CHAPTER SIX ARCHAEOBOTANICAL ASSEMBLAGES FROM THE ANAHULU ROCKSHELTERS by Melinda S. Allen T I AnANLANDscAPE at contact was the product of a lengthy and dynamic interaction between human populations and the environment. The extant Hawaiian flora in particular reflects three inter-related anthropogenic processes: (1) biotic dispersals by Polynesian colonists, both intentional and accidental; (2) reduction, extirpation, and disjunction of native species as a result of deforestation, anthropogenic dispersals, and direct human competition; and (3) creation and maintenance of anthopogenic environments. These three processes were repeated throughout Hawaiian prehistory at different scales, intensities, and rates in various localities, and they continue today. Their role in the transformation of one Hawaiian valley system, Anahulu Valley on the northwest coast of O'ahu Island, is the focus of this paper, as evidenced through the analysis of archaeobotanical assemblages from three inland rockshelters. Our understanding of changes in the native Hawaiian vegetation has been aided by several lines of paleobotanical inquiry including wood-charcoal analyses, flotation studies, ethnohistoric research, palynology, and phytolith work (see review in Allen 1984a). Analysis of fme-grained archaeological sedinmts has identified the prehistoric presence of a suite of weedy taxa associated with human occupations, some which were once thought to be European introductions; these taxa were probably inadvertant Polynesian introductions given their pan-Pacific occurrence and consistent agricultural associations (Allen 1981; see also St. John 1976). Wood-charcoal studies indicate both the removal of native forest and reduction in the number of species in the ecologically more fragile dryland zones (Murakami 1983a,b,c; see also McEldowney 1983). Similarly, ethnohistoric research suggests that most of the lowland forest had been replaced with agricultural fields by the time of European contact (cf. McEldowney 1976). The present research centers on archaeobotanical materials from a series of archaeological sites within the Anahulu Valley. Prior work in the area indicated that the relatively dry rockshelters favored botanical preservation (Kirch 1979:46-7). In light of this, special care was taken to recover plant materials during excavation and several bulk sediment samples were taken for flotation. This study considers temporal trends in the mid-valley vegetation as reflected in three column samples from Kuolulo (D6-60) and Ke'eke'e Iki (D6-36) rockshelters. Secondly, prehistoric plant utilization is addressed through a series of discrete feature samples from the Ke'eke'e Nui (D6-58) rockshelter. The vegetation within Anahulu Valley today is a mixture of introduced and native elements, with the former predominating in the lower valley and the latter in the upper valley (see also Kirch 1979:5-7; chapter 1). Vegetation zonation within Anahulu is comparable to that described by Hosaka (1937) for Kipapa gulch, another large leeward O'ahu valley system. As in Kipapa, the lower Anahulu Valley is characterized by a xerohytic flora, including koa-haole (Leucaena leucocephala Lam.] de Wit), prickly pear (Opuntia 84 Anahulu TABLE 6.1 VOLUME, WEIGHT, NUMBER OF SEEDS, AND NUMBER OF TAXA BY SAMPLE Sample Volume Weight No. Carbonized No. Carbonized (Ml) (gm) Seeds Taxa Site D6-60 (Kuoluluo Shelter), Column A *0-5 cm 350 366 0 0 5-10 cm 350 364 5 2 10-15 cm 300 280 7 3 15-20 cm 150 162 23 7 20-25 cm 250 279 18 4 25-30 cm 150 172 18 4 30-35 cm 200 232 27 4 35-40 cm 300 318 31 5 40-45 cm 350 430 16 3 45-50 cm 300 346 24 4 Site D6-36 (Ke'eke'e Shelter), Column B *05 cm 250 318 0 0 5-10 cm 300 342 4 3 10-15 cm 300 343 11 4 15-20 cm 100 132 4 2 20-25 cm 150 140 3 3 25-30 cm 75 62 12 4 30-35 cm 50 61 3 2 *35-40 cm 30 72 0 0 40-45 cm 150 180 2 1 *45-50 cm 75 90 0 0 Site D6-60 (Kuolulo Shelter), Column C *0.5 cm 300 347 0 0 *5-10 cm 450 551 0 0 10-15 cm 400 481 8 2 15-20 cm 450 455 27 8 20-25 cm 450 481 44 8 25-30 cm 450 478 33 8 30-35 cm 500 471 27 6 35-40 cm 500 471 23 7 40-45 cm 400 443 27 7 45-50 cm 450 451 30 7 50-55 cm 500 597 26 2 55-60 cm 500 493 26 4 60-65 cm 450 492 26 5 65-70 cm 500 591 30 3 70-75 cm 500 584 48 5 *75-80 cm 100 166 0 0 *80-85 cm 200 294 0 0 * These samples are not included in fig. 6.2 or in the reported statistics. a The number of carbonized taxa for each sample does not include non-diagnostic unidentified specimens which may potentially include one or more taxa. b NA-information not available. Archaeobotanical Assemblages TABLE 6.1, Continued Lab Designation Volume Weight No. Carbonized No. Carbonized & Provenience (ml) (gm) Seeds Taxaz Feature Samples Site D6-58 Unit 011 (A) Fe 6 (oven) 15-18 cmbs 500 501 10 1 (B) Fe 6 (oven) 17-21 cmbs 500 605 3 2 (D) Lens 72-74 cmbs 500 643 19 2 (E) Fe 9 (ash lens) 74-77 cmbs 500 576 13 2 (F) Fe 9 (ash lens) 80-85 cmbs 500 594 55 11 Unit K10 (G) Fe 1 (sm hearth) 17-28 cmbs NAb NAb 65 10 (H) Fe 5 (oven) 55-65 cmbs 500 444 19 1 a The number of carbonized taxa for each sample does not include include one or more taxa. b NA=information not available. megacantha Salm-Dyck), lantana (Lantana camara L.), and a host of other weedy exotics. The middle valley corresponds to Hosaka's "Guava Zone" and is more mesic in composition. Koa-haole continues and Christmas berry (Schinus terebinthifolius Raddi) becomes an important element. Candlenut (Aleurites molucanna [L.] Willd.) frequents the wetter areas, such as gully bottoms. In this zone, several economic plants, including both traditional Polynesian cultigens and historic introductions, are closely associated with the archaeological structural remains. Coffee, bananas, sweet potato, and taro are feral, while formerly cultivated lime, orange, mango, and coconut trees persist. Only the upper valley has a predominantly native vegetation cover, comparable in composition to Hosaka's "Koa Zone" (Acacia koa Gray); this grades into a higher elevation "Ohia Zone," dominated by the native Metrosideros collina (J.R. and G. Forst.) Gray. METHODS In the field, the three column samples were taken in 5 cm increments, varying in volume from 30 to 500 ml (table 6.1). Samples of more variable sizes were taken from discrete features; in all but two cases, 500 ml of each feature sample were analyzed. All samples were processed in the laboratory using a water sepearation device adapted from Hommon (1983:19-20, 23), a smaller and simpler version of the one described by Watson (1976). Whenever possible a sample of at least 500 ml was processed; unprocessed non-diagnostic unidentifiedspecimens which may potentially portions of several samples are currently conserved in the Archaeology Division of the B.P. Bishop Museum in Honolulu. Floated materials were passed through standard geologic wire mesh screens of 4, 2, 1.5, and .5 mm size during flotation and then air-dried separately. This portion of the analysis was conducted by B.P. Bishop Museum archaeological laboratory assistant Mary Riford. Dried flotation samples were sorted under a binocular microscope at 8X. All material from each size fraction was sorted, the bulk of which consisted of unidentifiable charcoal fragments. The heavy fraction (i.e., that which did not float) was not analyzed. Identifications are based on comparisons with B.P. Bishop Museum Herbarium collections and a vouchered reference collection of seeds and other plant parts. The term "seed" is used here in a general sense to include a variety of seed/fruit types including caryopses, arils, and mericarps. All Latin names follow St. John (1973) unless otherwise noted. In the following discussion, the designation "cf." indicates that the cited taxon is the most likely candidate. However, because of the small amount or poor condition of the materials recovered, or the absence of appropriate materials (i.e., reproductive parts), a conclusive identification is not possible. Use of the designation "?" indicates that the specimen resembles the cited taxon, but may resemble other taxa as well. In the following tables, counts are provided for individual seeds. Other materials which cannot be identified as "natural units" are treated as attributes 85 Anahulu rather than variables (see Grayson 1981). Problems inherent in quantification of plant materials and the relationship of these measures to the parameters of interest are discussed in more detail below. RESULTS Forty-four samples, totaling approximately 14 liters of sediment, were processed. Twenty taxa were identified at least to the family level and over half of these were carbonized (see table 6.2). In addition, 32 unidentified but diagnostic carbonized taxa were recorded. The present study focuses on the carbonized specimens because of the difficulty of determining the origins of non-carbonized plant remains, particularly small seeds (see Lopinot and Brussell 1982; Keepax 1977; Minnis 1981). Modem seeds can be introduced into archaeological contexts by several mechanisms, including soil-burrowing insects and rodents, aerial contamination during field work, and during the flota- tion process. Herein, except in the case of traditional cultigens, a conservative approach is taken and non- carbonized materials are considered recent contaminants. For carbonized taxa which could be identified to the family level, and for traditional cultigens, the preferred habitat, phytogeographical distribution, previous archaeological occurrences, and ethnographically documented uses are discussed below. The distribution of carbonized specimens and traditional cultigens by sample units is provided in tables 6.3 to 6.6. SYSTEMATIC REVIEW MONOCOTYLEDONS Family CYPERACEAE Eleocharis sp. (kohekohe) Eleocharis is a small sedge represented in the Hawaiian Isands by six species: two native, three adventive, and one recent introduction. This genus is frequently associated with wet environments such as freshwater marshes and cultivated wetlands (Stemmermann 1981:42-3). The bristles at the base of the fruit aid in animal dispersal. There are no ethonographically recorded uses for the genus. An archaeological example from Anahulu is illustrated in fig. 6.1. In addition to Eleocharis, seeds of two unidentified sedges were recovered from several samples; one of these may be a species of Carex. Family GRAMINEAE Eragrostis sp. ('emoloa, kalamalo) This genus includes twelve native Hawaiian species, all of which are low herbs. The recovered seeds most closely resemble E. variabilis (Gaud.) Hbd., a perennial grass which frequently occurs under mesic conditions. Once widespread in the islands, Hitchcock (1974:103) recorded it as the dominant species on the foothills behind Honolulu in the early 1900's. Archaeologically, Eragrostis has also been recovered Figure 6.1. A carbonized Eleocharis achene with bristles partially intact, Kuolulo shelter (80X). 86 Archaeobotanical Assemblages TABLE 6.2 SUMMARY OF IDENTIFIED TAXA Taxonomic Carbonized or Not Plant Part Identification Pteropytza Unidentified NC stem Monocotyledon Arceae (?) NC corm cf. Colocasia esculenta Cyperaceae Eleocharis sp. C seed Unid cf. Carex C seed Unidentified sp. (spp.) C seeds Gramineae Eragrostis sp. C seed Unidentified NC/C seed Dicotyledon Amarantacee NC seed Chenopodiaceae Chenopodium cf. oahuense NC/C seed Compositae (?) NC seed Cucurbitaceae Lagenaria siceraria C seed/fruit Sicyos sp. NC/C seed Euphorbiaceae Aleurites cf. molucanna NC/C testa Leguininosae Unidentified C seed Malvaceae Sida sp. C seed Mrytacae Psidium sp. NC seed/fruit Oxalidaceae Oxais cf. corniculata NC seed Rubiaceae (?) cf. Morinda citrjfolia NC seed Sapindacee Melia azearach NC fruit Solanaceae Solanum sp. NC/C seed Sterculiaceae Waltheria cf. americana C seed 87 88 Anahadu TABLE 6.3 DISTRIBUTION OF CARBONIZED SEEDS BY LEVEL FOR COLUMN A SITE D6-60 (KUOLULO SHELTER) TAXAa LEVELS 1 2 3 4 5 6 7 8 9 10 Chenopodiw,eae Chenopodum 3 5 13 11 12 21 26 13 21 Cucurbitacea Lagenaria rind Sicyos 1 2 1 1 Euphorbiaceae Aleurites testab - - x x x x x x x Gramineae Eragrostis Unidendfied rhizome/stem Leguminosae Unidentified Seed 1 -_ - 1 - _- - 1 2 Sterculiaceae Waltheria - 1 Unidentified Seed F Unidentified Seed H Unidentified Seed I Unidentified Seed J Unidentfied Seed K Unidenified Seed L Unidentified Seed M - -_- - - 1 1 3 1 1 1 1 3 1 Non-diagnostic Unid. Seeds 1 - - - - 1 1 Total No. of Taxac Total Carbonized Seeds 1 x 3 4 Malvaceae Sida 4 x 1 x 1 Solaaceae Solanum 1 0 0 2 5 3 7 7 23 4 18 4 18 4 27 a All specimens are seeds unless otherwise indicated. b Both caibonized and uncarbonized testae are included here. c These totals do not include non-diagnostic unidentified specimens which could represent one or more taxa. 5 31 3 16 4 24 1 1 89 Archaeobotanical Asmblages TABLE 6.4 DISTRIBUTION OF CARBONIZED SEEDS BY LEVEL FOR COLUMN C SITE D6-60 (KUOLULO SHELTER) TAXAb ,EVELSa 2 3 4 5 6 7 8 9 10 1 1 12 13 14 15 Chenopodiaceae Chenopodium Cucurbitaceae Lagenaria seed Lagenaria rind Sicyos Cyperaceae Eleocharis Unid cf. Carex Unidentified Euphorbiaceae Aleurites testac Gramineae Eragrostis Unidentified seed Unidentified rhizome/stem Leguminosae Unidentified Seed Malvaceae Sida Solanaceae Solanum Sterculiaceae Waltheria Unidentified Seed A Unidentified Seed N Unidentified Seed 0 Unidentified Seed P Unidentified Seed Q Unidentified Seed R Unidentified Seed S Unidentified Seed T Unidentified Seed U Unidentified Seed W Unidentified Seed X Unidentified Seed Y Non-diagnostic Unid. Seeds Total No. of Taxad Total Carbonized Seeds - 7 14 25 19 21 15 18 - - - 1? - - ? ? ? - X 2 6 2 1 1 1 - 1 -- 1 . x- xx x xx x 1 6 2 4- 3 2 - - - 1 - 1 1 - x x x x X 19 21 20 x 3 1 20 27 42 x 2 x x x 2 x 5 x 3 1 x 3 1 3 1 2 1 - x x x - - - - 1 1 1 1 - 1 3 3 2 1 1 - - - - - 1 2 1- - - -- - 1 - - - 1 - - ~~1 - - - 1 _ _-2--2? 1 -1 1 -1 1 - - - 3 1 -- 2 1 0 2 8 8 8 6 7 7 7 o 8 27 44 33 27 23 27 30 - - 1 2 26 4 26 5 26 - 1 3 30 1 5 48 a No specimens were recovered from levels 1, 16, and 17. b All specimens are seeds unless otherwise indicated. c Both carbonized and uncaibonized testae are included here. d These totals do not include non-diagnostic unidentified specimens which could represent one or more taxa. 90 Anahulu TABLE 6.5 DISTRIBUTION OF CARBONIZED SEEDS BY LEVEL FOR COLUMN B SITE D6-36 (KE'EKE'E IKI SHELTER) TAXAa LEVELS 1 2 3 4 5 6 7 8 9 10 Chenopodiaceae Chenopodium - 2 6 3 1 8 2 2 Cucurbitaceae Lagenaria seed - 1 Cyperaceae Unidentfied cf. Care- - 1 1 - Unidentified seed - 1 Euphorbiaceae Aleurites testab X X X Gramineae Eragrostis 1 - 1 Unidentified rhizome/stem - X X Unidentified Seed A 1 - - - - Unidentified Seed C - - 1 Unidentified Seed D - 1 - 1 Unidentified Seed E - 1 - Unidentified Seed K - - 2 Non-diagnostic Unid. Seeds 1 1 Total No. of Taxac 0 3 4 2 3 4 2 0 1 0 Total Carbonized Seeds 0 4 11 4 3 12 3 0 2 0 a All specimens are seeds unless otherwise indicated. b Both carbonized and uncarbonized testae are included here. c These totals do not include non-diagnostic unidentified specimens which could represent one or more taxa. from cultural deposits on Kaho'olawe (Allen 1984b) and at Kuakini, North Kona (Allen 1984c). There are no ethnographically known uses for the genus. Unidentified sp(p). In addition to Eragrostis, rhizomes, stems, and seeds of one or more unidentified grass species were recovered. Family CHENOPODIACEAE Chenopodiwn cf. oahuense (Meyen) Aellen ('aweoweo, 'aheahea) Seven species of Chenopodium occur in the Hawaiian Islands. Of these, five are European introductions and two are endemic. C. oahuense is the more common endemic species, while C. pekeloi Deg., Deg. and Aellen. is restricted to Moloka'i. The archaeological specimens compare well with C. oahuense on the basis of seed coat sculpturing. C. oahuense has been recorded from sea level to 2,515 m elevation (Hart and Neal 1940:264). Although generally found as a shrub today, arborescent individuals were apparently formerly more common. Hillebrand (1981:380), for example, observed a 12 to 15 foot tree in the upper woods of Mauna Kea. Selling (1948) used C. oahuense as a key indicator species in his Hawaiian pollen sequence. He suggested it moved into areas of previously mesophytic vegetation when the local climate became increasingly arid. ArchaeobotawucalAssemblages 91 TABLE 6.6 CARBONIZED TAXA FROM CULTURAL FEATURES SITE D6-58 (KE'EKE'E NUI SHELTER) TAXAa LAB DESIGNATION A B D E F G H Chenopodiaceae Chenopodium 9 10 8 34 23 19 Cucurbitaceae Lagenaria seed 2 Lagenaria rind - X Sicyos - 1 Cyperaceae Eleocharis - 10 Unidentified sp(p). 1 - 1 Unidentified cf. Carex 4 4 1 Euphorbiaceae Aleurites testab X X X X X X Gramineae Eragrostis - - - 13 Unidentified stems - - X Leguminosae Unidentified seed 1 Malvaceae Sida 4 1? Solanaceae Soanum- 5 4 Unidentified Seed A 2 Unidentified Seed K - 1 Unidentified Seed Z - - 1 Unidentified Seed AA - - 1 Unidentified Seed BB - 1 Unidentified Seed CC 1 Unidentified Seed DD 3 Unidentified Seed EE - - - 2 Unidentified Seed FF - - 3 Non-diagnostic Unid. Seeds 1 1 5 1 2 4 TotalNo.ofTaxaC 1 2 2 2 11 10 1 Total Carbonized Seeds 10 3 19 13 55 65 19 a All specimens are seeds unless otherwise indicated. b Both carbonized and uncarbonized testae are included here. c These totals do not include non-diagnostic unidentified specimens which could represent one or more taxa. Anahu A single individual of Chenopoduwn produces thousands of seeds on an annual basis. Each seed is enclosed within a glandular utricle which aids in dispersal. It is the most ubiquitous seed in Hawaiian archaeobotanical assemblages; Allen (1983b) suggests it may have been a common agricultural fallow component in drier localities. Chenopodiwn wood- charcoal has also been recovered from several archaeological contexts (Murakami 1983a,b,c; chapter 7, this volume) which could in part account for the presence of seeds within features. An alternative possibility is that Chenopodium was used as a food (see discussion in Allen 1983b:164-7), as has been recorded ethnographically (Malo 1951:23; Hillebrand 1981:380; Buck 1964:6; Handy and Handy 1972:235). Family CUCURBITACEAE Lagenaria siceraria (Mollina) Standl (ipu) One complete seed and two fragments of Lagenaria siceraria were recovered from the column samples; several pieces of possible fruit exocarp were also found. Lagenaria was a Hawaiian cultigen typically grown in drier localities (Handy and Handy 1972:212-22). Gourd rinds are commonly encountered in archaeobotanical assemblages, particularly those from rockshelters. Ethnographic accounts indicate that gourds were used for water and storage containers; immature specimens were edible. The occurrence of Lagenaria seeds in the Ke!eke'e Iki shelter may indicate that gourds were grown in the vicinity or that gourds were cleaned and prepared for use in the rockshelter. Sicyos sp. ('anunu) Sicyos is a native Hawaiian genus with several representatives. All are climbing vines which tend to inhabit drier localities. Archaeologically, the seeds have also been recovered from Kaho'olawe Island (Allen 1983b, 1984b). There are no recorded uses for Sicyos; although not generally recognized as edible, the leaves of many cucurbits are palatable (cf. Purseglove 1968 100-36) and its use as a dietary supplement should not be excluded. Family EUPHORBIACEAE Aleurites cf. molucanna (L.) Willd. (kukui) The hard seed testae of Aleurites were distributed tiroughout the samples and one kemal (the endosperm) was identified. Aleurites, or candlenut, is a common component of Hawaiian forests in areas which have been previously inhabited. It is considered to be a Polynesian introduction to the islands (St. John 1973:210). Although this has not been decisively documented, at present there is no evidence to the contrary. Aleurites thrives in mesophytic forests. Ethnographically, the oily kernals were burnt as a source of light; several kemals strung on a coconut mid- rib or other foundation provided a reasonably strong light with each kernal buring for two to three minutes (Buck 1964:107). The kernals are also part of a traditional Hawaiian condiment (inamona). The hard seed coats are the most ubiquitous, and often dominant, botanical component of Hawaiian archaeological sites. The kemals have also been recovered archaeologically from the Mauna Kea Adze Quarry (Allen 1981). Family LEGUMINOSAE Unidendfied sp. These seeds measure ca. 2 to 3 mm in length; the location of the hilum places them within the subfamily Papilionoidea. Unfortunately, within legume subfami- lies the seeds are often morphologically quite similar, differing primarily in size and surface coloring. The archaeological seeds compare reasonably well with Tephrosia, a common native weed which occurs in mesic areas, but may resemble other taxa as well. Similar legume seeds have been recorded from other Hawaiian archaeobotanical assemblages (table 6.7). Family MALVACEAE Sida sp. ('ilina) The genus Sida includes seven extant Hawaiian representatives, three of which are native. Sidafallax is the most common native species, occurring in drier localities from sea level to an altitude of more than 610 m (Neal 1965:552). The seeds (actually the endosperm) of these species are virtually indistinguishable; in contrast the outer seed cases, which are not present on the archaeological specimens, are quite distinctive. In addition to known species, the possibility of extinct types should be considered. Botanist David Nelson of the 1778 Cook Expedition collected two species of Sida from the Kona Distict, Hawai'i Island which have not been observed since (St. John 1976). Other regions of Hawai'i may also have harbored unique species which were never collected. The Hawaiian representatives of this genus are common in dry areas and frequently occur as weeds under disturbed conditions. Several ethnographic uses have been recorded for Sida. These include use of the roots and flowers as medicinals, woody stems for slats in house construction, and supple stems for "temporary baskets" (Neal 1965:553). 92 Archaeobotanical Assemblages Family SOLANACEAE Solanum sp. Several probable Solanum seeds were recovered; many seeds of this family are morphologically similar. Of those species examined, however, the archaeological materials compare most closely with S. nigrum L. (popolo). S. nigrum is a herbaceous weed frequently associated with cultivation. The taxon may have been inadvertantly introduced into the Hawaiian Islands by Polynesian colonizers on rootstocks such as taro (Colocasia esculenta [L.] Schott). Alternatively, the plant could have been a purposive introduction given its ethnographically recorded medicinal properties (Neal 1965:744) and its edible berries and leaves (Hillebrand 1981:307; Buck 1964:6; Parham 1943). Family STERCULIACEAE Waltheria cf. americana L. (hi'aloa, 'uhaloa) Two native Hawaiian species are recorded for this genus but one is rare (W. pyrolaefolia Gray) and its status as a distinct species is debatable (C. Lamoureux, pers. comm., 1982). W. americana is a perennial which occurs in arid and disturbed contexts, generally at lower elevations. The plant has been ethnographically recorded as a medicinal (Neal 1965:575; Territorial Board of Health 1922). EXPECTATIONS AND LIMITATIONS OF THE STUDY A Model of Temporal Change in the Valley Vegetation A simple model derived from the ecological literature (Mueller-Dombois and Ellenberg 1974; May 1978) predicts the sequence of vegetation changes which would accompany the gradual conversion of a previously unoccupied Hawaiian valley into a productive agricultural system. Archaeological evidence suggests that initial use of the Anahulu rockshelters was temporary and intermittent (chapter 2, this volume). During this early period, some weedy lowland taxa, particularly those adapted to animal dispersal, might first appear in the rockshelter deposits. Early activities in the middle valley might also have included the creation of small "forest gardens" of the kind described in the ethnographic liteture (Handy and Handy 1972) for famine relief and for use during forays into the island interior. The impact of these activities on the native vegetation, however, would have been quite localized and would be difficult to identify archaeologically. Larger scale agriculturl endeavors should be more clearly recognizable in the archaeobotanical record. Clearance of native vegetation would result in a decrease in the abundance of native taxa. The richness values (i.e., number of taxa) for native species might also decline as rar forest elements were removed. Concomi- tantly, weedy taxa would increase in both abundance and richness. These patterns would be expected to occur contemporaneously in a number of sites. If shifting cultivation were practiced in the valley, successional sequences might be observable. The succession literature (e.g., Tilman 1984:213-300; May 1978; Smathers and Mueller-Dombois 1974) predicts a pattern of few weedy taxa early in a succession sequence, followed by increasing species enrichment and increasing arboreal elements. Given that shifting cultivation follows a rotational sequence of cropping and fallow, the valley vegetation would consist of a mosaic of different successional progressions at any given time. This latter pattem might not be archaeologically detectable, however, given the resolution of current dating techniques. A shift to permanent agriculture, specifically irrigated field systems, would be reflected in a further reduction, or possibly the disappearance, of native taxa from the archaeobotanical record. The development of permanent irigated fields might be accompanied by the establishment of a new and distinctive suite of hydrophilic weeds. Using the results from the three column samples, the Anahulu archaeobotanical assemblages were compared with the sequence of vegetation changes outlined above. Deriving from two separate shelters, the columns potentially offered two independent monitors of major vegetation changes within the valley over a period of roughly 500 years (chapter 2, this volume). The earliest date, A.D. 1280-1430 (corrected), derives from the basal ash lens of the Ke'eke'e Nui shelter. A second sample from the same shelter indicates continued use into the period A.D. 1500 to 1700, while a third sample is indistinguishable from modern. In the Kuolulo shelter a radiocarbon sample from near the base of Column C provided a corrected date of A.D. 1412- 1465, indicating that this column represents paleoenvironmental conditions which postcede initial middle valley use. Similarly, a Ke'eke'e Iki shelter sample yielded a date in the A.D. 1600-1700 range; thus Column B also reflects valley conditions after several hundred years of occupation and probably after the establishment of shifting cultivation. Dates from the pondfield features (Kirch and Spriggs, in press), in combination with the ethnohistoric literature, indicate that these irrigated systems were not established until the historic period. 93 Anahulu Cultural Plant Utilization Hearth and oven features are likely Waps for discard materials from food processing, or for food remains themselves. In the Anahulu case, such features were regarded as possible sources of information on cultigens being grown in the vally, or on the exploitation of native plant resources. Also of interest was a pattem of taxonomic richness identified in an earlier analysis of Kaho'olawe Island archaeobotanical materials (Allen 1984b). On Kaho'olawe, samples from large earth ovens (imu) tended to have richer seed assemblages than those from small hearths. Allen suggested this difference might be related to feature functions, specifically the use of plant materials in the oven steaming process. The small hearths on Kaho'olawe, on the other hand, were comparatively richer in bone and shell but contained few seeds. Consideration of Sample Size The measurement of relative abundances and taxonomic richness are critical to the monitoring of vegetation trends and analysis of other archaeobotanical patterns. These measures have recendy received considerable attention in the zooarchaeological literature (e.g., Grayson 1979, 1981, 1984), but their importance for archaeobotanical studies has been neglected. The statistical relationships between these two measures (relative abundance and richness) and sample size (number of identifled specimens or NISP) are considered here in some detail before further interpretation of the Anahulu evidence. The absolute number of any one taxon in the Anahulu assemblages is generally so small that a consideration of relative abundance is not meaningful. As discussed below, this is a clear indication that larger samples must be processed in the future. The exception is Chenopodium. which occurs in all three column samples and in six out of seven features. As a potential indicator of disurbance, temporal and spatial variation in the relative abundance of this taxon relative to others could be ecologically significant. Alternadvely, vari- ability in Chenopodium abundances might be nothing more than a reflection of variability in NISP. Looking at Column C, the column with the greatest temporal duration and the largest NISP values, Spearman's rho rank-order correlation coefficient indicates that NISP and the relative abundance of Chenopodium are not significandy correlated (rs=-.198, p=.517). A similar situation holds for Column A (rs=.477, p=. 195). When the levels with very small NISP values are removed (cf. Grayson 1984:121), the correlations are further reduced (Column C: rs=-.078, p=.809; Column A: rs=.378, -p-.403). Column B, unfortunately, has such small NISP values that the relative abundance of Chenopodium in that column can not be meaningfully evaluated. Thus, changes in the relative abundance of Chenopodium may signify important ecological changes (see discussion in next section). Changing taxonomic richness (NTAXA) values was a second parameter of interest here. Grayson (1984:131-67) also discusses the potential statistical relationship between NISP and NTAXA. As with relative abundance, if variation in NTAXA cannot be deomonstrated to be independent from NISP, then cultural or ecological interpretations of variability in NTAXA are of questionable validity. For Column C, the correlation between loglO NTAXA and loglO NISP is not highly significant (r=.554, p=.025), and the correlation is gready reduced when level 3 with its small NISP value is removed (r=.186, p= .28 1). Conversely, loglO NTAXA and loglO NISP are highly correlated in the case of Colum A (r=.8095, p=.004), but the correlation is significantly reduced when the strata with small NISP values are removed (r=.6273, p=.048). When the feature samples are considered as a group, the correlation between NISP and NTAXA is even stronger (r=.945, p>.00l), with NISP accounting for over 87% of the variation in NTAXA. The small number of cases, however, makes interpretation of this correlation somewhat problematic. The foregoing suggests that only in the case of Column C can we discuss changes in richness values and be confident that these changes are independent of changing sample sizes. In this study, sample size was also considered in terms of the volume of sediment processed. This param- eter was of interest from two perspectives. First, in a given environment can we predict the volume of sediment needed so as to avoid sample size problems? Second, does the relative homogeneity (or heterogenei- ty) of seed densities across samples inform on accumu- lation mechanisms, including those of a cultural nature. The volumes of sediment processed from the column samples were statistically compared with NTAXA and NISP recovered (see fig. 6.2). NISP is moderately correlated with sample volume (r=.648, p>.00l), with the latter accounting for 40% of the variation in NISP. This suggests that seed densities are fairly homogeneous within the deposits. Independent analysis of sedimentological characters (chapter 3, this volume) further suggests continuous, even sediment accumulation of a relatively unchanging nature throughout the occupation of sites D6-36 and -60. Thus, the seeds present may represent natural seed "rain" rather than cultural activities. Methodologically, the results suggest that under similar conditions of constant sediment deposition it may be possible to predict 94 Archaeobotanical Assemblages adequate sample volumes for processing after some initial trial runs. For NTAXA, there is a low correlation with sample volume (r=.435, p<.0l), suggesting that other factors are more important in determining NTAXA. In the case of the feature samples, sample volume was a constant. Nevertheless, the feature samples were quite variable in NISP and NTAXA, even more so than the colulmn samples which were derived from different sample volumes. This suggests that other factors are controlling NISP and NTAXA. Potential sources of variation include functional differences between the features or variable preservation contexts (see further discussion below). The evidence suggests that assessing representative sample volumes for processing from features may be more problematic and analysis of the entire feature rather than samples may be necessary. The Role of Accumulation Mechanisms Knowledge of the accumulation mechanisms involved in deposit formation is important in terms of: (1) understanding the quantitative relationship between the recovered sample and the target population (usually the environment and/or subsistence); and (2) identifying the agent responsible for the materials deposited. Figure 6.3 models the more important factors which may intervene between two common target variables and the archaeobotanical sample (additional factors may enter during the analysis itself). The left hand side of the diagram emphasizes cultural biases, while the right hand side oudines natural factors. At each step these biases alter the quantitative relationship between the target population and the archaeological sample. Species may be differentially affected depending on their economic importance or associations with economic plants, natural modes of dispersal, morphological characteristics relevant to physical transport, and physio-chemical characters which affect preservation potential. Furthermore, differential seed production may skew the record in favor of heavy producers as is the case with pollen (Davis 1963). Given that these mechanisms are difficult to reliably identify, the relationship between the quantitative structure of the target variables and the sample remains imprecise (see Grayson 1981). For example, variability in natural seed abundances is difficult to impossible to separate from that which might reflect changing ecological conditions or cultural factors. Grayson (1981) suggests that for single site assemblages, taxa should be treated as attributes and presence/absence analysis used. While presence/absence studies are not without inherent hazards, they employ fewer assunptions than techniques based on taxonomic abundances. If ordinal or greater level of measurement is desired, then it is necessary to obtain parallel sequences of changing taxonomic frequencies from other localities within the same region (cf. Grayson 1984). Only then can the vagaries of preservation and deposition be discounted. A second critical factor in interpretation of archaeobotanical assemblages lies in discerning between natural and cultural depositional agents. Many seeds may be incorporated into archaeological sediments through natural seed "rain." The depositional agent(s) responsible for small seeds, such as those considered here, would be the most difficult to identify. Furthermore, even if a cultural agent can be demonstrated, it is quite difficult to determine what the economic importance of a given taxa might be on archaeological criteria alone. This issue is considered in more detail below. DISCUSSION OF THE ANAHULU ASSEMBLAGE Vegetation Change The sequence of vegetation changes modeled above predicts a tnd of decreasing abundance and richness values for native taxa, and increasing values for weedy and exotic taxa through time. Unfortunately, the Anahulu seeds generally occur in such small numbers that changing abundances cannot be assessed. Furthermore, no native taxa which might represent natural forest conditions were identified. Significantly, the basal ash feature in Site D6-58 (Sample F) has a high richness value and some (or all) of the unidentified taxa may be native forest elements. However, even in this basal feature, aLagenaria seed and the weedy Solanum were recovered, suggesting the possibility of agricultural activities within the middle valley at this time (A.D. 1280-1430). Turning to the chronologically later column samples, weedy taxa are again present from the earliest levels. Taxa which are thought to represent disturbance of the basis of their ecological habits include Solanum, Sida, and Waltheria. Although these are native genera, they would not normally occur within a mature mesic forest. They might, however, grow in forest openings or on the exposed, rocky valley walls. These three taxa are present throughout the column samples. Unfortunately, their numbers are too small to evaluate changing frequencies and no trends are evident on nominal assessment either. Aleurites, often associated with native Hawaiian activities and environmental modification, is also quite common in the samples. It appears throughout Column 95 Anahu SCATTERGRAM OF NISP AND VOLUME FOR ALL SAMPLES 60so I1W 200 VOLUME 400 OF SEDIMENT Figure 6.2. Scattergram of the number of identified specimens (NISP) and volume for all samples. C, sporadically within Column A, but is absent from Column B. Aleurites nuts could have been brought to the shelters for lighting, however, its presence in the wood-charcoal samples suggests it was growing in the valley (see chapter 7, this volume). By far the most abundant taxon is Chenopodium. When the taxa are rank-ordered, Chenopodium assumes a clear position of dominance in 28 out of the 29 column samples which contained carbonized seeds. Chenopodieum oahuense is an endemic shrub which could have formed part of the natural vegetation, although on the basis of its present distributions it would not be expected to occur under mesic forest settings. The proclivity for disturbance conditions of both the native Hawaiian species (Environmental Impact Study Corp. 1977:17-8), and of the family in general (Asch and Asch 1977:6), led Allen (1983b) to suggest that it may have been a common agricultural fallow component in prehistoric Hawaiian settings. In the Anahulu assemblages, Chenopodium ocurs in higher frequencies in the earlier levels of Colunm C and probably reflects human disturbance. It is also 50 40 (I, z LJ C-) CD, 0 LLJ I-4 z LU z 30 20 U m m . ~ ~ ~~~~~~m m m m m *2 m~~ m9 M~~ m m9 X9 m m .v i i i 4 i I i. i i i I 1o 0 0 sco 800 IO% 96 Archaeobotanical Assemblages ARCHAEOLOGICAL PROCEDURAL SAMPLE -CONTAMINANTS FLOTATION SAMPLE Figure 6.3. Factors intervening between archaeological sample and target variables. HUMAN SELECTION {SUBSISTENCE. TECHNOLOGY). TARGET VARIABLE NO. 2 ARCHAEOLOGICAL DEPOSIT 97 Anahilu interesting that the relative abundances of Chenopodium and taxonomic richness are negatively correlated (rs=- .901, p=.001), suggesting that as the number of other taxa increased, Chenopodium growth was suppresssed (fig. 6.4). One explanation is that early disturbance was limited in scale. The subsequent increases in the number of other weeds would thus reflect a change in the scale and intensity of human disturbance, i.e., an expansion of agricultural activities. In general, these samples suggest disturbed conditions throughout the time period represented in the columns and features. When the archaeobotanical 8 t 6+ x I- 0 .0 E z 4 assemblages alone are considered, the most parsimonious interpretation is that the disturbance was agricultural in nature. Alternatively, the samples might reflect disturbance in the immediate environment of the shelters. The wood-charcoal samples from Site D6-60 Column C (chapter 7, this volume), also suggest a human-modified environment in the vicinity of the shelter dating to the earliest sample levels. Wood- charcoal of two Polynesian introductions, Eugenia malaccensis (mountain apple) and Artrocarpus altilis, (breadfruit) occurs throughout the column. Although 7 I I 15 14 13 12 11 10 9 a Stratigraphic Levels 7 6 5 4 3 15 14 13 12 11 10 9 8 7 Stratigraphic Levels 6 5 4 3 Figure 6.4. Comparison of NISP, NTAXA, and relative abundance of Chenopodium by stratigraphic level for Column C. I.- . . E 'a .90 0. 0 0 -C or .70 - 0 (U m .50- I I I I I I a -- n a I I I I I a - I - -- I I - - -- I - - - n I 98 Archaeobotanical Assemblages elements of the native forest are represented, there are no distinct trends in the taxonomic composition of the samples. In contrast, the land snail assemblages show clear trends from completely native assemblages, to those including a Polynesian agricultural associate, to those with historic European introductions (chapter 4, this volume). The appearance of anthropophilic landsnails in levels 11/12 of Column C correlates with the increase in taxonomic richness of seeds noted above. This provides independent evidence for an increase in the scale and intensity of agricultural activities later in the sequence. The possibility of a new suite of hydrophilic weeds showing up in conjunction with the establishment of pondfield systems in the early historic period was noted above. No such changes, however, were noted. Subsistence and Plant Utilization Archaeobotanical remains of two traditional cultigens were recovered: Colocasia (taro) and Lagenaria (gourd). Both are poorly represented in the analyzed samples and their presence may, but does not necessarily, indicate local cultivation. Several European- introduced cultigens including rice, beans, tobacco, figs, various citrus fruits and coffee are known to have been grown in the valley since the mid-1800s. Significantly, none are represented in these rockshelter samples and hence the seed data coroborates other evidence which suggests minimal use of rockshelters during the historic period. The discussion of weedy taxa above assumes that these seeds were part of the natural seed "rain." Alternatively, they may represent plants directly utilized. Notably, the Anahulu seed assemblage is similar in species composition to those recorded for other Hawaiian localities (table 6.7). Chenopodium and Waltheria are the most common taxa, occurring in seven out of eight assemblages. The archaeological contexts represented here include a subalpine rockshelter at 3780 m elevation (Mauna Kea), coastal localities (Kalahuipua!a, Kaunakakai, Kaho'olawe), xeric leeward inland areas (Kawela, Kuakini, Waimea-Kawaihae, Kaho'olawe), and the mesic middle Anahulu Valley. The recurrence of these same taxa in several sites and across varied habitats is strong evidence that (1) they are part of an anthropogenic plant community which accompanied a variety of Hawaiian activities; or alternatively, (2) they are important components of prehistoric Hawaiian ethnobotany, dietary, medicinal, or otherwise. At present there are no unambiguous criteria for separating direct plant use from natural seed "rain" of the surrounding vegetation. While ethnographically documented cultigens are generally assumed to reflect direct use, such a claim is more difficult to make for small seeds of weedy and/or anthropophilic plants. In discerning between these two alternatives, contextual information will be important and sedimentary analysis in particular may be critical. One promising line of inquiry is grain size analysis, a technique used in detemining sediment source, mode, and agent of deposition. In conjunction with an archaeobotanical study of Easten Woodland earth ovens, morphologically and functionally similar to Hawaiian oven features, Whittaker (1985) analyzed the grain size characteristics of her sample matrices. The results provided strong evidence tiat the seeds recovered were deposited during post-use infilling and were not directly associated with feature use. Variability in Feature Samples The highly variable NISP and NTAXA values for the feature samples have ahready been commented on above. Keeping in mind that this variability in NTAXA has been shown to be highly correlaed with NISP, the following patterns were observed across feature types. Sample F (Unit 011, Feature 9), an ash deposit, has the highest species richness with 12 taxa. As noted earlier, this sample may reflect the pre-occupation vegetation of the area. Sample G (Unit KIO, Fe 1), a small scoop hearth, also has a high species richness with at least 10 taxa. Sample H (Unit Kl0, Fe 5), an earth oven, contained only Chenopodium seeds. Samples A and B from a single earth oven (Unit 011, Fe 6) are both poor in taxa. The remaining samples contained small numbers of seeds and were taxonomically poor. These include Sample D (a small lens) and Sample E (from the upper part of Ke'eke!e Nui ash deposit). In sum, no distinctive patterns enwrge on the basis of ineted functional differences, but the number of features considered here is quite small. In comparison with the Kaho'olawe samples, the relative taxonomic richness values for ovens versus small hearths are reversed. CONCLUSIONS This analysis of the Anahulu archaeobotanical assemblages raises several concems for future work. First, larger seed assemblages will be necessary if the problems of sample size effects are to be averted. This preliminary assessment suggests that increasing the volume of sediment to be processed will probably be necessary (seed density will undoubtedly vary with the particular environment), but a prescribed sample amount will be especially difficult to predict for cultunal features. 99 Anahulu TABLE 6.7 COMPARISON OF SEED ASSEMBLAGES FROM VARIOUS HAWAIIAN ARCHAEOLOGICAL LOCALITIES TAXA ARCHAEOLOGICAL LOCALITIESa KALA KK KAW KLWE KUA WK ANA MK Cyperaceae Eleocharis - - + Gramineae Eragrosds + + + Chenopodiaceae Chenopodium + + - + + + + + Convolvulaceae + + + + + + Cucurbitacem Lagenara + - + + + + + Sicyos - + _ + Euphorbiaceae Aleurites + + + + + + + Leguminosae - + + + - + Malvaceae Sida - + + + + Solanaceae Solanum - - + + + + Sterculiaceae Waltheria + + + + + + Total No. of Taxa 5 3 3 10 8 7 11 3 a Locality References: KALA: Kalahuipua'a, Hawai'i (Allen 1983a) KK: Kaunakakai, Moloka'i (Allen 1982a) KLWE: Kaho'olawe Island (Allen 1983b; 1984b) KAW: Kawela, Moloka'i (Allen 1982b) WK: Waimea-Kawaihae Road Corridor, Hawai'i (Allen 1983c) KUA: Kuakini Road Corridor, Hawai'i (Allen 1984c) ANA: Anahulu, 0'ahu (this volume) MK: Mauna Kea Adze Quarry, Hawai'i (Allen 1981) Second, the mechanisms involved in the accumulation of archaeobotanical assemblages must be considered. Sedimentological studies of grain size may be particularly helpful in assessing whether the seeds witiin a particular feature are related to feature function or post-depositional infiling. Similarly, identification of the mode(s) and agent(s) of deposition in deep deposits, where column samples might be obtained, can be useful. Given the problems of differing accumulation mechanisms, repetitive pattems from a series of localities are needed before ordinal (or greater) level of measurenmnt can be achieved. Third, if temporal trends are of interest, continuous column samples, rather than bulk feature samples, will be the most useful. Comparison of land snails, wood- charcoal, and archaeobotanical assemblages from the same columns, where sediment accumulation mechanisms may be reasonably comparable and can be evaluated, will allow for more fine-grained paleoenvironmental reconstructions. Fourth, the value of an interdisciplinary approach has again been demonstrated (see also Clark and Kirch 1983; Schilt 1984). The addition of sedimentological studies holds particular promise. The integration of 100 Archaeobotanal Assemblages varied lines of evidence in this case was gready facilitated by Kirch's use of a column sampling design wherein a series of samples were drawn from a single deposit of significant temporal duration. Finally, and most importantly, if we are to understand the significance of archaeobotanical assemblages in either cultural or ecological terms, field sampling will need to proceed with particular problem orientations in mind. In order to address questions of the relationship between plant remains and featre functions, for example, feature types will need to be rigorously defined and a sufficient number of featues from a given locality sampled. If we are to understand plant utilization in cultural contexts then it will be necessary to control for natural seed "rain" under similar sedimentological conditions-a problem possibly remedied by control samples from non-cultural contexts in adjacent localities. The presence of botanical materials in Hawaiian archaeological sediments has now been amply demonstrated and some interesting ecolog- ical and ethnobotanical issues raised. In the fuure, field sampling designs explicitly framed around paleobotan- ical concerns should greatly enhance our understanding of the recovered archaeobotanical material. ACKNOWLEDGEMENTS I would like to thank Donald Grayson, Terry Hunt, and Patrick Kirch for their valuable comments on earlier versions of this paper. REFERENCES CITED Allen, M.S. 1981. "An analysis of the Mauna Kea adze quarry archaeobotanical assemblage." M.A. thesis, University of Hawaii. Honolulu. . 1982a. Report on the archaeobotanical materials from Kaunakakai, Molokai. In K. Shun, Archaeological Reconnaissance Survey and Test Excavations of the Wastewater Treatment Facility Area, Kaunakakai, Molokai, Appendix, pp. 30-3. Department of Anthropology, Bishop Museum. Submitted to CH2M Hill. Maui, Hawaii. . 1982b. Preliminary report on archaeobotanical materials of Kawela Ti 13 rockshelter. Ms. on file at the Deparunent of Anthropology, Bishop Museum. Honolulu. . 1983a. Analysis of plant remains from 1982 excavations at Site 342. In RJ. Hommon, Archae- ological Data Recovery at Site 342 Kalahuipua'a, Hawaii, pp. 52-62. Science Management, Inc. Submitted to Mauna Lani Resort, Inc. Hawaii. . 1983b. Report on the palaeoethnobotanical remains recovered from Kaho'olawe Island. In RJ. Hommon, Kahoolawe Archaeological Excavations 1981, Appendix B, pp. 168-88. Prepared by Science Management, Inc. for U.S. Departnent of the Navy, Pacific Division, Naval Facilities Engineering Command. Pearl Harbor. . 1983c. Analysis of archaeobotanical material. In J.T. Clark and P.V. Kirch, eds., Archaeological Investigations of the Mudlane-Waimea-Kawaihae Road Corridor, Island of Hawaii, Report 14, pp. 384-400. Anthropology Dept. Report Series 83-1, Bishop Museum. Honolulu. 1984a. A review of archaeology and paleoethno- botany in Hawaii. Hawaiian Archaeology 1: 19-30. 1984b. Kaho'olawe archaeobotanical materials. Manuscnpt on file with Paul H. Rosendahl, Inc., Hilo, Hawai'i and with author. . 1984c. Analysis of selected paleoethnobotanical materials. In R. Schilt, Subsistence and Conflict in Kona, Hawaii, pp. 377-83. Anthropology Dept. Report Series 84-1, B.P. Bishop Museum. Honolulu. Asch, D.L., and N.B. Asch. 1977. Chenopod as cultigen: A re-evaluation of some prehistoric collections from eastern North America. Mid- Continental Journal of Archaeology 2:3-45. Buck, P.J. (Te Rangi Hiroa). 1964. Arts and Crafts of Hawaii. B.P. Bishop Museum Special Publication 45. Honolulu. Clark, J.T., and P.V. Kirch. 1983. Archaeological investigations of the Mudland-Waimea-Kawaihae Road Corridor, Island of Hawai'i: An Interdisciplinary Study of an Environmental Transect. Anthropology Dept. Report Series 83-1, B.P. Bishop Museunm Honolulu. Davis, M. 1963. On the theory of pollen analysis. American Journal of Science 261:897-912. Environmental Impact Study Corp. 1977. A report on the Botanical Survey for the Installation Environ- mental Impact Statement, U.S. Army Support Command, Hawaii, Pohakuloa Section. Prepared for the Department of Anny Corp of Engineers. Honolulu. Grayson, D.K. 1979. On the quantification of vertebrate archaeofaunas. In M.B. Schiffer, ed., Advances in Archaeological Method and Theory, vol. 2:199- 237. Academic Press. New York. . 1981. A critical view of the use of archaeological vertebrates in paleoenvironmental reconstruction. Journal of Ethnobiology 1:28-38. . 1984. Quantitative Zooarchaeology. Academic Press. New York. Handy, E.S.C., and E.G. Handy. 1972. Native Planters in Old Hawaii. B.P. Bishop Museum Bulletin 233. ''101 Anahulu Hart, C., and M.C. Neal. 1940. The Plant Ecology of Mauna Kea, Hawaii. Ecology 21:237-66. Hillebrand, W.I. 1981. Flora of the Hawaiian Islands. Lubrecht and Cramer. Monticello. (Facsimile of the 1888 edition). Hitchcock, A.S. 1974. The Grasses of Hawaii. B.P. Bishop Museum Memoir Vol. VIII, No. 3. (Reprint of the 1922 edition). Hommon, RJ. 1983. Kaho'olawe Archaeological Excavations 1981. Prepared by Science Management, Inc. for the U.S. Department of the Navy, Pacific Division, Naval Facilities Engineering Command. Pearl Harbor. Hosaka, E.V. 1937. Ecological and floristic studies in Kipapa Gulch, Oahu. BP. Bishop Museum Occasional Papers XI (17):175-232. Keepax, C. 1977. Contamination of archaeological deposits by seeds of modem origin with particular reference to the use of flotation machines. Journal of Archaeological Science 4:221-29. Kirch, P.V. 1979. Late Prehistoric and Early Historic Settlement-Subsistence Systems in the Anahulu Valley, O'ahu. Anthropology Dept. Report Series 79-2, B.P. Bishop MuseunL Honolulu. Kirch, P.V., and M.J.T. Spriggs. in press. A radio- carbon chronology for the upper Anahulu Valley, O'ahu. Hawaiian Archaeology. Lopinot, N.H., and D.E. Brussell. 1982. Assessing uncarbonized seeds from open-air sites in mesic environments: an example from southern Illinois. Journal of Archaeological Science 9:95-108. McEldowney, H. 1983. A description of major vegetation patterns in the Waimea-Kawaihae region during the early historic period. In J.T Clark and P.V.Kirch, eds. Archaeological Investigations of the Mudlane-Waimea-Kawaihae Road Corridor, Island of Hawaii, Report 16, pp. 40748. Anthropology Dept. Report Series 83-1, Bishop Museum. Honolulu. .1976. Archaeological and Historical Literature Search and Research Design. Lava Flow Control Study, Hilo, Hawai'i. Prepared for U.S. Army Engineer Div., Pacific Ocean, Honolulu by Dept. of Anthropology, B.P. Bishop Museum. Malo, D. 1951. Hawaiian Antiquities. B.P. Bishop Museum Special Publication 2. Honolulu. May, R.M. 1978. The evolution of ecological systems. Scientific American 239(3):118-33. Minnis, P.E. 1981. Seeds in archaeological sites: sources and some interpretive problems. American Antiquity 46:143-52. Mueller-Dombois, D., and H. Ellenberg. 1974. Aims and Methods of Vegetation Ecology. John Wiley and Sons. New York. Murakami, G. 1983a. Identification of charcoal from Kaho'olawe archaeological sites. In RJ. Hommon, Kahoolawe Archaeological Excavations 1981, Appendix b, pp. 168-88. Prepared by Science Management, Inc. for U.S. Department of the Navy, Pacific Division, Naval Facilities Engineering Command. Pearl Harbor. . 1983b. Analysis of charcoal from archaeological contexts. In J.T. Clark and P.V. Kirch, eds., Archaeological Investigations of the Mudlane- Waimea-Kawaihae Road Corridor, Island of Hawaii, Report 20, pp. 514-513. Anthropology Dept. Report Series 83-1, Bishop Museum. Honolulu. . 1983c. Identification of charcoal from Kaho'olawe. Manuscript on file with Paul H. Rosendahl, Inc., Hilo, Hawai'i and with author. Neal, M.C. 1965. In Gardens of Hawaii. B.P. Bishop Museum Special Publication 50. Honolulu. Parham, H.B.R. 1943. Fiji Native Plants. Polynesian Society Memoir no. 16. Purseglove, J.W. 1968. Tropical Crops: Dicotyledons. Longman. London. Schilt, R. 1984. Subsistence and Conflict in Kona, Hawai'i. Anthropology Dept. Report Series 84-1, B.P. Bishop Museum. Honolulu. Selling, O.H. 1948. Studies in Hawaiian Pollen Statistics. Part III. B.P. Bishop Museum Special Publication 39. Honolulu. Smathers, G.A., and D. Mueller-Dombois. 1974. Invasion and Recovery of Vegetation After a Volcanic Eruption in Hawaii. National Park Service Scientific Monograph Series No. 5. Stemmermann, L. 1981. A Guide to Pacific Wetland Plants. U.S. Army Corps of Engineers, Honolulu. St. John, H. 1973. A List of Flowering Plants in Hawaii. Pacific Tropical Botanical Garden Memoir 1. Lawai, Hawaii. _ . 1976. New species of Hawaiian plants collected by David Nelson in 1779. Pacific Science 30:7-44. Territorial Board of Health. 1922. Hawaiian Herbs of Medicinal Value. Charles E. Tutde Co. Rutland (Facsimile reprint). Tilman, D. 1988. Plant Strategies and the Dynamics and Structure of Plant Communities. Princeton Univ. Press. Princeton. Watson, PJ. 1976. In pursuit of prehistoric subsistence: a comparadve account of some contemporary flotation techniques. Midcontinental Journal of Archaeology 1:77-100. Whittaker, F. 1985. Archaeobotany from an Evolutionary Perspective. Paper presented at the 50th Annual Meeting of the Society for American Archaeology. Denver. 102