Increasingly, zooarchaeological data from multiple sites are synthesized and integrated as meta-analyses to help answer pressing contemporary questions in conservation biology and beyond (e.g., Buss et al. Reference Buss, van den Hurk, Falahati-Anbaran, Elliott, Evans, Frasier and Mulville2023; Fossile et al. Reference Fossile, Ferreira, da Rocha Bandeira, Dias-da-Silva and Carlo Colonese2020; McKechnie et al. Reference McKechnie, Lepofsky, Moss, Butler, Orchard, Coupland, Foster, Caldwell and Lertzman2014). In order for zooarchaeological data to be useful for applied studies, these data must be accurate and comparable. Recent work has taken a critical approach to “big data” in zooarchaeology, and archaeology more generally, emphasizing the need to generate data (and metadata) that are accessible and suitable for future study (Heilen and Manney Reference Heilen and Manney2023; Lau and Kansa Reference Lau and Kansa2018; LeFebvre and Sharpe Reference LeFebvre, Sharpe, Christina and Michelle2017; Neusius et al. Reference Neusius, Styles, Peres, Walker, Crothers, Smith and Colburn2019; Nicholson et al. Reference Nicholson, Kansa, Gupta and Fernandez2023; Nims and Butler Reference Nims and Butler2019). Zooarchaeologists are encouraged to be more explicit and transparent in their identification methods and tighten both quality control (QC) and quality assessment (QA) through blind tests and detailed protocols and descriptions (Driver Reference Driver2011; Wolverton Reference Wolverton2013). Accordingly, there has been much discussion of the challenges in taxonomic identifications (Bochenski Reference Bochenski2008; Fischer Reference Fischer2015; Gobalet Reference Gobalet2001; Hawkins et al. Reference Hawkins, Buckley, Needs-Howarth and Orchard2022; Lubinski et al. Reference Lubinski, Lee Lyman and Johnson2020; Nims and Butler Reference Nims and Butler2017; Sipilä et al. Reference Sipilä, Steele, Dickens and Martin2023), with some scholars suggesting we frame specific identifications as subjective or probabilistic (e.g., Hawkins et al. Reference Hawkins, Buckley, Needs-Howarth and Orchard2022; Lyman Reference Lyman2019; Nims and Butler Reference Nims and Butler2019; Wolverton Reference Wolverton2013). However, there has been little discussion of initial laboratory sorting of faunal remains. Given that major animal groups (fish, mammal, bird, invertebrate, reptile, amphibian) are often analyzed by different specialists, and analysts may be working with “legacy” collections sorted long ago, we argue that it is also important to consider error introduced in this early stage of analysis.
More than 20 years ago, Lyman (Reference Lyman2002:14) noted that taxonomic identification (typically genus- or species-level identification) occupied less than 5% of zooarchaeology textbooks (e.g., Chaplin Reference Chaplin1971; Davis Reference Davis1987; Hess and Wapnish Reference Hesse and Wapnish1985; Klein and Cruz-Uribe Reference Klein and Cruz-Uribe1984; O’Connor Reference O’Connor2000; Reitz and Wing Reference Reitz and Wing2000). Discussions of the very initial stages of faunal sorting into major taxonomic groups, normally conducted soon after screening, are even rarer in these texts and more recent works (Albarella, ed. Reference Albarella2017; LeFebvre and Sharpe Reference LeFebvre, Sharpe, Christina and Michelle2017; Lyman Reference Lyman2019).
One exception is O’Connor (Reference O’Connor2000:34), who recognized that “the sorting of sieved samples can be a logistical nightmare. It is one of the most important stages in the recovery of bones by sieving, yet it is seldom discussed in the literature.” He goes on to mention that sorting is often done by nonspecialist personnel who may missort unusual or unexpected elements, such as fish otoliths or bird tracheal rings (O’Connor Reference O’Connor2000:35). However, he asserts that even nonspecialists can sort with a high degree of success, speculating that “only one or two percent of fish bones in a rich sample will be mis-sorted as mammal or bird bones; more frequent is the mis-sorting of amphibian bones as fish” (O’Connor Reference O’Connor2000:35). More recently, Gifford-Gonzalez (Reference Gifford-Gonzalez2018) has devoted a chapter in her introductory text to “Identification: Sorting Decisions and Analytic Consequences,” which includes some discussion of initial sorting. She notes that even novices can conduct a preliminary sort accurately, provided they have adequate training and oversight (Gifford-Gonzalez Reference Gifford-Gonzalez2018:173).
Other authors (e.g., Beisaw Reference Beisaw2013; Broughton and Miller Reference Broughton and Miller2016; Reitz and Wing Reference Reitz and Wing2000) provide useful descriptions or illustrations and photographs of major animal groups, but these frequently occur in the context of background biology—how differences in locomotion, feeding, or protection relate to different anatomical features—rather than specific tips for sorting the faunal remains themselves. Beisaw’s (Reference Beisaw2013) manual (and online appendix: https://aprilbeisaw.com/identifyingbones/) is an exception, with tables and lists of characteristics for distinguishing elements of different animal groups. Yet she also asserts that different classes of animals are relatively easy to distinguish (Beisaw Reference Beisaw2013:18). Overall, it seems that the basic sorting of major taxonomic groups is regarded as being so obvious that little discussion or critical evaluation of the process are warranted.
The enormous faunal analysis we undertook for the Čḯxwicən Project offered an opportunity to assess the frequency of error in this early stage of analysis. Was the impact of laboratory sorting error minimal, as is often assumed? Čḯxwicən (45CA523) is a traditional village of the Lower Elwha Klallam Tribe (LEKT) located on the Strait of Juan de Fuca near Port Angeles, Washington (USA; Figure 1). In 2012, our team began a project to study a large sample of Čḯxwicən’s faunal remains to evaluate long-term human-environmental change (human ecodynamics) at the site (Butler, Bovy et al. Reference Butler, Bovy, Campbell, Etnier and Sterling2019; Butler, Campbell et al. Reference Butler, Campbell, Bovy and Etnier2019; Fitzhugh et al. Reference Fitzhugh, Butler, Bovy and Etnier2019). While working on the analysis, we frequently encountered remains that had been incorrectly sorted in the field laboratory and systematically tracked the patterns of missorting. We analyze the trends, including common elements and taxa that were missorted, provide photos of common misidentifications, and make specific recommendations for mitigating this error in future research. While our example comes from a Pacific Northwest Coast shell midden, the issue of sorting error is not unique to this region, site type, or even artifact class, and is important for scholars working in both academia and cultural resource management (CRM).
Map of Northwest Coast showing location of Čḯxwicən. Dashed line outlines the Salish Sea watershed. Figure drafted by Kendal McDonald and used with permission.
Material and Methods: Čḯxwicən Village Faunal Assemblage
Čḯxwicən Village was extensively excavated in 2004 by Larson Anthropological Archaeological Services (LAAS) as part of a transportation construction project (Butler, Bovy et al. Reference Butler, Bovy, Campbell, Etnier and Sterling2019). The excavated sample is immense, with 518 m2 area and 261 m3 volume of material removed. Multiple occupations spanning the last 2,700 years were documented with exceptionally fine geostratigraphic control, and field sampling was explicitly designed to facilitate integration of all faunal data to allow for a detailed reconstruction of animal use over this period. Matrix was excavated from each uniquely defined deposit into 10-liter buckets, which were water-screened through graded mesh 1″ (25.6 mm), ½″ (12.8 mm), and ¼″ (6.4 mm); every twentieth 10-liter bucket recovered from a uniquely defined deposit was water-screened down to ⅛”, or 0.32 cm, mesh (Kaehler and Lewarch Reference Kaehler, Lewarch and Larson2006). Over 38,000 10-liter bags of water-screened matrix were recovered; faunal remains were recorded in situ as well.
The initial sorting was completed by dozens of different lab technicians, with varying experience in zooarchaeology and levels of supervision. Laboratory staff processed (dried, sorted, counted) bags of water-screened matrix, which was sorted into “mammal, avian and fish bone, charcoal, wood, shell, artifacts, and unmodified rock” (Kaehler and Lewarch Reference Kaehler, Lewarch and Larson2006:6-6). Sorting of each ⅛” (0.32 cm) sample typically took between 8 and 24 hours, depending on the density of the midden material. Lab personnel were under pressure to sort this vast assemblage in a timely manner, which may have precluded time for extensive training or consultation of comparative specimens.
All of the archaeological materials from Čḯxwicən sorted by LAAS personnel were sent to the Burke Museum of Natural History and Culture (Seattle), where they were formally accessioned into the permanent research collections. Our project targeted study of selected areas at the large site to represent a range of time periods and cultural activities (e.g., house structures, extramural midden; see Butler, Campbell et al. Reference Butler, Campbell, Bovy and Etnier2019; Butler et al. Reference Butler, Bovy, Campbell, Etnier, Butler, Bovy, Campbell, Etnier and Sterling2020). In total we studied faunal remains from 8.2 m3 from 820 distinct contexts. Faunal remains from our selected areas were borrowed from the Burke Museum and delivered or shipped to our various research institutions. Our project team included four primary analysts, each of whom specializes in a different kind of fauna: Kristine Bovy (University of Rhode Island; birds), Virginia Butler (Portland State University; fish), Sarah Campbell (Western Washington University; invertebrates), and Michael Etnier (WWU; mammals). The researchers were aided by dozens of undergraduate and graduate students working under close supervision in their respective laboratories. As analysis progressed, missorted faunal material was pulled and transferred to the appropriate analyst as it was encountered.
Here we focus on all the specimens from our selected sample that were identified to major taxonomic group; specimens identified only as “vertebrate” are not included. We quantify the assemblage using NSP and NISP (number of identified specimens). To be considered “identified,” the specimen had to be identified at least to taxonomic Order. Further details on our analysis decisions can be found in our Open Context reports (Bovy Reference Bovy, Butler, Bovy, Campbell, Etnier and Sterling2018; Butler et al. Reference Butler, Hofkamp, Molenhoff, Nims, Patrick Rennaker, Syvertson, Butler, Bovy, Campbell, Etnier and Sarah2018).
Results
Overall Patterns in Sorting Error
Over 1.2 million faunal specimens are included in this analysis (Table 1). Of these, 99.6% were correctly sorted in the laboratory; this positive result, however, is driven by the immense size of the invertebrate assemblage (>1,100,000 NSP), which had relatively few missorted specimens (0.1%). Sorting error had much greater impact on other taxonomic groups, especially birds (Figure 2). Table 1 shows the number of bones correctly sorted in the field laboratory for each taxonomic group, the number of transfers by taxonomic group, and the identification rate for both the transfers and total specimens. Birds had the highest proportion of missorted specimens: nearly one-fourth (22.6%) of the bird bones were missorted (n = 1,358). In addition, birds had the highest rate of identification for the transferred specimens (42.5%). Mammals were less frequently missorted (n = 754; 11.3%) than birds, and the missorted specimens were less identifiable (9.9%). Fish had the highest absolute numbers of missorted specimens (n = 2,049), but this accounted for a small percentage (2.0%) of the overall total because of the enormous size of the fish assemblage (>100,000 NSP).
The frequency (percent) and total number of transferred specimens by major taxonomic group.
Summary of Transfers by Major Taxonomic Group.
Note: “Identified specimens” were identified to at least taxonomic Order.
Trends for Major Taxonomic Groups
Birds. A total of 1,358 bird bones were initially missorted (Table 1); of these, 785 (57.8%) were initially identified as mammal, while 450 (33.1%) were transferred from fish. A smaller proportion of bird bones was initially sorted as invertebrate (n = 123; 9.1%), 63% of which were calcined (white in appearance; Figure 3).
Examples of missorted bird elements transferred from shell, which were calcined (white in appearance): (a) mandible fragment of loon (Gaviidae; catalog #: WS-2639.99.04.22); (b) carpal of gull (Laridae; catalog #: WS-122.99.08.22).
Elements from all parts of a bird skeleton were missorted (Table 2). The elements that were most commonly missorted (by percent of element identified) were tracheal rings, carpals, patellae, coracoids, sternae, hyoids, furculae, and vertebrae. Vertebrae were commonly transferred from both fish and mammal; other common transfers included skulls, sternae, and tracheal rings from fish (Figure 4), and humeri, coracoids, and carpals from mammals (Figure 5). No attempt was made to identify the bird vertebrae, ribs, rear phalanges, or tracheal rings to taxon below Aves, so the fact that these were missorted did not affect our understanding of taxonomic abundance at the site.
Commonly missorted bird elements transferred from fish: (a) tracheal rings (catalog # WS-15788.99.08.22); (b) quadrates of Common Murre (Uria aalge; catalog # WS-20815.99.04.22); (c) premaxilla of Common Murre (catalog # WS-12211.99.04.22); (d) sternum fragment of shearwater (Procellariidae; catalog # WS-0290.99.04.22).
Commonly missorted bird elements transferred from mammal: (a) carpals of Common Murre (Uria aalge; catalog # WS-15484.99.08.22); (b) proximal humerus of Red-Throated Loon (Gavia stellata; catalog # A4-245.01.01); (c) proximal humeri fragments of loon (Gaviidae; catalog # WS-14474.99.04.22).
Initial Sorting of Bird Skeletal Elements by Major Animal Group.
Note: “% Missorted” is the frequency of missorted bird bones by skeletal element, as a frequency of the overall total NSP for that element.
* No attempt was made to identify these elements more specifically than “bird.”
Turning to patterning in missorting of birds by higher taxon, the most commonly missorted bird taxa were very large birds, such as eagles (Accipitridae) and albatrosses (Phoebastria spp.) and diving birds with dense, more spongy-looking bones, such as cormorants (Phalacrocoracidae), loons (Gaviidae), and grebes (Podicipedidae; Supplementary Table 1; Figure 5b and c), which most resemble mammal bone to the nonspecialist. In two cases, rare taxa—eagle and woodpecker (Picidae)—would not have been recorded without transferred specimens. Nearly 43% (n = 577; 42.5%) of the transferred bird bones were identified at least to Order (Table 1).
Fish. The greatest number of fish transfers came from the invertebrate sample (n = 905; 44.2%; Table 1). While much of the transferred fish material was unidentifiable to taxonomic Order, a few very distinctive fish elements were commonly missorted, including cod (Gadidae) otoliths, Pacific spiny dogfish (Squalus suckleyi) spines, spotted ratfish (Hydrolagus colliei) dental plates, and skate (Rajidae) dermal denticles (Figure 6). Significantly, 80.0% of all cod otoliths (n = 55), which are useful for aging and seasonality studies, were initially missorted; the otoliths were initially sorted as invertebrate (n = 31) and mammal (n = 13). The most frequently missorted fish was the ratfish (17.6%; 51 out of 289 NISP), followed by dogfish (5.9%; 114 out of 1935 NISP) and skate (4.5%; 20 out of 447 NISP).
Commonly missorted fish elements: (a) otolith of Pacific cod (Gadus macrocephalus; catalog # WS-9306.99.04.23); (b) spine of Pacific spiny dogfish (Squalus suckleyi; catalog # WS-9139.99.08.23); (c) dental plate of spotted ratfish (Hydrolagus colliei; catalog # WS-9306.99.04.23); (d) dermal denticle of a skate (Rajidae; catalog # WS-8358.99.04.23). Photography by Anthony R. Hofkamp and used with permission.
Mammal. The overall number of missorted mammal bones was low (n = 754; 11.3%; Table 1). Only 9.9% of the transferred mammal bones (75 out of 754) were identifiable to taxonomic Order (Table 1). The most frequently missorted mammals were rodents (23 out of 50; 46.0%) and carnivores (38 out of 285; 13.3%), with the latter dominated by tooth fragments (n = 13) and phalanges (n = 9).
One unexpected set of mammal remains that was commonly missorted is “bone chips” or shavings (Figure 7; Bovy et al. Reference Bovy, Etnier, Butler, Campbell and Deo Shaw2019). These items, which were cut/shaved from robust mammal bones or antler as part of bone-tool production, show distinctive curls and step-fractures along the length of the curl analogous to wood chips. There is little discussion of bone chips in the literature of the Pacific Northwest Coast, so it is not surprising that they were missorted as often as they were. Out of a total of 194 bone chips ultimately recorded in our sample, 53 (27.3%) were originally missorted, with most initially assigned to fish (n = 32), then invertebrate (n = 20) or bird (n = 1). While the condition of these remains made it impossible to assign them to taxon beyond mammal, they are an important record of bone-tool manufacture.
Example of a mammal bone chip or shaving (catalog # WS-10825.04.01), a byproduct of tool production; these were frequently missorted as fish. Photography by Anthony R. Hofkamp and used with permission.
Invertebrates. The analyzed invertebrate sample from Čḯxwicən is enormous (>1,100,000 NSP), so sorting error (n = 1,077) had little impact on the results. Not surprisingly, invertebrates are easier to distinguish from bone owing to the distinctive color and texture. One of the most commonly missorted invertebrate elements was the sea urchin (Strongylocentrotus sp.) rotula (n = 62; Figure 8; see also Campbell Reference Campbell2008: Figure 2). Given that 2,813 rotulae were correctly sorted to invertebrate, this error hardly affects results. On the other hand, for smaller-scale projects, systematic missorting of sea urchin or any invertebrate elements could affect our understanding of animal use.
Example of sea urchin (Strongylocentrotus sp.) rotulae (catalog # WS-1573.99.08.10, left and # WS-10456.99.08.10, right), a commonly missorted invertebrate element. Photography by Oliver Jue and used with permission.
Discussion and Conclusions
So, what was the potential impact of laboratory sorting error on our understanding of the Čḯxwicən fauna? The answer depends on the question being asked and sample sizes of the taxa under study, but it is clear the sorting error was systematic, not random. The relatively small bird assemblage was the most affected of all the groups, with a high frequency of sorting error: 22.6% of the bird bones in the sample were initially sorted as mammal, fish, or even invertebrate. Larger birds (eagles, albatrosses) and diving birds (cormorants, loons, grebes) with dense bones were frequently confused for mammals. The relative importance of birds at the site would have been less apparent without the missorted specimens (Table 1). After the initial sort, mammal specimens (n = 5,924) were considerably more abundant than bird (n = 4,644), but when the transferred specimens are included, the number of bird specimens (n = 6,002) is closer to the final mammal total (n = 6,678), which is unusual in the region (Butler and Campbell Reference Butler and Campbell2004). In addition, the relative frequency of certain taxa was affected; for example, the relative frequency of loons in the assemblage increases from 6.4% to 9.7% when transfers are added. In two cases, rare bird taxa (eagle, woodpecker) would not have been recorded without transferred specimens. In contrast, the vast fish assemblage was less affected by sorting error, although the relatively less abundant spotted ratfish (Hydrolagus colliei) would have been underrepresented without the transfers.
The impact of sorting error might be more significant when investigating finer temporal trends or in studies that rely on measures of taxonomic richness or diversity, especially for small assemblages. In a similar study, Graesch (Reference Graesch2009:771) discusses how the importance of bias in field identification varies depending on your question; for example, he found that while the overall percentage of artifact classes may not change significantly from field to lab, the error did hinder understanding of specific artifact classes. By extension, this would also be true for rare animal taxa. Moreover, sorting errors are more likely to occur for rare animal taxa, since analysts are less familiar with these species. The missorting of even a few specimens might affect interpretations of rare and/or threatened taxa.
Unusual elements are more likely to be missorted or unrecognized. For example, ratfish dental plates, dogfish spines, skate dermal denticles, and cod otoliths were frequently missorted in the Čḯxwicən assemblage, even though these taxa are common in the region. Likewise, Hawkins and others (Reference Hawkins, Buckley, Needs-Howarth and Orchard2022:698) found that certain fish elements of even relatively common taxa were missorted in blind tests. Other elements (e.g., artiodactyl stylohyoid) may be underreported, since they are omitted from published skeletal guides or lacking in reference specimens (Lubinski et al. Reference Lubinski, Lee Lyman and Johnson2020). For example, at Čḯxwicən, the most frequently missorted bird elements—tracheal rings and carpals—are rarely featured in skeletal guides. In addition, rarely reported elements, such as a grebe patella and a duck hyoid, were missorted. Systematic missorting of skeletal elements may skew results in studies focused on skeletal part frequencies, such as questions related to butchery or transport behavior, and may also affect calculations of the minimum number of individuals (MNI).
Although our study comes from a large and complex coastal shell midden, the impacts of laboratory processing are relevant for other regions and environments. Birds are often understudied compared with mammals, perhaps due to challenges in analysis because of “the high species diversity and high level of osteological similarity among taxa” (Broughton and Miller Reference Broughton and Miller2016:129) or simply the prevalence of mammals at many sites resulting from dietary preference or preservation. Since they are less frequently studied, smaller birds may be confused with Leporid and Sciurid bones, and larger and/or juvenile bird specimens mistaken for mammals (Bovy Reference Bovy2011a:30). The presence or absence of even small numbers of juveniles may be important if investigating biogeographic questions. Missorting of faunal material may also be important beyond questions of diet or biogeography. For example, fish otoliths, used in seasonality and aging studies, and the remains of amphibians or reptiles, which are so important for environmental reconstructions, may be unrecognized. In addition, evidence for bone-tool production, such as the bone chips from Čḯxwicən (Figure 7), may be overlooked.
Sorting error is especially important to keep in mind for legacy collections or any large collaborative zooarchaeological project where major taxonomic groups are sent to separate analysts. As more researchers choose to reuse existing collections rather than excavate to obtain new assemblages, we need to be cognizant of best practices for working with these assemblages. Collections that were excavated, sorted, and cataloged in the past may have had certain materials removed and stored separately, or even discarded (Hesse and Wapnish Reference Hesse and Wapnish1985:71). For example, at the Watmough Bay site (45SJ280) in Washington State (USA), juvenile cormorant bones, which are relatively large and with a spongy texture, were initially sorted as mammal, so it was necessary to look through every mammal bag to get a complete sample of birds (Bovy Reference Bovy2011b). Faunal remains of interest may also have been relegated to “unidentifiable” bags (Gifford-Gonzalez Reference Gifford-Gonzalez2018:170). However, as Lyman (Reference Lyman2019:1403) notes, whether a specimen is “identifiable” depends on the research question being asked, so all specimens should be made available for future analysis. In addition, formal bone tools or modified bones and shells, some of which may be taxonomically identifiable, may also be curated separately and not readily available to a given analyst unless requested.
Reducing error in the initial sorting of faunal remains would have many benefits, including to help increase the quality of the faunal data used in synthetic studies and stored in databases for future researchers. In addition, accurate faunal sorting at the start of a project would increase efficiency. Better estimates of staffing needs could be created, less time spent on transferring specimens between analysts, and analysts would have relevant material at the start, which is particularly helpful if museum visits are needed to refine taxonomic identifications.
We conclude with a series of recommendations for sorting and training protocols to help reduce initial sorting error.
Recommendations
Sorting/Transfer Protocols
• Assess the degree of missorting during the proposal stage when working with legacy collections. As our study demonstrates, it is best not to assume that original laboratory sorting was accurate; this is especially important for those analyzing bird bones. We suggest that investigators create a pilot project to analyze a fraction, perhaps a 5%–10% random sample, of each major taxonomic group to assess error. A systematic evaluation of sorting error at the start would allow for changes in methodology, as needed, and for a better estimate of budgetary needs (e.g., transfer and reanalysis by different analysts) when crafting research proposals. In addition, the analyst may also want to look at formal tools, which are often curated separately.
• Create a good working list of possible taxa at the start of the project (Driver Reference Driver2011:27; Lyman Reference Lyman2019:1393; Wolverton Reference Wolverton2013:386). Lab supervisors should work collaboratively with zooarchaeologists to generate such lists, so everyone is aware of the possible universe of taxa. These lists should include all taxonomic groups, including reptiles and amphibians.
• Consider potential sources of error before beginning to sort. Thus, analysts should “anticipate difficult to separate taxa” (Wolverton Reference Wolverton2013:387) and “develop a set of rules about how identifications are to be made” (Driver Reference Driver2011:27). Although Wolverton (Reference Wolverton2013) and Driver (Reference Driver2011) were discussing more specific-level identifications (e.g., Bos/Bison), protocols should also be used for the initial stages of sorting (see Nims and Butler [Reference Nims and Butler2017:760] and Gifford-Gonzalez [Reference Gifford-Gonzalez2018:173–174] for suggestions on developing effective sorting protocols). To increase the quality of faunal data, laboratory protocols should also consider faunal remains that might be mistaken for plant tissues—or plant tissues which might appear to be animal remains. For example, Beisaw (Reference Beisaw2013:22) notes that the bone caps of amphibians may be mistaken for seeds, and the bone chips in the Čḯxwicən samples superficially resembled wood shavings (Figure 7).
• Have protocols in place for transferring materials between analysts. While error will hopefully be reduced with planning, initial sorting decisions should still be viewed as hypotheses. Therefore, principal investigators should consider this issue at the start of a project and anticipate the need to transfer missorted materials between analysts (and budget accordingly), while lab managers should establish protocols during a project to facilitate these transfers. The transfer protocols might include specifications for the types of materials (elements) to be transferred, as well as steps for maintaining accountability for collections borrowed from museums or other agencies. Additional discussions at the beginning of the Čḯxwicən project would have streamlined the transfers, and we might have avoided sending specimens across the country that were not included in taxonomic analysis; for example, bird vertebrae, ribs, and tracheal rings, which were not identified beyond “Aves,” might have simply been retained and recorded by the original analyst.
Training
• Increase discussion of initial sorting in zooarchaeological textbooks and guides, rather than concentrating solely on distinguishing closely related species, such as sheep and goat. A few zooarchaeology textbooks or guides do discuss differences between major animal groups or provide photos or images (Adams and Crabtree Reference Adams and Crabtree2012; Beisaw Reference Beisaw2013; Broughton and Miller Reference Broughton and Miller2016; O’Connor Reference O’Connor2000; Reitz and Wing Reference Reitz and Wing2000), and some include images of rare elements, such as fish otoliths and dental plates (Beisaw Reference Beisaw2013:34–35) and duck syringeal (tracheal) bulla (Broughton and Miller Reference Broughton and Miller2016:145). However, the formats of these publications could be improved for ease of use in the lab (see Lyman [Reference Lyman2019] for suggestions on creating effective illustrated skeletal guides). Lyman (Reference Lyman2019:1405) and Lubinski and others (Reference Lubinski, Lee Lyman and Johnson2020:791) advocate for the development of an open-access database of identification criteria, which could be continually tested and refined; such a resource could also include criteria and images related to initial sorting decisions.
• Train sorters with reference collections and/or photos of frequently missorted elements. As Lyman pointed out (Reference Lyman2011:33), the notion that anyone can identify bones with minimal training is not true. While considerable time may be spent teaching students how to identify stone flakes or other artifact types, less time may be devoted to faunal identification (in our experience). Graesch (Reference Graesch2009) compared screening data for two consecutive years of field schools and found that the artifact recovery rates significantly increased in the second year, probably due to expanded and explicit training on artifact identification provided to students. Likewise, quality control in the laboratory would probably be enhanced if experienced supervisors worked closely with lab personnel who discussed and checked their identifications (Gifford-Gonzalez Reference Gifford-Gonzalez2018:173; Lyman Reference Lyman2002:18). To minimize error, CRM firms and academic departments may need to expand reference collections to include multiple species from all major taxonomic groups, keeping in mind applicable state and federal laws regarding the collection of wild animals. While skeletal reference collections are often preferred, digital files and virtual tools are now available and are increasingly sophisticated (Albarella Reference Albarella and Albarella2017; LeFebvre and Sharpe Reference LeFebvre, Sharpe, Christina and Michelle2017:Table 1; McKechnie et al. Reference McKechnie, Kansa and Wolverton2015).
• Train zooarchaeology students in recognizing all kinds of animals. More training is also needed for zooarchaeologists, not just for lab technicians and students. Many zooarchaeologists specialize in one type of fauna (birds, fish, mammals, invertebrates), which is clearly beneficial for making refined identifications. In his blind test of fish-bone identification, Gobalet (Reference Gobalet2001:385) found differing opinions on taxonomic identifications, even among highly trained researchers, and concluded, “A specialist with a narrow expertise, rather than one who claims a wide-based knowledge (e.g., all vertebrates or just ‘faunal analysis’) will probably be more credible.” While this is true for more specific taxonomic analysis, generalized knowledge and training are still important. For example, although our project team included four specialists, each with a PhD and years of experience in zooarchaeology, there were still things we did not recognize or may not have sorted correctly without consultation among ourselves. One solution may be to increase exposure and training in all types of animals in zooarchaeology courses, rather than focusing primarily (or even exclusively) on mammals. In our experience, training in identifying reptiles and amphibians is particularly lacking, which may account for frequent missorting of these remains (O’Connor Reference O’Connor2000:35).
Supplementary Material
The supplementary material for this article can be found at https://doi.org/10.1017/aap.2024.43.
Supplementary Table 1. Frequency of missorted bird bones by taxon (in order of most to least missorted), as a percentage of the overall total NISP for that taxon.
Acknowledgments
We are grateful to the Lower Elwha Klallam Tribe for their ongoing support. Bill White, LEKT Tribal Historical Preservation Officer, provided guidance and assistance at various times. Laura Phillips (Burke Museum, University of Washington) facilitated the loan of all the materials used in analysis, helped with the photo of the sea urchin rotulae (Figure 8), and provided guidance on many aspects of the project. The Burke Museum Ornithology and Mammals divisions provided comparative specimens on loan, and the Museum of Comparative Zoology (Harvard) provided access to additional bird skeletons. No permits were needed for this research, but the Washington State Department of Transportation granted permission for the study and publication of images. Kristina Dick, Laura Syvertson, Adam Freeburg, and Kendal McDonald helped create and maintain the Access database over the years. We acknowledge the enormous efforts of LAAS, especially project leaders Dennis Lewarch and Lynn Larson, who directed site excavations; and LAAS lab personnel, who did a great job under challenging circumstances. Finally, we acknowledge the passing of Sarah K. Campbell, who was a driving force behind our project and was keen to explore the impacts of sorting error on our results, among countless other things. She is sorely missed, and we dedicate this article to her.
Funding Statement
Most of the funding for Čḯxwicən analysis came from the National Science Foundation (Grant Numbers 1219468, 1353610, and 1663789 to Portland State University; 1219483 to University of Rhode Island; 1219470 to Western Washington University), through the efforts of Anna Kerttula de Echave, whose support we gratefully acknowledge. Washington State Department of Transportation wrote letters of support for funding and helped subsidize loan costs.
Data Availability Statement
The Čḯxwicən collection is curated at the Burke Museum of Natural History and Culture (Seattle, Washington), in trust for the Washington Department of Transportation and the Lower Elwha Klallam Tribe. Complete zooarchaeological data are available on the Open Context data management platform (https://opencontext.org; https://doi.org/10.6078/M7Q52MQ7).
Competing Interests
The authors declare none.
Open access
