Volume 91 1982 > Volume 91, No. 1 > Samoan village patterns: four examples, by J. D. Jennings, R. Holmer and G. Jackmond, p 81-102
SAMOAN VILLAGE PATTERNS: FOUR EXAMPLES
This paper 1 deals with further attempts to determine whether identifiable social factors are reflected in the layout of Western Samoan villages from about A.D. 1200 to the present. The data summarised in the few pages were generated during the four seasons of the University of Utah Samoan Archaeological Program. Most of the findings of the Program have been reported (Jennings et al. 1976, 1980). This paper reports research that was not part of the original research design.
Already reported, in the reports cited, is the detailed mapping of portions of two prehistoric Samoan villages (one on Upolu; one on Savai'i) dating from about A.D. 1350 to 1600, and the conclusion from archaeological, statistical, and ethnohistoric data that the village plan mirrored or reinforced many aspects of Samoan social organisation. After the final season of excavation (1976), however, work continued. The continuation included the mapping of another large prehistoric village (Nelson Plantation) near Letolo and the study — by mapping, through interviews, and a brief study of village politics — of a modern (1978–79) village (Fa'aala), both on Savai'i. The reason for the prehistoric village mapping was to determine if the previously observed archaeological pattern was valid at other settlements; the modern village study was to find out if our social and behavioural interpretation of that pattern was valid. Another question to be answered was whether village layout changed with the alleged abandonment of inland villages around 1750–1800 in favour of the historic coastal locations, leaving the interior of the island uninhabited (Davidson 1969). That the settlement locations had shifted between A.D. 1500 to 1800 was obvious; we were curious as to whether there were comparable changes in the organisation of social space, distance, or household relationships within the village.
The basic data, already analysed and reported, should first be recapitulated. The University of Utah Samoan Archaeological Program was inaugurated in 1973 by the study of evidence suggesting that a beachside Lapita village had once been located at the Mulifanua ferry berth, on the - 82 island of Upolu (Figure 1). Submerged some 3m today, the site was discovered during the enlargement of the interisland (Upolu to Savai'i) ferry berth. Investigation conclusively demonstrated that there had been a village at the location, large amounts of both decorated and undecorated Lapita pottery having been recovered. The village was dated by C14 at 3000 B.P. During this visit, a brief coastal and inland search for other sites was also conducted. The two 1973 activities led to the development of a two-part research programme. One goal was to find and excavate other early sites; the other was to investigate in greater detail the vast, abandoned inland villages described and discussed by Davidson (1974d). No sites as early as the 3000-year-old Mulifanua ferry berth were found, but a number of locations dating to approximately 2600 B.P., yielding only plain ceramics, were discovered and sampled. Thus, the overall object of the Programme was to expand, and possibly modify, by intensive excavation and mapping, the earlier findings of Green and Davidson (1969, 1974).
The settlement mapping programme on Upolu was confined to the Mt Olo Plantation on the west end of Upolu where the 1973 survey had located an extensive area of ruined platforms and tumbled walls which lay near an area that Davidson had briefly examined earlier, although the survey team was unaware of her work there when the site was found. At Mt Olo an area of 2km2 was mapped. The survey covered a total of three years, one month being devoted to the task during each of the three major seasons of work during August, September and October of 1974, 1976 and 1977. The Mt Olo archaeological settlement extends beyond the mapped section in all directions, but as it is covered in bush or rank fern growth, our efforts to find its limits failed, except to the north and west where two of its primary paths were traced to the coast, approximately two kilometres away. The aggregation of structures was called a village, even though it is only a sample of a larger settlement. The ambiguity of the term “village” is recognised. It is here nothing more than an adoption of the Samoan practice where all settlements, regardless of size, are called villages, although named subvillages often occur within the large villages. The sample mapped is believed to be adequately representative of the whole settlement. By inspection of the map, units of use which were called Household Units (HHU) were recognised. Larger units called Wards were detected visually and confirmed statistically. The Household Units consisted of an area more than 75% enclosed by walls and paths containing one large or two small stone platforms with one stone-free area within the enclosure. They were interpreted as single family residences, an area for family domestic activities. The HHUs have an interesting Samoan name — fua i ala — ‘measurement along a path’.- 83 - 84
Certainly, one side of most HHUs is bounded by a path. The sides along a path tend to be uniform in length, but the depth varies considerably. The consequent variation in square metres of enclosed area may mark some social distinction but this is only conjecture, no additional archaeological evidence being noted. A total of 120 HHUs were detected. They were by no means identical. Some contained three or more stone platforms, of large size — from 250–400m2 — and are regarded as the households of chiefs. While larger platforms were the normal clue to a high-status HHU, the presence of a huge earth oven — the umuti — in association, was an additional attribute to the chiefly HHU, five of the seven high-status clusters having associated ovens. Thus, two kinds of domestic areas were noted, the large platforms clusters interpreted as chiefs' domiciles, with smaller platforms marking residences of lower class families.
It became obvious that the walkways, raised on fill as in a causeway (8.8km) and walled (3.0km) at ground level or possibly sunken or entrenched a few centimetres, represented the organisational structure of the village. All HHUs face a pathway; in effect, the domicile “fronts” towards the walkway or path. Each section of walkway had numerous associated platforms and HHUs; their average sizes proved to be significantly different among adjacent walkway sections. Six clusters of HHUs were generated by the statistics and each included a previously identified high-status group. The clusters were called residential Wards, the name being drawn from the Samoan pitonu'u, a lineage that resides together in a grouped domiliary area. The lineage head — a high chief or subchief — is described ethnohistorically as having a larger house platform and a larger domestic use area than the average village resident.
The many kilometres of walkway are believed to have been of extreme importance in marking and reinforcing such basic elements of the society as social status, land tenure and use, lineage solidarity, and probably other more subtle nuances of ancient Samoan society. It should be mentioned, too, that the walkways were of two types: primary (long, continuous and generally directional) and secondary (short, connective). The primary ones are wider and of better workmanship and are of either the walled or raised variety although near chiefs' houses they appear always to be walled. This may be archaeological testimony to the still honoured custom of approaching the chief or passing his house with a crouching posture of submissiveness. Even a slightly bowed head, as a lower ranked person passed behind a wall, would have effectively preserved unobtrusiveness when such individuals passed the chief's house and, in passing, showed the respect due to one of high rank.
On the tract were four star mounds — stone structures of unknown use - 85 which are 1 to 2m high. They are characterised by projecting masonry piers, extending 1m or more from the central circular, straight-walled mound. The arms vary in number from 5 or 6 to 10; the shape is vaguely like a starfish, or cog wheel, hence the name. Their location does not form a pattern within the village; they may even be younger than the village itself. If, as locally believed, the star mounds were used for snaring wild pigeons in the forest, they may not have been functional until the village was abandoned and returned to bush.
More detail could be added to this account, but the above serves to illustrate the results of our efforts to learn village layout as part of an original curiosity about the island settlement pattern. The interpretation of the social aspects of the architectural patterning in the Mt Olo ruins made the need for comparable cartographic data very obvious. Even with the conclusions summarised above, there was still no certainty whether the archaeological pattern of HHUs and Wards was a widespread cultural reality; and even more important, whether our initial social and behavioural assumptions that led to the pattern interpretation were valid.
Data for comparison with Mt Olo were unexpectedly generated through the volunteer work of Gregory Jackmond, a member of the Peace Corps. Having a strong interest in archaeology, he had begun to search for sites known to lie inland from Sapapali'i on the island of Savai'i (Buist 1969). During his search he discovered an area of extensive ruins, largely obscured by bush, that had not been noted previously. After consultation with the University of Utah personnel he undertook a survey of that site (Appendix, Jennngs et al. 1980). The area surveyed was 20 hectares, a small part of the suspected total settlement. The data from Sapapali'i tended to replicate those from Mt Olo, with interesting exceptions. The architectural forms are the same — walls, platforms showing great range in size, earth ovens, both walled and raised walkways again with the raised ones more common, and identifiable high-status HHUs. Two Wards, each with one high-status household, were identified. There were differences, however. Walls were 10 times more common in linear metres per hectare than at Mt Olo, and walkways were three times as frequent in linear metres per hectare. Earth ovens, especially in one of the wards, were very common, there being a total of 17 in the small mapped area as opposed to eight in the much larger Mt Olo segment.
The difference in frequency of earth ovens is not understood. The observed association at Mt Olo of ovens with high-status households may be more a statistical illusion than real. Certainly there is no such association at Sapapali'i, where they are abundant. If the ovens are not - 86 true associations, not used during village occupancy but earlier or later, then the C14 dates apply only to the ovens themselves and we are left without an age ascription for the villages. However, at Letolo (see below) the association of umuti with high-rank HHUs is noted, so the umuti dates are assumed to give the correct age of the villages and are used as the basis for establishing age differences between the villages.
More important may be a density of platforms that is three times greater per hectare than at Mt Olo. As a result, the average HHU area was smaller and the “measurements along a path” were shorter than the 90m average at Mt Olo. It was also true that on the stated criteria, the HHUs were harder to identify at Sapapali'i, because some enclosures did not enclose stone platforms. No star mounds occurred in the surveyed tracts, but they do occur on the island. The lack of star mounds means little. They were not significant in delineating Wards at Mt Olo and the lack of star mounds in the Sapapali'i Wards does not detract from the identification of Wards there. The density of structures can perhaps be explained as an artefact of greater human population and the resultant scarcity of land, but there are no solid data upon which to base such an inference.
Chronological facts are available, however. Three Mt Olo umuti yield C14 dates that are contemporary, averaging A.D. 1635. Another from the Cog Oven is A.D. 1355, outside the one standard deviation that would suggest contemporaneity. Other C14 dates from structures other than umuti at Mt Olo tend to support a late 16th/early 17th century ascription. The charcoal from the Sapapali'i ovens averages A.D. 1450, about 200 years earlier. If the dates are taken at face value, the apparent time difference, and the different densities between the sites may be evidence of a general loss of population long before the postulated decline ascribed to imported European diseases. That matter cannot be settled here. Thus, except for the discrepancy in earth oven frequency, Sapapali'i provides strong support for the longevity of the village pattern first observed at Mt Olo, establishing the same pattern at another inland village some 200 years earlier. From the similarities over time a stability of social organisation can be inferred.
The data summarised above were at hand, if not fully understood, in 1976. Then came an opportunity to conduct a similar mapping project, at the Nelson Plantation (Letolo) on Savai'i. Arrangements were made to extend Jackmond's Peace Corps duty for two years; he was appointed archaeologist to the Secretary of Government, and began the survey. The ancient village at Letolo was several square kilometres in extent. It is operated as both a cattle and coconut plantation, so there was neither fern nor bush to impede the survey. When approximately 2km2 had been - 87 surveyed, it was decided that the best way to verify or disprove the social conclusions, and to test the notion of stability of social organisation as arrived at archaeologically, was to map completely a modern village where all relevant behaviour, including status, land use and control, could be determined with some exactitude. The village of Fa'aala, 2km from the Nelson Plantation was selected. There, with the aid of a college educated assistant, a member of a high-status family of Fa'aala, Jackmond conducted the survey. The remainder of this paper provides the new data and a comparison of the four data sets — three archaeological and one ethnographic.
The survey of the prehistoric settlement in the Nelson Plantation covers an area that extends from the coast inland for approximately 3km (Fig. 2). The area is bounded on both east and west by the two major drainages in the district: the Faleata River on the east and the Seugagogo River on the west. Numerous intermittent drainages feeding the Seugagogo watercourse transect the survey area giving the terrain a gently rolling appearance. In general, the land slopes to the south from an elevation of 135m at the northern edge of the survey area. The plantation land is currently divided into paddocks with boundaries marked by modern stone walls. The borrowing of stones for this and other recent plantation facility construction has noticeably altered the form of many prehistoric features. (It should be noted that permission to work at Letolo was contingent upon no excavation. Therefore no charcoal from umuti or other structures was available for C14 assay.)
Within the 198.8 hectare (less than 2km2) survey area, approximately 3000 features of human manufacture were recorded. They consist of stone fences, walkways, foundation platforms, earth ovens and refuse piles; and the repetitive patterns that they form are again interpreted as the basic village structural units: the familiar Wards, centres of local authority, and individual family HHUs. The skeletal structure of the Letolo Settlements is formed by 3.75km of walkways. Two parallel walkway systems traverse the length of the upper half of the survey area. These “primary” walkways apparently represent the principal access routes into and through the village wards. Secondary walkways branch out providing access to the more remote residential units.
Additional features in the settlement are the 64.6km of stone fences. They consist today of stone rubble linear mounds that segregate the area into roughly rectangular household units. These collapsed fence remains vary considerably in width and height. Examining their cross-sectional area gives a good indication of their original size (Jennings et al.- 88 - 89
1976:51). From a total of 1200 observations, the mean cross-sectional area is 61.45 ± 44.93cm2 resulting in a reconstructed fence height of 84 ± 6cm. There are approximately 300 HHUs in the survey area of which 50 are clearly defined (i.e., entirely enclosed by fences or walkways and containing platforms). The area of the 50 fully enclosed HHUs is 6460 ± 2957m2; they each contain 3.4 ± 2.2 platforms. The correlation coefficient relating enclosed area to platform quantity (r = .81) indicates a moderate relationship between the number of platforms and the size of the HHU.
A total of 1059 platforms were recorded during the survey. They vary considerably in size and quality of construction; the largest is the Pulemelei mound reported by Scott (1969:77). It measures 50 by 61m and is over 12m high. There are other large platforms, but none approaches Pulemelei in size. All platforms are constructed of basalt stones and boulders. Many have retained their original form with either vertical or sloping sides and smoothly paved top surfaces. Others have deteriorated and their original forms are difficult to reconstruct. A few clearly modern, but abandoned, platforms occur towards the southern (coastal) end; there is little doubt that most others are prehistoric. Disregarding Pulemelei, platforms average 236 ± 251m2 in basal area and 46 ± 52cm in height. Their average size is larger than those recorded in the other survey areas on Savai'i and Upolu. As is apparent from Figure 2, the platforms are rectangular to oval in shape with two exceptions. These are two small star mounds near the Seugagogo River. They measure 7.8 and 10.2m in diameter with 7 and 5 arms respectively, radiating outwards approximately 1m. Both are 50cm high and are constructed of stone as are other platforms. Two additional types of features occur; earth ovens and stone piles. Forty-one raised rimmed ovens were recorded; they average 9.09 ± 2.72m in diameter and 47 ± 18cm high. Numerous small stone piles dot the areas between platforms, particularly where the soils contain more stone. They could result from clearing the land for agricultural purposes but there is only ethnological evidence of gardening inside the HHUs; none was generated archaeologically, here or elsewhere.
The inferences regarding rank at the Letolo settlement are again based on platform size. As suggested by the ethnohistorical literature and observations at Mt Olo, the larger platforms form clusters that are interpreted as indicating centres of local authority and high status. Five such clusters are readily apparent in the Letolo survey area (Fig. 3). A few other large platforms occur, however, that appear to have a random distribution differing from the highly clustered nature of high-status areas. This distribution is not understood.- 90 - 91
The question of what should be considered a large platform is largely a subjective decision. However, histograms of platform basal area can assist in the decision (Fig. 4). Following the Mt Olo interpretation, approximately 5% of the Letolo platforms are expected to be “large”. At Letolo, the 5% expectation results in 48 platforms that are larger than 750m2 in basal area which conveniently corresponds with a break on the histogram. Presumably, the smaller platforms are the domiciles, kitchen, latrines, etc. of lower ranked villagers.
As in Mt Olo the five distinct clusters of large platforms fall along primary walkways and probably represent the domiciles of the local chiefs along with their community meeting and guest houses (Fig. 3). Ethnographically, the chief presided over a village ward; archaeologically, the wards consist of approximately 50–75 fence-enclosed HHUs. They represent the pig pens, houses, kitchens, and storage facilities of the non-titled residents of that village ward.- 92
Associated with each of the high-status clusters are earth ovens as at Mt Olo. However, a significant difference does occur; at Letolo numerous ovens also occur away from the high-status clusters as at Sapapali'i. The ovens not associated with high-status cluster are mostly located in Ward B, while only three of the 11 ovens in other wards are distinctly separate from high-status clusters. The role of ovens as indicators of high-status areas, therefore, remains questionable.
The village pattern that emerges from the analysis of the Letolo data is similar to that observed at Mt Olo, although a few differences should be noted. The settlement appears to consist of five village wards, each possessing certain characteristics: a cluster of 2 to 5 large platforms located adjacent to a primary walkway accompanied by one or more earth ovens, and 50–75 household units largely enclosed by walkways and fences and containing at least one recognisable platform of suitable size to support a dwelling.
The average household unit at Letolo is slightly larger and contains a slightly greater quantity of platforms than its counterpart at Mt Olo, although the differences are not statistically significant (as determined by a t-test). But, a significant difference does occur in the average size of platforms; at Letolo they are approximately 33% larger in basal area although 20% lower than those at Mt Olo, making the platform volume slightly greater at Letolo.
We tend to agree that the very large mounds date to the 16th or 17th century. Another mound of a size comparable to Pulemelei is Laupule at Vailele on Upolu which is geneologically dated to the early 17th century (Green 1969). The period was one of acquisition of considerable political power by a few individuals and the residences of those few reflect a manpower investment far greater than either the earlier or later historic periods. The fact that the platforms at Letolo are significantly larger than at Mt Olo (dated to the early 17th century) and that Pulemeli is part of Letolo, leads us to favour the 17th century date for the site. In further support of the point is the evidence of the increase of HHU size from early to late. The smallest are at Sapapali'i (15th century). At Mt Olo (17th century) they are larger, so the even larger HHU at Letolo lead us to ascribe a later date — perhaps the late 17th century.
The mapping of Fa'aala was conducted in the same way as for the prehistoric villages, but more data were, of course, available. All structures were mapped and data were recorded concerning the division of land into HHUs and the exact functions of individual structures were ascertained. Particular attention was given to learning the relative ranks- 93 - 94
of the resident chiefs and where their households were located, as well as the size, number and function of the associated structures.
Fa'aala consists of approximately 20 hectares of village land distributed in a narrow band on each side of the coastal highway, which forms the primary “path”. Some household construction has taken place along the inland road which provides access to the plantation lands (Fig. 5). The village land is clearly separated from plantation land by a large stone fence, but all household units constructed outside of this fence are completely surrounded by their own fence system.
Although all existing structures were mapped, the discussion here will deal only with dwellings, guest houses, kitchens, latrines, and storage structures. Some of the structures are traditional in form and construction while others reflect a strong European influence. Most, however, combine features of both traditions. The most common attribute borrowed from the Europeans is galvanised iron replacing palm frond roof thatching. Also common are rectangular house floor plans instead of the traditional oval shape. Most structures, however, retain the stone basal platform with only a few sitting on concrete foundations and floors.
A total of 314 structures are currently being used and maintained in the village. There are an additional 92 abandoned platforms (23% of the total) that at one time supported superstructures but these have decayed or removed, leaving their original function unknown. Table 1 lists the functional types of structures giving the average basal area and volume of the supporting platform. The dwellings of chiefs are combined with other high-status structures such as guest and meeting houses. A point of interest is that 36 dwellings are residences of persons holding chiefly titles even though there are only 11 titles in the village. This burgeoning number of chief's dwellings has resulted from the sharing of a title among two or more, often related men. Historical events have made it advantageous to split a title because the village then controls more votes in the election of members of Parliament, since only titled persons can vote. The proliferation of titles disrupts the pattern of the location of the house of the ranked individuals in the village, in that today many chiefs live apart from their title's guest/speaking house in Fa'aala and probably elsewhere.- 95
Descriptive Statistics of Fa'aala Structure Types (in text)
One of the principal objectives of this survey was to test the assumption that platform size significantly correlates with the function of the supported structure. A median test was selected to test the relationship because the basic assumption of normality or equality of variance could not be met for other more powerful techniques. The results of the tests are demonstrated in the contingency tables in Figure 6. All functional categories, as in the prehistoric villages, except for storage structures and kitchens, are significantly different from each other. Of primary importance is that the high-status structures (chiefs' houses, guest houses, and meeting houses) are again statistically distinguishable from all other structure types. Other structure categories might also be distinguishable by size. The histogram in Figure 7 clearly shows modes that correspond to structure function although interpretation would be difficult without the prior knowledge provided by the interview data.
Additional village data were collected for comparison with the surveyed prehistoric settlements. Of particular interest is the concept of household units, because it has played an important role in interpreting the archaeological settlements (Fig. 8). In Fa'aala the HHUs are not always physically delineated except that occasionally the boundaries are marked by a few stones, a row of banana trees, or a few fence posts, although the families know where their use-area ends and the next family's begins. In the village proper (i.e. along the coastal highway) which is entirely enclosed by the encircling wall, HHUs surrounded by stone fences are absent. This is contrary to the pattern observed in the prehistoric villages. However, HHUs outside of the village proper are surrounded by stone fences. Local village informants explain that the fenced HHUs are outside of the “pig wall” (Fig. 5) where the fences are needed to keep the free-roaming pigs out of the garden areas. It is also explained that the practice of keeping the pigs outside of the village in the plantation area is the product of missionaries, who felt it was improper for humans to share their living quarters with the pigs.- 96
Descriptively the 55 HHUs in Fa'aala are 3565 ± 2198m2 and contain 6.5 ± 4.0 occupied and abandoned structures. More specifically, HHUs contain 1.0 ± 0.8 latrines, 1.5 ± 1.2 kitchens, 0.3 ± 0.6 storage sheds, and 2.5 ± 1.7 dwellings for a total of 5.4 ± 3.4 used structures. Of course, the large standard deviation indicates a considerable variation in actual quantities of structures per HHU.
Contingency tables of Fa'aala platform basal areas.- 97
The village pattern observed at Fa'aala may be taken as typical. In fact, the villagers take pride in their observance of tradition. It consists of 314 occupied structures within the 20 hectares of village land. The village is divided into 55 household units each containing a few domiciles, 1 or 2 kitchens, and a latrine. Most HHUs front the main coastal highway or secondary plantation road. Most often the domiciles are the structures closest to the road with the kitchens, latrines and storage sheds behind the house.
The concept of village ward does not apply to Fa'aala because the village is but a single ward. Elsewhere in the islands, of course, larger villages today have several named wards, each being approximately the same size as Fa'aala. As already noted, the distribution of large platforms does not conform to the archaeological pattern; the only significant concentration being near the western end of the village where four of the five highest ranked chiefs reside (Fig. 8).- 98
Map of the household units and the residences of chiefs in Fa'aala. (Although 36 active titles are living inside the mapped area, only 35 locations are shown because 2 title-holders live in the same house.)
A comparison of the Fa'aala survey data with those collected at the prehistoric settlements indicates that the basic village layout and organising social principles have changed little over the last 500 or more years, except for the difference in location of chiefly residences noted at Fa'aala. There have also been adjustments in the size of structures and houshold units but the data from the four villages — Mt Olo, Sapapali'i, Letolo and Fa'aala — consistently reveal the same use of space, the importance of pathways, the clustering of ranked individuals, keen interest in the boundaries of the HHU, etc.
Although the modern HHUs at Fa'aala are not as clearly delineated as their prehistoric counterparts, there are known reasons for that; at least, informants state that stone fences are not needed now that pigs are - 99 collectively kept in the adjoining plantation land. Modern HHUs contain more structures than is apparent for the prehistoric settlements. That finding is probably deceptive because many of the smaller structures, such as latrines and kitchens, do not have foundations of sufficient size or quality of construction to be visible after several hundred years, and were therefore probably not observed in the mapping of the prehistoric sites. The largest modern platforms are smaller, being of only the average size of the unranked domiciles of the prehistoric settlements. This may indicate a change in ability to amass human labour which may relate to the changing social structure or the actual reduction of the labour force through population decline as postulated by many.
The differences are minor when compared to the similarities. Villages consist of one or more wards that are political units under the authority of one or more local chiefs and subchiefs. Each ward consists of the numerous household units of the village citizens and domiciles, guest houses, and meeting houses of the local chief(s). The chief's facilities are usually located directly adjacent to the main access into the village and the modern examples often have an associated open area (malae). Prehistorically, the chief's facilities may have included large earth ovens presumably for the preparation of the ti root or large quantities of food to feed guests. The modern counterpart is that chief's residences have more kitchen facilities than the average, as indicated in the following contingency table (a chi square test indicated significance at the .01 level). As is widely recognised, lavish feeding of guests is still an important attribute of Samoan life.
Each household unit consists of one or more domiciles, a kitchen and a latrine. Quite possibly the distribution and kinds of latrines have changed since the arrival of missionaries and health officers, but the domiciles and kitchens appear to have remained much the same.
Through this comparative exercise a few new interpretive possibilities have been developed. C14 dates and trends in village attributes allow the arrangement of the four villages in a chronological sequence. The earliest is Sapapali'i, next is Mt Olo followed by Letolo (which is believed to be later on the basis of the large platform Pulemelei, and the larger HHUs) - 100 and modern Fa'aala. We now know that, through time, the HHUs became larger, the number of ranked individuals per ward increased, and the location of ranked domiciles within the wards tended towards the diffuse as opposed to the tight clusters noted earlier. It is noted that some of our data, and conclusions, particularly the matter of increase of HHU size from early to late, contradict Davidson's (1974a:11) view that the densest occurrence of houses falls in the early historic period in the Falefa Valley. Other minor differences were also noted (see Davidson 1974b:155–62), but on most major points our conclusions agree closely with Davidson's (1974c:242–3) analysis of settlement distribution.
Another interesting refinement of the earlier interpretations of the Mt Olo and Sapapali'i data is that none of the three archaeologically examined villages are isolated inland villages, but are large inland extensions of villages located in desirable coastal spots. Witness, the oldest, Sapapali'i, is .5km inland from modern Sapapali'i, which is probably the remainder of the prehistoric village. Mt Olo, 3km + from the coast can be traced (despite the bush) to the edge of modern Satuimalufilufi; a major prehistoric pathway is actually still in use as a track from the centre of that village. And Letolo, now entirely abandoned, extends 3km inland from the marshes where the streams enter the sea. The recognition of the inland location as part of large villages originating on the coast does not change the fact of inland abandonment but does give a different perspective on the nature of the abandonment, which seems to have been a gradual coastward shrinkage rather than a total removal of the inhabitants of an entire village. (Letolo, however, was eventually entirely abandoned).
The evidence seems to support the idea that coastal areas have always been preferred and occupied since initial settlement of the islands. With a growing population, once all suitable coastal areas had been inhabited, villages expanded in the only direction available — inland. During the dramatic reduction of population, usually ascribed to the introduction of European diseases, the inland locations were slowly emptied as the remaining villaters concentrated themselves in the coastal portions of the villages.
Perhaps the most interesting finding is that the use of space, the importance of HHU boundaries, the importance of rank in the disposition of households along the paths, and other organising principles appear to have been stable for 500 to 600 years on both Savai'i and Upolu and to have survived large population fluctuations and changes in village size. Finally, we believe that our original interpretations (Jennings et al, 1976, 1980) derived from archaeological and ethnohistoric data were accurate. The Fa'aala data support those findings on all major points.- 101
Support for the research came from the National Science Foundation (BNS 74–03909 and BNS 76–02682) and the National Endowment to the Humanities (RO-23841–76–365).
1 The data reported in the body of this paper were gathered by Gregory Jackson. Although he was unable to participate in the preparation of this report, full credit for the field work goes to him. His notes and the computer printouts are with the other University of Utah Samoan Archaeological Program data in the Auckland Institute and Museum in Auckland, New Zealand. Thanks are due to Joe Annandale, manager of Nelson Plantation on the island of Savai'i, who gave permission for the Letolo Survey, reported in this paper.