The Salish Sea and Salishan Subsistence

“Salish Sea” was recently adopted as the official name for the waters of the Strait of Juan de Fuca, Strait of Georgia, and Puget Sound, in tribute to the Salishan people who lived on these shores for millennia. Our study area (Figure 1) in the central part of the Salish Sea is (and was) occupied by people speaking Central Coast Salish (the Downriver and Vancouver Island Halkomelem on the eastern and western shores of the Strait of Georgia, respectively), Northern Straits Salish (the Sooke, Songish, and Saanich on southern Vancouver Island; the Samish and Lummi on the San Juan Islands; and the Semiahmoo on the mainland coast), and Klallam (the S’Klallam on the Olympic Peninsula). Each group had a distinctive dialect and a discrete territory, but they interacted extensively and often intermarried. Boyd (Reference Boyd and Suttles1990) estimated the total population of these groups in the immediate precontact era to be about 16,000 people. This estimate, he notes, is likely a minimum value but includes the people living in the riverine villages, as well as those on the coast.

Figure 1.

(A) “Salmosphere” of the northeast Pacific Ocean; (B) the Salish Sea (JStr = Johnstone Strait) and the Fraser River watershed, showing the migration routes of juvenile and adult sockeye salmon; (C) the central segment of the Salish Sea, showing locations mentioned in the text. BI = Burrard Inlet, BB = Boundary Bay, HS = Haro Strait, RS = Rosario Strait; GI = Gulf Islands, SJ = San Juan Islands, Bell = Bellingham, Nan = Nanaimo, R = Point Roberts, So = Sooke, Van = Vancouver, Vic = Victoria; archaeological sites: CPt = Cattle Point, Cx = Čḯx^wicən, EC = English Camp, PL = Parry Lagoon, Sq = Sequim, WB = Watmough Bay. (Color online)

Shell deposits, often hundreds of meters in length and several meters deep, mark the locations of the ancestral shoreline villages and camps. Mollusk shells and fish bones are the dominant organic constituents in these deposits. The fish-bone assemblage is generally diverse, with salmon and herring as prominent components (McKechnie and Moss Reference McKechnie and Moss2016).

Although salmon stocks have declined substantially over the course of the last century, about ten million salmon still attempt to return each year to their natal streams around the margins of the sea. Seven million of them have the Fraser River and its tributaries as their goal.

Indigenous fishers were familiar with the cyclicity of the local salmon runs, but long-term variability caused by changes in the North Pacific environment would have been difficult to anticipate. Persistent weak salmon runs would undoubtedly lead Salishan people to expend more effort in harvesting other resources, because even in years when salmon were plentiful, a diverse array of fish (e.g., herring, eulachon, cod, and flounder), plants (roots, shoots, nuts, and berries), and animals (mollusks, deer, and waterfowl) were seasonally harvested, and, if not consumed immediately, preserved by drying or smoking as a winter food supply.

The ethnographic record indicates that virtually all the coastal Salishan groups engaged in seasonal movements that tracked the availability of these disparate food sources. In the nineteenth century, the Nanaimo, for example, left their winter village on the east coast of Vancouver Island in April, after the herring harvest, to fish, hunt, and dig clams on an island in the Strait of Georgia. In August they fished for salmon on the lower Fraser River, moving to the mouth of the Nanaimo River to intercept the local runs of salmon in late fall and returning to their home village in December (Bouchard Reference Bouchard and Wilson1994). If the main salmon runs diminished, however, the mainstream fishing camps might only be occupied for a brief period each summer, with a concomitant increase in the length of time spent fishing, harvesting plants and shellfish, or hunting in the satellite areas of their territories. And if the local salmon runs collapsed, then people might be forced to move to places with a more reliable resource base.

People from a single village often dispersed in spring, coordinating their movements with other relatives and in-laws, with whom they fished, harvested, hunted, exchanged food, traded goods, and socialized (Suttles Reference Suttles1960). Exogamous marriage resulted in individuals having family ties to many local villages: a tradition of food sharing with relatives was a buffer against hunger during challenging times.

Residential flexibility, at least on a local scale, was one of the defining features of Salishan society. Family members could move, sometimes permanently, to any village in which they had kin, thereby creating “a sophisticated, far-reaching and versatile social system that . . . helped its participants cope with change, giving most of them more than one possible place to belong” (Harmon Reference Harmon and Miller2007:48). Higher-status individuals had the most extensive social ties, but because they were also the most invested in maintaining their inherited property rights, they were less likely to leave their natal village. “Commoners,” however, had no individual property rights and were more likely to be transients. Resettlement in a nearby village, and subsequent acts of food sharing across the familial network might be prime examples of the potential cultural responses to resource scarcity that Wayne Suttles (Reference Suttles, Lee and DeVore1968:60) had in mind: “I have dwelt at length on the existence of times of scarcity on the Northwest coast precisely because recognizing their existence may be essential to our understanding of the complexities of Northwest coast culture.”

Sockeye at Sea and the Homing Migration

The adults of all five salmon species must find their way home to the Fraser River through the straits that connect the Pacific Ocean to the Salish Sea and through the channel mazes in the archipelagos at the landward ends of both straits (Figure 1). Homing fish initially swim at depth in the broader sections of the straits, but the schools rise closer to the surface (Quinn et al. Reference Quinn, Terhart and Groot1989) and closer to shore as they approach the archipelagos, using the strong tidal flows in the narrow channels to facilitate their return.

At the beginning of the twentieth century, it was believed that adult Fraser River sockeye migrated almost exclusively through the southern strait. Babcock (Reference Babcock1903:5) was particularly adamant: “All of the sockeye which enter the Fraser come in from the sea through Juan de Fuca Strait. There seems to be no evidence that any portion of the sockeye run comes through the north through Johnstone Strait.” In the following decades it became clear, however, that a minority (5%–20%) migrated through Queen Charlotte-Johnstone Strait (Gilhousen Reference Gilhousen1960). That corridor therefore became known as the “northern diversion” route. Since the 1970s about 50% of the returning sockeye, on average, have used this northern corridor (Hamilton Reference Hamilton1985), and 96% used it in 2014. That tendency has also been mirrored in pink runs; most Fraser pink now return via the northern route. Given that chum, chinook, and coho spawners also predominantly use the northern route, decades-long switches in migration routes by sockeye and pink in the past undoubtedly had drastic consequences for Indigenous communities along the southern corridor.

The Southern Migration Corridor

All the Pacific salmon species continue to feed until they reach freshwater; the energy they expend on their upriver journeys, however, is not replaced, and those reaching the more remote headwaters of the Fraser River watershed have lost more than half their nutritive value (Gilhousen Reference Gilhousen1980). Salmon harvested along their maritime migration routes, although far less accessible than those migrating upriver (Kew Reference Kew and Hayden1992), are thus far more nutritious.

Sockeye and pink salmon return to the Fraser River in mid- to late summer, after the peak flows of the spring freshet. Verhoeven and Davidoff (Reference Verhoeven and Davidoff1962) analyzed the catch distribution of sockeye recovered from those tagged annually at Sooke from 1938 to 1948. After congregating on the Salmon Banks area off the southwest coast of San Juan Island, most of the fish headed home through Rosario Strait, rather than Haro Strait (Figure 2). The fish that did migrate through Haro Strait appeared to hug the eastern shore and then rejoined the Rosario schools off Point Roberts. Tag recoveries on the eastern shore of Vancouver Island and in the Canadian Gulf Islands accounted for <0.5% of the total. That was also the conclusion reached by O’Malley and Rich (Reference O’Malley and Rich1920), who noted that the number of homing sockeye in the channels between Haro and Rosario Straits was “insignificant”.

Figure 2.

Flood tidal velocities (open arrows show current directions) through the Strait of Juan de Fuca and southern Strait of Georgia (redrawn from Crean’s fine-grained tidal model; Thomson Reference Thomson1981) and percentage recoveries of sockeye salmon en route to the Fraser River tagged at Sooke [So] from 1938 to 1948 (Verhoeven and Davidoff Reference Verhoeven and Davidoff1962), excluding tags recovered in the open waters of Georgia Strait (6.6% of the total), the Fraser River fishery (47.5%), and from carcasses on the spawning grounds (5.1%). The broad arrows denote the salmon migration “mainstream” through Rosario Strait. (Color online)

If this was a long-standing pattern, then the sea fishery in antiquity should have comprised those locations where salmon migrating through the Strait of Juan de Fuca were funnelled along a “mainstream” by topography and bathymetry—from the Salmon Banks off Lopez Island and the southwest coast of San Juan Island, through Rosario Strait, along the eastern shore of the Strait of Georgia, and then past Point Roberts to the mouths of the Fraser River. Long-term variations in salmon abundance should therefore be most evident in the occupational history of sites that lie abreast of that route. The archaeological record from sites in Haro Strait may be concordant with those on Rosario Strait but more muted. As for the salmon heading to natal streams elsewhere in the Salish Sea, they were undoubtedly locally important as food sources, but their numbers were relatively small, and they may be regarded as satellites to the Fraser River runs. We expect that these areas were particularly important in the periods of declining salmon runs.

Salmon in the Archaeological Record

Spatial Variability in the Salmon Harvest

The diet of coastal Salishan peoples in the past can be reconstructed from the faunal and floral remains in the shell middens on the shores of the Salish Sea. More than a million fish bones have been sampled from about 100 archaeological sites in the central reaches of the Salish Sea. Fish bones commonly account for at least half and occasionally more than 90% of the vertebrate fauna. About 40% of the fish bones (around 610,000) have been identified to family, genus, or species, of which an estimated 235,000 (38%) were identified as “salmon” (Supplementary Material 3). Figure 3 shows the spatial distribution of the sampled sites, the total NISP (number of identified specimens), and the salmon percentage. Note that some of the sites are undated, and the age of others is only weakly constrained, so many of the faunal samples likely pre- or postdate our 2400–1200 cal BP time frame.

Figure 3.

Number of fish bones sampled and percentage frequency of salmon at archaeological sites in the central Salish Sea and on the lower reaches of the Fraser River (limited to samples of more than 30 bones). (Color online)

The salmon bones in the assemblages are predominantly vertebrae, which are quite distinctive and therefore are readily identified. Even fragments of salmon vertebrae are easily identifiable, which likely inflates the relative abundance of salmon in the faunal assemblages. The vertebrae of each salmon species are morphometrically very similar, and although identification of the individual species is consequently almost impossible without resorting to DNA or ZooMS testing (e.g., Richter et al. Reference Richter, Wilson, Jones, Buckley, Van Doorn and Collins2011; Yang et al. Reference Yang, Cannon and Saunders2004), two morphometric subgroups of species can be differentiated, at least in modern stocks. The vertebrae of adult chinook, chum, and coho salmon in Washington State hatcheries exceed (on average) 6.5 mm in height and length (Huber et al. Reference Huber, Jorgensen, Butler, Baker and Stevens2011:Table 2); those of sockeye and pink are significantly smaller (mean height and length <5 mm). The latter are therefore not likely to be retained by a ¼-inch (6.35 mm) screen. The relative abundance of salmon in the midden assemblage may still be overestimated, however, because the bones of small fishes such as herring, anchovy, eulachon, and sandlance will also be lost.

Even though a bone count may inflate the relative abundance of salmon remains, it may still devalue the contribution of salmon to the local diet. This is most likely to be true of counts from the middens of winter villages, because the salmon that were consumed there were commonly filleted. According to Duff (Reference Duff1952), the general practice at the fishing camp was to remove the head of the salmon, separate the backbone from the flesh while leaving it attached only at the tail, and then hang it on a drying rack or, later in the season, in a smokehouse. The detached bones would be thrown back into the river or the sea. If the village was near the fishing camp, then salmon heads and backbones might be saved to be subsequently boiled or roasted (Barnett Reference Barnett1955), which might account for the observation that salmon vertebrae tend to be far more common in house-floor deposits than in neighboring village middens (e.g., Walker Reference Walker2003).

Salmon remains are not uniformly distributed in the middens around the Salish Sea (Figure 3). They commonly account for at least 80% of the fish bones at the mouths of rivers hosting spawning populations and more than 90% of the NISP total at some upriver sites (e.g., Derian Reference Derian2022; Langemann Reference Langemann1987). They are far less abundant in relative terms (because of the far greater diversity of marine than riverine fish species) along their oceanic migration route, but their relative abundance parallels the tag recovery pattern shown in Figure 2.

They are a major and, in some cases, the dominant component of the fish fauna at archaeological sites on the north shore of the Strait of Juan de Fuca, the “outer” shores of the San Juan Islands (but often completely absent at sites in the maze of channels between islands), and along the eastern shore of the Strait of Georgia. In contrast, they generally comprise less than 10% of the fish fauna at archaeological sites on the eastern shore of Vancouver Island (except at the mouths of the larger rivers) and in the Gulf Islands. The only significant exception to this overall pattern is a site in the Gulf Islands that is located at the western end of a narrow pass through the archipelago.

Temporal Volatility in Salmon Consumption

As we noted earlier, many of the sites along the salmon migration mainstream remain undated, and the utility of other faunal catalogs is often limited, either by a sparsity of radiocarbon dates or by a preponderance of dates on marine shell. In some cases, the faunal lists are from horizons that pre- or postdate our temporal limits or are based solely on ¼-inch screen samples. In other cases, the associated radiocarbon dates are either of questionable validity, were considered unreliable by the original investigators, or were derived from different excavation units than those that yielded the faunal remains. In one case the midden ridge appears to have been subject to severe slumping, and the faunal sequences are consequently a mélange (Supplementary Material 2).

In only three of the candidate sites do we consider the faunal analyses both fine-grained enough and well-dated enough to allow reconstruction of the temporal variability in salmon consumption (Figure 4). However, these reconstructions should be considered as approximations not only because of the inherent uncertainty of the radiocarbon dates but also because some segments of the age–depth relationships were of necessity adduced from dates derived from neighboring excavation units. These three cases, presented sequentially along the southern migration route, are Čḯx^wicən (45CA-523), Watmough Bay (45SJ-280), and Parry Lagoon (DgRv-6).

Figure 4.

Century-scale variation in fish (including salmon) NISP values during the Marpole phase at Čḯx^wicən (45CA-523), Watmough Bay (45SJ-280), and Parry Lagoon (DgRv-6). See Supplementary Materials 5–7 for data. (Color online)

Čḯx^wicən (45CA-523)

Čḯx^wicən, one of the home villages of the S’Klallam, lies at the western end of Port Angeles harbor in the lee of Ediz Hook on the south shore of the Strait of Juan de Fuca (Figure 1). The original excavators grouped the cultural deposits into seven “chronozones” that varied in length from 150 to 450 years (Larson Reference Larson2006). The fish faunal assemblages, grouped by chronozone, are described by Butler et alia (Reference Butler, Campbell, Bovy and Etnier2019). Herring (Clupea pallasii) remains comprise about half the total assemblage; salmon bones account for only 4%.

We reevaluated the published faunal counts by calculating bone depositional rates (i.e., NISP per liter, for 100-year intervals) in the midden strata that were directly dated. Our analytical procedures and results are summarized in Supplementary Material 5.

The low accumulation rate of fish bone (mean NISP <4 bones per liter) in the oldest midden strata suggests that Čḯx^wicən functioned primarily as a shellfish-gathering camp before about 1700 cal BP (Figure 4). The rate of deposition increased to more than six bones per liter from 1600 cal BP to 1500 cal BP and declined to fewer than two bones per liter in the following century. Salmon bones echo this overall pattern; they comprised about 13%–18% of NISP in the period from 1700 cal BP to 1500 cal BP.

The peak depositional phase for salmon residues at Čḯx^wicən is therefore essentially coincident with the inferred upsurge in fishing activity on the Thompson River starting at about 1700 cal BP. Given, however, that almost all (94%) of the salmon bones are vertebrae, and 87% of the intact vertebrae were recovered from the ¼-inch screens, we conclude that the fish were caught and processed elsewhere, and the targeted runs were predominantly local chinook, chum, or coho stocks—not the migrant sockeye and pink salmon in the Strait of Juan de Fuca. The most likely salmon source is the mouth of the Elwha River, which lies less than 10 km to the west of the village. That river had the largest salmon runs on the south shore of the Strait of Juan de Fuca before construction of dams in 1913, and chinook, chum, and coho were the most numerous of the salmon species.

Watmough Bay (45SJ-280)

The Watmough Bay site is located on a peninsula on southern Lopez Island, about 1 km from known salmon reef-netting locations on the opposite side of the peninsula (Figure 1; Suttles Reference Suttles1974). The excavation history of the site and the faunal contents of cultural deposits at one end of the site close to the modern shoreline are described by Bovy and Kopperl (Reference Bovy and Kopperl2007). Salmon bones account for 78% of the identifiable fish remains in one of their excavation units, at densities ranging from fewer than one to more than 36 bones per liter of midden material. Some of the bones were cranial elements and were burned, which led Kopperl to conclude that the fish were caught nearby and processed at this location.

Century-scale variations in fish-bone accumulation rates from 2400 to 1200 cal BP (Figure 4) are inferred from an age–depth model based on 11 radiocarbon samples on wood fragments from the excavated units, plus 3 ages on charcoal from a neighboring excavation (Supplementary Material 6). The modest rate of accumulation from 2200 to 1600 cal BP suggests that only a few families fished here at that time (Supplementary Material 5). The following two centuries appear to have been a period of much more intense fishing activity (for both salmon and nearshore fish). A steep decline in the fish-bone accumulation rate after 1400 cal BP is followed by a modest upsurge shortly after 1300 cal BP.

Parry Lagoon

Parry Lagoon (DgRv-6) is located at Dionisio Point on the northern tip of Galiano Island (Figure 1), adjacent to DgRv-3 (formally known as the Dionisio Point site). We reconstruct the temporal variation in the fish fauna at Parry Lagoon from data presented by Hopt (Reference Hopt2015), based on two charcoal samples supported by two ages on marine shell from the central part of the midden (Units 14–16). The shell ages are corrected for the local component of the marine radiocarbon reservoir effect (∆R = 190 ± 40 years) from dates on paired charcoal and shell samples from the midden-stained sands at the base of the cultural deposits (Supplementary Material 7).

Herring bones account for 73% of the 11,091 fish bones in the faunal samples, an indicator that this was a spring fishery. Salmon bones account for only 8% of the total. In excavation units 14–16 their frequency decreases from about 15% in the basal levels (about 1600 cal BP) to about 2% in the upper levels (around 1400 cal BP; Figure 4C).

Ewonus et alia (Reference Ewonus, Cannon and Yang2011) and Grier et alia (Reference Grier, Flanigan, Winters, Jordan, Lukowski and Kemp2013) sampled the DNA in salmon vertebrae from DgRv-6 and an adjacent site (DgRv-3); they identified 35 specimens as sockeye, 25 as chum, 15 as pink, three as coho, and one as chinook. These fish may have been obtained locally, but given the proximity of Galiano Island to the mouth of the Fraser River, it is likely that people fished there from mid-summer (for sockeye and pink) to late fall (for chum). If the catch was processed on the Fraser riverbanks and only the fillets were transported back to the island, the midden contents (as at Čḯx^wicən) are unlikely to reflect the importance of salmon in the islanders’ winter diet.

Demographic Responses to Long-Term Changes in Salmon Productivity

Given that so few of the analyses of faunal assemblages in the middens surrounding the Salish Sea are matched by chronological assays of equivalent resolution, are there other ways of assessing the potential variability of salmon catches? We propose that the “dates as data” approach (Rick Reference Rick1987) is such a metric and that by comparing activity at sites on the salmon migration mainstream with those in satellite areas we may be able to delimit periods when salmon were bountiful from those when they were relatively scarce.

Several investigators have previously used this approach to reconstruct the occupational histories of regions, villages, or households on the margins of the Salish Sea (e.g., Morin et al. Reference Morin, Lepofsky, Ritchie, Porčić and Edinborough2018; Ritchie et al. Reference Ritchie, Lepofsky, Formosa, Porčić and Edinborough2016; Rorabaugh Reference Rorabaugh2023). The underlying assumption in all these cases is that the radiocarbon-date catalogs are random samples of the age profiles of the local cultural deposits; the temporal density of radiocarbon dates is therefore a population proxy or, if the dates are from disparate environmental contexts, a gauge of the relative intensity of subsistence activities over time.

Our analysis of the occupational history of the coastal sites is based solely on dates on charcoal in the interval between 2500 to 1100 ¹⁴C BP (Supplementary Material 8A), with analytical uncertainties ≤150 years. This expanded time frame minimizes edge effects in our analyses. We eliminated the following dates: those on charcoal from the Gakushuin laboratory because of doubts about their validity (Maschner Reference Maschner2004); those on forest detritus because they may mark episodes of site abandonment, rather than site activity, and those on marine shell given the uncertain magnitude of the local marine reservoir effect (Hutchinson Reference Hutchinson2020). We also omitted dates on wooden fish traps because they are likely to be derived from brief episodes of construction (see the later discussion), not the persistent occupation of nearby settlements.

We divided the resultant regional catalog into six subareas based on the location of sites in relation to the southern migration route (Figure 5). The Juan de Fuca subarea encompasses the northern and southern shores of that body of water as far east as Haro Strait and the entrance to Puget Sound. Rosario Strait lies on the mainstream salmon migration route: it extends from the Salmon Banks off the south coast of San Juan and Lopez Island north to the Lummi Peninsula. Haro Strait, which encompasses the southern Gulf Islands, lies abreast of the secondary salmon migration route. Three subareas—Bays, Burrard Inlet, and Eastern Vancouver Island—are peripheral to the migration route, but people from all three areas potentially had access to Fraser River salmon, either at Point Roberts (Figure 1) or at the mouth of the river. The Bays subarea incorporates sites on three semi-enclosed bodies of water: Boundary Bay, Semiahmoo Bay, and Drayton Harbor, plus the intervening shorelines. Burrard Inlet includes sites on the shores of the fjord that lies immediately north of the Fraser River delta. Eastern Vancouver Island incorporates the shoreline of that island, plus some of the northern Gulf Islands.

Figure 5.

SPDs (100-yr bins, normalized to unity) of radiocarbon ages from cultural deposits on the banks of the Thompson River, the plateau of the Thompson River watershed, the six subareas of the Salish Sea (see map inset for the number of sites and dates in each subarea), and the Sequim hunting camp (Sq on map). (Color online)

In several subareas the radiocarbon catalog is dominated by assays from a single site. In the Haro Strait area, for example, 60% of the 48 ages are derived from English Camp (45SJ-024) on San Juan Island, and 40% of the 38 ages in the Juan de Fuca catalog are from Čḯx^wicən. To provide a more balanced representation of sites in each area, we limited the number of dates to a maximum of eight from each site (with Parry Lagoon and the adjacent Dionisio Point site treated as a single site). The eight dates with the highest precision—that is, the lowest “lab error”—were retained in each case (Supplementary Material 8B).

The residual catalog of dates clearly represents a tiny fraction of all the datable remains in the middens in each area. The resultant summed probability distributions (SPDs) of radiocarbon ages in each catalog should therefore be considered as first approximations, not definitive representations of local temporal patterns. The broad (100-year interval) binning procedure that we use generalizes the SPD patterns in acknowledgment of this fact.

We divided the resultant SPDs into three groups. Three coastal areas display patterns of activity that are broadly congruent with the SPD from the Thompson riverbank sites: the “river” pattern (Figure 5). This group displays moderate to high levels of activity in the interval from 2400 to 2000 cal BP, lower levels of activity in the interval from 2000 to 1600 cal BP, a rapid increase in the following two centuries, and, with the exception of Haro Strait, a dramatic decline after 1400 cal BP. The second group displays maximal activity in the 2000–1600 cal BP interval and minima at about 2200 cal BP and 1400 cal BP, which is broadly congruent with the SPD from the uplands in the Thompson watershed: the “plateau” pattern (Figure 5).

The closest parallels to the “river” pattern are the SPDs from Rosario Strait and Eastern Vancouver Island. The most intensively dated sites in the Rosario Strait—Watmough Bay and Cattle Point—are adjacent to the Salmon Banks, the nexus of the sockeye migration route. Activity at these two sites is essentially limited to two periods—2400–2000 cal BP and 1550– 1350 cal BP—which coincide with the episodes of maximal activity on the Thompson riverbanks. The discrepancies between the Rosario Strait and Thompson SPDs may be attributable to two causes. Lower activity in the 2400–2200 cal BP interval in the Thompson watershed may represent a local exodus in the aftermath of a volcanic eruption in the Coast Mountains at or about 2360 cal BP (Clague et al. Reference Clague, Evans, Rampton and Woodsworth1995). Tephra deposits from the Mt. Meager eruption extend northeastward across the Thompson River watershed and into Alberta (Leonard Reference Leonard1995). The higher value in the 2000–1600 cal BP interval in the Thompson riverine SPD likely occurred because local radiocarbon ages are derived from a broader array of archaeological contexts than those represented by the summer fishing camps in Rosario Strait.

The SPD from Eastern Vancouver Island also matches the Thompson riverine pattern, at least in its later phases. Given that this area does not overlap the sockeye migration routes and that salmon bones are generally uncommon in local archaeofaunas, this finding came as a surprise. The ethnographic record, however, indicates that the two largest groups of Vancouver Island Halkomelem (the Nanaimo and the Cowichan) encamped on the banks of the lower Fraser River each summer and fished for sockeye (Duff Reference Duff1952). They almost certainly discarded the salmon backbones on the riverbank, ferrying only the edible parts back across the Strait of Georgia to their winter villages. The Haro Strait SPD is also a relatively close match to the initial phases of the river pattern, but the SPD does not decline after 1400 cal BP, unlike the other members of this group.

The Juan de Fuca SPD also deviates from all the above in more recent centuries. The SPD indicates moderate levels of activity in the area in the 2400–2000 cal BP interval, a nadir at around 1800 cal BP, with a peak of activity (primarily at sites on the northern shore of the strait) at about 1600 cal BP, followed by a lull, and a second peak (primarily from the Čḯx^wicən radiocarbon catalog) at around 1300 cal BP (Figure 5).

The Bays area—encompassing Semiahmoo territory and the northern parts of Lummi territory—displays an SPD pattern that closely resembles that from the plateau sites in the Thompson River watershed, with an apex at around 1900 cal BP, and minima at about 2200 and 1500–1400 cal BP (Figure 5). That activity pattern, with a somewhat later peak, is also displayed by sites in Burrard Inlet and characterizes the occupational history of the Sequim Bypass site, the only well-dated hunting camp in the Salish Sea hinterland (Morgan Reference Morgan1999). The camp is located about 10 km inland from the Strait of Juan de Fuca in S’Klallam territory.

These plateau-pattern SPDs are the antitheses of those from Rosario Strait, Haro Strait, and Eastern Vancouver Island: we ascribe this bipolarity to century-scale fluctuations in the size of the salmon runs on the Fraser and other local rivers. Activity in the river areas tracks the salmon runs; if those runs consistently fall below a subsistence threshold, then people move to areas where other resources are available.

Discussion/Conclusions

The extant zooarchaeological data offer only modest support for our contention that the variations in salmon harvests along the coastal mainstream migration route are concordant with those in upstream areas of the Fraser River between 2400 and1200 cal BP. The primary limitation in testing that hypothesis is that even though many of the published archaeofaunal records from sites on the shores of the Salish Sea are exhaustive, only three sites are reliably dated with faunal counts derived from fine-mesh screening of deposits. Only one of these (Watmough Bay) displays a marked upsurge in salmon utilization centered on 1500 cal BP. The concurrent increases in salmon bone abundance at Čḯx^wicən and Parry Lagoon are much more modest, likely because the salmon were butchered elsewhere.

The high degree of concordance between the Rosario Strait and Thompson River SPDs and Eastern Vancouver Island and the Thompson River after 1600 cal BP implies that the demography of those coastal areas reflects the relative success of the salmon fishery. If that inference is correct, then the salmon runs on the Fraser River started to recover from a multi-century low at about 1700 cal BP, increased dramatically about 1600 cal BP, and collapsed shortly after 1400 cal BP.

The success of the Indigenous fishermen in Rosario Strait and from Eastern Vancouver Island at those times reflects their several advantages over their counterparts elsewhere. Fishing camps in Rosario Strait border the salmon migration mainstream and the shallow waters adjacent to the entrance of the strait and the promontories in the narrow strait that trigger eddies in the tidal current that bring the fish closer to shore. Indigenous fishermen from Eastern Vancouver Island took advantage of the fact that the mouth of the Fraser River was the first chokepoint on the final, riverine stage of the migration. In contrast, the schools of migrating sockeye and pink salmon were likely to be too deep (in the Strait of Juan de Fuca) or too dispersed (in both the Strait of Juan de Fuca and the Strait of Georgia) to have been readily accessible to Indigenous fishermen.

The concordance between the Thompson riverine pattern and that from Rosario Strait suggests that switches in sockeye and pink salmon migration paths to the “northern diversion” route, if they occurred, must have been relatively modest or brief. Analyses of occupational histories and fish faunas at coastal archaeological sites in the archipelagos along the northern migration corridor are clearly needed to test this assertion.

If activity levels at the villages in the areas that display the river pattern echo salmon abundance on the Fraser River, then the areas that display the plateau pattern appear to have functioned as havens during periods of reduced salmon runs and depressed marine productivity. This is most evident in the contrasting patterns between about 1900 and 1700 cal BP, when activity at the sites along the mainstream declines and activity at sites in the satellite areas intensifies.

Century-scale changes in the availability of marine resources were likely to affect a broader region than the Salish Sea, particularly in those areas where salmon were the most important source of protein (Coupland et al. Reference Coupland, Stewart and Patton2010): this potentially led to differential migration of affected populations and the abandonment of some settlements. People might move, as appears to have happened in the Salish Sea, to places offering a broader and less volatile resource base. Alternatively, they might adapt by modifying their seasonal round, expending more effort on hunting or harvesting roots and other plant foods. The adaptive response to salmon scarcity in the Thompson River watershed appears to have been a greater focus on root harvesting and deer hunting on the upper terraces and hillsides in late summer and fall; the predominant response in the central Salish Sea, in contrast, appears to have been more permanent changes in residence.

Shifts in population were likely predominantly irruptive; that is, people moved to live with nearby kin. They may also have been eruptive, involving incursions into the territories of neighboring groups. Those incursions may not always have been welcomed, however; fortified defensive sites in the central Salish Sea were first built in the century following 1350 cal BP, coincident with the downturn in marine productivity (Angelbeck Reference Angelbeck2009:Table 4, excluding the single Gakushuin date). A further coincidence is that radiocarbon ages on wooden stake complexes in intertidal areas in Burrard Inlet and at several locations on the east coast of Vancouver Island indicate that fish traps were initially constructed at or about 1350 cal BP, suggesting that Coast Salish households responded to the salmon shortfall (and, perhaps, an influx of kin) by attempting to intensify their marine harvests.

The first studies of the material remains in the shell middens on the margins of the Salish Sea prompted local archaeologists to define (and reify) cultural-historical periods based on changing tool kits. The variability of local activity patterns linked to subsistence that we describe lends support to critiques of this typological approach to Coastal Salishan history by archaeologists who have stressed the geographic variability of those tool kits (e.g., Clark Reference Clark2010) and the dynamic nature of trade networks and settlement patterns (e.g., Morin Reference Morin2016; Ritchie et al. Reference Ritchie, Lepofsky, Formosa, Porčić and Edinborough2016).

Early authors also posited large-scale (e.g., Borden Reference Borden1954) or, in later decades, medium-scale (e.g., Burley and Beattie Reference Burley, Beattie, Auger, Glass, MacEachern and McCartney1987) migrations of people to account for the cultural transitions between cultural-historical periods. Those notions have largely been replaced by an evolutionary perspective that emphasizes continuity of occupation and the sustainable management and use of marine and intertidal resources (e.g., Campbell and Butler Reference Campbell and Butler2010). This longue durée view (Ames Reference Ames1991) stresses millennial-scale stability but veils the potential social impacts of courte durée fluctuations of the marine resource base. Both perspectives are incorporated in Central Coast Salish oral histories, which stress continuity of occupation and describe famines that resulted in loss of life and residential movements.

The demographic effects of famine—delayed female puberty, deferred marriage, amenorrhea, lowered birth rates, and increased infant mortality—have been well documented by social historians (e.g., Le Roy Ladurie Reference Le Roy Ladurie1979). All those effects may have also been evident in fisher-hunter-gatherer societies on the west coast of North America in antiquity; subsistence crises are known to have triggered wholesale changes in the diet, economy, and trade networks of some groups (e.g., Arnold and Martin Reference Arnold and Martin2014). But in Coastal Salishan society the response to extreme changes in resource availability may have been more straightforward: people moved to neighboring areas with a less volatile resource base during the crisis periods. It is therefore tempting to consider the possibility that the residential flexibility and ceremonial feasting that were characteristic of Coastal Salishan society were outgrowths of and adaptive responses to the variability of the salmon runs.

Finally, we note that although it may be a natural tendency for archaeologists to categorize and label the “ages” of human history, simply attaching these labels to archaeological site components impedes, rather than enriches, our understanding. The spatial and temporal variability that we have documented in this study is evidence of that: the 1,200-year spans encompassed by the Marpole and Plateau labels should no longer be viewed as essentially uniform periods of Salishan history. Our dichotomous river and plateau patterns represent broad-scale demographic changes that are unlikely to be revealed by analyses of artifacts in site-specific investigations—yet, like these earlier studies that label periods of human history based on artifact typologies, we are also guilty of compartmentalization and generalization. We represent intra- and interregional population dynamics by century-scale trends, but we should not lose sight of the fact that they are the composite product of the decisions, often prompted by scarcity, that generations of people made daily.