Journal of Archaeological Science (2000) 27, 725–737doi:10.1006/jasc.1999.0498, available online at http://www.idealibrary.com on

Assessing Variability in Northwest Coast Salmon and HerringFisheries: Bucket-Auger Sampling of Shell Midden Sites on theCentral Coast of British Columbia

Aubrey Cannon

Department of Anthropology, McMaster University, Hamilton, Ontario, Canada L8S 4L9

(Received 28 May 1999, revised manuscript accepted 8 September 1999)

The results of bucket-auger sampling at shell midden sites show that this is an accurate and efficient method forassessing the focus and intensity of local fisheries. The results confirm long-term trends in salmon production identifiedpreviously on the basis of excavations at the Namu site, and show the extent of variability in the salmon and herringfisheries at other sites in the region. Comparison of the relative density of fish bone in auger samples serves as a usefulmethod for distinguishing between long-term habitation and specific-purpose campsites, and provides a basis forassessing the extent and implications of temporal and inter-village variability in salmon productivity.

� 2000 Academic Press

Keywords: SHELL MIDDENS, BUCKET-AUGER SAMPLING, SALMON, HERRING, NORTHWESTCOAST.

7250305–4403/00/080725+13 $35.00/0

Figure 1. Location of Namu on the central coast of BritishColumbia.

Introduction

C oring and augering are increasingly beingused as fast and efficient methods of obtainingdetailed information about archaeological site

structure and content (Stein, 1986; Whittaker & Stein,1992; Hoffman, 1993; Canti & Meddens, 1998). How-ever beyond their value as methods for preliminary siteinvestigation, core and auger samples can also be aneffective basis for problem-oriented regional researchprograms. The capacity of these methods is clearlyillustrated by the results of recent investigation into thehistory of settlement and subsistence in the vicinity ofthe Namu site, which is located in the traditionalterritory of the Heiltsuk Nation on the central coast ofBritish Columbia (Figure 1). The main purpose inobtaining auger samples in this case was to measure thedensity of fish remains in site matrices as a way toassess regional and temporal variability in the intensityof salmon fishing. The results of preliminary testingconducted at Namu in 1994 show that auger samplesare a valid basis for monitoring variability in salmonand herring fisheries, and application of this method toa broader range of sites in 1996 and 1997 is providingan initial indication of regional as well as temporalvariability in these fisheries.

The ultimate goal of this research is to integrate thewell-documented 10,000-year history of the Namu site(Carlson, 1979, 1996; Cannon, 1991) into a broaderregional context. Up until now, a key impediment todeveloping this type of regional context has been thesmall number of excavated sites in the area, and the

inherent difficulties of mounting regional research pro-grams involving excavations in remote, often-deep, andstratigraphically-complex shell midden sites. This is aproblem common to all regions of the Northwest Coast

� 2000 Academic Press

726 A. Cannon

(Lyman, 1991; McCartney, 1992; Ames, 1998). Despitea 60-year history of archaeological investigations in theapproximately 10,000 km2 area of Heiltsuk territorialclaims, as of 1995 only 16 of 334 recorded shellmiddens in the region had ever been the subject of evenlimited test excavation (Millennia Research Ltd, 1996),and little information concerning fish remains is avail-able from most of these few sites. A program of sitetesting based on limited coring and augering thereforeseemed the best available means of putting what wasknown about the history of the Namu subsistenceeconomy into a broader regional context. The coreswere an effective means of monitoring stratigraphicvariation in midden deposits and of obtaining uncon-taminated samples of material for radiocarbon dating(Cannon, 2000a). The bucket-auger samples providedlarger samples of matrix that could be examined toassess their shell and vertebrate fauna content. Thiscombination was a practical alternative to what wouldotherwise have been a prohibitively expensive programbased on traditional methods of test excavation.

Table 1. Frequency and proportion of salmon remains among identifiedfish from the 1977–1978 Namu excavations

PeriodDate range

(14C years ) Frequency % of fish

6 1880–980 380 675 3500–2185 9509 854 4390–3825 58,940 973 4775–4540 17,272 972 6060–5170 5720 891 9720–6550 (no faunal remains preserved)

The Namu FaunaNamu is well known as the oldest continuously-occupied site on the coast of British Columbia. Theoldest deposits date to 9720�140 (radiocarbonyears) (Carlson, 1979, 1991, 1996), and a well-preserved record of vertebrate faunal remains docu-ments the history of the local subsistence economyfrom c. 6000 up until the European-contact era(Cannon, 1991). Excavations conducted by theUniversity of Colorado in 1968–1970 (Hester &Nelson, 1978) yielded a well-documented record of themammal and shellfish remains from the site, but noinformation concerning the fish remains. Analysis ofall vertebrate fauna recovered from excavations con-ducted by Roy L. Carlson of Simon Fraser Universityin 1977–1978 yielded further information on themammalian fauna, but also provided the first detailedevidence of the focus and intensity of the local fishery(Cannon, 1991).

Despite some difficulties in the interpretationresulting from variation in recovery methods, whichincluded the use of 1/4� and 1/8� screens and dry- aswell as water-screening, the fish fauna from Namushowed a clear and dominant focus on salmon amongthe larger fish. Examination of limited matrix samplesalso showed an abundance of herring and greenlingbones, which were not recovered in representativequantities in the field-screened deposits (Fawcett,1991). To compensate for some of the variablerecovery problems and to obtain a sense of temporaltrends in faunal-class abundance, raw counts ofidentified specimens were standardised as proportionsof all fish, bird, or mammal specimens recovered fromfauna-bearing deposits representing each of the fivelatest periods in the site occupation. These temporal

components were defined on the basis of stratigraphy,fauna and artefact content, and radiocarbon dating(Carlson, 1991). Since the remains of most fish, birds,or mammals from a given period were subject toroughly similar recovery biases, the period propor-tions are reasonable indicators of relative abundance.With the exception of small fish, such as herring andgreenling, these period proportions could be comparedto show trends in relative abundance over time. Theresults showed that salmon were clearly dominantamong the recovered fish remains, making up to 97%of all fish from some periods (Table 1). The data alsoindicated clear trends in the intensity of the salmonfishery. The proportional representation of salmonshows a modest increase to reach a peak in the periodsbetween 4775 and 3825 , followed by a declinebeginning in the period after 3500 . This declinefinally becomes quite dramatic by the last period of thesite’s occupation.

Taken at face value, the Namu faunal data arerelatively unambiguous, but a number of factors otherthan variability in the production and consumption ofsalmon are capable of producing similar patterns. Theincrease in salmon that coincides with an initialincrease in shell deposition, for example, could be dueto enhanced preservation of salmon remains. Salmonbones are semi-cartilaginous and are particularlyvulnerable to chemical and mechanical breakdown(Butler & Chatters, 1994). The apparent increase insalmon could be due to the acid-neutralising effects ofincreased shell deposition as much as to any indepen-dent increase in the consumption of salmon. The earlyvertebrate fauna appear to show greater use ofmarginal resources such as dogfish, which decline inimportance with the later increase in salmon (Cannon,2000b), but these data alone are insufficient to dispelthe ambiguity introduced by potentially variablepreservation factors. Even the apparent increase inshell deposition could be attributable to differentialpreservation (Stein, 1992), though preserved lenses ofshell in pre-5000 deposits and nitrogen isotopetrends consistent with increasing shellfish consumption(Cannon et al., 1999) support the interpretation ofincreasing shell deposition over time.

The inference of a later decline in salmon is based ona reduction in its proportion of the recovered fish

Variability in Northwest Coast Salmon and Herring Fisheries 727

remains. Since proportions are closed arrays (Grayson,1984: 19), however, this trend could equally representan increase in the abundance of other fish species,possibly resulting from diversification of fishingstrategies, rather than any decrease in salmon. Impres-sions of a later increase in diversity among therecovered but unreported fish remains from theUniversity of Colorado excavations, and observationsof a coincident increase in bone artefact diversity,thought to represent an elaboration of fishing gear, hadbeen the basis for inferring a late period peak in thelocal fishing economy (Hester, 1978: 102; Luebbers,1978: 62).

The reliability of the salmon trends evident from the1977–1978 excavations could also be questioned on thegrounds that these excavations represent only a limitedvolume of the overall site. Based on extensive analysisof fauna recovery from shell middens on the Oregoncoast, Lyman (1991: 313) recommended an excavationvolume of 100 m3 per 1000 years of site occupation asthe minimum needed for reliable interpretation ofvariability in the fauna and artefact content of middendeposits. Sampling may not affect the recovery ofcommonly-occurring fish remains as much as it affectsthe recovery of other classes of fauna, but the potentialfor sampling effects cannot be discounted. The totalvolume of the 1977–1978 excavations, on which morethan 5000 years of faunal trends are based, is onlyjust over 100 m3. Although this represents intensivesite investigation by most standards, and exceedsthe volume of excavation undertaken at any othersite in the region, it is only 20% of the minimumrecommended by Lyman.

The Namu faunal trends could therefore be subjectto very different interpretations depending uponwhether the observed patterns are assumed to be dueprimarily to the original patterns of consumption anddeposition or to the effects of subsequent preservation,sampling, or quantification. The potential for variedinterpretations, despite the large number of faunalremains and the apparent robustness of the trends theyindicate, has been the impetus for undertaking effortsto find independent lines of evidence to support theinterpretation that the Namu salmon fishery actuallydid experience a pattern of modest growth from itsearliest beginnings, to reach a period of peak pro-duction, followed by a subsequent decline. Theseefforts have now yielded a wide variety of supportingevidence.

Periods before and especially after the peak in pro-duction, for example, show increased use of moremarginal fish resources such as dogfish and ratfish, anddeposits dating after 3500 also show increasedabundance of deer and shellfish remains (Cannon,1995, 2000b). All of these resources can reasonably beinterpreted as evidence of efforts to compensate for alower level of salmon production. A study of deerphalanges from Namu also indicated a significantincrease after 3500 in bones exhibiting medial–

lateral fractures (Zita, 1997), which is a pattern ident-ified by Binford (1978: 148) as characteristic of humanefforts to break open phalanges to obtain marrow, incontrast to the longitudinal fractures that are moretypical of bones broken by dogs (Binford, 1981: 51).The implication is that people were increasingly forcedto undertake extraordinary efforts to obtain dietary fatduring periods when the salmon fishery was in overalldecline.

A recent isotopic study of dog bones (Cannon et al.,1999) provides still further evidence in support of theapparent trends in the salmon fishery. These resultsshowed high levels of marine protein consumption inall periods, but much lower variability in �13C levelsduring the period of peak salmon production andmuch higher variability during the period of decline.The implication is that dogs, and presumably humans,were more likely to rely periodically on terrestrialsources of protein (primarily deer) after 3500 , whenthe salmon fishery was in decline. The �15N valuesfrom the dog bones also showed significantly higherlevels after 3500 , which would be consistent withincreased consumption of shellfish, also evident in thelarge amounts of unbroken clam shell and thick layersof almost pure shell in deposits dating to between 2880and 980 (Conover, 1978: 98). All of these multiplelines of evidence point to the conclusion that the Namusalmon fishery did decline significantly after 3500 from its earlier peak levels, and that this decline causedsome degree of periodic economic hardship resultingin greater reliance on more marginal alternatives,including minor fish species, shellfish, and deer.

Seasonal indicators suggest that the nature of theNamu settlement as a winter village, occupied at leastfrom late autumn through late spring, remained con-sistent throughout most of the last 6000 years. Salmon,which would have been harvested at the time of theautumn spawning runs, and herring, which were thefocus of intensive fisheries in the late winter/earlyspring, when they congregate near shore to spawn,were abundant in all periods. Neonatal harbour seals,which must have been hunted sometime around themid-June peak in the pupping season, are also presentin all of the fauna-bearing components (Cannon,1991). Variability in the productivity of the localsalmon fishery, however, clearly had implications forthe prosperity of the residents and possibly on the sizeand intensity of the occupation as well. The period inwhich the salmon fishery was at its peak, for example,is also marked by the presence of the majority ofexcavated burials, several of which contain beads andother finely-crafted artefacts that can be interpreted asmarkers of status and wealth. This association sug-gests a linkage between economic prosperity and theelaboration of social distinctions. The subsequentdecline in the salmon fishery may have resulted in somecontraction in the size of the village, as indicated byabandonment of the area in the vicinity of theRivermouth trench at around 2530 (Figure 2)

728 A. Cannon

Figure 2. Excavation unit and auger sample locations at the Namu site [the N. and W. Main trenches (1977) and Rivermouth trench (1978)excavations produced the overwhelming bulk of the faunal remains recovered in 1977–1978 (Cannon, 1991), the Front trench excavation (1970)is located in deposits that date entirely to the period 1840–480 (Luebbers, 1978: 25)].

(Carlson, 1991: 93). The range and intensity of activi-ties at the site, however, remained consistent duringthis and the subsequent period, at least up untilc. 980 , when stratigraphic, faunal, and artefact evi-dence shows some restriction in the range of activities,which might indicate more intermittent or morerestricted seasonal use of the site (Conover, 1978: 98).

While the size and prosperity of the Namu settle-ment appear to have been closely tied to the produc-tivity of its salmon fishery, the regional implications ofvariation in local salmon production are not at all clearfrom this one example. The decline in the Namusalmon fishery could be the result of local conditionsaffecting the spawning habitat of the Namu River(Cannon, 1991), or could represent a broader regional

trend of declining salmon populations. The impli-cations for human populations in the region areequally difficult to infer. If the peak in economicprosperity was the basis for population growth, thenthe result might have been a process of regionalsettlement expansion. Alternatively, regional expan-sion might have been predicated on the decline in theNamu fishery, as segments of the local population wereforced to establish settlements in more productivelocations. With little or no indication of the chronol-ogy of site settlement in the vicinity, and no indicationof the focus or intensity of fisheries at other locations,such questions remain impossible to resolve.

The obvious strategy for resolving these ques-tions would be to mount a regional program of site

Variability in Northwest Coast Salmon and Herring Fisheries 729

investigation. But traditional excavation in deep shellmidden deposits is both costly and time consuming.Reliable evidence of variability in fishing economieswould also require a considerable volume of exca-vation or the availability of multiple lines of confirmingevidence, as at Namu. Limited test excavation, evenif that option was practical and affordable, would notbe certain to produce either result. The insurmountableproblems posed by the necessity of regional-scaleinvestigation led me to explore the possibility ofobtaining sufficient relevant information throughmuch more limited site testing. I therefore undertooka small pilot project at Namu in 1994 to deter-mine whether trends in the density of fish bones inbucket-auger samples would match the overalltrends evident on the basis of the earlier full-scaleexcavations.

Figure 3. Removal of a matrix sample from the bucket-auger.

Bucket-Auger Sampling at Namu

Core sampling has long been advocated as an efficient,cost-effective alternative to excavation (Casteel, 1970;Stein, 1986), and a simple hand-operated bucket-auger,which enables quick and easy sampling from a range ofdifferent locations, with minimal site disturbance, hasbeen used as an effective method for sampling matrixand for mapping the depth and extent of shell-middendeposits (Whittaker & Stein, 1992). Casteel (1976) hadalso long ago demonstrated that the focus and intensity

of fisheries could as effectively be determined on thebasis of small samples of matrix as on the basis ofbone collected through large-scale field-screening ofexcavated matrix. The object of the pilot study atNamu was to determine whether the quantities of bonein very small bucket-auger samples could also be usedto monitor subtle variation in the intensity of salmonfisheries over time.

A bucket-auger measuring 7 cm in diameter wasused to take samples at ten locations adjacent toprevious excavation units (Figure 2). One set ofsamples, from location ‘‘A’’, was retained unwashedfor the possibility of future sediment and micro-debitage analysis. Sampling at location ‘‘B’’ wasdiscontinued when a large rock was encountered.Sampling at the remaining eight locations (C–J) wascontinued to a depth below the shell-bearing middendeposits, which ranged from 1·5 to 5·1 m below thesurface. The samples were obtained by twisting theauger into the midden matrix until the bucket waseither full or the compaction of material inside thebucket made further downward progress impossible.The sample was then brought to the surface, removedfrom the bucket (Figure 3), and bagged. The depth ofthe shaft was measured and the auger was then loweredto take the next sample. The vertical extent of eachsample varied, and was impossible to control with anyprecision, but was generally 7–15 cm. With some effort,the auger penetrated through or past most potentialimpediments, including a disintegrated granite boulder,

730 A. Cannon

Figure 4. Salmon remains from a bucket-auger sample (scale in mm).

layers of whole clam shell, and pieces of fire-brokenrock.

Each sample’s volume was measured and the matrixwas washed through 2 mm mesh screen. The residuewas then dried and bagged for later analysis. Thesamples were examined under a low-powered micro-scope, and all fragments of bone were recovered. Themajority of identifiable bone consisted either of frag-ments of salmon vertebrae, which could be positivelyidentified on the basis of even the smallest recoveredfragment (Figure 4), or whole herring vertebrae(Figure 5). Other salmon and herring elements werealso recovered, and the remains of ten additional fishtaxa, including anchovy and eulachon, were identified.A very small number of identifiable mammal bonesand artefacts including worked bone, lithic debitage,and tiny shell disk beads were also recovered.

One of the first results of the analysis was thedemonstration of the early and consistent importanceof the local herring fishery. More than 2200 wholeherring vertebrae were recovered from the 63 l ofmatrix obtained at the eight sample locations, whichoverall was 30 times the recovery rate for whole salmonvertebrae. The vast majority of herring vertebraewould easily pass through 1/8 inch mesh screen, andmost were not even visible as the samples were washedin the field. By every measure, therefore, the field-recovery methods of the full-scale site excavation were

unsuitable for assessing the importance of herring tothe Namu subsistence economy.

To assess variability in the herring and salmonfisheries over time, the samples were assigned to majortemporal components that had been defined on thebasis of previous excavations. This was done by match-ing the depth of the samples to the depth of thetemporal components identified in the adjacent exca-vation units. In cases where temporal components inthe excavation units had also been distinguished on thebasis of readily-identifiable breaks in stratigraphy,comparable changes in the matrix from the augersamples were also used to aid in the assignment ofsamples to their appropriate temporal component.This process of assigning samples to time periodscomparable to those used in the earlier analysis andinterpretation of faunal trends undoubtedly resulted insome minor errors. Major changes in matrix com-position, however, were easy to recognise, and wherethese were associated with temporal componentidentifications, no more than half a sample was likelyto be mis-assigned to a later or earlier component. Thedepths of less readily visible stratigraphic boundariesmight have varied between the excavation and auger-ing locations, but the auger samples were taken fromlocations close to the adjacent excavations, and it isunlikely that the depth of deposits representing theboundaries of temporal components varied enough to

Variability in Northwest Coast Salmon and Herring Fisheries 731

Figure 5. Herring remains from a bucket-auger sample (scale in mm).

Table 2. Density of salmon and herring remains in auger samples from Namu (density as number of identifiedspecimens per litre of matrix)

Temporal period(14C years )

No. ofsamples

Matrix volume (l) Density/total volume Density/<2 mm volumeTotal <2 mm Salmon Herring Salmon Herring

1880–480 56 19·2 16·3 41·4 40·1 48·7 47·13500–2185 42 12·2 8·8 31·7 26·6 44·1 36·94390–3825 51 15·0 11·8 61·9 42·5 78·5 53·94775–4540 18 7·9 6·6 72·2 41·4 86·3 49·56060–5170 24 8·4 7·6 64·3 40·5 71·3 44·9

result in the mis-assignment of more than a singlesample either way.

The abundance of identified faunal remains fromeach of the five major fauna-bearing components wasstandardised by calculating the numbers of identifiedbones and bone fragments per litre of sample matrix(Table 2). The figures in Table 2 show the densities forthe total volume of recovered matrix and for thevolume of matrix finer than 2 mm. The latter figure isprobably a better basis for comparison between timeperiods since it controls somewhat for variability inthe rate of shell deposition. The presence of largequantities of whole and coarsely-broken shell has theeffect of disproportionately increasing the overallmatrix volume, and consequently decreasing the appar-ent density of fish remains. As the total and less than

2 mm volumes show, the volumes of coarse and wholeshell vary considerably between periods at Namu. Thedensity of herring (vertebrae and cranial elements) andsalmon (whole vertebrae, vertebra fragments, andteeth) per volume of the finer matrix can be morereliably compared across components to monitortrends in abundance over time.

The results show a roughly consistent emphasis onherring throughout the period 6060–480 , but a clearincrease and later decrease in salmon consistent withtrends identified from the full-scale excavations. Com-parison of salmon trends based on auger-sampling andfull-scale excavation (Figure 6) show a very strongmatch (Spearman’s rho=0·87, p<0·05); the only excep-tion is the period between 3500 and 2185 . Apartfrom this one anomaly, however, and despite probable

732 A. Cannon

errors in assigning samples to temporal com-ponents, the auger sampling results show remarkablecorrespondence to the trends in salmon identifiedpreviously.

The auger sampling results are all the more remark-able since they are based on only 3000 specimens, mostof which are tiny bone fragments, recovered from atotal of 63 l of matrix, while the full-scale excavationresults are based on more than 90,000 salmon bonesfrom more than 100 m3 of site matrix. The overallresults suggest that auger sampling is an effectivealternative to full-scale excavation for monitoringtrends in the abundance of at least the salmon andherring fisheries. Salmon is probably particularly well-suited to this type of sampling because even very smallfragments of vertebrae are readily identifiable. Theycan be used to compare the relative abundance ofsalmon between sites and site components, providedthat preservation conditions and the degree of frag-mentation are comparable. Herring is also suitablebecause vertebrae and various cranial elements aresmall and are normally well preserved and oftenextremely abundant in Northwest Coast shell middensites.

01880–480

100

Period (years BP)6060–5170

80

60

40

20

3500–21854390–38254775–4540

Figure 6. Comparison of salmon trends based on auger samples (�) and full-scale excavation (�) (excavation values are the percentage ofsalmon among all identified fish remains, auger sample values are the number of identified salmon elements and vertebra fragments per litreof matrix finer than 2 mm).

Regional Site Testing

Based on the success of the Namu pilot project,funding was obtained for an expanded program of site

testing in the Namu vicinity. Over the course of two4-week field seasons in 1996 and 1997, a total of 16sites were tested (Figure 7). The sites ranged frommajor village middens, often with extant remains ofhouses or house platforms on the surface, to smallermiddens probably representing specific-purpose camp-sites, to very small midden deposits on rocky islets. Allbut two of the sites (ElSx-17 and ElSx-18) were pre-viously recorded as the result of surveys conducted inthe 1960s and 1970s (Pomeroy, 1980). The goal inselecting sites for investigation was to sample withinthe context of a limited field program a range of sitetypes and locations extending from the relatively pro-tected waters of Fitz Hugh Sound to the more exposedreaches of the outer coast. Site settings included majormainland rivers, small island coves and streams, tidalchannels, and rocky islets. Although the choice of siteswas not based on systematic sampling, they include anapproximate cross-section of the range of shell middensites in the area.

Each of the sites was mapped, and sampled using acombination of coring and bucket-augering. The coresamples were obtained with a subsoil probe capable ofextracting 0·8 inch (2 cm) diameter cores in 3-foot long(91·44 cm) sections. The stratigraphically-intact coreswere retained in clear plastic liners, and were usedto obtain uncontaminated samples of shell or charcoalfor AMS radiocarbon dating (Cannon, 2000a). Thebucket-auger samples were obtained by the samemethods used at Namu. The number of sampling

Variability in Northwest Coast Salmon and Herring Fisheries 733

Figure 7. Location of auger- and core-tested sites in the Namu vicinity.

locations varied from as little as one to as many as sixdepending on the area and depth of the site and theconstraints of time. The auger samples were returnedto the laboratory, where they were gently washedthrough 2 mm mesh screen. The matrix retained in thescreen and the less than 2 mm residue were dried andretained for further analysis. All of the larger matrixwas examined under a microscope. Analysis of thevertebrate faunal remains showed that although avariety of fish species and a small number of mammaland bird remains are present in the samples, theoverwhelming majority of vertebrate fauna from allsites consists of salmon and herring. The density of fishbone from these sites, however, varies widely.

The figures in Table 3 show the density of herringand salmon remains in matrix from each of the sitestested. Table 3 also shows the period of occupation at

each of the sites as determined from radiocarbon datesobtained from samples taken from the cores and fromsamples of shell obtained from surface or erodedforeshore exposures. The results clearly show a con-trast in salmon and herring bone density between mostof the smaller sites and three of the larger sites tested[ElSx-3 (Kisameet Bay), EkSx-12 (Koeye River), andElTb-1 (Hurricane Island)]. Each of these latter threesites shows evidence of having been a major village site,at least in its most recent occupation. Surface depres-sions and clearings indicate recent habitation at theKisameet Bay site (Luebbers, 1978: 17). The remains offallen post and beam, cedar plank houses are evidenton the surface of the Hurricane Island site, and largerectangular house platforms are present on the surfaceof the Koeye River site. This evidence, together withthe size of the sites and the density of fish remains,

734 A. Cannon

Table 3. Density of salmon and herring remains in auger samples from sites in the Namu vicinity (density as number of identified specimens per litreof matrix)

SiteDate range

(14C years )

Sitearea(m2)

No. ofsamplinglocations

No. ofsamples

Matrixvolume (l)

Density/(total volume)

Density/(<2 mm volume)

Total <2 mm Salmon Herring Salmon Herring

EkSx-12 2030–230 2500 4 88 26·2 23·0 61·1 7·6 69·4 8·6ElSx-3 2290–contact 1630 6 110 30·7 22·9 50·7 52·4 67·9 70·3ElTa-3 770–110 50 1 8 2·4 2·0 35·5 6·6 43·8 8·1ElSx-18 3110–40 1310 3 43 13·2 8·5 20·8 20·5 32·3 31·9ElSx-5 5770–310 2190 2 58 21·4 14·4 17·0 28·4 25·2 42·1ElSx-10 5270–130 840 6 83 27·6 20·0 17·4 32·4 24·0 44·7ElSx-6 240 260 1 2 0·5 0·4 20·0 35·6 23·7 42·1ElTb-1 2540–contact 1110 4 77 29·8 24·3 14·2 139·0 17·4 170·6ElTa-25 3720–150 675 3 60 16·6 6·0 3·0 2·2 8·3 6·1ElSx-16 1240–100 300 2 15 4·1 2·5 3·9 37·1 6·3 59·7ElSx-4 2620–360 140 2 16 6·0 4·5 4·5 2·8 6·0 3·8ElTa-21 1730–170 130 3 15 6·2 5·3 2·3 0·6 2·7 0·8ElTa-18 9940–190 670 3 23 12·1 10·7 2·2 0·1 2·5 0·1ElTb-2 1880–630 60 1 9 3·0 2·5 1·7 12·2 2·0 14·9ElSx-17 990 320 1 3 0·8 0·7 1·2 0·0 1·5 0·0ElSx-8 1690–120 930 3 16 4·3 2·8 0·0 1·2 0·0 1·8

argues in favour of all of these sites having been majorvillage settlements throughout most if not all of theiroccupation history.

The small size and relatively low density of fishremains in the matrix from most of the other sitesargues in favour of their interpretation as specific-purpose campsites. The majority (ElSx-4, 5, 8, 10, 16,18, ElTa-25) qualify as true shell midden sites, in whicha major part of the midden matrix consists of shellfishremains (see Lyman, 1991: 39–40). These sites prob-ably served primarily as campsites for the specificpurpose of collecting shellfish (e.g. ElTa-25) or for acombination of activities including intensive herringfishing (e.g. ElSx-5, 10, 16). Three of the remainingsites are better described as shell-bearing middens. Thedensity of both shell and fish remains is relatively low,and they may have served a variety of specific func-tions, including campsites used for hunting or forless-intensive fishing of a variety of species other thansalmon or herring (e.g. ElTb-2), or possibly for moreintensive fishing of salmon (e.g. ElTa-21) which wastaken away for processing and consumption at otherlocations. One other small, shell-bearing site (ElTa-3)exhibits a relatively high density of salmon remains, aswell as small fragments of mammal bone, most ofwhich are charred or calcined, but almost no herringbones. This site is located at the mouth of a smallstream, where a two-stage stone fish trap was con-structed. It probably served as a campsite for intensiveshort-term catching and processing of salmon. Theremaining two sites (ElSx-6, 17) consist of small,thin patches of midden deposit on rocky islets inNamu harbour. They may have served as refuges,lookouts, or some other specific purpose for theresident population at Namu.

The densities of herring and salmon bone seem todistinguish clearly between longer-term habitation and

specific-purpose campsites. The combination of inten-sive autumn salmon fisheries and early spring herringfisheries, with an assumed reliance on stores of salmonand other foods over the winter period, also provides aclear indication of probable winter village sites (Namuand Kisameet Bay). The high density of salmon, lowdensity of herring, and indications of large houseplatforms (n=5, x̄=101·4 m2) on the surface at theKoeye River site suggest a winter village that may havebeen at least partially abandoned in the late winter/early spring in favour of residence at a location withgreater access to herring. Conversely, the low densityof salmon, very high density of herring, and indicationsof relatively small, but permanent cedar plank houses(n=7, x̄=c. 49·7 m2) on the surface of the HurricaneIsland site suggest a spring/summer base camp orvillage which would have been occupied beginning inthe late winter or early spring of each year at the startof the intensive herring fishery.

Beyond these indications of site function, the resultsof the auger sample analysis also indicate temporalvariability in the productivity of local salmon fisheries,and suggest one possible consequence of the decline ofthe Namu fishery. The earliest dates at the KisameetBay [2290�110 (Luebbers, 1978: 25)] and KoeyeRiver sites (2030�55 ) place their initial occupationnear the end of the Namu Period 5 occupation (3500–2185 ), which witnessed the earliest clear signs ofdecline in the salmon fishery. Although it may be nomore than mere coincidence that these villages and thevillage on Hurricane Island (2540�50 ) were allestablished at around the same time, and at around thetime that the Rivermouth trench area at the Namusite was abandoned (2530�160 ), this initial evi-dence strongly suggests that expansion of village settle-ment in the region was at least in part a consequenceof decline in the Namu salmon fishery. The further

Variability in Northwest Coast Salmon and Herring Fisheries 735

possibility is that these villages were founded by someof the former Namu residents. The overall density ofsalmon remains in the midden deposits at the KisameetBay and Koeye River sites is also much higher than inthe post-3500 deposits at Namu, indicating thatlower productivity of the Namu fishery was likely aconsequence of local environmental circumstancesrather than regional decline in salmon populations.

Whatever the precise relationship between events atNamu and at other sites in the vicinity, which onlyfurther, more intensive investigation will be able toresolve, the results of bucket-auger sample analysisclearly show the value of this method of limited sitetesting for identifying the focus and intensity of localfisheries. At a minimum these data can be used to inferbasic differences in site usage, but they also offerfurther evidence of temporal and spatial variability insalmon fisheries that may be a key to understandingthe settlement and economic history of this and otherregions of the Northwest Coast. To establish the extentand significance of variability in fishing intensity, how-ever, requires more than simple counts of bone frag-ments and calculations of overall bone density. In thisregard the auger samples can also be used as a basisfor statistical comparison of densities between sitesand between components within sites by taking intoaccount the degree of variability in sample densities.

Intra- and Inter-Site Comparisons ofSalmon-Fishing IntensityTable 4 shows the average per sample density ofsalmon bones and bone fragments from each of thefauna-bearing components from Namu and from theKisameet Bay and Koeye River sites. When calculatedin this manner the overall trend for Namu is roughlythe same, and the average per sample densities for theKisameet Bay and Koeye River sites remain higherthan those of the post-3500 samples from Namu.The advantage of calculating bone density on this persample basis, however, is that it gives a measure ofvariability between samples (indicated by the standarddeviations shown in Table 4), which can be used as a

basis for statistical comparison of the sample means.This can be used to show not just that the overalldensity of bone is higher or lower in one site orcomponent in comparison to another, but also whetherthat apparent difference is statistically significant, or nomore than the kind of difference one might expect onthe basis of sampling variability.

Using the mean per sample density as the basis ofcomparison, the various components at Namu and atthe Kisameet Bay and Koeye River sites were com-pared to one another. Table 5 shows the results ofStudent’s t-test comparisons of the means fromselected sites and components. The results clearly showthat although the average density in the Namu Period5 samples is not significantly lower than that of thepreceding Period 4 samples, which represent the peakin the salmon fishery, the samples from the followingPeriod 6 do show a significantly lower density ofsalmon bone. The average of the combined samplesfrom the peak periods in the Namu fishery (4775–3825 ) is significantly higher than the average of thecombined post-3500 samples. The average densitiesof salmon bone in the samples from the Kisameet Bayand Koeye River sites are also significantly higher thanthe average density in samples from the final period ofthe Namu occupation. Bone densities at these sites arenot significantly lower than the density in the peakNamu Period 4 samples.

The auger sample results therefore not only confirmin an overall sense the general trends observed in theresults of the full-scale excavation at Namu, they alsoprovide yet another independent line of evidence thatstatistically confirms the reality of the apparentdecline in the Namu salmon fishery. Furthermore,these results verify that the decline was a local ratherthan a regional phenomenon, and provide additionalevidence to suggest that one of the consequences forthe human populations of the region may have beenthe establishment of alternative village settlements.

Table 4. Average per sample density of salmon per litre of <2 mmmatrix

Site/component Date range n x ..

NamuPeriod 6 1880–480 56 47·4 34·3Period 5 3500–2185 42 54·6 71·3Period 4 4390–3825 51 75·4 60·8Period 3 4775–4540 18 90·2 66·3Period 2 6060–5170 24 61·1 71·8Periods 5 and 6 3500–480 98 50·5 53·2Periods 3 and 4 4775–3825 69 79·3 62·1Kisameet Bay 2290–contact 110 60·7 64·0Koeye River 2030–230 88 69·6 83·6

Table 5. Student’s t-test results for comparison of mean sampledensities of salmon from selected sites and Namu components

Site/component(s)

Comparison pairsNamu

period(s) tSignificance

(1-tailed)

Namu period 6 4 �2·9 0·002Namu period 5 4 �1·5 0·07Namu periods 5 and 6 3 and 4 �3·1 0·001Kisameet Bay 6 1·7 0·04

4 �1·4 0·09Koeye River 6 2·2 0·02

4 �0·5 0·32

Summary and ConclusionThe application of limited core and auger sampling tosites in the Namu vicinity effectively resolved the

736 A. Cannon

problems involved in mounting a regional program ofsite investigation to assess the extent and consequencesof variability in local salmon fisheries. As shown incomparisons with the results of earlier full-scale exca-vation at Namu, the contents of limited auger samplesare an accurate indicator of overall trends in theintensity of salmon fishing. Auger sampling also pro-vides an accurate measure of the relative importance oflocal herring fisheries, which were a mainstay of latewinter/early spring subsistence economies in manyparts of the Northwest Coast. Auger sample indica-tions of the relative intensity of fall salmon fisheriesand spring herring fisheries can also be used to infer theseasonality of site occupation, and can be used as oneline of evidence in distinguishing between habitationand specific-purpose campsites. More importantly,from the perspective of the present study, variabilityin salmon bone densities in auger samples can beused to monitor long-term trends in the productivity oflocal site fisheries and to compare the intensity offisheries between sites within a region. The ability tomake these types of comparisons has important conse-quences on our ability to assess the role of subsistencevariability in the overall history of the NorthwestCoast region.

A number of current interpretations of NorthwestCoast archaeology rest on assumptions and the resultsof investigations at a very small number of sites tosuggest that intensification of salmon fishing was agradual, long-term process culminating in the highlevels of production that supported dense populationsand complex social organisations present in the regionat the time of European contact (Matson & Coupland,1995; Coupland 1998). Evidence from a small numberof sites, such as Namu, seemingly contradict this pointof view by demonstrating an early dependance onstorage-based salmon-fishing economies. Still furtherevidence from a few sites that show the equal or evengreater importance of fish other than salmon have beenused to argue that the importance of salmon in prehis-toric Northwest Coast fishing economies has beenoverstated (Monks, 1987). Although often phrased interms of mutual contradiction, all of these results areconsistent with expectations of seasonally-variable siteuse and long-term variability in the productivity oflocal salmon fisheries (Cannon, 1996, 2000c). Theopportunity for contradictory interpretations has beenenhanced by the small number of site investigations,and inadequate appreciation of the extent of local andtemporal variability in salmon and other fisheries.Although theoretical models can offer alternative inter-pretations of variable site data, nothing can replace themuch more extensive regional site investigationsneeded to resolve these apparent contradictions. Inthe face of the prohibitively high costs of large-scaleshell midden excavation on even a local, much less aregional scale, auger sampling provides one effectivealternative specifically well-suited to the investigationof variability in salmon and herring fisheries.

AcknowledgementsAuger sampling at Namu was conducted with thefinancial support of a grant from the Arts ResearchBoard, McMaster University. I want to thank Roy L.Carlson for making it possible for me to conduct the1994 pilot project in conjunction with the SimonFraser University field school he was directing atNamu. I am also grateful to the field school students,who contributed to the success of the project throughtheir direct participation. I especially want to thankCarl Humchitt, representative of the Heiltsuk CulturalEducation Centre, for his help in 1994. The investi-gation of sites in the Namu vicinity in 1996 and 1997was supported by a research grant from the SocialSciences and Humanities Research Council of Canada,and was carried out in cooperation with the HeiltsukCultural Education Centre. The assistance of JenniferCarpenter, Program Manager of the CulturalEducation Centre, is gratefully acknowledged. Thefield work was carried out by crews consisting ofAubrey Cannon, Paul Prince, and Lisa Rankin ofMcMaster University, and Bennie Humchitt, JessicaHumchitt, and Mike Windsor representing theHeiltsuk Cultural Education Centre.

ReferencesAmes, K. M. (1998). Economic prehistory of the northern British

Columbia coast. Arctic Anthropology 35, 68–87.Binford, L. R. (1978). Nunamiut Ethnoarchaeology. New York:

Academic Press.Binford, L. R. (1981). Bones: Ancient Men and Modern Myths. New

York: Academic Press.Butler, V. L. & Chatters, J. C. (1994). The role of bone density

in structuring prehistoric salmon bone assemblages. Journal ofArchaeological Science 21, 413–424.

Cannon, A. (1991). The Economic Prehistory of Namu: Patterns inVertebrate Fauna. Department of Archaeology, Simon FraserUniversity, Publication 19. Burnaby, British Columbia: SimonFraser University.

Cannon, A. (1995). The ratfish and marine resource deficiencies onthe Northwest Coast. Canadian Journal of Archaeology 19, 49–60.

Cannon, A. (1996). Scales of variability in Northwest salmon fishing.In (M. G. Plew, Ed.) Prehistoric Hunter–Gatherer FishingStrategies. Boise State University Monograph Series. Boise: BoiseState University, pp. 25–40.

Cannon, A. (2000a). Settlement and sea-levels on the central coast ofBritish Columbia: evidence from shell midden cores. AmericanAntiquity 65, 67–77.

Cannon, A. (2000b). Faunal remains as economic indicators on thePacific Northwest Coast. In (P. Rowley-Conwy, Ed.) AnimalBones, Human Societies. Oxford: Oxbow, pp. 49–57.

Cannon, A. (2000c). Was salmon important in Northwest Coastprehistory? In (S. C. Gerlach & M. S. Murray, Eds) People andWildlife in Northern North America: Essays in Honor of R. DaleGuthrie. BAR International Series, in press.

Cannon, A., Schwarcz, H. P. & Knyf, M. (1999). Marine-basedsubsistence trends and the stable isotope analysis of dog bonesfrom Namu, British Columbia. Journal of Archaeological Science26, 399–407.

Canti, M. G. & Meddens, F. M. (1998). Mechanical coring as anaid to archaeological projects. Journal of Field Archaeology 25,97–105.

Carlson, R. L. (1979). The early period on the central coast of BritishColumbia. Canadian Journal of Archaeology 3, 211–228.

Variability in Northwest Coast Salmon and Herring Fisheries 737

Carlson, R. L. (1991). Appendix B: Namu periodization and C-14chronology. In (A. Cannon, Ed.) The Economic Prehistory ofNamu: Patterns in Vertebrate Fauna. Department of Archaeology,Simon Fraser University, Publication 19. Burnaby, BritishColumbia: Simon Fraser University, pp. 85–95.

Carlson, R. L. (1996). Early Namu. In (R. L. Carlson & L.Dalla Bona, Eds) Early Human Occupation in British Columbia.Vancouver: University of British Columbia Press, pp. 83–102.

Casteel, R. W. (1970). Core and column sampling. AmericanAntiquity 35, 465–467.

Casteel, R. W. (1976). Fish remains from Glenrose. In (R. G.Matson, Ed.) The Glenrose Cannery Site. National Museumof Man, Mercury Series, Archaeological Survey of Canada PaperNo. 52. Ottawa: National Museum of Man, pp. 82–87.

Conover, K. (1978). Matrix analyses. In (J. J. Hester & S. M.Nelson, Eds) Studies in Bella Bella Prehistory. Department ofArchaeology, Simon Fraser University, Publication 5. Burnaby,British Columbia: Simon Fraser University, pp. 67–99.

Coupland, G. (1998). Maritime adaptation and evolution of thedeveloped Northwest Coast pattern on the central NorthwestCoast. Arctic Anthropology 35, 36–56.

Fawcett, I. (1991). Appendix C: faunal analysis of matrix samples.In (A. Cannon, Ed.) The Economic Prehistory of Namu: Patternsin Vertebrate Fauna. Department of Archaeology, Simon FraserUniversity, Publication 19. Burnaby, British Columbia: SimonFraser University, pp. 97–100.

Grayson, D. K. (1984). Quantitative Zooarchaeology: Topics in theAnalysis of Archaeological Faunas. Orlando: Academic Press.

Hester, J. J. (1978). Conclusions: early tool traditions in NorthwestNorth America. In (J. J. Hester & S. M. Nelson, Eds) Studies inBella Bella Prehistory. Department of Archaeology, Simon FraserUniversity, Publication 5. Burnaby, British Columbia: SimonFraser University, pp. 101–112.

Hester, J. J. & Nelson, S. M. (Eds) (1978). Studies in Bella BellaPrehistory. Department of Archaeology, Simon Fraser University,Publication 5. Burnaby, British Columbia: Simon FraserUniversity.

Hoffman, C. (1993). Close-interval core sampling: tests of a methodfor predicting internal site structure. Journal of Field Archaeology20, 461–473.

Luebbers, R. (1978). Excavations: stratigraphy and artefacts. In (J. J.Hester & S. M. Nelson, Eds) Studies in Bella Bella Prehistory.Department of Archaeology, Simon Fraser University, Publi-cation 5. Burnaby, British Columbia: Simon Fraser University,pp. 11–66.

Lyman, R. L. (1991). Prehistory of the Oregon Coast: The Effects ofExcavation Strategies and Assemblage Size on ArchaeologicalInquiry. San Diego: Academic Press.

Matson, R. G. & Coupland, G. (1995). The Prehistory of theNorthwest Coast. San Diego: Academic Press.

McCartney, A. P. (1992). Along the coast: regional archaeology insouthern Alaska. Arctic Anthropology 29, 192–204.

Millennia Research Ltd (1996). Heiltsuk Traditional TerritoryArchaeological Overview Assessment. Unpublished report on file,Archaeology Branch, Government of British Columbia, Victoria.

Monks, G. G. (1987). Prey as bait: the Deep Bay example. CanadianJournal of Archaeology 11, 119–142.

Pomeroy, J. A. (1980). Bella Bella Settlement and Subsistence. Ph.D.Thesis. Simon Fraser University, Burnaby, British Columbia.

Stein, J. K. (1986). Coring archaeological sites. American Antiquity51, 505–527.

Stein, J. K. (1992). Sediment analysis of the British camp shellmidden. In (J. K. Stein, Ed.) Deciphering a Shell Midden. SanDiego: Academic Press, pp. 135–162.

Whittaker, F. H. & Stein, J. K. (1992). Shell midden boundaries inrelation to past and present shorelines. In (J. K. Stein, Ed.)Deciphering a Shell Midden. San Diego: Academic Press,pp. 25–42.

Zita, P. (1997). Hard times on the Northwest Coast: deer phalanxmarrow extraction at Namu, British Columbia. In (A. Dolphin &D. Strauss, Eds) Drawing Our Own Conclusions, Proceedings of the1997 McMaster Anthropology Society Students Research Forum.Hamilton: McMaster Anthropology Society, pp. 62–71.