Style and function in East Polynesian fish-hooks.

Allen, Melinda S.

The concepts of style and function are theoretically defined from a neo-Darwinian perspective and the expected spatial-temporal distributions of each kind of trait outlined. Fish-hook assemblages from Aitutaki, Cook Islands, are examined using this framework and related to previously studied collections. Emerging stylistic patterns support notions of interaction between certain East Polynesian archipelagos around the 14th century Ad.

Traditional Polynesian fish-hooks, found in an array of sizes, shapes, and raw materials, have been used primarily for chronological purposes and to illustrate cultural affinities (e.g. Davidson 1967; Emory et al. 1959; Sinoto 1962; 1967; 1968; Skinner 1942). Fish-hooks may also inform on the process of adaptation, ecological relationships, fishing strategies and tool performance (e.g. Kirch 1980; Reinman 1970), but these issues have received comparatively little archaeological attention in the Pacific. Before fish-hooks can be reliably used to elucidate cultural patterns and evolutionary processes, stylistic and functional traits must be explicitly separated. Failures to do so in the past have too often provided ambiguous or inaccurate conclusions. This paper looks at Polynesian fish-hooks from a theoretical framework that explicitly defines the conditions which give rise to stylistic as opposed to functional traits and the expected spatial-temporal distributions of each. Traits hypothesized to be of one kind or the other are empirically evaluated against the theoretical model. Previous studies of Polynesian fish-hooks, and new data from the Cook Islands, are discussed in relation to this framework.

The model: stylistic and functional traits

The term 'style' is often treated in a casual and emic sense as characteristic features, artistry, or habits. As Neiman (1995: 29-30) observes, these abstract and poorly defined notions of style are problematic for systematic archaeological analysis and fail to consider the mechanisms responsible for stylistic variation. In the early- to mid-20th-century, culture historians used a fairly rigorous, but implicit, definition of style in developing the seriation method (Teltser 1995). They observed that certain kinds of artefact traits waxed and waned through time, displaying short-lived unimodal frequency distributions. These historical traits proved useful for establishing relative chronologies, provided that certain conditions were met, as formalized in the seriation model (Dunnell 1970; 1981). The mechanism underlying these distributions in time was identified as 'popularity', and in space as different forms of interaction, including diffusion, migration and exchange (e.g. Deetz & Dethlefsen 1965; Rouse 1939).

Dunnell (1978a) subsequently observed that stylistic traits have a decidedly stochastic character, similar to patterns generated during random clade simulations (see Gould et al. 1977). This led him to propose that stylistic forms could be defined in an neo-Darwinian framework as those which have no detectable selective value (i.e. are selectively neutral). While style in general may have a selective component, functioning as a reservoir of variability (Dunnell 1978a), or to mark social identities and boundaries (Meltzer 1981: 314), particular styles are not directly explicable by natural selection. If stylistic traits are not subject to selection, by definition, then their transmission is accounted for in probabilistic terms, with the rate and spread of change attributable to characters of the transmission systems, the frequency of interaction and the rate of innovation (Dunnell 1978a; Neiman 1990; 1995; O'Brien & Holland 1990).

Artefact similarities may also result from analogous processes, arising independently in response to similar environmental conditions (including similar raw materials). Archaeologists became fully aware of the importance and qualitative difference of functional variability in the 1960s (Binford 1973; Binford & Binford 1966; Jelinek 1976). This was perhaps best demonstrated by the well-known Mousterian tool debate, where Bordes (1961) argued that assemblage variability reflected distinct cultural groups, while Binford (1973) maintained it related to functionally diverse tools. The evolutionary model (Dunnell 1978a; 1978b) specifies that functional forms are those which directly affect the Darwinian fitness of the populations in which they occur. The distribution of these traits arises from the absolute difference in fitness between one variant and that (those) it replaces. Functional traits typically persist for long periods of time, then decline rapidly when selective conditions change, or a new and more successful variant appears. Functional traits are also predictable, correlating with particular environments, patterns of tool use, activities and other functional complexes, and ahistorical, cross-cutting regional and temporal boundaries (Dunnell 1978a; Meltzer 1981: 316).

The dichotomy between stylistic and functional traits is theoretical, mutually exclusive and exhaustive in principle, and leads to specific expectations as to how these two classes of traits will behave through time and across space [ILLUSTRATION FOR FIGURE 1 OMITTED]. The identification of a particular trait as stylistic or functional begins as a hypothesis, which is then empirically evaluated by comparing the spatio-temporal patterns of particular traits with those of the model. Notably, something that is stylistic in one context may become functional in another, as function is driven by selection and selective conditions may change across time and space. In Polynesian fish-hook studies, distinguishing between stylistic and functional traits aids in understanding the origin of specific features of fish-hook design and how they functioned in capturing prey. At another scale, these traits are a critical tool for tracking historical and adaptive processes in Polynesian prehistory (Allen 1992a; 1992b; Cachola-Abad 1993).

Stylistic variability in Polynesian fish-hooks

In East Polynesia [ILLUSTRATION FOR FIGURE 2 OMITTED], where ceramics are lacking, fish-hooks have been used for relative dating and for assessing historical relationships. The successes, and shortcomings, of these past efforts demonstrate the importance of separating stylistic and functional traits and the utility of an explicit theoretical model for evaluating the empirical results of a given analysis.

The seriation method (after Dunnell 1981) was first applied to Polynesian materials by Emory et al. (1959). They examined the temporal distribution of two-piece hook point and base lashing devices (see below) and the occurrence of barbs (which showed little temporal variation). Green (1961) subsequently pointed out the chronological potential of fish-hook heads (functionally the line lashing device), leading Sinoto (1962) to develop his well-known classification of head morphologies [ILLUSTRATION FOR FIGURE 3 OMITTED]. Sinoto applied the head classification to assemblages from Nu'alolo Shelter, Kaua'i (N=202) and three South Point, Hawai'i sites (N=1056) but even with these large assemblages, he had difficulty temporally ordering the basal layer of a key site, H-8. Given that Sinoto's (1962) analysis has been a cornerstone of efforts to order the South Point sites chronologically (Dye 1992; Green 1971), cross-date Hawaiian sites and determine historical relationships with other Polynesian islands (Davidson 1967; Goto 1986; Sinoto 1967; 1970), it is important to understand why his seriation was not entirely successful. Dunnell's (1970; 1981) theoretical formulation of the seriation model was particularly useful in this regard, enabling Green (1971) to identify the underlying causes. First, certain types were not continuously distributed through time, suggesting inadequate samples of rarer forms (Green 1971: 171). More importantly, Green argued that the analytic units (e.g. the stratigraphic zones) were not of equivalent temporal duration, ranging from intervals of 100 to 400 years. Drawing on illustrations of the H-8 stratigraphy (in Emory & Sinoto 1969: figures 6b & 8), Green observed that the lowest cultural layer was 'in reality composed of a series of discontinuous deposits accumulated intermittently over a long span of time', resulting in 'a group whose age duration is no longer minimal'. Importantly, Green's (1971) analyses supported the other chronological interpretations of Sinoto (1962).

Elsewhere in Polynesia, use of the seriation method and attempts to identify historical types have been troubled by the inclusion of inappropriate (e.g. functional) traits, with the result that no clear or consistent order appears (Allen 1992a). Suggs' (1961) Marquesan fish-hook types, for example, conflate stylistic and functional traits; much of the identified variation may, in fact, be functional, including the rotating versus jabbing distinction (see below), hook thickness and size and the occurrence of barbs. Variation in Suggs' assemblages also could reflect different cultural traditions (spatial variation) or ecological differences (functional variation) rather than time, as he compares sites from varied environments and localities across Nuku Hiva Island. For example, three distinctive types (Curved Shank, Rotating Hook and Bent Upper Shank) could represent early types, environmental variability or, a particular cultural tradition, as they are restricted to two early occupations, both from the same site within a large relatively sheltered bay. Only two types - Jabbing and Compound - provide the unambiguous unimodal frequency distributions which signal a successful seriation of stylistic types, while another four types are not continuously distributed through time: Obtuse Recurved Point, Acute Recurved Point, Heavy Shank and Open Jabbing. These discontinuous distributions (at Sites NH4, NH1I and NHo3I) suggest sample-size problems, an issue also raised by Anderson et al. (1994: 42) at the Marquesan site of Hane, Uahuka Island. Finally, the lack of correspondence between the frequency distributions of Site NHo3 Stratum II and those of other sites suggests that the units may not be of comparable duration. As Anderson et al. (1994: 42) observe for Hane, the chronological significance of most fish-hook variation, as thus far reported, is ambiguous.

Another potential stylistic attribute is the relationship between fish-hook shank and point length (expressed as the shank/point ratio). Archaeologically, Sinoto (1967: 354-5) typically found longer points on rotating hooks than on jabbing forms - suggesting the trait is functional at this scale. He also noted geographic variation which might be stylistic. On rotating forms, ratios averaged 1.35 in the Society Islands, 1.46 in the Hawaiian Islands, and 1.43 in the Marquesas Islands. Goto's (1986: 286) study on larger Hawaiian assemblages found shank/point ratios on early rotating hooks ranging from 1.47 to 1.59, those on late hooks from the same area ranging from 1.54 to 1.64. Comparison of hook ratios from diverse environments revealed that the Hawaiian collections were internally homogenous relative to those from other East Polynesian localities (Goto 1986: 279), suggesting that shank/point relationships are selectively neutral within the broader functional categories of rotating and jabbing. Potential functional and technological aspects of shank/point ratio also bear consideration. Johannes (1981:117) suggested that long points work well for species with narrow, deep mouths or for vigorous fighters which might easily dislodge a short point. Reinman (1970) observed that a short point would more evenly distribute mechanical stresses.

Classification issues

Earlier studies also raise the issue of how best to describe and analyse assemblage variation. Regional scale studies require an analytical structure that is explicit, replicable, accommodates variability and facilitates comparisons. As Rolett (1990: 248) observes, Suggs' Marquesan typology is a non-classificatory arrangement (after Dunnell 1971), wherein types are described rather than defined based on groups of specimens. The types thus produced are historical, contingency-bound and 'defined' by the assemblages at hand, constraining application to new assemblages. Types so defined also tend to ignore or suppress variability, rather than seeing it as a source of change. Rolett (1990: 248-9; 1993: 40) suggests that such a non-classificatory arrangement is preferable when working with fragmentary materials and erroneously implies that paradigms (see below) require whole specimens. In fact, it is the attributes used in developing a typology, not the structure of the analytical device, which determines whether whole or partial specimens are needed. Moreover, as Green (1961: 142) observes and Sinoto (1962) demonstrates, classifications based on portions of objects can be quite useful, as are hybrid paradigms (Dunnell 1971). Suggs' types, defined on the basis of both shank and point features, do require complete, or nearly complete, hooks. Unless the appropriate parts are present, assigning fragments to Suggs' types can only be an intuitive judgement of whether or not enough 'necessary' features are shared.

Sinoto's (1962; 1991) analysis of fish-hook heads is a taxonomy (after Dunnell 1971: 7784), mistakenly identified by Rolett (1993: 40) as a paradigm with a 'taxonomic aspect'. While taxonomies are a kind of classification (sensu Dunnell 1971), they are non-dimensional in character. In taxonomies, classes are created through a series of hierarchical oppositions. The resulting classes are not equivalent, as the oppositions leading to particular classes are not necessarily parallel in number or kind. Thus, classes may overlap in certain traits, leading to errors in the assignment of specimens if the hierarchical order is not strictly followed. Although more rigorous and replicable than Suggs' (1961) approach, the non-equivalent classes of taxonomies, and their non-permutable order, makes inter-assemblage comparisons difficult (see Dunnell 1971).

Paradigms (after Dunnell 1971), in contrast to the above, are a particularly parsimonious means of creating analytic units. The inclusion of a particular trait in a paradigm begins as a hypothesis related to the purpose of the classification, as for example the notion that point curvature relates to hook function. The result is that class definitions are tied to a particular problem, rather than to a specific assemblage. This gives the paradigm an ahistorical character, which facilitates application to new assemblages. Classes are formed by the intersection of various attributes. More specifically, paradigms have the following features:

* each class is defined by the same set of criteria: if size is an attribute, then it is considered for all types not just some;

* the attributes are not weighted: no feature is considered more important in type separation than any other;

* the modes are mutually exclusive: only one value can be displayed at a time; and

* the modes are exhaustive: one value must be displayed.

Given these features, the assignment of specimens to classes is not open to the kind of ambiguities of taxonomies. Moreover, all classes are comparable (defined by a common set of attributes) to all other classes in the same classification, a feature which is critical when trying to determine the origins and significance of traits on a regional scale. Finally, the structure of a paradigm is such that it is easily expandable and able to accommodate infinite variability, again contrasting with taxonomies. Overall, paradigmatic classifications provide systematic information on variability, distributions and abundances and lay a firm foundation for comparative studies. Two paradigmatic classifications are used herein, one for fish-hook heads and a second for fish-hook shape.

Background to the Aitutaki analysis

Aitutaki is an almost-atoll (after Stoddart 1975: 31-3) in the southern Cook Islands. The roughly triangular reef surrounds a relatively shallow 50-sq. km lagoon. A 16-sq. km volcanic mainland lies on the western edge of the lagoon, flanked to the west by a shallow, c. 0.8 to 1.7 km wide, reef flat; beyond this the ocean floor drops dramatically. Two tiny volcanic islets (Moturakau and Rapota) are found within the lagoon and a series of coralline islets are located along the eastern reef platform. Biotically, the richest marine habitats are the outer reef flats and the coral-studded lagoon floor (see Allen 1992a: 79-84).

Until recently, angling was thought to have played a relatively minor role in traditional Aitutaki fishing strategies. Buck (1927: 306-7) describes two traditional hooks: a one-piece barracuda (Sphyraena) or toko hook made from forked branches of iron-wood (Casuarina) and a large wooden two-piece shark hook. A metal hook used for catching bonefish (Albulidae) or kiokio may have been a traditional form rendered in iron, but one high chief opined that the only true native hook in the southern Cook Islands was the toko hook (Buck 1927: 307). In 1792, Captain Bligh observed turtle-shell hooks on Aitutaki (Oliver 1988: 267), while ethnographic accounts from other southern Cook Islands (Buck 1944: 236-41; Gill 1880) report hooks of Turbo, coconut-shell, turtle-shell and Pandanus barbs. Neither wood nor turtle-shell hooks were found in the well-preserved cultural strata of Moturakau Rockshelter, suggesting these materials were never of much importance locally.

Most of the archaeologically recovered hooks are made from pearl-shell (Pinctada margaritifera), a bivalve now rare in the local lagoon; a smaller number were made from Turbo (probably T. setosus), a locally abundant gastropod. Four Aitutaki sites [ILLUSTRATION FOR FIGURE 4 OMITTED] yielded fish-hooks: 151 from Moturakau Rockshelter (MR-1); 5 from Ureia (AIT-10); 6 from Hosea (AIT-50); and 1 from Aretai (AIT-49) (Allen 1992a; Allen & Schubel 1990; Allen & Steadman 1990). While Ureia is the oldest site, dating to 1000 BP, the bulk of the fish-hooks come from the lengthy Moturakau Rockshelter sequence. Given the small sample, the Moturakau fish-hooks are compared by two broad chrono-stratigraphic units: Zones H-E date from the mid 13th to 15th century and Zones D-A date from the late 15th century to the historic period. From the 16th century (Zone C) onward, fish-hooks became scarce. At European contact (AD 1773) shell fish-hooks were rare to non-existent, and other fishing techniques dominated (Allen 1992b; Buck 1944).

Aitutaki stylistic analysis

Typically, the fishing line is attached to the most proximal end of the fish-hook, the 'head' (after Sinoto 1991: 86), although the shank area immediately below the head may also be involved (see Davidson 1967; Kirch 1993). Despite this important functional role, the particular morphologies of these line-lashing devices have proven to be good chronological indicators, consistent with the definition of style used here. Building on Sinoto's (1962; 1991) seminal study, morphological features of the Aitutaki fish-hook heads are explicated as a paradigm and supersede earlier efforts (i.e. Allen 1992a; Allen & Schubel 1990). The chronological sensitivity of fish-hook heads is taken as a given; the objective is to capture more effectively the variation in head morphology, while drawing upon the advantages of a paradigmatic structure. The resulting classes are defined by the intersection of three dimensions ([ILLUSTRATION FOR FIGURE 5 OMITTED]; TABLE 1). For example, Class 613 has an stepped proximal end (Mode 6), a straight inner edge (Mode 1), and a protrudent outer edge (Mode 3). Of the 96 potential stylistic classes, 16 are filled by the Moturakau assemblage of 69 specimens with finished intact heads. While the small sample does not allow a rigorous test of the historical distribution of these classes, at least three head classes (Classes 223,213, and 613) have frequency distributions suggestive of a stylistic origin [ILLUSTRATION FOR FIGURE 6 OMITTED]. As a whole, the hooks reflect a shift from fish-hook heads with straight or concave proximal ends, reduced inner edges and protrudent outer edges (e.g. Class 123,223); to ones with concave proximal ends (sometimes accentuated on one side), unmodified inner shanks, and protrudent outer edges (e.g. Class 213); to those where a stepped proximal end combines with a protrudent outer edge to produce a prominent knob (e.g. Class 613).

Shank/point ratios are treated here as a stylistic attribute. Ratios were determined for each hook by calculating maximum shank length divided by maximum hook width. Shank/point ratios for the 13 measurable Aitutaki hooks, all pearl-shell, ranged from 0.7 to 1-51 (mean=1.1, [Sigma]=0.24).

Regional comparisons of stylistic traits

The trends in fish-hook head styles seen on Aitutaki have parallels with hooks recovered from the 14th-century Anai'o Site on nearby Ma'uke Island (Walter 1989; 1990). Comparison of the Aitutaki and Ma'uke assemblages is facilitated by Walter's (1989) use of a paradigmatic classification, even though the specific features defining the classes are not isomorphic. At Anai'o, Class 223 (following TABLE 1) is the most abundant form (N=8), while on Moturakau the frequency of this class peaks during Zone F (late 14th-early 15th century AD). Class 223 also appears in the 16th-century occupation of Hosea (AIT-50). Classes 113 and 213, also at Anai'o, occur on Moturakau between the late 14th and mid 15th century. Another well-represented head style at Anai'o (N=4) is Class 133, a form not found in the Moturakau Rockshelter. The consistency in fish-hook heads in small samples from these two sites suggests a degree of interaction between Aitutaki and Ma'uke at this time, as does other independent evidence (Allen 1992a; in press; Walter 1989; 1990). Indeed, pearl-shell is exotic to Ma'uke and could have been an import from Aitutaki, the closest source.

Gross hook morphology, in contrast, varies significantly between the two assemblages, with jabbing forms better represented on Ma'uke. These functional differences in hook morphology undoubtedly reflect differences in the respective marine environments - the lagoon and extensive reef fiats of Aitutaki as opposed to the narrow, fringing reef and deep near-shore waters of Ma'uke. The association of the same head forms, with varied hook shapes believed to be functional, underscores the stylistic character of the former.

Comparisons with fish-hook assemblages from the northern Cook Islands are also appropriate, as these atolls were a potential source of pearl-shell for the southern Cook Islands. Except for Aitutaki and Manuae Islands, pearl-shell is an import, or rare, elsewhere in the southern Cook Islands. Nevertheless, pearl-shell hooks are found in early (c. AD 1000 to 1500) contexts on the islands of Aitutaki (Allen 1992b), Ma'uke (Walter 1990), Mangaia (Kirch et al. 1992) and Rarotonga (Bellwood 1978). Subsequently, use of pearl-shell declines throughout the southern group, and Turbo hooks increase on at least two islands, Aitutaki and Mangaia. On Aitutaki, Allen's (1994) geomorphic studies suggest terrigenous sedimentation may have altered the lagoon environment and led to local reductions in indigenous pearl-shell. Over-exploitation could also have been a contributing factor (Allen 1992a: 91). Intes et al. (1982) observe that pearl-shell is especially unstable under intensive exploitation, given the accessibility of the stocks, the ease with which they can be harvested, and slow reproductive rates (typically not until the second or third year of life). Based on his marine studies, G. Paulay (pers. comm. 1995) suggests that Aitutaki's lagoon may simply have been too shallow and all-accessible to over-harvesting.

Alternatively, pearl-shell declines could reflect a breakdown in exchange between the northern and southern Cook Islands. If fish-hook head morphology measures interaction (as argued herein), and the northern atolls were a pearl-shell source for the southern group, then we should see similar styles in the two areas at comparable time periods. However, this is not the case, at least on Pukapuka where the dominant style consists of a straight inner shank, an angled proximal end, and an outer edge that is notched or sometimes stepped, therein creating a small projection which usually does not extend beyond the shank (see Chikamori 1987; 1988). Information on early head styles in the other northern atolls is lacking; Yamaguchi (1995: 49) reports a single late prehistoric pearlshell fish-hook from Tongareva with a Class 132 head form (see also undated hook from Manihiki in Skinner 1942: 278).

The earliest East Polynesian assemblages are found in the Marquesas Islands, whether one accepts the recent revised settlement estimates of c. AD 700 (Anderson et al. 1994), or a pre-AD 300 date (Kirch 1986). Detailed comparisons await application of comparable classifications and quantitative information on the frequency of particular forms. Casual observation, however, suggests that the somewhat later assemblages reported here, and from other southern Cook Islands (Y. Igarashi pers. comm. 1994; Kirch et al. 1992; Walter 1989), bear many resemblances to those of the Marquesas Islands (Sinoto 1970; Suggs 1961), suggesting a common origin. Early southern Cook Island fish-hooks heads also are similar to those from the Society (Sinoto & McCoy 1975; Sinoto 1988) and Hawaiian (Sinoto 1962) Islands, New Zealand (Duff 1956; Leach 1979) and, to a lesser extent, Tuamotu Islands (Sinoto 1976), Mangareva (Green 1960), the Austral Islands (Verin 1969) and Rapa Nui (Easter Island) (Ayres 1979). Of particular note are forms with straight and slightly concave proximal ends, protrudent outer edges, and reduced inner edges (e.g. Classes 223 and 123).

The Aitutaki Class 613 form, typically a late style in the Hawaiian (Sinoto 1962) and Society Islands (Beasley 1928), first appears in the Moturakau Rockshelter in the mid-15th century and by late prehistory is the sole head style on Aitutaki. A single example of a comparable form is also known from a post mid-14th-century context on Maupiti Island (Emory & Sinoto 1964: 155, 158). Green (1971) suggests the change from notched to protrudent (after this classification; 'knobbed' sensu Sinoto 1962) forms in the Hawaiian Islands occurred after the mid 14th century, while Dye's (1992) re-analysis of the radiocarbon evidence suggests a' post-AD 1650 shift - a date which seems too late. On Rapa Nui, where archaeologists have traditionally assumed isolation following colonization, Ayres (1979: 75, 85) reports late hooks with a 'distinctive, protruding posterior knob' which he equates with Sinoto's HT4 class, but none of his illustrated examples appear fully comparable.

Weaker stylistic linkages are seen with fish-hooks from West Polynesia. One of the largest assemblages (N=28) comes from the To'aga site, Manu'a Islands, dating to roughly 2500 BP (Kirch 1993: 160-63). Fashioned primarily from Turbo setosus, the heads on these hooks are in several cases not differentiated from the shank proper and small notches or grooves on the proximal end of the outer shank apparently served for line attachment. One Aitutaki specimen is broadly similar (Class 122), with two shallow notches on the outer shank, but also a straight proximal end and a reduced inner shank. Two other Manuan hooks have projections on the inner shank and would be classified herein as Class 541. A similar style has recently been recovered on Mangaia by Y. Igarashi (pers. comm. 1994) and heads with notched inner shanks are common in Marquesan sites (Ottino 1992a; 1992b; Suggs 1961).

In sum, both Hawaiian seriations and regional distributional patterns suggest that the morphology of fish-hook heads is stylistic, and by extension, chronologically sensitive. In the southern Cook Islands, the distribution of certain head classes across two hook shapes of distinct mechanical properties (rotating and jabbing forms) further supports this stylistic interpretation. The appearance and increased frequency of 'knobbed' (Class 613) heads on several islands across the region at broadly comparable times suggests that interaction between the Cook-Society and Society-Hawaiian Islands continued at least through the 14th century. This corroborates independent evidence for interaction between these three island groups (e.g. Allen in press; Allen & Johnson in preparation; Cachola-Abad 1994; Green pers. comm. 1995).

TABLE 2. Available data on shank/point ratios in rotating hooks.

(See text for sources.)

island group X range [Sigma] N material

Pukapuka 1.09 0.86-1.30 0.10 24 PS
Mangareva 1.10 0.62-1.60 0.32 12 PS
S. Cook Islands
 Ma'uke 1.13 1.10-1.20 ? 3 PS
 Aitutaki 1.14 0.70-1.51 0.24 13 PS
Austral Islands 1.31 1.04-1.56 0.15 9 Turbo
Society Islands 1.36 ? ? ? PS/Turbo
Rapa Nui 1.41 ? ? 16 Bone
Marquesas Islands 1.43 ? ? ? PS/Bone
Hawaiian Islands(*) 1.53 1.47-1.59 ? 33 Bone

* Early forms only.

TABLE 2 compares shank/point ratios for several islands, based on published analyses (Ayres 1979; Goto 1986:280 for Hawaiian Islands; Sinoto 1967:354-5 for Marquesas and Society Islands; Walter 1989) and measurements of published illustrations (Chikamori 1988; Green 1960; Verin 1969). While the available assemblages are quite small, not necessarily contemporaneous, and metric data is often incompletely reported, some interesting geographic patterns emerge. Low shank/point ratios typify central East Polynesia islands, but are also found in early New Zealand assemblages (see examples in Davidson 1984: 67). The highest shank/point ratios characterize the Marquesas and Hawaiian Islands, both relatively isolated archipelagos, while the Austral Islands, Society Islands and Rapa Nui are intermediate in this regard. Three alternatives are possible:

* shank/point variation is stylistic; variation between archipelagos will exceed that within;

* variation is functional and related to hook performance; variation will correlate with marine habitats or prey and intra-archipelago heterogeneity should be expected; or

* variation is functional and related to raw material; shank/point ratios will co-vary with raw material and geographic patterning will reflect raw material distributions. Discerning between the three alternatives will require more data than is presently available.

Functional variability in Polynesian fish-hooks

Functional dimensions of fish-hook morphology are more poorly demonstrated. Theoretical discussions (e.g. Allen 1992b; Reinman 1970) have dominated with few practical or quantitative demonstrations of specific functional relationships. Ethnographic studies offer some insights as to traditional Pacific Island fishing strategies, their relative productivity, and relevant features of marine environments (e.g. Ayres 1979; Buck 1927; Conte 1988; Davidson 1967; Johannes 1981; Kirch & Dye 1979; Nordhoff 1930; Rolett 1990). However, in no case has systematic catch data been directly related to hook morphologies. Johannes' (1981) Palauan study provides the most thoughtful analysis of the relationships between mechanical properties of hooks and physical parameters of fish-feeding behaviours and marine environments. While cultural practices may vary between Micronesia and Polynesia, the physical relationships he identifies should hold across similar fish morphologies and marine conditions; the applicability of his observations would be improved by quantitative analyses and cross-regional studies (see also Johannes 1081: 120).

Faunal assemblages also may be helpful in determining hook function, as specific prey types should co-occur archaeologically with particular hook morphologies. However, use of fauna in this manner will require constructing classifications that relate directly to fish-feeding behaviours, mouth morphologies or habitats, rather than relying on biological classifications (see Allen 1992a: 330-31; 364-76; 420-28). Even thus classified, it is more likely that probabilistic rather than deterministic relationships will hold between particular hook features and prey types, as exemplified by modern fishing practices (e.g. Lewers 1978).

Several functional categories of fish-hooks and related lures are recognized in Polynesia. Fish-hooks designed for specific prey include the large, typically wooden, shark hooks (e.g. Buck 1944: 172-94; Grudger 1927); bonito hooks and lures used in trolling (e.g. Buck 1932); and hooked multi-component octopus lures (Pfeffer 1995). The focus here is on functional variation in the more generalized one- and two-piece fish-hooks, most commonly made from shell and bone.

Nordhoff (1930: 155-6) first drew attention to a key functional dimension of Polynesian hooks, the relationship between hook curvature and the mechanical action produced when stress is applied to the fishing line. Hooks with curved shanks, in-curved points, or both, rotate in the prey's mouth when the bait is nibbled or swallowed (rotating hooks), while hooks with parallel shank and point axes must be pulled sharply to secure the fish (jabbing hooks). Sinoto (1991) suggests the two types can be separated by extending the fish-hook point with an imaginary line: in a rotating form the point intersects the hook shank, in a jabbing form it does not. Sinoto (1991: 97) and others (Coutts 1975; Goto 1986) recognize variability in these two categories, but no one has explored the spatio-temporal importance of this to any degree.

The practical choice of a rotating or a jabbing form depends on both the desired prey and local marine conditions. Rotating hooks are particularly useful in deep waters, or where the currents are strong and setting a hook by jerking is difficult (Johannes 1981: 113-14; Reinman 1970). They are also useful in shallow waters when the intended prey is large or coral heads abundant (Johannes 1981: 116). Jabbing forms are more commonly used in shallow waters and under other conditions where a taut line can be maintained, as for example, trolling.

The greatest stress on a hook is at the bend. The bend can be strengthened by:

* thickening the bend;

* reducing the length of the hook shank; and

* in-curving the shank and point limbs, making the hook more circular, to disperse the stresses (Reinman 1970: 54).

In this respect, rotating forms can have advantages over jabbing forms, as stresses are more equally distributed and less stress is generated in setting the hook.

Within the rotating form, variations in the 'gape' (the distance between the hook point and shank) also affect performance. The strongest rotating hook is one where the gape is narrow. However, the gape distance is also a compromise between securely hooking a fish, which demands a narrow gape, versus speedy removal of a fish once caught, which calls for a wide gape (Johannes 1981: 115). Jabbing forms (especially without barbs) are often preferred when fish must be removed rapidly, as in trolling (Johannes 1981; see also Goto 1986). In rotating hooks, Johannes (1981: 114) noted a functional relationship between gape size and water depth; hooks with narrower gapes were used in the deepest waters where the prey are often large and of high quality. These theoretical etic and emic relationships between hook form and catch need to be systematically evaluated before we can confidently use them to understand the archaeological record.

In the Hawaiian Islands, Emory et al. (1959: 36) found the relative frequencies of rotating and jabbing hooks to be fairly stable through time, consistent with a functional, as opposed to stylistic, origin. In the Marquesas Islands, in contrast, jabbing hooks became more common in late prehistory (Dye 1990; Kirch 1980; Rolett 1990; Anderson et al. 1994). These changing frequencies could reflect a change in fishing environments (i.e. abandonment of offshore fishing) (Dye 1990; Rolett 1990); development of adaptations that made jabbing hooks a better choice (e.g. a change in selective value of this form) (Kirch 1980); or a combination of the two. Dye (1990; also Rolett 1990) uses the archaeo-fauna to support the first alternative.

A second potential functional dimension, hook size, relates to the size and the kind of prey that can be captured. Hook size has been measured by hook width (Owens & Merrick 1994), point length (Coutts 1975; Goto 1990) and hook length (Kirch & Yen 1982). Some generalizations include: hook width is not closely related to fish mouth width (within a species), but is almost always smaller than mouth width (Owens & Merrick 1994); small hooks are best for fish with small mouths and fish which nibble rather than rapidly strike; and large hooks are more effective for large, voracious carnivores such as lutjanids (snappers), scombrids (tunas) and carangids (jacks, trevallys) (Johannes 1981). Johannes (1981: 117) also observed that wide bends catch deeply in the fish's jaw and are more difficult for the fish to spit out. They work well with large-mouthed species such as groupers (Serranidae) and most Holocentridae, and in small sizes for the more cautious nibblers like the smaller-mouthed wrasses (Labridae). Hooks with narrow bends are used for fish with small mouths, such as Holocentrus.

Archaeologically, Emory et al. (1959: 14-18) and Ayres (1979: 85) demonstrate two distinct size classes in Hawaiian and Rapa Nui assemblages respectively. However, these are not related to particular prey or fishing strategies. Kirch & Yen (1982: 239-43) comparing modern metal hooks of known use with those from Tikopian archaeological contexts, identified three size groups. The intermediate group, in both modern and archaeological cases, was the most variable and currently associated with a variety of fishing techniques. Modern hooks of less than 20 mm in length were used for inshore angling, while those greater than 60 mm in length were used for tow-line and bottom-line fishing of large prey.

Raw material must also be considered a functional trait to the degree that it affects the cost and success of a particular hook form. Production costs include raw material acquisition, ease or difficulty of manufacture, and the possibility of breakage during manufacture. Performance costs include the success of prey capture and the likely use-life of the hook. Based on archaeological findings and structural variables, Allen (1992b) hypothesized that pearl-shell is more easily worked, relative to Turbo, allowing more varied shapes and fewer manufacturing failures. Shell and bone also have been compared, with Reinman (1970: 51, 55) arguing that shell is brittle and has little resiliency, and others (Green 1961: 143; Kirch 1985: 201) opining that shell's cross-laminated structure makes it stronger at the critical area, the hook bend. Engineering comparisons of pearl-shell, Turbo and bone are now underway.

By the archaeological evidence, pearl-shell was the raw material of choice for shell fishhooks in Polynesia (Allen 1992b); other species, such as Turbo and Trochus, were most frequently used when pearl-shell was rare or unavailable, as in West Polynesia. In addition to its potential structural advantages, the sheen of pearl-shell may have enhanced hook performance, as variations in lustre, colour, and intensity were recognized and valued by native fishermen (Kennedy 1930:102-3; Nordhoff 1930: 140; 240-43). Green (1961; also Reinman 1970: 57) observed that in East Polynesian islands where pearl-shell was uncommon, hooks were typically made of bone, as in the Hawaiian Islands, New Zealand and Rapa Nui; a small number of bone hooks also occur early in the Hane sequence (Sinoto 1967: 347-8). On islands where bone was the most common raw material, two-piece hooks are often found. Kirch (1982: 74) attributes this development to raw material constraints, suggesting that bone is less resistant than shell to shear stress. With two-piece hooks, the weak bend region is effectively 'broken' and then reinforced with a flexible lashing.

Fish-hook assemblages from South Point, Hawa'i (Emory et al. 1959) provide an example of how style can serve as a 'reservoir of variability', some which may acquire adaptive value with changing selective conditions (Dunnell 1978b: 199; Dunnell & Feathers 1991: 34). Sinoto (1962) identified eight Hawaiian fish-hook head styles, of which five are represented in FIGURE 3. Knobbing is one such style (HT-4) and first appears in Site H-1 Layer II-U. As a head form, it persists along with two to three other styles into the late prehistoric period [ILLUSTRATION FOR FIGURE 3 OMITTED]. Knobbing later appears as a basal fastening devices for two-piece hooks (Site H-1 Layer I-U). Here, in contrast to fish-hook heads, knobbed bases replace the single earlier base form, consistent with the model of function [ILLUSTRATION FOR FIGURE 1 OMITTED]. Moreover, the notched to knobbed transition is gradual in one-piece fish-hook heads but rapid in two-piece hook point bases (Sinoto 1968: 61), again supportive of the idea that the knobbed basal morphology positively affected hook performance, and rapidly assumed a selective advantage over the older notched form. The new design probably strengthened the connection between hook shank and point. The reinforced bend also apparently allowed larger hooks, as Goto (1986: 287, 291) observed an increase in the size of two-piece hooks at South Point after the appearance of knobbed bases. Additionally, two-piece hooks increase in abundance while large one-piece hooks decline, suggesting that the former is functionally superior.

Aitutaki functional analysis

The foregoing discussion identifies potential relationships between aspects of hook morphology and particular marine environments and prey, while the evolutionary model outlines the expected temporal patterning of functional traits in general. Fully understanding functional variability in Polynesian fish-hooks will require a number of assemblages from varied environments and lengthy archaeological sequences. The Aitutaki data contribute to these discussions with new evidence from a lagoon marine environment and a 1000-year sequence of marine exploitation. Three dimensions of hook morphology are considered here: gross hook shape, length (a measure of size) and width (a second measure of size which may be relevant to certain prey types).

A simple paradigmatic classification based on two dimensions of morphology describes the shape of 13 complete hooks (TABLE 3). As both the shank and point may be curved, hooks are oriented along an axis that bisects the hook, into roughly equal proportions, through the bend. Of the four potential classes generated by this paradigm, three have members [ILLUSTRATION FOR FIGURE 7 OMITTED]. Shape Class 12 (N = 6), the best-represented, consists of hooks with straight shanks and strongly in-curved points. Class 21 (N = 3) hooks, with straight points but in-curved shanks, would rotate under stress from the line, but less so than members of Class 12. Some Class 21 hooks would be considered 'jabbing' forms by traditional definitions. The Class 21 shape apparently incorporates the advantages of better penetration associated with curved forms, and the functional requirements of an open form, most useful for rapidly striking fish or when quick removal and re-casting is required. Class 22 (N = 4) hooks, where both the shank and the point are in-curved (especially so in the Aitutaki assemblages) would produce a strongly rotating action. Rotating shapes like these would penetrate well, be stronger than less curved shapes, and hold the fish more securely (Reinman 1970: 51-4).

Straight shank fragments (N = 10) could be from either Class 11 ('jabbing' hooks in traditional typologies) or Class 12 type hooks. The former was not represented among the complete Aitutaki hooks, while the latter was the most common hook form, suggesting that most (or all) of these straight shanks fragments are from Class 12 members. Seven fragments with strongly in-curved points were recovered; these could be members of Class 12 or 22. Three fragments with marked curvature may represent Class 21 or 22. Overall the fragments, as with the whole specimens, indicate a predominance of rotating shapes.

Three potential functional types were identified within the general rotating morphology, as expressed in the Aitutaki shape classification; those with narrow gapes are most common. The pattern of hook shape, especially those with narrow gapes, suggest that deep-water angling (including bottom-line fishing) was favoured, potentially both within the lagoon and off the outer reef. This interpretation is broadly corroborated by the Aitutaki faunal evidence where relatively large, deep-water fish, such as lutjanids and carangids, are more common early in time when angling is well-represented (Allen 1992a). The lack of jabbing (Class 11) hooks is consistent with the use of netting and spearing technologies in the shallow near-shore zone, as was the case at European contact.

Hook length (shank height) was evaluated for general distributional patterns rather than in relation to specific prey. The Aitutaki hooks are quite variable in size, ranging from a large specimen of 71.6 mm length to a diminutive bend fragment of 9.7 mm width. Early forms are, on the average, slightly longer than late forms ([ILLUSTRATION FOR FIGURE 8 OMITTED], TABLE 4), a difference that is not statistically significant (t = 1.47; p = 0.150) and probably had little impact on hook performance. A bimodal distribution is indicated in the early assemblage, suggesting at least two functional types. Turbo hooks are on the average smaller and less variable than those of pearl-shell (TABLE 3).

Comparison of hook length by shape class reveals no distinct patterns, and comparison of the three shape-class members with one-way analysis of variance indicates no significant difference in hook length (F = 0-757; p = 0.494). Members of Class 12 are the most variable, with shank lengths ranging from 17.5 to 42.6 mm; this variability could reflect the larger sample. Class 21 hooks range from 23.3 to 27.9 mm in length and Class 22 from 18.2 to 29.2 mm.

Hook width was measured at its widest point, with each hook oriented vertical to an axis running perpendicular to the bend. Hook width is a second measure of hook size, providing an overlapping but non-equivalent sample to compare with the specimens measured for length [ILLUSTRATION FOR FIGURE 9 OMITTED]. As with hook length, the two means are not significantly different (t = 1.52, p = 0.143). The hook widths have a unimodal distribution, suggesting a single generalized hook in terms of width.

As expected, hook length and width are positively correlated (r2 = 0-651; p = [less than]0.001). Two clusters in the regression plot [ILLUSTRATION FOR FIGURE 10 OMITTED] reflect the bimodal distributions of hook length. The four residuals, three above the regression line and one below, could be functionally specialized forms, but this is speculative on present evidence.

In sum, little temporal change is apparent in hook size or shape during the roughly 350-year period when fish-hooks are common on Aitutaki. After the 16th century, hooks decrease in abundance, confounding assessments of morphological trends. As the materials are mostly from a single site, the relation of particular functional characters to environmental parameters can not yet be assessed. A change in raw materials is seen later in the sequence, with Turbo apparently an unsatisfactory, or perhaps more costly, substitute for pearl-shell. The preference for rotating hook forms may have combined with the difficulty of producing these forms from Turbo, leading to the abandonment of shell hooks on Aitutaki, and a shift from angling to other fishing methods.

Regional comparisons of functional traits

Previously (Allen 1992b), I suggested that the abundance and diversity of fish-hook forms found with entry into East Polynesia is attributable to the increased availability, and superior mechanical properties, of pearl-shell over common nacreous West Polynesian taxa like Turbo and Trochus. Certainly few West Polynesian fish-hooks have been found and the 28 one-piece Manuan hooks (Kirch 1993) are the largest assemblage available for comparison. These Turbo hooks are relatively small, varying in length from 13.1 to 30.4 mm. The few classifiable specimens are of two kinds: Shape Classes 21 and 22, both forms represented on Aitutaki. The Manu'a specimens indicate that Turbo fish-hooks have a long history and the related technology initially developed in West Polynesia.

Relative to other East Polynesian assemblages, the Aitutaki hooks are notable for the lack of jabbing forms (i.e. Class 11). Class 11 hooks are not only common elsewhere in East Polynesia, but also in other southern Cook localities (e.g. Walter 1989). Detailed shape comparisons are not possible at this time because other assemblages have been primarily described in terms of the basic rotating versus jabbing dichotomy.

In-depth regional comparisons of size are also not possible at this time. Metrical information is in many cases altogether lacking, or incompletely reported. Moreover, hook size needs to be considered in relation to raw material and hook shape before functional interpretations relating to use are possible. In general, the Aitutaki hooks appear neither exceptionally large nor small relative to other regional assemblages. The majority fall within the intermediate group of Kirch & Yen (1982: 239-43), a size category ethnographically used on Tikopia for tow-line and bottom fishing from a canoe. Relative to Ma'uke Island hooks (X = 19.7 mm; range: 10-39 mm) (Walter 1989), those from Aitutaki are larger and more varied (but the latter sample is also larger). The Aitutaki hooks are similar to the mean size and ranges of Rapa Nui one-piece bone rotating hooks (X = 30.1 mm; range: 14-73 mm) (Ayres 1979), and generally consistent with limited metric data for Marquesan assemblages (Suggs 1961) and means of various Hawaiian one-piece hooks assemblages (19.3 to 31.3 mm) (Goto 1986: 230-34). Variability in hook size warrants closer examination in conjunction with data on the related marine environments and faunal assemblages.

TABLE 4. Hook measurements (mm).

 mean N range
[Sigma]

hook length

late pearl-shell 27.3 21 16.7-53.2 9.3
early pearl-shell 32.7 17 20.8-71.6 12.4

Turbo hooks 22.9 4 16.7-29.1 5.1

hook width

early pearl-shell 20.7 17 9.7-33.0 6.9
late pearl-shell 17.8 12 14.2-24.7 3.2

Concluding thoughts

A consideration of the spario-temporal patterning of fish-hook head morphologies reinforces earlier notions that line lashing devices are stylistic (selectively neutral), with specific morphologies reflecting ancestral relationships and patterns of interaction. Although not appropriate for seriation, the Aitutaki hooks show homologous similarities with early East Polynesian collections from the Marquesan Islands, and with pene-contemporaneous assemblages from elsewhere in central East Polynesia, New Zealand, and Rapa Nui. The appearance of a new head form (Class 613) in three widely separated geographic locales suggests continued inter-island contact after initial settlement. Class 613 appears on Aitutaki and in the Society Islands by roughly AD 1400, and in the Hawaiian Islands after AD 1350 (after Green 1971). This interpretation fits well with emerging evidence both for contact between these three archipelagos and for post-colonization inter-island voyaging within the East Polynesia region (e.g. Allen & Johnson in preparation; Green 1981; 1994; Irwin 1992; Sutton 1987; Walter 1990).

Shank/point ratios have also been investigated as a potential stylistic character, with inconclusive results. Hooks found in several central East Polynesia islands have low shank/point ratios, suggesting that the trait may be ancestral, but also raising the possibility that it is associated with pearl-shell which is common here. If the latter, high shank/point ratios in the Hawaiian Islands could reflect the use of bone, but then high ratios seen in Marquesan pearl-shell hooks remain difficult to explain. For the moment, the underlying cause of variability in shank/point ratios remains ambiguous.

Attempts to derive chronology and historical associations from other features of hook morphology have been more problematical. Rolett's (1993: 42) search for similarities in three roughly contemporaneous Marquesan assemblages using Suggs' types was forced to a similar conclusion, albeit on different grounds. Hook shape, barbs and thickness are all traits that most likely affected hook use, were subject to selection and thus are inappropriate as temporal markers.

Functional aspects of Polynesian hook morphology are more poorly understood. In part this stems from our fascination with tracing migration routes and homelands, but there are also methodological issues. Whereas lithic analysts have profitably used tool wear as an empirical marker of function (e.g. Dunnell 1978b; Meltzer 1981), shell hooks from tropical Pacific Islands are often too degraded for use-wear to be reliably identified. Functional studies would, however, greatly benefit from research on fish behaviours relative to hook morphology, mechanical properties of raw materials and engineering aspects of hook performance. Spatio-temporal data from archaeological assemblages also is critically needed to identify regular and predictable relationships between marine prey, habitats and hook morphologies.

In this paper, I have sought to offer an integrated and explicit theoretical framework for examining variability in Polynesian fish-hooks, one which might clarify the origins, spatio-temporal distributions and cultural relevance of fish-hook characteristics. While the model presented here cannot be fully evaluated at this time, my hope is that I have demonstrated the potential of the approach. My second goal was to bring together a diverse and somewhat scattered body of regional evidence on the use and prehistoric distribution of one-piece fish-hooks. In examining previous studies, I have attempted to evaluate the stylistic and functional potential of various hook attributes, loosely based on previous demonstrations of historical patterning, traditional cultural practices and potential mechanical relationships. We are at a juncture where it will be quite profitable to return to old collections with new questions and reconsider variability in Polynesian fishhooks in both methodological and substantive terms.

TABLE 1. Stylistic classification of fish-hook head morphology.

I. SHAPE OF PROXIMAL END (as viewed in profile)

1. straight: a linear or flat surface with no additional modifications

2. concaved: a rounded arc with a curve into a flat surface; the concavity is not necessarily symmetrical

3. pointed: the proximal end comes to an apex or peak; differs from angled in that the point is centered on the proximal end; point may be rounded or sharp

4. notched: a flat surface cut by grooves or angled concavities

5. angled: a linear or flat surface that slopes distally towards either the outer or inner edge, forming an offset point

6. stepped: a 90 [degrees] or greater shelf or step, with the step being perpendicular to the main axis of the shank

II. SHAPE OF INNER EDGE (as viewed in profile)

1. straight: a linear or flat surface with no additional modifications; very slightly excurved edges are also considered straight

2. reduced/stepped: inner edge thinned towards the proximal end of the hook, varying from slightly to more sharply reduced (as in a 90 [degrees] step)(a)

3. notched: an acute to obtuse concavity cut into the shank, often but not necessarily at an oblique angle; although the intersection of this edge and the proximal end may appear to form a 'projection' or knob, it does not extend beyond the shank proper

4. protrudent: proximal end of the edge projects outward beyond the edge of the shank proper; the protrudent may extend out from the shank at a sharp right angle or may be formed by a gentle outward flaring(b)

III. SHAPE OF OUTER EDGE (as viewed in profile) 1. concaved: a rounded arc with a curve into a flat surface

2. notched: an acute to obtuse concavity cut into the shank, often but not necessarily at an oblique angle; although the intersection of this edge and the proximal end may appear to form a projection or knob, it does not extend beyond the shank proper

3. protrudent: proximal end of the edge projects outward beyond the edge of the shank proper; the protrudent may extend out from the shank at a sharp right angle or may be formed by a gentle outward flaring(b)

4. straight: a linear or flat surface with no additional modifications

a The classification currently makes no distinction between these two morphological variants, which on present evidence do not appear to be chronologically distinct from one another.

b Eventually it may be useful to separate angled protrudent edges from curved, flaring ones but in the present small assemblage the two forms have comparable temporal distributions and are associated with similar modes in the other dimensions.

TABLE 3. Functional fish-hook shape classification.

I. ALIGNMENT OF SHANK

1. straight or excurved: shank parallel to or curved away from main axis of hook

2. incurved: shank curved towards from main axis of hook

II. ALIGNMENT OF POINT

1. straight or excurved: point parallel, nearly parallel, or curved away from main axis of hook

2. incurred: point curved towards main axis of hook, sometimes strongly so

Acknowledgements. I thank Robert Dunnell, Terry Hunt and Yosi Sinoto for comments and discussions, Roger Green and Helen Reeves Lawrence for their thoughtful reviews and Brad Evans for the illustrations. Jadelyn Moniz and Michael Graves were especially helpful in refining the stylistic classification in the course of applying it to Hawaiian assemblages. The Aitutaki research was supported by the hospitality of the Cook Island people, National Science Foundation (BSN-8822768 to M.S. Allen and BSR-8607535 to David Steadman), Wenner-Gren Foundation for Anthropological Research (Grant 5079), a University of Washington W.W. Stout Fellowship, Sigma Xi, The Scientific Research Society, and the field assistance of Lisa Nagaoka, Sue Schubel and Algernon Allen.

References

ALLEN, M.S. 1992a. Dynamic landscapes and human subsistence: archaeological investigations on Aitutaki Island, southern Cook islands. Ph.D dissertation, University of Washington. Ann Arbor (MI): University Microfilms International.

1992b. Temporal variation in Polynesian fishing strategies: the southern Cook Islands in regional perspective, Asian Perspectives 31: 183-204.

1994. The chronology of coastal morphogenesis and human settlement on Aitutaki, southern Cook Islands, Polynesia, Radiocarbon 36: 59-71.

In press. Patterns of interaction in southern Cook prehistory, Indo-Pacific Prehistory Association Bulletin. Proceedings of the 15th Indo-Pacific Prehistory Congress, Chiang Mai, Thailand, 1994.

ALLEN, M.S. & K.D. JOHNSON. In preparation. Tracking ancient patterns of interaction: recent geo-chemical studies, in Marshall W. Weisler (ed.), Prehistoric long-distance interaction in Oceania.

ALLEN, M.S. & S.E. SCHUBEL. 1990. Recent archaeological research on Aitutaki, southern Cooks: the Moturakau Rockshelter, Journal of the Polynesian Society 99: 265-95.

ALLEN, M.S. & D.W. STEADMAN. 1990. Excavations at the Ureia Site, Aitutaki, Cook Islands: preliminary results, Archaeology in Oceania 25: 24-37.

ANDERSON, A., H. LEACH, I. SMITH & R. WALTER. 1994. Reconsideration of the Marquesan sequence in East Polynesian prehistory, with particular reference to Hane (MUH1), Archaeology in Oceania 29: 29-52.

AYRES, W.S. 1979. Easter Island fishing, Asian Perspectives 22: 61-92.

BEASLEY, H.G. 1928. Pacific island records, fishhooks. London: Seeley Service.

BELLWOOD, P. 1978. Archaeological research in the Cook Islands. Honolulu (HI): Department of Anthropology, Bernice P. Bishop Museum. Pacific Anthropological Records 27.

BINFORD, L.R. 1973. Interassemblage variability - the Mousterian and the 'functional' argument, in C. Renfrew (ed.), The explanation of culture change: 227-54. London: Duckworth.

BINFORD, L.R. & S.R. BINFORD. 1966. A preliminary analysis of functional variability, in the Mousterian of Levallois facies, American Anthropologist 68(2): 238-95.

BORDES, F. 1961. Mousterian cultures in France, Science 134: 803-10.

BUCK, P. 1927. The material culture of the Cook Islands (Aitutaki). New Plymouth, NZ: Board of Maori Ethnological Research. Memoirs of the Board of Maori Ethnological Research 1.

1932. Ethnology of Manihiki and Rakahanga. Honolulu (HI): B.P. Bishop Museum. Bulletin 99.

1944. Arts and crafts of the Cook Islands. Honolulu (HI): B.P. Bishop Museum. Bulletin 179.

CACHOLA-ABAD, C.K. 1993. Evaluating the orthodox dual settlement model for the Hawaiian Islands: an analysis of artefact distributions and Hawaiian oral traditions, in M.W. Graves & R.C. Green (ed.), The evolution and organization of prehistoric society in Polynesia: 13-32. Auckland: New Zealand Archaeological Association. Monograph 19.

CHIKAMORI, M. 1987. Archaeology on Pukuapuka Atoll, Man and culture in Oceania, Special Issue: 105-15.

1988. Archaeological investigations, in M. Chikamori & S. Yoshida (ed.), An archaeological survey of Pukapuka Atoll, 1985 (preliminary report): 4-39. Tokyo: Department of Archaeology and Ethnology, Keio University.

CONTE, E. 1988. L'exploitation traditionnelle des ressources marines a Napuka (Tuamotu, Polynesie francaise). Unpublished Ph.D dissertation, Universite de Paris-I.

COUTTS, P.J.F. 1975. Marine fishing in archaeological perspective: techniques for determining fishing strategies, in R.W. Casteel & G.I. Quimby (ed.), Maritime adaptations of the Pacific: 265-306. Paris: Mouton.

DAVIDSON, J. 1967. An archaeological assemblage of simple fishhooks from Nukuoro Atoll, Journal of the Polynesian Society 76: 177-96.

1984. The prehistory of New Zealand. Auckland: Longman Paul.

DEETZ, J.F. & E.S. DETHLEFSEN. 1965. The doppler effect and archaeology, South western Journal of Anthropology 21: 196-206.

DUFF, R. 1956. The moa hunter period of Maori culture. Wellington: Government Printer. Canterbury Museum Bulletin 1.

DUNNELL, R.C. 1970. Seriation method and its evaluation, American Antiquity 35: 305-19.

1971. Systematics in prehistory. New York (NY): Free Press.

1978a. Style and function: a fundamental dichotomy, American Antiquity 43: 192-202.

1978b. Archaeological potential of anthropological and scientific models of function, in R.C. Dunnell & E.S. Hall (ed.), Archaeological essays in honor of Irving B. Rouse: 41-73. The Hague: Mouton.

1981. Seriation, groups, and measurement, in G. Cowgill, R. Whallon & B.S. Ottaway (ed.), Manejo de datos y metodos matematicos de arqueologia: 67-90. Mexico: Union Internacional de Ciendias Prehistricas y Protohistoricas.

DUNNELL, R.C. & J.K. FEATHERS. 1991. Late Woodland manifestations on the Malden Plain, southeast Missouri, in M.S. Nassaney & C.R. Cobb (ed.), Stability, transformation, and variation: the late Woodland Southeast: 21-45. New York (NY): Plenum Press.

DYE, T. 1990. The cause and consequences of a decline in the prehistoric Marquesan fishing industry, in D.E. Yen & J.M.J. Mummery (ed.), Pacific production systems: approaches to economic prehistory: 70-84. Canberra: Department of Prehistory, Australian National University.

1992. The South Point radiocarbon dates thirty years later, New Zealand Journal of Archaeology 14: 89-97.

EMORY, K.P., W.J. BONK & Y. SINOTO. 1959. Hawaiian archaeology: fishhooks. Honolulu (HI): B.P. Bishop Museum. Special Publication 47.

EMORY, K.P., & Y.H. SINOTO. 1964. Eastern Polynesian burials in Maupiti, Journal of the Polynesian Society 73: 143-60.

1969. Age of the sites in the South Point Area Ka'u, Hawaii. Honolulu (HI): B.P. Bishop Museum. Pacific Anthropological Records 8.

GILL, W.W. 1880. Historical sketches of savage life in Polynesia. Wellington: Government Printer.

GOTO, A. 1986. Prehistoric ecology and economy of fishing in Hawaii: an ethnoarchaeological approach. Ph.D dissertation, University of Hawaii, 1986. Ann Arbor (MI): University Microfilms International.

1990. Prehistoric Hawaiian fishing lore: an integrated approach, Man and Culture in Oceania 6: 1-34.

GOULD, S.J., D.M. RAPP, J.J. SEPKOSKI JR., T.J.M. SCHOPF & D.S. SIMBERLOFF. 1977. The shape of evolution: a comparison of real and random clades, Paleobiology 3: 23-40.

GREEN, R. 1960. The archaeology of the Mangarevan archipelago. American Museum of Natural History Expedition. Unpublished manuscript, on file, Department of Anthropology, University of Auckland.

1961. Review of Emory et al. (1959), Journal of the Polynesian Society 70:139-44.

1971. The chronology and age of sites at South Point, Hawaii, Archaeology and Physical Anthropology in Oceania 6: 170-76.

1981. Location of the Polynesian homeland, a continuing problem, in J. Hollyman & A. Pawley (ed.), Studies in Pacific languages and cultures in honour of Bruce Biggs: 133-58. Auckland: Linguistic Society of New Zealand.

1994. Changes over time: recent advances in dating human colonization of the Pacific Basin area, in D. Sutton (ed.), The origins of the first New Zealanders: 19-51. Auckland: Auckland University Press.

GREEN, R.C. & M. KELLY (ed.). 1970. Studies in Oceanic culture history. Honolulu (HI): Department of Anthropology, Bishop Museum.

GRUDGER, E.W. 1927. Wooden hooks used for catching sharks and Ruvettus in the South Sea: a study of their variation and distribution. New York (NY): American Museum of Natural History. Anthropological Papers 28(3).

INTES, A. 1982. La nacre en Polynesie francaise: Pinctada margaritifera Linne, Mollusca, Bivalvia: evolution des stocks naturels et de leur exploitation. Tahiti: ORSTOM. Notes et Documents, Oceanographie, 16.

IRWIN, G. 1992. The prehistoric exploration and colonization of the Pacific. Cambridge: Cambridge University Press.

JELINEK, A.J. 1976. Form, function, and style in lithic analysis, in C.E. Cleland (ed.), Cultural change and continuity: essays in honor of James Bonnet Griffin: 19-33. New York (NY): Academic Press.

JOHANNES, J.E. 1981. Words of the lagoon: fishing and marine lore in the Palau District of Micronesia. Berkeley (CA): University of California Press.

KENNEDY, D.G. 1930. Field notes on the culture of Vaitupu, Ellice Islands, Journal of the Polynesian Society 39: 101-24.

KIRCH, P.V. 1980. Polynesian prehistory: cultural adaptation in island ecosystems, American Scientist 68: 39-48.

1982. Advances in Polynesian prehistory, World Archaeology 1: 51-97.

1985. Feathered gods and fishhooks: an introduction to Hawaiian archaeology and prehistory. Honolulu (HI): University of Hawaii Press.

1986. Rethinking East Polynesian prehistory, Journal of the Polynesian Society 95: 9-40.

1993. Non-ceramic portable artifacts from the To'aga Site, in P.V. Kirch & T.L. Hunt (ed.), The To'aga site: three millennia of Polynesian occupation in the Manu'a Islands, American Samoa: 157-166. Berkeley (CA): Contributions of the University of California, Archaeological Research Facility, 51.

KIRCH, P.V. & T.S. DYE. 1979. Ethnoarchaeology and the development of Polynesia fishing strategies, Journal of the Polynesian Society 88: 53-76.

KIRCH, P.V., J.R. FLENLEY, D.W. STEADMAN, F. LAMONT & S. DAWSON. 1992. Ancient environmental degradation. Prehistoric human impacts on an island ecosystem: Mangaia, central Polynesia, National Geographic Research and Exploration 8: 166-79.

KIRCH, P.V. & D.E. YEN. 1982. Tikopia: The prehistory and ecology of a Polynesian outlier. Honolulu (HI): B.P. Bishop Museum. Bulletin 238.

LEACH, F. 1979. Excavations in the Washpool Valley Palliser Bay, in B.F. Leach & H.M. Leach (ed.), Prehistoric man in Palliser Bay: 67-136. Wellington: National Museum of New Zealand. Bulletin 21.

LEWERS, D. 1978. Understanding fishing tackle. Sydney: Reed.

MELTZER, D.J. 1981. A study of style and function in a class of tools, Journal of Field Archaeology 8: 313-26.

NEIMAN, F.D. 1990. An evolutionary approach to archaeological inference: aspects of architectural variation in the 17th century Chesapeake. Ph.D dissertation, Yale University. Ann Arbor (MI): University Microfilms International.

1995. Stylistic variation in evolutionary perspective: inferences from decorative diversity and inter-assemblage distance in Illinois Woodland Ceramic assemblages, American Antiquity 60: 7-36.

NORDHOFF, C. 1930. Notes on the off-shore fishing of the Society Islands, Journal of the Polynesian Society 39: 221-62.

O'BRIEN, M.J. & T.D. HOLLAND. 1990. Variation, selection, and the archaeological record, in M.D. Schiffer (ed.), Archaeological method and theory 2: 31-80. Tucson (AZ): University of Arizona Press.

OLIVER, D. 1988. Return to Tahiti: Bligh's second breadfruit voyage. Honolulu (HI): University of Hawaii Press.

OTTINO, P. 1992a. Anapua: Abri-sous-roche de pecheurs, etude des hamecons ([1.sup.re] partie), Journal de la Societe des Oceanistes 94: 57-79.

1992b. Anapua: Abri-sous-roche de pecheurs, etude des hamecons ([2.sup.e] partie), Journal de la Societe des Oceanistes 95: 201-26.

OWENS, J.F. & J.R. MERRICK. 1994. Analysis of coastal middens in south-eastern Australia: selectivity of angling and other fishing techniques related to Holocene deposits, Journal of Archaeological Science 21: 11-16.

PFEFFER, M. 1995. Distribution and design of Pacific octopus lures: the Hawaiian octopus lure in regional context, Hawaiian Archaeology 4: 47-56.

REINMAN, F. 1970. Fishhook variability: implications for the history and distribution of fishing gear in Oceania, in Green & Kelly (ed.): 47-60.

ROLETT, B.V. 1990. Hanamiai: changing subsistence and ecology in the prehistory of Tahuata (Marquesas Islands, French Polynesia). Ph.D dissertation, Yale University, 1989. Ann Arbor (MI): University Microfilms International.

1993. Marquesan prehistory and the origins of East Polynesian culture, Journal de la Societe des Oceanistes 96: 29-47.

ROUSE, I.B. 1939. Prehistory in Haiti, a study in method. New Haven (CT): Yale University. Publications in Anthropology 21.

SINOTO, Y.H. 1962. Chronology of Hawaiian fishhooks, Journal of the Polynesian Society 71: 162-6.

1967. Artifacts from excavated sites in the Hawaiian, Marquesan, and Society Islands: a comparative study, in G. Highland, R. W. Force, A. Howard, M. Kelly & Y.H. Sinoto (ed.), Polynesian culture history: 341-62. Honolulu (HI): Bishop Museum Press.

1968. Position of the Marquesas Islands in East Polynesian prehistory, in I. Yawata & Y.H. Sinoto (ed.), Prehistoric culture in Oceania: 111-18. Honolulu (HI): Bishop Museum Press.

1970. An archaeologically based assessment of the Marquesas Islands as a dispersal center in East Polynesia, in Green & Kelly (ed.): 105-32.

1976. Preliminary report of archaeological survey on Reao Atoll, Tuamotu Archipelago, French Polynesia, in A study of the Polynesian migration to the eastern Tuamotus: 102-181. The University of Kanazawa 1976 Polynesian Scientific Expedition, Japan.

1988. A waterlogged site on Huahine Island, French Polynesia, in B. Purdy (ed.), Wet site archaeology: 113-30. Caldwell (NJ): Telford Press.

1991. A revised system for the classification and coding of Hawaiian fishhooks, Bishop Museum Occasional Papers 31: 85-105.

SINOTO, Y.H. & P. McCOY. 1975. Report on the preliminary excavation of an early habitation site on Huahine, Society Islands, Journal de la Societe des Oceanistes 41: 143-86.

SKINNER, H.D. 1942. A classification of the fish-hooks of Murihiku with notes on allied forms from other parts of Polynesia, Journal of the Polynesian Society 51: 208-21, 256-86.

STODDART, D.R. 1975. Almost-atoll of Aitutaki: geomorphology of reefs and islands, in D.R. Stoddart & P.E. Gibbs (ed.), Almost-Atoll of Aitutaki: reef studies in the Cook Islands: 31-58. Washington (DC): Smithsonian Institution.

SUGGS, R.C. 1961. The archaeology of Nuku Hiva, Marquesas Islands. east Polynesia. New York (NY): The American Museum of Natural History. Anthropological Papers 49(1).

SUTTON, D.G. 1987. A paradigmatic shift in Polynesian prehistory: implications for New Zealand, New Zealand Journal of Archaeology 9: 135-55.

TELTSER, P.A. 1995. Culture history, evolutionary theory, and frequency seriation, in P.S. Teltser (ed.), Evolutionary Archaeology, Methodological Issues: 51-8. Tucson (AZ): University of Arizona.

VERIN, P. 1969. L'ancienne civilisation de Rurutu (Isles Australes, Polynesie Francaise). Paris: ORSTOM. Memoires 33.

WALTER, R. 1989. An archaeological fishhook assemblage from the southern Cook Islands, Man and Culture in Oceania 5: 67-77.

1990. The southern Cook Islands in eastern Polynesian prehistory. Unpublished Ph.D dissertation, Anthropology Department, University of Auckland.

YAMAGUCHI, T. 1995. Archaeological monograph of Tongareva (Penrhyn) Atoll in M. Chikamori, S. Yoshida, and T. Yamaguchi, Archaeological Studies on the Cook Islands, Series 1:17-106. Tokyo: Department of Archaeology and Ethnology, Keio University. Occasional Papers 10.