Did ciguatera prompt the late Holocene Polynesian voyages of discovery?

Authors


*Teina Rongo, Department of Biological Sciences, Florida Institute of Technology, 150 West University Blvd., Melbourne, FL 32901, USA. E-mail: trongo@fit.edu

Abstract

The famous Polynesian voyages characterized an intensive network of cultural exchange and colonization that was particularly active from ad 1000 to 1450. But, why would large groups of people leave their homelands to voyage into the unknown? Oceanic voyages are risky, albeit less so today than in the past. Landfalls were not guaranteed improvements over ports of departure. Taking the Cook Islands as an example, we ask whether harmful algal blooms that result in ciguatera fish poisoning in humans prompted past and present emigration pulses of peoples from within Polynesia. We take a multipronged approach to examine our hypothesis, involving: (1) archaeological evidence, (2) ciguatera fish poisoning reports since the 1940s, and (3) climate and temperature oscillations using palaeodatasets. The archaeological records of fish bones and hooks show abrupt changes in fishing practices in post-ad 1450 records. Sudden dietary shifts are not linked to overfishing, but may be a sign of ciguatera fish poisoning and adjustment of fishing preference. While fishes form the staple diet of Polynesians, such poisoning renders fishes unusable. We show that ciguatera fish poisoning events coincide with Pacific Decadal Oscillations and suggest that the celebrated Polynesian voyages across the Pacific Ocean may not have been random episodes of discovery to colonize new lands, but rather voyages of necessity. A modern analogue (in the 1990s) was the shift towards processed foods in the Cook Islands during ciguatera fish poisoning events, and mass emigration of islanders to New Zealand and Australia.

Introduction

The period between ad 1000 and 1450 was characterized by extensive cultural exchanges in Polynesia, which declined once marginal regions of Polynesia (e.g. Rapa Nui, Hawaii and New Zealand; Fig. 1a) were colonized (Hunt & Lipo, 2006; Kirch & Kahn, 2007; Wilmshurst et al., 2008). The collapse of the voyaging network between island archipelagos led to complete isolation of marginal regions, and the development of distinct cultures. However, low-level inter-island communications continued until ad 1800 (Rolett, 1998).

Figure 1.

 (a) Waves of migration (shaded grey arrows) originating from East Polynesia (dotted region; French Polynesia) with approximate migration dates to Rapa Nui (Hunt & Lipo, 2006), Hawai’i (Kirch & Kahn, 2007) and New Zealand (Wilmshurst et al., 2008). A possible subsequent migration from the southern Cook Islands (boxed region) to New Zealand after ad 1450 is noted. (b) Southern Cook Islands. (c) Map of Aitutaki indicating archaeological sites (Allen, 1992, 2002).

The early ‘great fleet theory’ (Smith, 1921), suggested that a single wave of seven canoes colonized New Zealand around ad 1350. It is now generally accepted that multiple waves of migrations occurred, originating from the East Polynesian ‘homeland’ (Rolett, 2002), with the first canoe arriving in New Zealand around ad 1280 (Wilmshurst et al., 2008). But why would large groups of people leave their homelands to voyage into the unknown? Oceanic voyages were risky. Landfalls were not guaranteed improvements over ports of departure. Explanations for migration range from overpopulation, resource overexploitation, warfare, environmental change, advances in canoe technology (e.g. double-hulled canoe) and the adventurous spirit of Polynesians combined with favourable winds associated with the El Niño–Southern Oscillation (ENSO; Bridgman, 1983; Kirch, 1996; Finney, 1998; Rolett, 2002; Anderson et al., 2006). However, factors affecting marine resources, the main source of protein in Polynesia, have been ignored. We suggest that some migrations occurred when fish resources were unusable.

For millennia Polynesians have had the technology to capture large pelagic and coral-reef fishes. However, a heavy reliance on a fish-based diet makes Pacific Islanders particularly vulnerable to harmful algal blooms that cause fishes to become toxic. Examining palaeoclimate and archaeological records, we ask if harmful algal blooms that result in ciguatera fish poisoning could have prompted waves of migrations of peoples across Polynesia between ad 1000 and 1450, a period when climatic conditions may have been favourable for ciguatera fish poisoning in the ‘homeland’. We suggest that after the collapse of cultural exchanges around ad 1450, societal adaptations to ciguatera fish poisoning resulted in dietary shifts away from ciguatoxic predatory fishes. We draw a parallel with the modern 1990s migration of Cook Islanders to New Zealand and Australia during ciguatera fish poisoning events.

Overview of ciguatera

Early historical accounts of ciguatera fish poisoning in the Pacific Ocean are found in Captain James Cook’s journal. While moored at Vanuatu in 1774, Captain Cook recorded a fish poisoning incident which has been interpreted as ciguatera fish poisoning (Doherty, 2005). To date, ciguatera fish poisoning is the most common form of toxic seafood poisoning globally (Baden et al., 1995), affecting between 50,000 and 500,000 people annually, in tropical to subtropical regions (Quod & Turquet, 1996). Yet, for cultural reasons, about 80–90% of ciguatera fish poisoning cases go unreported (Dalzell, 1991).

Ciguatera fish poisoning is generally considered a consequence of fishes inadvertently ingesting toxic dinoflagellates (e.g. Gambierdiscus toxicus Adachi and Fukuyo, 1979; Yasumoto et al., 1977) that are epiphytic to macroalgae (see Cruz-Rivera & Villareal, 2006). Recently, cyanobacteria have also been proposed to have a role as an inducer of ciguatera fish poisoning (Laurent et al., 2008). Ciguatoxins bioaccumulate within the food web and are highly concentrated in large, carnivorous fishes (Lewis, 2006). Because these toxigenic dinoflagellates rarely bloom, and only a low percentage of the strains produce detectable levels of toxin (Holmes et al., 1991), the distribution of ciguateric fishes is patchy, and therefore difficult to quantify in the field (Lewis, 2006). Ciguatera fish poisoning is characterized by gastrointestinal (i.e. vomiting and diarrhoea), neurological (i.e. temperature dysthesia and chronic fatigue) and cardiovascular symptoms (Lehane & Lewis, 2000). Extreme cases can lead to paralysis or death. During ciguatera fish poisoning events, most large reef fishes are rendered unusable in the food supply (Lehane & Lewis, 2000). Because these toxigenic dinoflagellates are limited to shallow coastal waters, pelagic migratory species, such as bonito, are not affected.

Effects of ciguatera in the southern Cook Islands

Anecdotal reports of ciguatera fish poisoning were noted from Aitutaki in the mid-1940s, but the first recorded cases in the southern Cook Islands (Fig. 1b) were in 1984 (Losacker, 1992). Based on the resident population on Rarotonga for 2001 (Cook Islands Statistics Office), the average incidence of ciguatera reported for Rarotonga from 1993 to 2006 was 17.6 cases per 1000 persons per year (the highest in the literature to date). Because most reef fishes are potentially ciguatoxic, about 71% of Rarotonga residents have excluded fish from their diet (Hajkowicz, 2006). Residents who still eat fish eat only pelagic species, which are costly, and to a lesser extent also eat those reef fishes considered ‘safe’. Occasionally, residents import reef fishes from the northern Cook Islands, which according to locals are currently unaffected by ciguatera. This restriction, incidentally, has resulted in an increase of cardiovascular diseases because of a heavy reliance on processed foods (Li et al., 1994), but also in part led to the emigration of 18% of the resident Cook Islands’ population (from 1994 to 2000; Cook Islands Statistics Office, 2007) to New Zealand and Australia during the 1990s, partly as a consequence of the high cost of processed foods. Most emigrant families were from the lower-income strata of Rarotonga, many of whom supplemented their daily protein with coastal marine resources (R. Crocombe, personal communication). In addition, the status of free association between the Cook Islands and New Zealand facilitated relocation. A similar movement in the past is not improbable because the areas least likely to be influenced by ciguatera were (and are) the islands amid cool waters, such as Rapa Nui and New Zealand.

Palaeoclimate in the Pacific

A variety of proxy records based on ice cores, lake records, oceanic clastic deposition and coral cores shed light on Pacific Ocean sea surface temperatures (SSTs) during the period of Polynesian migrations. The largest single cause of interannual variability in Pacific climates is the ENSO. El Niño tends to bring cool SSTs to the western and central Pacific while La Niña episodes promote warm SSTs and higher rainfall. However, regional differences in SSTs during El Niño have been noted (Hales et al., 1999); for example, the southern Cook Islands are cooler during El Niño conditions and warmer than usual during La Niña, while the opposite is apparent in the northern Cook Islands and French Polynesia (i.e. the Society, Marquesas, and Tuamotu Islands) (Fig. 2a,b).

Figure 2.

 The South Pacific region, with red areas indicating warmer sea surface temperatures (SSTs) and blue areas indicating cooler SSTs during (a) El Niño and (b) La Niña (modified from Hales et al., 1999, with permission from Blackwell Publishing). These regional differences noted during the El Niño–Southern Oscillation are maintained during the Pacific Decadal Oscillation (PDO), but are less intense, with positive PDO similar to El Niño (a) and negative PDO similar to La Niña (b). Dotted lines delineate regions of interest.

Palaeorecords consistently indicate that the period from ad 900 to 1400 was one of relatively low El Niño activity (Thompson et al., 1984, 2000; Hendy et al., 2002; Moy et al., 2002; Cobb et al., 2003) with sporadic bursts of La Niña activity. Of the proxy records available, the Quelccaya ice cap provides the most direct quantification of ENSO intensity, with low precipitation corresponding to strong La Niña events (Thompson et al., 1984Thompson et al., 2000). Between c. ad 1250 and 1300 drought reduced Andean precipitation at the Quelccaya ice cap by > 20% compared with the long-term mean (Thompson et al., 1984). The period from ad 1300 to 1720 was a time in which La Niña activity subsided and El Niño became dominant.

Using wavelets we reanalysed the c. 270-year Strontium/Calcium (Sr/Ca) raw dataset derived from a Rarotonga coral by Linsley et al. (2000). The Sr/Ca ratio is an accurate proxy of water temperature (McCulloch et al., 1994). The analysis revealed significant multidecadal anomalous temperature variability recurring every 30 years (Fig. 3). Although results indicate some ENSO-like signatures around ad 1800 in the southern Cook Islands, clearly multidecadal oscillations related to the Interdecadal Pacific Oscillation cycle (IPO; Power et al., 1999) dominated temperature changes over the last 300 years (Linsley et al., 2000; Ren et al., 2002). The IPO and Pacific Decadal Oscillation (PDO; Mantua et al., 1997) are highly correlated and equivalent in describing Pacific-wide variations in ocean climate (Power et al., 1999; Verdon & Franks, 2006), therefore we used the PDO to describe interdecadal cycles in the South Pacific region, because of the extended time-series data available. Although the link between ENSO and PDO is poorly understood, Verdon & Franks (2006) indicated a coupling effect in which El Niño is frequent during the positive PDO and La Niña is frequent during negative PDO. Indeed, regional SST differences noted during ENSO are maintained during PDO but are less intense (see Fig. 2a,b).

Figure 3.

 (a) Changes in estimated sea surface temperature anomalies spanning from ad 1726 to 1996 for Rarotonga using coral Sr/Ca ratios (Linsley et al., 2000). (b) Local time-frequency spectrum, where dark-red shading indicates high sea surface temperature variance and black contours enclose regions that are significantly above red noise at the 95% confidence level. (c) The results of a chi-square test that displays the return period (y-axis) in relation to red noise; above-noise return periods are significant when displayed to the right of the global red-noise (indicated as a dashed line), which is only evident for Rarotonga at the multidecadal timescale (Torrence & Compo, 1998).

Climate and ciguatera fish poisoning events

Environmental factors affecting ciguatera fish poisoning remain complex, and no single factor seems to consistently drive these events. For example, ciguatera fish poisoning cases were positively correlated to SST anomalies on Rarotonga from 1973 to 1994 (Hales et al., 1999). However, our examination from 1993 to 2006, using reported cases provided by the Cook Islands Ministry of Health, and SST anomalies from the Cook Islands Meteorological Service, showed no significant correlation (Pearson’s correlation, = 0.26; = 0.372). Interannual cycles such as the Southern Oscillation Index (SOI) have been shown to be negatively correlated with ciguatera fish poisoning cases (Hales et al., 1999); however, our analysis from 1993 to 2006 for Rarotonga, using SOI data from the National Climate Centre (2009), showed no significant association (= −0.13; = 0.653). While the link between ciguatera fish poisoning events and interannual cycles remain unclear, we suggest possible links with long-term climatic oscillations. Based on the literature, anecdotal reports and available hospital records of ciguatera fish poisoning events from the Cook Islands and French Polynesia, we examined long-term relationships of ciguatera fish poisoning events with climate forcing. For our purposes, we refer to periods when ciguatera fish poisoning begins to cause nuisance to the human population as the ‘initial’ ciguatera fish poisoning events, which we consider unbiased because humans tend to remember these periods.

Over the last 70 years, we note that ciguatera fish poisoning cases in regions examined coincided with different PDO phases. For example in the southern Cook Islands, anecdotal reports were noted in Aitutaki in the mid-1940s (Losacker, 1992) during the positive PDO (1925–46). There were no reports in the southern Cook Islands during the 1950s when PDO shifted to a negative phase (1947–76), but extensive ciguatera reports were evident in the 1980s during the recent positive PDO (1977–98; Fig. 4a). In the northern Cook Islands (Penrhyn Atoll), (initial) ciguatera fish poisoning events occurred in the late 1950s (Losacker, 1992; J. Williams, personal communication), during the negative phase, but declined in the mid 1990s (T. Rongo, personal communication with local inhabitants of the northern Cook Islands) during a predominantly positive phase (see Fig. 4a). Early reports of ciguatera fish poisoning in French Polynesia were noted in the early 1960s during the negative PDO (1947–76), but generally declined toward the end of the study period in 1984 (Bagnis et al., 1985), as PDO shifted to positive. Therefore, French Polynesia and the northern Cook Islands experienced ciguatera fish poisoning events during the negative PDO, when conditions were similar to those in the southern Cook Islands during a positive PDO (see Fig. 2). In the light of these examples, we suggest that PDO played a role in driving ciguatera fish poisoning events in these regions. Drawing from our proposed link to ciguatera fish poisoning events in East Polynesia, we examine whether PDO phase shifts during the late Holocene (Fig. 4b) coincided with dietary and technological shifts noted in southern Cook Island archaeological records (after ad 1450).

Figure 4.

 (a) Pacific Decadal Oscillation (PDO) index from 1930 to 2007 with an 11-year running average (data from http://jisao.washington.edu/pdo/PDO.latest). Blue shaded regions: poisoning events in the southern Cook Islands. Green shaded region: poisoning events in the northern Cook Islands and French Polynesia. Horizontal bars: the predominant PDO phase (+ or −). (b) PDO index from ad 993 to 2007 with an 11-year running average (ad 993–1996, data taken from MacDonald & Case, 2006, and ad 1997–2007 from part a). Green shaded region: heightened voyaging in Polynesia (ad 1000–1450). Blue shaded region: collapse of voyaging network and dietary shifts in archaeological records (ad 1450–1550).

Evidence of cultural adaptations

Archaeological evidence and primary dietary indicators derived from middens and site excavations show temporal changes in both diet and technology. While other explanations have been offered, these archaeological findings are consistent with our hypothesis of ciguatera fish poisoning shaping Polynesian cultures. The colonization of East Polynesia was marked by a rapid decline of terrestrial resources such as bird populations (Steadman, 1989; Steadman & Kirch, 1990; Steadman & Rolett, 1996; Grayson, 2008). Midden records reveal that as terrestrial resources became scarce, coastal people relied more on fishes as their primary protein resource (Broughton, 1994). Archaeological indicators of dietary shifts are drawn from middens and site excavations. Archaeo-fish assemblages provide direct evidence of food remains, and hooks provide an indication of the size of the target prey. Various fishing methods such as traps, nets, spears and poisoning were practised throughout the Pacific. Angling was not important until people reached East Polynesia, where use of the pearl-shell (Pinctada margaritifera Linnaeus) hook became widespread, primarily because of its abundance, strength and efficacy as a lure (Allen, 1992).

Archaeological evidence of bonito lure shanks from Mo’orea, French Polynesia, suggests that offshore fishing was also important in this region (Green et al., 1967). Furthermore, bone assemblages recovered from Huahine, Society Islands, indicated that pelagic fishes formed a high proportion of the islanders’ protein source (Leach et al., 1984). While the use of other shell material such as Turbo was limited to small one-piece hooks, pearl-shell was used for most fish hooks (Green et al., 1967). Extensive use of pearl-shell suggests that angling or trolling for pelagic fishes were important to past populations in East Polynesia.

In the southern Cook Islands, the collapse of the voyaging network after ad 1450 was marked by a shift from the preferred pearl-shell to the weaker Turbo hooks on Aitutaki, Mangaia, Mitiaro and Ma’uke (Steadman & Kirch, 1990; Allen, 1992; Walter & Campbell, 1996) (see Fig. 1b; Table 1). The shift also coincided with dietary shifts in Aitutaki (Allen, 1992) and Mangaia middens (Steadman & Kirch, 1990; Butler, 2001), indicating a transition from large to small inshore fishes [e.g. small cirrhitids (hawkfish) and small serranids (grouper) were consistent throughout the middens; Butler, 2001; Allen, 2002] (see Table 1). Fishing in general declined on Mangaia and Aitutaki (at Ureia and Moturakau; Fig. 1c) after ad 1450; declines were most noticeable in lutjanids (snapper), lethrinids (emperor), belonids (crocodile needlefish) and sphyraenids (barracuda) (Butler, 2001; Allen, 2002) (Table 1), species that are considered potentially highly ciguatoxic or ‘high-risk’. While carangids also declined in both Ureia and Moturakau middens on Aitutaki after ad 1450, there was a slight increase around the European contact period (post-ad 1650). In contrast, muraenids (moray eels; a high-risk family) increased in general after ad 1450 at Moturakau (Table 1).

Table 1.   Summary of fish middens from archaeological records (limited to potentially ciguatoxic fishes taken by angling), and terrestrial resources consumed. Records for Aitutaki were taken from Allen (2002), and the Mangaia record from Steadman & Kirch (1990) and Butler (2001). Contact is defined as post-ad 1650.
IslandTime (ad) and dominant hook materialFish targeted by anglingCiguatoxic fish avoided (declined* or absent between 1450 and 1650)Terrestrial resources
Offshore and reef edgeInshore
  1. *Indicates fish that declined in the record after ad 1450.

  2. †Indicates an increase in abundance.

  3. Bold font indicates consistency throughout record.

Aitutaki (Ureia)950–1450
Pearl-shell
Belonids (large)
Carangids (large)
Lutjanids (large)
Lethrinids (large)
Serranids (small)
 Domesticates
Non-fish vertebrates
1450– Contact
Turbo
 Serranids (small)Carangids (large)*
Lutjanids (large)*
Lethrinids (large)*
Belonids (large)
Domesticates†
Non-fish vertebrates†
Aitutaki (Moturakau)1250–1450
Pearl-shell
Belonids (large)
Carangids (large)
Lutjanids (large)
Sphyraenid (large)
Lethrinids (large)
Muraenids (large)
Serranids (small)
 Domesticates
Commensals
1450– Contact
Turbo
Carangids (large)†Muraenids (large)†
Serranids (small)
Belonids (large)*
Lutjanids (large)*
Lethrinids (large)*
Sphyraenid (large)
Domesticates
Commensals†
Mangaia1000–1450
Pearl-shell
Muraenids
Serranids (large)
Serranids (small)
Carangids
Lutjanids
Cirrhitids (small)
  Land birds (rapidly depleted)
Domesticates
Commensals
Anguillids (large)
Eleotris
1450–1650
Turbo
Serranids (small)
Cirrhitids (small)
 Muraenids*
Carangids
Lutjanids
Serranids (large)
Domesticates†
Commensals†
Anguillids (small)†
Eleotris

Fishes became eclipsed in the diet by domesticates (e.g. chickens, dogs and pigs) and commensals (e.g. rats) on Aitutaki and Mangaia after ad 1450 (Steadman & Kirch, 1990; Allen, 2002; see Table 1). Anguillids (freshwater eels) and eleotrids (sleeper gobies) also became conspicuously common in the Mangaia midden (cal. yr ad 1000–1700). Most interesting was the seemingly sequential reduction in the relative abundance of anguillids that followed the initial increase (Steadman & Kirch, 1990; Butler, 2001), suggesting that this terrestrial resource (unaffected by ciguatera fish poisoning) was subjected to progressive overharvesting.

Discussion

Although ciguatera fish poisoning cases were linked to elevated SST anomalies on Rarotonga from 1973 to 1994 (Hales et al., 1999), we found no significant association with SST from 1993 to 2006. The inconsistency may be attributed to the time periods examined, where SST anomalies were important from 1973 to 1994, but not from 1993 to 2006. While under-reporting is likely to bias both analyses, data obtained from 1973 to 1994 included the entire Cook Islands (including the northern group where ciguatera fish poisoning was chronic during that period). According to the Cook Islands Ministry of Health, cases prior to 1993 were not separated by island and were therefore excluded from our analysis. We note clear differences between the climatic conditions in the northern and southern Cook Islands under ENSO and PDO conditions (see Fig. 2a,b). While the southern Cook Islands are bathed in cool waters during El Niño conditions, the northern Cook Islands are warm. Therefore, combining ciguatera cases without considering locality may mask regional ciguatera interpretations.

Studies linking ciguatera fish poisoning events to interannual cycles have been ambiguous; perhaps relationships may be better understood at interdecadal cycles. The impact of interdecadal cycles has been noted in the Pacific Ocean, where variability in fisheries was linked to PDO phase shifts (Lluch-Belda et al., 1989; Chavez et al., 2003). For example, dramatic declines in the production of Alaskan salmon in the north-east Pacific Ocean occurred during the negative phase of the PDO, but increased during the positive phase (Hare & Francis, 1995; Mantua et al., 1997). Similarly, recruitment of yellowfin tuna in the eastern Pacific Ocean declined during the negative phase of the PDO and increased during the positive phase (Maunder & Watters, 2001). Based on initial ciguatera poisoning events that coincided with different PDO phases, we suggest that positive phases are linked to ciguatera fish poisoning events in the southern Cook Islands, while events in the northern Cook Islands and French Polynesia are linked to negative phases, although some ciguatera fish poisoning cases were reported outside these climatic windows (T. Rongo, personal communication with local inhabitants of the northern Cook Islands; Chateau-Degat et al., 2007). Most important, however, is that these different phases (i.e. negative PDO in the northern Cook Islands and French Polynesia, and positive PDO in the southern Cook Islands) elicit the same climatic circumstances (see Fig. 2). Determining the importance of PDO and the timing of initial ciguatera fish poisoning events in different regions of the South Pacific will be important to further examine this possible link.

The heightened voyaging network noted from ad 1000 to 1450 in eastern Polynesia may have been prompted by ciguatera fish poisoning. As terrestrial resources were depleted, Polynesians resorted to fishing. But when ciguatera fish poisoning was encountered, few alternatives were available. Advanced canoe technology and steady north-easterly La Niña winds (Anderson et al., 2006) during this negative PDO period may have increased the probability of voyaging in the direction of New Zealand. These favourable winds were tested by a modern replicate of a traditional double-hull long-distance voyaging canoe (Hokule’a), where a crossing from Rarotonga to New Zealand was achieved in 16 days (Finney, 1998). The adventurous spirit and superb navigational skills of Polynesians, who always had return routes planned (as observed by Captain Cook in Tahiti; see Finney, 1998) when venturing into uncharted waters (Michael Tavioni, personal communication), also made voyaging less daunting.

Though much of the native forest of Aitutaki was removed by ad 1100, the ‘tamanu’ tree (genus Calopyllum), important for canoe-building, remained throughout the record (Allen, 1992), indicating that long-distance voyaging and offshore fishing was still possible after ad 1100. The increase in offshore fishing in eastern Polynesia has been attributed to the availability of pearl-shell (see Allen, 1992). We agree that pearl-shell trading was critical to offshore pelagic fishing, but alternatively suggest that offshore fishing was influenced by ciguatera fish poisoning. The shift to offshore fishing may have been driven by an understanding that migratory pelagic fishes were ‘safe’ to eat. Modern examples of the shift to pelagic species because of ciguatera fish poisoning are noted on Raivavae, Austral Islands (M. Chinain, personal communication) and in the southern Cook Islands. Consequently, the collapse of voyaging and of the pearl-shell trade after ad 1450 would have reduced offshore fishing in the southern Cook Islands. Such circumstances would have been particularly consequential when PDO became positive around ad 1450 (see Fig. 4b), increasing the likelihood of ciguatera fish poisoning events in the southern Cook Islands.

The collapse of the voyaging network after ad 1450 would have led to the isolation of populations in the southern Cook Islands where adjustment to ciguatera fish poisoning was inevitable. Consequently, the dietary shift from large carnivorous fishes to smaller fishes and terrestrial resources may indicate a human behavioural adjustment (see Table 1). Small serranids and cirrhitids were consistent throughout middens despite the decline in overall fishing. Even today, in the southern Cook Islands, small serranids and cirrhitids are caught in the lagoon and off the reef edge using bamboo rods (takiri) with baited hook (T. Rongo, personal observation). However, for most of the 1990s, because of the ciguatera fish poisoning scare, takiri and fishing in general declined. Recently, Cook Islanders have learned, or as it may turn out relearned, that small serranids (e.g. Epinephelus merra and Epinephelus hexagonatus) and cirrhitids (Cirrhitus pinnulatus) are among the few reef fishes considered ‘safe’. The decline in lutjanids from middens on Aitutaki has been attributed to the loss of pearl-shell hooks (Allen, 2002). Alternatively, we suggest that lutjanids were purposely avoided. In the southern Cook Islands today, it is common knowledge that lutjanids (i.e. Lutjanus fulvus and Lutjanus monostigma) are off-limits because they are considered high risk.

Reasons offered for the shift towards small fishes in Aitutaki and Mangaia middens were shifts in fishing technology, environmental change, overfishing or a preference for smaller fishes (Butler, 2001; Allen, 2002); we support the latter scenario. While contemporary signatures of overfishing include sequential reductions in average size frequency, this is not apparent on Rarotonga today; rather the size-frequency distributions are skewed to the left (T. Rongo, unpublished data), an apparent result of preference for small, low-risk fishes. The interpretation of archaeo-fish assemblages from the Pacific has been challenging because certain fish species do not preserve well (particularly small species; Nagaoka, 1994). In addition, reference collections are lacking (Nagaoka, 1994). Consequently, the highest taxonomic resolution has been to fish family [for the Aitutaki (Allen, 2002) and Mangaia middens (Butler, 2001)]. Identification to species level and information on size-frequency distribution may facilitate the interpretation of ciguatoxic prevalence, as large individuals bioaccumulate toxins while small and juvenile individuals are often ‘safe’. Clearly, archaeological evidence in the southern Cook Islands points to a shift to ‘safer’ fishes. Because ciguatera exists on these islands today, we suggest that instantaneous shifts in fish use were in the past probably a consequence of ciguatera fish poisoning.

Conclusions

Counter to the general trend across the Pacific Ocean, where fishing was important over time (Allen, 2002), a general decline in fishing was noted in the southern Cook Islands after ad 1450. This can be compared with recent ciguatera fish poisoning incidents in the 1980s that led to a sudden decline in fishing in the 1990s. However, ciguatera fish poisoning may have had a greater impact in the past, when imported alternatives were unavailable. Terrestrial resources were limited on small islands, and ciguatera fish poisoning rendered marine resources unusable. The impact on past human populations may have been life-threatening and options were few.

Clearly, ciguatera fish poisoning influenced the migration of Cook Islanders to New Zealand and Australia in the 1990s, and may have also induced late Holocene human population migrations. Notwithstanding the adventurous spirit of people of the distant past, we suggest that when ciguatera fish poisoning became chronic, people migrated out of necessity. Our inference that migration originating from French Polynesia was associated with ciguatera fish poisoning events and negative PDOs is consistent with the timing of heightened voyages and the colonization of marginal regions of Polynesia (see Fig. 1a). We suggest that the importance of offshore fishing and the dietary shifts noted in middens were indications of adjustments to ciguatera fish poisoning. We also infer that the collapse of voyaging after ad 1450 possibly marked the decline of ciguatera fish poisoning in French Polynesia, and its increase in the southern Cook Islands when the PDO shifted from negative to positive. Therefore, any migration originating from the southern Cook Islands to New Zealand could have occurred after ad 1450 (see Fig. 1a) when ciguatera fish poisoning became chronic.

Although our inferences are circumstantial and drawn from available climate and archaeological records, the possibility of ciguatera fish poisoning affecting late Holocene human populations throughout the Pacific Basin should not be overlooked, and future archaeological studies should consider its impact. Our examination used the longer PDO reconstruction available from the northern Pacific Ocean (MacDonald & Case, 2006); perhaps a reconstruction from the South Pacific region would allow a rigorous test of our proposed hypothesis. Furthermore, we need a more thorough biogeographical examination of the impact of interdecadal climate fluctuations that may trigger ciguatera fish poisoning events, which in turn may have critically influenced the behavioural patterns of past human populations. Detecting ciguatera fish poisoning events that happened in the past will not only aid our understanding of the influence of climate on human history, but will also guide us in the future.

Acknowledgements

We thank Teariki Rongo, Ma’ara Maeva, Peter Houk and Keri Jones for their valuable suggestions. Thanks to the three referees and editor, Lawrence Heaney, for their contributions to improving this manuscript. Meitaki maata to Tearoa Iorangi of the Cook Islands Ministry of Health, Arona Ngari of the Cook Islands Meteorological Service, and Davina Hosking of New Zealand’s National Institute of Water and Atmospheric Research for their assistance. A special thanks to Sandra van Woesik for editorial assistance and to Jackalyn Rongo for research assistance and discussions throughout the preparation of the manuscript. This is Contribution Number 2 from the Institute for Adaptation to Global Climate Change at the Florida Institute of Technology.

Biosketches

Teina Rongo is a Cook Islands Maori, born on Rarotonga and raised in a family where subsistence fishing was a normal practice. This cultivated his passion for the marine environment, motivating him to pursue a BA and MS in Biology at the University of Guam. He is currently a PhD student in the Department of Biological Sciences at the Florida Institute of Technology.

Mark Bush is a Professor of Biology at the Florida Institute of Technology. His research focuses on fossil pollen analysis of Neotropical settings, environmental reconstructions of past climates and vegetation communities, and palaeoecological evidence of human responses to climate change.

Robert van Woesik is a Professor of Biology at Florida Institute of Technology. His research focuses on understanding coral-reef processes associated with climate change, historical and modern thermal stresses, coral bleaching and modelling population trajectories.

Editor: Lawrence Heaney

Ancillary