Table 4). Bones were much more common than at THK, and not simply because of the larger sample size. The layer 3 midden showed distinct concentrations of bone, reaching densities as high as 1698 NISP per m2 in square D2, nearly four times greater than the value for THK (487). High values were also recorded in PPT squares C2 (568) and C3 (992), and extremely low values in B1 (62), A1 (11) and A2 (4). Fish bones were the most common items (Table 5), although they contributed less than 30% of the individual animals identified. Vertebrae outnumbered the single most common head bone (left dentary, n=18) by a ratio of 20:1, demonstrating that whole carcasses of fish, mainly barracouta, had been dumped in the midden.
Unlike the THK assemblage, sea birds made up most of the avifauna, the main species being penguins, shearwaters and shags. Pigeons and parakeets were the most common forest birds. Reconstruction of butchery patterns for - 145 the Fiordland crested penguin showed that crania, as well as bones from the lower wings and lower legs, were poorly represented (Fig. 7), indicating deliberate removal of lower limbs before arrival of the bird carcasses at the site.
Mammal bones made up 18.3% of total NISP, more than 20 times greater than the value observed at THK (0.8%). Three-quarters of the identifiable bones were from dogs and fur seals, each of which was represented by four individuals. Two sea-lions and an elephant seal were also present, along with four rabbit bones.
The dog remains were aged on the basis of epiphyseal fusion and tooth eruption (Silver 1969, Sisson 1930), indicating that one pup (<6 months), one subadult (6-18 months), and one adult (>18 months) were present, along with another individual that could not be aged. Crania occurred more frequently- 146
than other body parts (Fig. 8), which might suggest that parts of the trunk and limbs were taken elsewhere on the site for consumption or the extraction of bone for artefact manufacture. However, the sample size is too small to be confident of this interpretation.
The fur seal remains were assigned to age-sex classes on the basis of epiphyseal fusion and bone dimensions (Smith 1985), showing the presence of a juvenile, a subadult male, an adult female and an adult male. This diverse range suggests exploitation of a breeding population. Examination of proportional body parts representation (Fig. 8) showed high numbers of upper forelimb and upper hindlimb bones, while those from the lower limbs and trunk were scarce. While the sample size must again qualify our conclusions, this pattern seems to indicate that most of the fur seals were butchered elsewhere, with only their easily detached meat-bearing upper limbs being brought to this part of the site.
Moa remains were seven times more common in PPT (8.7% of total NISP)- 147
than they were at THK (1.2%). The most notable feature of these remains was their fragmentary nature. Only 22 of the 214 specimens could be confidently assigned to a particular skeletal element. These included vertebrae, ribs, pelves and phalanges, as well as complete or relatively complete long bones. Most (72%) of the remaining items appeared to be small fragments from the shafts of long bones, suggesting that these bones were regularly being broken up for artefact manufacture. Because of the fragmentary nature of the bone, only one of the five individual moas represented could be positively identified to species.
Comparison with Earlier Excavations. Several direct comparisons can be drawn with the faunal evidence from TT1, particularly with regard to the upper shell layer which equates with our layer 3 midden at PPT. The first significant point to note is the variation in midden density. We recorded this during excavation (Fig. 6), and provided further evidence of it above in relation to bone concentrations. In addition, the molluscan MNI in our three modest bulk samples (641) exceeds the total value (607) for the entire 4.5 m2 of upper shell midden in TT1 (Hamel 1977:Table 4:8). Together, these observations demonstrate that our excavation sampled a portion of this deposit considerably richer in faunal remains.
There are some similarities in midden composition, especially among molluscs, except that mussels were more prominent at THK and PPT than TT1 and, while we identified only blue mussels, Hamel reported only green mussels (Perna canaliculus). Bone remains show greater differences. Fish made up half the TT1 upper shell MNI, but less than 30% of that at PPT; moas a quarter of the former, but only 6.4% of our sample. Also, birds dominated at PPT but contributed only 16.7% at TT1.
Two factors may explain these differences. Firstly, the smaller size of the TT1 assemblage must call into doubt the reliability of MNI derived from it (Grayson 1978). We do not know exactly how many bones and fragments were present in the TT1 upper shell layer, but with total MNI only 15% of that calculated for PPT, and only 68 moa bones and fragments (Hamel 1977:Table 4:8) to our 214, there can be little doubt there were considerably fewer than the 2459 NISP at PPT. Secondly, even with comparably sized samples, variations of this order would occur. Our observation during excavation and test pitting was that both shellfish and bone midden density and composition varied horizontally within the same stratum. Although we cannot equate the THK midden stratigraphically with TT1, comparison of their similarly sized bone assemblages provides a numeric example of just such variation.
Finally, we note that observed variations in midden contents between TT1 upper shell and PPT layer 3 make only a limited difference to the relative - 148 importance of species when they are considered as sources of food. Meat weight values for the TT1 fauna were calculated by Smith (1985:161-6, Table 30), and the same procedures have been employed here to derive values for PPT (Table 6). In the latter case, no figures could be calculated for shellfish because their total abundance in the deposit was not measured. However, it is unlikely that they would have contributed more than 2-3% of total meat weight. Table 6 shows that the relative contributions of fish, birds and dogs were very similar in the two assemblages, and that, in each case, their combined contribution was only about one-quarter of total meat weight. The only notable differences are in the values for seals and moas, which reverse their relative positions in the two assemblages. However, meat weight values for large animals such as these are particularly susceptible to the variations in MNI induced by small sample size (Grayson 1978), suggesting that the observed difference in Table 6 is unlikely to be significant. Indeed, we note that the addition of just one moderately sized fur seal to the TT1 assemblage would produce values almost identical to those from PPT. The significant point that we take from these calculations is that in both assemblages big-game animals contributed about three-quarters of total meat weight.
The major limitation of the faunal evidence recovered during our excavations is that it did not include any material from the lower cultural layer, thus precluding a quantitative assessment of the faunal changes hypothesised by previous investigators. As indicated above, we have no reason to doubt the general trends that they identified, although, as our meat weight data show, big game still dominated the diet in the upper layer despite the abundance of fish, small birds and shellfish. However, the point with which we do take issue is whether these changes necessarily took place over an extended period of time. Resolution of this question rests not with the faunal remains themselves, but with other evidence for the length of occupation at the site.
Only 69 artefacts were found. Forty-seven stone flakes were recovered from PPT layer 3;20 of argillite, probably from Southland sources, of which five have the colour and texture of the Riverton source. A waterworn cortex was observed on four flakes and two had signs of hammer-dressing. All are probably debris from adze manufacture. There were 13 porcellanite flakes, five of which showed signs of use, and 12 silcrete flakes, three of which had minor use-damage. There was one flake of rock crystal and one of chert. Of 11 flakes from THK, eight were porcellanite, two argillite and one chalcedony.
A hammerstone (Fig.9a) of grossular garnet, the nearest source of which would be Orepuki beach near Riverton (Mason, personal communication), - 149 came from PPT. It had been heavily used. There were also five polished fragments of Southland argillite adzes (e.g., Fig.9b). Two small pieces of polished nephrite were recovered from PPT. One (Fig. 9c) is bevelled and probably from an adze. The stone source is either south Westland or Otago, whereas the other fragment is from south Westland, according to Hooker (personal communication).
Area THK yielded a fishhook tab-core (Fig.9d), and hook point (Fig. 9e),
Figure 9. Artefacts from the 1990 excavations, a. hammerstone(PPT14-SO-1), b. adze fragment (PPT14-SO-2), c. adze fragment (PPT15-SO-2), d. tab core (THK53-BW-1), e. fishhook point (THK53-BW-2), f. worked tooth (PPT15-BW-1).
both made from moa bone. The point appears to be the reworked point leg of a one-piece hook. It retains the basal notch as in other one-piece hooks (e.g., Hjarno 1967:36,54,59,61). The attachment facet is bevelled from both sides into a chisel form (type B in Gumbley 1988:15). No comparable pieces are illustrated by either Lockerbie (1959) or Hjarno (1967). The final piece (Fig. 9f), from PPT, is part of the canine tooth of a sea-lion or elephant seal in which a notch has been filed to produce two “legs”. One of these has broken. The form suggests no obvious function, though it might once have been a hook.
In summary, there is nothing typologically distinctive about the artefacts and they offer no evidence of long-term or especially early occupation.
Three charcoal samples were submitted for radiocarbon dating. Two were from THK layer 5, the black lower cultural layer in the main site area. Sample THK 1 (WK 1761), from the top of this layer, was selected from a much larger sample of totara and matai charcoal. It consisted of 28 pieces (15.0 g), described as twig charcoal by Rod Wallace (personal communication, who identified all the samples), of which 13 were manuka (Leptospermum scoparium), three each were five finger (Pseudopanax arboreus), Pseudowintera colorata and ribbonwood (Hoheria sp.), two were beech (Nothofagus sp.) and one each were Coprosma sp., Pittosporum sp., Archeria traversi and bark. Sample THK 9 (WK 1762), from the base of layer 5, consisted of approximately 5% of a larger charcoal sample of which the balance was from conifers. The 25 pieces (10.0 g) in the dated sample consisted mainly of manuka, 14 pieces, followed by six of raukawa (Pseudopanax edgerleyi), three of mapau (Myrsine australis) and one each of ribbonwood and Dracophyllum sp.
The third sample was from the PPT layer 3 midden. Sample PPT 11 (NZA 1415), consisted of only three pieces (1.0 g), one each of manuka, five finger and putaputaweka (Carpodetus serratus).
These are good samples in that they consist of pieces from relatively short-lived species, but if there is any doubt about the identification of twig wood, as in other cases (Anderson: in press), then they might still exhibit some significant in-built age from the predominance of manuka, as well as from the small quantities of beech and bark. However, we can still accept the results as maximum estimates of age.
The results (Table 7) show almost identical ages for the two samples from the lower layer at THK, and one only slightly younger for the upper layer at PPT. When compared with the less reliable dates from previous investigations (Table 1), our data suggest that the lower black layer in the main site area is much younger than was earlier thought, although the upper layer remains - 151 about the same age. Many more radiocarbon results would be needed before we could be confident about that conclusion for the site as a whole, but a hypothesis of younger, shorter occupation is clearly indicated.
DISCUSSION AND CONCLUSIONS
Earlier interpretations of the archaeological evidence at Papatowai proposed a general model for the site, the Catlins and, indeed, for cultural change in southern New Zealand as a whole. The basis of the model was a uniform sequence of stratigraphy which was regarded as the result of lengthy occupation. That argument was supported by evidence of changes in fauna and material culture, and by a series of radiocarbon dates.
We have argued that the stratigraphic model is not actually supported by evidence derived from earlier excavations. The “black” layer seems to have occurred in only the north-east sector of the site. The “working floor” is not a stratigraphic component: it seems to have been identified at various levels in the stratigraphic column, usually by the occurrence of artefactual remains. There is no evidence that the “upper shell layer” identified in the main site area and the “shell and bone layer” in peripheral areas are actually stratigraphically distinct. The pattern of change in frequency of faunal remains with stratigraphy was observed as a decline in big game and a rise in fish, shellfish and small birds, but, even in the relatively restricted part of the site where this was evident, it does not follow that the explanation requires an extended passage of time. The argument that there were significant changes in material culture assemblages correlated with stratigraphy cannot be substantiated by the data, and the radiocarbon dates which supported this general model do not withstand modern scrutiny.
Our excavations disclosed only two cultural layers. In both THK and PPT, the upper layer was a shell midden and the lower a charcoal-stained sand. There were variations in contents within each layer and between the excavation areas. It is possible that this stratigraphy reflects two phases of occupation, but, since the lower layer clearly results from cooking activities (without much dumping) and the upper from dumping of food and other refuse (without much cooking, or at least construction of large umu), it is more likely that they represent simply two activities of a single occupation. Nor would this interpretation be inconsistent with the radiocarbon dates from our excavations, which showed that the two layers were almost identical in age. Even if a two-phase interpretation were preferred, our data indicate that little time could have separated the occupations. Importantly, it is apparent that big game are highly predominant in the estimated meat weight from the upper layer, and might be even more so if the moa bone was not so fragmented.
Our interpretation of Papatowai, therefore, is that it was a large camp, - 152 rather than a village, that was occupied over a few years, perhaps a generation, and probably discontinuosly, at which butchery and consumption of moas and seals was the main subsistence activity throughout.
These revisions clearly impact on thinking about the course and causes of economic change in southern New Zealand, including the extinction of moas (Lockerbie 1959; Anderson 1983, 1989) and, therefore, on Maori palaeoeconomics as a whole, but those are issues beyond the scope of this paper. It is sufficient to note here that the basic trend, from a relatively greater emphasis on big game to greater reliance on small game including fish and shellfish, must still exist in southern prehistory. What should change, we argue, is the belief that this process took place over an extended period of time at sites such as Papatowai.
For permission to excavate at Papatowai, we thank Mr Huata Holmes (Tautuku Waikawa Maori Lands Board), Mr Jeff Connell (Department of Conservation, Dunedin) and the New Zealand Historic Places Trust for permit number 1990/6. For assistance in the field, we thank Mr Blair Townshend (Department of Conservation, Owaka), Dr Jill Hamel, Dr David and the late Jacqui Pilditch, Matthew Campbell, David Cassaidy, Nigel Chang, Tom Higham, David Hood, Caroline McGill, Rick McGovern-Wilson and Pat Wells. For specialist assistance with identification of excavated materials, we thank Ray Hooker (Department of Conservation, Hokitika), Graeme Mason (University of Otago), Dr Rod Wallace (University of Auckland) and Trevor Worthy (Nelson). The excavation and analysis were funded by the University of Otago, to which we are grateful.
Table 1. Radiocarbon Dates from Previous Investigations.
Notes: W.F. = Working Floor, U.S. = Upper Shell.
Calibration of charcoal dates after Stuiver and Reimer (1986) with offset of 30 radiocarbon years (Stuiver and Pearson 1986), and of shell dates according to Stuiver, Pearson and Braziunas (1986) with geographic offset Delta-R set at 0 radiocarbon years and conventional ages shifted by -31 radiocarbon years (McFadgen and Manning 1990).
Table 2. THK Layer 3: Shellfish Identifications.
mni= minimum number of individuals- 155
Table 3. THK Layer 3: Bone Identifications.
nisp = number of identified specimens
mne = minimum number of elements
mni = minimum number of individuals
Table 4. PPT Layer 3 Midden bulk samples: Shellfish Identifications.
Table 5. PPT Layer 3 Midden: Bone Identifications.
Table 6. Meat Weight Estimates.
Table 7. Radiocarbon Dates from the 1990 Excavations.
Note: for calibration conventions, see notes to Table 1.
1 On Hamel's (1977: Fig. 4.8) map, the location of TT1 is shown about 8 m south of its actual position, as established by our instrument survey (Fig. 3).