CHAPTER NINE CONCLUSIONS by Patrick V. Kirch T IHE TBREE ANAHULU RocKsHTERsS reported in this volume have yielded one of the largest samples yet available of prehistoric occupations in a truly interior valley setting in the Hawaiian Islands. They thus provide an opportunity to investigate the chronology and sequence of prehistoric human penetration and use of the interior regions of 0'ahu Island, and to compare this sequence with that from coastal or lowland areas. The shelters likewise offer evidence for the course of local environmental change in the inland parts of the Anahulu Valley over the past six to seven centuries, and of the probable role of humans in initiating such change. These are the principal themes to be summarized in this concluding discussion. The perspective on Anahulu and Waialua prehistory contributed by the rockshelter evidence is a viewfrom the periphery, that is, from an interior, hinterland region. Unfortuately, we as yet know little of the prehistory of coastal Waialua, much of which has been subject to major land use changes and disruptions in the historic period. Sporadic surface finds of early artifact types in the vicinity of Haleiwa (A. Anderson, pers. comm., 1982), however, hint that occupation in this area may date to the Colonization Period (A.D. 300- 600) (Kirch 1985:298-302). Certainly the environmental setting and resources of coastal Waialua-its sheltered bays, rich marine life, and especially its well-watered fertile alluvial plains-would have been attractive to early Polynesian settlers. There can be little doubt that this part of 0'ahu must have been settled relatively early, although an effective archaeological research program to locate and excavate sites dating to the Colonization and Developmental Periods has yet to be mounted. Nonetheless, there is no reason to doubt that the occupation sequence revealed in the Anahulu rockshelters reflects the gradual expansion into a peripheral, hinterland zone some centuries (and perhaps as long as a millennium) after initial human colonization of the region. The large Ke'eke'e Nui rockshelter (D6-58) encapsulates the longest occupation sequence for the interior valley. Feature 9, a lens-like deposit of ash and charcoal immediately overlying the sterile alluvial base in Unit 0 1, and probably representing burning of vegetation on the site prior to first use, yielded a calibrated calendar age range of AD. 1280-1430. The shelter appears to have been used, more-or-less continuously, from the fourteenth century up through the early historic period (ca. A.D. 1810). Although our excavated sample from the site is relatively small, limited to 3 m2, we were able to identify major differences between lower and upper cultural deposits. These deposits were first observed during excavation as depositional units with contrasting color differences, representing varying quantities of included ash and charcoal (see chapter 2). The hypothesis was raised that a major shift in the intensity of occupation in the shelter might be represented by the transition from lower to upper cultural deposit. Additional support for this hypothesis comes largely from the analysis of both vertebrate and invertebrate faunal assemblages (see chapter 4.) With almost equivalent excavated volumes, Anahidu the upper cultural deposit contains more than three times the number of vertebrate faunal specimens (NISP), especially of pig and dog, domestic animals likely to have been raised in the interior valley. This pattern is replicated by the invertebrate fauna, particularly marine shell, which displays a roughly three-fold increase in density from lower to upper cultural deposit. Thus, assuming that the physical conditions of sedimentation and deposition within the rockshelter remained constant (and there is nothing from the geoarchaeological analysis of sediments to suggest otherwise, see chapter 3), these significant increases in the density of faunal remains (as of charcoal and ash content of the sediments), can only signal a shift to greater intensity of human occupation. The question remains, howerver, whether thefunctional activities carried out in the shelter actually changed, or whether the same range of activities was merely carried out more frequently, leading to a higher density of cultural materials in the upper deposit. The portable artifact assemblage from Ke'eke'e Nui shelter helps to discriminate between these alternative explanations for the differences between lower and upper cultural deposits. In terms of the presence/absence of discrete artifact classes represented, the upper cultural deposit contains a significantly more diverse array than the lower cultural deposit. The latter contains primarily flakes of both basalt and volcanic glass, some adz flakes, small one-piece fishhook fragments, a few possible coral abraders, and one piece of worked bone. The upper cultural deposit contains not only all of these classes present in the lower deposit, but in addition a grindstone, bone awls, dog tooth ornament, tattooing needle, polished basalt file, complete quadrangular adz, and an ulumaika gaming stone. The awls suggest some kind of domestic activity not carried out earlier in the shelter, while the basalt file and grindstone may indicate new kinds of tool manufacturing activity. More telling may be presence of the dog tooth ornament and tattooing needle, as well as of the gaming stone. These classes of artfact are generally regarded as associated only with permanent residential sites. Based on the results of both the faunal and artifactual analyses, the interpretation of a change in the nature of occupation in the D6-58 shelter, first generated on the basis of field observations, can not only be confirmed but further elaborated. While the lower cultural deposit was accumulating, a relatively restricted range of activities was taking place in the site, including the flaking of volcanic glass and basalt (probably also adz bevel re-sharpening), bone working, and use of one-piece fishhooks (in the nearby stream, for freshwater fish?). During the period when te upper cultural deposit was accumulating, this range of activities expanded to include other domestic work (using awls and other tools), bodily ornamentation, and leisure activities. Further, the frequency or intensity with which fires were lit in the shelter (for light and cooking) increased substantially, indicated by the differences in ash/charcoal content of the sediments. These changes would be most parsimoniously explained by a change in the use of the shelter from a temporary camp to a permanent residential site. The pattern evidenced by the lower cultural deposit would fit well with the model outlined by Kirch (1979:51), and summarized in chapter 1, in which the shelter would have been visited periodically by small groups residing in the interior valley for periods of a few days. Presumably during such visits, these groups would have exploited particular interior resources, such as nesting seabirds (in the cliffs above the shelters) or native landbirds (both of which are indicated in the vertebrate faunal assemblages), and the endemic gobiid fish. Clearing of small shifting cultivations on the alluvial flats adjacent to the shelter may also have been inidated during this early phase. During the later phase, represented by the upper cultural deposit, one household group evidently took possession of the shelter as its permanent place of residence. Exploitation of wild resources continued, evidenced in the faunal assemblages, but in addition the substantial increases in domestic pig and dog bone are highly suggestive of animal husbandry in the vicinity of the site. Based on ethnographic descriptions of Hawaiian pig and dog husbandry, which document the regular feeding of these animals on vegetable crops (such as taro and sweet potato), the existence of permanent gardens can also be inferrde The timing of this shift in the use of Ke'eke'e Nui rockshelter from temporary to pennanent residence is approximately indicated by the radiocarbon sample from Feature 5, which was cut from near the base of the upper cultural deposit (see chapter 2, table 2.4). This sample yielded two calibrated calendar age ranges, but most likely dates to the first half of the 17th century (ca. A.D. 1600-1650). It is probably significant that this age is only slightly younger than the date from Feature 4 in nearby Site D6-36, which was presumably first used at about the same time that Site D6-58 became permanently occupied. The archeological record itself is mute as to the cause(s) that may have precipitated the shift from temporary use to permanent occupation of Ke'eke'e Nui shelter, but it may be instructive to briefly consider some possibilities. The most deterministic explanation would be to invoke the pressure of a burgeoning lowland Waialua population on coastal lands and resources, thus forcing certain members of society 126 Conclusions (those without access to pme lands ?) to move inland. Alternatively, we might suppose that a group which had been regularly exploiting the interior valley on a seasonal or periodic basis chose to specialize fuiltime on exploitation of the valley's interior resources (such as the opportunitites presented by development of pig and dog husbandry). This might be labelled the "ecological opportunism" scenario. Yet a third possibility would invoke force and chiefly control, by means of the forced dispossession of a group or household of its coastal lands, perhaps occasioned by conquest of Waialua and redistribution of its lands, as is known to have occurred repeatedly in late prehistoric Hawaiian society (Kamakau 1961; Sahlins 1958; Goldman 1970). Indeed, none of these scenarios need be mutually exclusive, and all three factors could have played a role in the decision of a particular group to take up permanent residence some 6-7 km inland. Significantly, this removal of a group of people from lowlands to interior, indicated by the Ke'eke'e Nui rockshelter sequence, appears to have been repeated later, especially in the early historic period following the conquest and occupation of O'ahu in 1804 by Kamehameha I. Interior, hinterland areas such as the middle and upper reaches of the Anahulu Valley offered opportunities for agricultural expansion to individuals and groups dispossessed of, or for other reasons without access to, more desirable coastal land and resources. This is a theme which forms a dominant subject of a longer work on Anahulu ethnohistory and archaeology (Sahlins and Kirch, in prep.). Having pressed the interpretation of the evidence from Site D6-58 to its limits, how does the sequence compare or contrast with that from Kuolulo shelter? Initial use of Site D6-60 is indicated by the radiocarbon date from the base of excavation Unit D20, of the mid 15th century AD., about one hundred years after the first use of Ke'eke'e Nui shelter. The relatively deep and continuous stratigraphic sequence provided by Unit D20 does not evidence a significant shift in the intensity or nature of use/occupation in this site. The vertebrate fauna, for example, shows no major shifts, although sample size is admiedly small (chapter 4, table 2.5). Nor do the deposits in Site D6-60 show the same density of vertebrate or invertebrate faunal remains as in the upper cultural deposit at D6-58 (chapter 4, tables 4.2, 4.6). While the deep stratigraphic column from Unit D20 does not indicate a shift in the intensity of occupation or changes in the density of faunal remains, other evidence from Site D6-60 can be adduced to suggest a rather late change in the nature of human occupation at this site. Most important is the construction of both the inner and outer terraces in the westen part of the site, which indicate the subdivision of the shelter floor into discrete activity spaces (probably domestic activity and sleeping on the inner terrace, certainly earth oven cooking on the outer terrace). Two radiocarbon dates (B- 5170 and B-5168, see chapter 2, table 2.4) reveal that the construction of these formal activity areas took place fairly late in the site's occupation sequence, probably not prior to A.D. 1700, and possibly quite late in the eighteenth century. As in Site D6-58, the range of portable artifact classes recovered from excavation of these late, formalized constructions is suggestive of a broad range of activities, most likely to have resulted from permanent residence in the site. These include several classes of tools (adzes, grindstones, whetstones, files, polishing stones, abraders, hammerstones, and awls), fishhooks and line sinkers, a dog tooth ornament, and a large number of cone-shell beads, as well as the more ubitiquous basalt and volcanic glass flakes present in the deeper D20 sequence as well. Although their density was less than in D6-58, the inner and outer terrace deposits also yielded 61 dog and pig bones, again suggestive of the local husbandry of these animals, probably in conjunction with permanent gardening on the alluvial flats. This evidence suggests that there was also a shift from temporary to permanent occupation at Kuolulo Shelter, but that this shift occurred quite a bit later than that at Ke'eke'e Nui shelter. Indeed, the radiocarbon evidence would not mitigate against the permanent occupation of D6-60 having been confined to a few decades immediately prior to the Historic Period (i.e., the latter part of the eighteenth century). Such a short phase of permanent occupation would readily account for a lack of accumulation of deposits with a high density of faunal materials, as in Site D6-58. Turning lastly to the evidence from Site D6-36, Ke'eke'e Iki shelter, it will also be recalled that this site is so close to the larger Ke'eke'e Nui shelter that the two may well have been utilized by the same social group. The radiocarbon age determination from Feature 4 indicates that the shelter was first used no earlier than the mid-seventeenth century, and possibly not until the later half of the eighteenth century. (The multiple calendar ages produced by radiocarbon age calibration in this time period make a more precise temporal placement impossible; see Stuiver and Becker 1986.) Thus, D6-36 was not utilized until after the shift from temporary to permanent residence in D6-58, and quite possibly not until sometime late in the final century prior to European contact As with D6-60, the Ke'eke'e Il cultural deposit has densities of vertebrate and invertebrate faunal materials somewhat lower than in D6-58. However, D6-36 did yield 39 pig and dog bones, again fitting well with a 127 Anabulu pattern of late prehistoric animal husbandry in the interior valley. The portable artifact array (chapter 8, table 8.1) is the least diverse of the three rockshelter assemblages, but does contain several classes of stone tools (adz flakes, grindstone, polishing stones, coral abraders, and hammerstones), as well as bone awls, fishhooks, and the ubiquitous volcanic glass and basalt flakes. The most likely interpretation of this evidence is that Ke'eke'e Iki rockshelter was utilized permanently for a relatively short period of time, most probably by the same social group occupying nearby Ke'eke'e Nui rockshelter. The individual occupation sequences of the three rockshelters are depicted graphically in figure 9.1. Human penetration of the interior Anahulu Valley must be understood, not as a sudden event, but as a gradual process of increasingly intensive use occuring over at least four centuries. The earlier phases at D6-58 and -60 appear to represent only intermittent use of the shelters, associated with a pattern of exploitation of inland resources by groups still residing permanently in the lowlands. By the early seventeenth century, at least one group had shifted permanendy inland to reside at D6-58, but the permanent occupation of D6-60 did not take place until sometime in the eighteenth century. In both sites, the shift to permanent occupation appears to be associated with a greater emphasis on husbandry of pig and dog, arguably a reflection of some form of agricultural expansion or intensification in the interior valley. This prehistoric sequence from the inland Anahulu Valley may be briefly compared with my proposed archipelago-wide phase sequence (Kirch 1985:298-308), and also with Hommon's (1976, 1986) model of "inland expansion" in Hawaiian prehistory. The inland Anahulu sequence coincides with the Expansion Period (A.D. 1100-1650) and Proto-Historic Period (A.D. 1650-1795) of the sequence proposed on the basis of an archipelago wide synthesis of the several hundreds of excavated and dated sites. The Expansion Period was "the most significant period of cultural change in the entire sequence of Hawaiian prehistory" (Kirch 1985:303), witnessing explosive population growth, expansion of population into previously uninhabited ecological zones (especially leeward valleys and slopes), intensification of production, and reorganization of socio-political structure. The dating of initial human penetration into the Anahulu interior coincides well with the time span of the Expansion Period, and documents yet another case of movement into a new ecological zone, at first on a low-intensity, intermittent basis, and later through permanent occupation. By the Proto-Historic Period, the interior valley seems to have been permanendy occupied by resident inland groups (probably not only at the D6-58 D6-60 D6-36 A.D. 1800- A.D. 1700- A.D. 1600- A.D. 1500- A.D. 1400- A.D. 1300- I I = Permanent Occupation = Temporary Occupation __ = Occupation Uncertain Figure 9.1. Occupation sequences of the Anahulu Valley rockshelters. rockshelters excavated by our project, but at other, as yet untested sites as well). Hommon (1976, 1986) has put forward a slighdy different, although fundamentally similar model in which the major phase of "inland expansion" is dated to ca. A.D. 1400-1600 (his Phase II, Hommon 1986:61-65). This phase coincides well with the initial use of both Sites D6-58 and -60. The process of increasingly intensive human use, and eventual permanent occupation of the inland Anahulu Valley was reflected by a parallel sequence of changes and modifications to the natural environment, both physical and biotic. A major objective of the rockshelter excavations was to track this sequence of environmental change, as recorded in the rockshelter sediments, and in the floral and faunal materials contained within those sediments. Application of a range of interdisciplinary techniques for investigating prehistoric environmental change in Hawai'i-such as palynology, phytolith analysis, landsnail analysis, charcoal identification, geoarchaeological analysis of sediments, and archaeobotanical studies-has been largely a development of the 1980s, building upon the earlier settlement pattern-ecological approaches of the 1970s. Thus, at the time of the Anahulu Project fieldwork in 1982, most of the approaches which we planned to bring to bear on the retrodiction of environmental change in the valley were still experimental, and in some cases (especially with the idendfication of charcoal) hampered by a lack of adequate reference collections. A further limitation to our objective of investigating environmental change is that the ,128 Conclusions rockshelters themselves can be expected to yield evidence of a particularly local sort. The physical sediments and landsnails derive from very localized catchments in the immediate vicinity of the shelters. Carbonized seeds and charcoal probably reflect somewhat larger catchments, but cannot be expected to provide a comprehensive picture of valley vegetation pattems. The faunal assemblages reflect the broadest spatial range (extending indeed to the coastal and marine sections of the valley), but are biased by the fiter of cultural patterns of food choice, as well as of processing and disposal. The varying spatial scales of these catchments from which particular sets of archaeological materials derive must be taken into account in any interpretation of the evidence from the rockshelters. Ideally, a comprehensive study of environmental change in the Anahulu Valley would incorporate evidence from a variety of other depositional contexts aside from the rockshelters, such as alluvial stream sections (for both geoarchaeological sediment analysis and identification of incorporated charcoal), agricultural features (both dryland and irrigation systems), open habitation sites, and bottom sediments from the large coastal ponds (Ukoa and Loko'ea). Such a comprehensive project, however, was well beyond either the objectives or resources of the Anahulu Project. Hunt's study (chapter 3) is, to our knowledge, the first effort to analyze in detail the sedimentological characteristics of any Hawaiian rockshelter deposits. The lack of comparable data from other sites makes interpretation of his results difficult, as we cannot yet know whether the Anahulu shelter sediments differ from those of other sites in similar, or different, geomorphological contexts. This is a situation that will hopefully be remedied as geoarchaeological analyses become accepted as a standard part of excavation programs. The sediment profiles from D6-60 and -36, and individual unit samples from D6-58, certainly do not display any radical changes in sediment characteristics that would imply major shifts in the geomorphological regimes of which the shelters were a part. This evidence thus argues for a relatively low degree of human impact on the slopes surrounding the shelters, the primary source environment for sediment influx to the sites. However, the valley walls in the vicinity of all three sites are quite steep (incorporating substantial cliffs), and were thus unlikely to have been the focus of any agricultural activities. It is rather on the alluvial terraces immediately below the shelters (and thus lying outside of the source catchment for sediments) that any forest clearing and agricultural activities would have been focussed. The landsnail sequences from D6-60 and -36 (chapter 5) present a clear and consistent picture of microenvironmental changes in the immediate vicinities of these sites. The deep D20 column from D6-60 displays three temporal zones or phases, as discussed in detail in chapter 5. The earliest zone represents an undisturbed mesic native forest containing several endemic, aroreal and terrestrial landsnail taxa, such as Auriculella, Achatinella, and Amastra. Initial clearance of this vegetation is reflected by the change to Zone 2, maked by the absence of arboreal taxa and other species requiring relatively wet micro-habitats. Also significant is the appearance, in Zone 2 and also in the D6-36 column, of the anthropophilic snail Lamellaxis gracilis. This species very likely reflects gardening activity in the vicinity of these sites. Finally, Zone 3, which corresponds with the Historic Period (after European contact), represents a new phase of microenvironmental changes, with the disappearance of additional native land taxa, and the introduction of exotics such as Gastrocopta and Hawaiia. In sum, the landsnail evidence is consistent with our interpretation that the occupants of the rockshelters also opened up the pre-human mesic forest canopy through clearing for agricultural activity. Truly radical changes in the local vegetation that supported these endemic snail populations did not occur, however, until the Historic Period. Allen (chapter 6) has thoroughly discussed the problems and limitations of the archaeobotanical evidence obtained by flotation of column sediment samples and from features. Her cautious interpretation of these data is that they reflect disturbed conditions in the valley's vegetation throughout the period that the shelters were occupied, and that such disturbance was most likely due to agricultural activity. The archaeobotanical evidence thus meshes well with that of the landsnail analysis. It is also worth remarking that the absence of seeds of hydrophilic taxa (especially Ludwigia octovalvis, Scirpus juncoides, or Cyperus spp.), even though negative evidence, supports the archaeological field evidence that pondfield irrigation systems were not constructed in the upper valley until early in the nineteenth century (Kirch and Sahlins, in prep.). The analysis of charcoal from the Site D6-60 column, by Murakami (chapter 7), was substantially limited by the incomplete state of available reference collections, thus requiring the use of "surrogate taxa" for which actual taxonomic identifications are not available. Certainly the charcoal data do not support a picture of radical environmental changes in the interior valley during the period of prehistoric human occupation. Neither are they inconsistent with our interpretation of agricultural activities on the valley floor and lower slopes, based on the landsnail and archaeobotanical data. Perhaps the most significant 129 Anahulu result of the charcoal analysis is the demonstration that two species of cultivated trees, beadfruit (Artocarpus altilis) and the Malay apple (Eugenia malaccensis), were present in the valley throughout the period of prehistoric occupation of D6-60, and presumably in the interior valley itself, as it seems unlikely that filrewood would have been carried 6 km from the coast to these shelters. The potential of charcoal identification to yield significant evidence on the nature of prehistoric Hawaiian cultivation systems, as well as on changes in the native vegetation communities, is clearly considerable, once problems of adequate reference materials have been overcome. The marine fauna in the rockshelters tells us nothing of environmental change within the inland part of the valley, but the presence of the diadromous snail Neritina granosa is significant, as this species would have required permanent flow of the Anahulu Stream to survive. Historic Period changes in the valley's hydrologic regime (particularly the use of streamflow for sugarcane irrigation) have led to the exipation of N. granosa from the watercourse. Finally, there is the evidence of the cultura1ly- deposited faunal assemblages (both vertebrate and invertebrate) which, as we have noted, were certainly derived from an extensive catchment area, and may therefore not necessarily reflect conditions in the interior valley per se. Of great interest is the range of indigenous and endemic birds represented in all rockshelter deposits. No less than five species of procellarids are present, and surely indicate that nesting populations of these seabirds were present within the Anahulu Valley environment until quite late in the prehistoric period. Also notable is the Bonin Petrel (Pterodroma hypoleuca), whose historically-documented range does not even extend to the main Hawaiian Ishlands. Thus, the Anahulu rockshelter evidence reinforces the interpretations of Olson and James (1982) that there were massive reductions in the resident populations of seabirds in the main Hawaiian Islands during the period of Hawaiian occupation. Also of note are the several species of native land birds represented in the rockshelter assemblages, most of which are today either extinct on O'ahu, or threatened with extinction. It would appear that exploitation of these forest birds was one activity of the rockshelter occupants throughout the prehistoric period. REFERENCES CITED Goldman, I. 1970. Ancient Polynesian Society. University of Chicago Press. Chicago. Hommon, R.J. 1976. "The Formation of Primitive States in Pre-Contact Hawaii." Ph.D. Dissertation, Univ. of Arizona. (University Microfilms.) . 1986. Social evolution in ancient Hawaii. In P.V. Kirch, ed., Island Societies: Archaeological Approaches to Evolution and Transformation, pp. 55-68. Cambridge University Press. Kamakau, S. 1961. Ruling Chiefs of Hawaii. Kamehameha Schools Press. Honolulu. Kirch, P.V. 1979. Late Prehistoric and Early Historic Settlement-Subsistence Systems in the Anahulu Valley, O'ahu. Anthropology Dept. Report 79-2. B.P. Bishop Museum. Honolulu. . 1985. Feathered Gods and Fishhooks: An Introduction to Hawaiian Archaeology and Prehistory. University of Hawaii Press. Honolulu. Olson, S., and H. James. 1982. Prodromus of the Fossil Avaifauna of the Hawaiian Islands. Smithsonian Contributions to Zoology 365. Washington, D.C. Sahlins, M. 1958. Social Stratification in Polynesia. American Ethnological Society. Seattle. Sahlins, M., and P.V. Kirch. in prep. Anahulu: The Archaeology of History in the Early Sandwich Islands Kingdom. Stuiver, M., and B. Becker. 1986. High-precision decadal calibration of the radiocarbon time scale, A.D. 1950-2500 B.C. Radiocarbon 28(2B):863- 910. 130