Validating archaeological oyster morphometric studies and expanding their potential role in socioecological research

Abstract

Over the last decade, there has been a surge in archaeological studies of oyster shell morphometrics published in high-impact, high-profile scientific journals. Using large datasets of oyster size measurements yielded through archaeological investigations, these studies have aimed to contribute to inter- and trans-disciplinary bodies of work centered on issues of conservation, sustainability, ecosystem stability, and resource management. Indeed, teams of archaeologists, ecologists, geoscientists, and community partners have produced key insights into long-term trends in human-ecosystem dynamics, especially the anthropogenic impacts of oyster harvesting on this keystone species. In this paper we argue, through a series of quantitative validations and suggestions, that there remains untapped potential for archaeological studies to expand their impact on and relevance to broader socioecological studies. Critically, few studies have aimed to validate the meaningfulness of these extant archaeological findings about oyster height. Further, we continue to lack important interpretive frameworks guided by modern validations by which to evaluate the actual meanings of different oyster morphometrics as they relate to key ecological, climatic, and anthropogenic variables. To these ends, we quantitatively establish the robustness and reliability of recent eastern oyster (Crassostrea virginica) morphometric studies and suggest new analytical avenues that may be more fruitfully employed in future studies. We end with a series of suggestions to improve the archaeological study of oyster shell morphometrics for both archaeological interpretation and to increase the impact and relevance of these studies for broader interdisciplinary socioecological research.

Introduction

Over the last decade, there has been a surge in archaeological studies of oyster shell morphometrics published in high-impact, high-profile scientific journals^1–10. Using large datasets of oyster size measurements yielded through archaeological investigations, these studies have aimed to contribute to inter- and trans-disciplinary bodies of work centered on issues of conservation, sustainability, ecosystem stability, and resource management. Indeed, teams of archaeologists, ecologists, geoscientists, and community partners have produced key insights into long-term trends in human-ecosystem dynamics, especially the anthropogenic impacts of oyster harvesting on this keystone species. Importantly, and admirably, these studies have made their way beyond the bounds of archaeology, and the wide majority of their citations continue to come from outside of the discipline, from studies spanning the ecological, economic, and even applied sciences.

To date, however, archaeological studies have focused primarily on a single dimension to ascertain broader insights into the human-ecosystem-oyster nexus: oyster shell heights. In this paper we argue, through a series of quantitative validations and suggestions, there remains untapped potential for archaeological studies to expand their impact on and relevance to broader socioecological studies. Critically, few studies have aimed to validate the meaningfulness of these extant archaeological findings about oyster height. Further, we continue to lack important interpretive frameworks guided by modern validations by which to evaluate the actual meanings of different oyster morphometrics as they relate to key ecological, climatic, and anthropogenic variables. We, of course, believe these archaeological studies are critical for informing a range of modern and contemporary challenges and research programs. Archaeological studies and archaeological data, however, are often met with skepticism, rejected outright, or simply ignored. We argue it is thus imperative for archaeologists to contextualize their data within robust interpretive frameworks to produce meaningful and useful results to contemporary ecosystem scientists and applied practitioners.

Archaeologists, in particular zooarchaeologists, have written extensively about survivorship, or taphonomic, biases in the interpretations of archaeological faunal material. Recent work has pointed to eastern oyster (Crassostrea virginica) specifically as a species for which little attention has been paid to the effects of such biases, nor to potential solutions to account for the potential impacts of taphonomic processes¹¹. Indeed, shell assemblages are often highly fragmented, while archaeologists rely on whole valves (usually left valves) on which to make key measurements. A range of studies using allometric models to estimate whole shell characteristics from fragmented hinges have been published, yet studies specifically aimed at oyster remain sparse (see Wallis et al. 2025 for review; e.g., ^12–15. Recently, Wallis et al. (2025) argued, based on mean weight comparisons, survivorship biases may account for significant variation in oyster size measurement studies. While we demonstrate below that mean assemblage weights remain unreliable indicators of survivorship biases, we agree there are key studies that must be undertaken to evaluate, and justify, the relevance, reliance, and applicability of archaeological oyster morphometric studies. Without such work, the value and importance of this archaeological research to broader fields within the ecosystem and applied sciences remains open to valid skepticism.

To these ends, in this paper we quantitatively establish the robustness and reliability of recent eastern oyster morphometric studies, demonstrate that survivorship bias (in our particular case along the Georgia Coast, USA (Fig. 1)) is a non-issue, and suggest morphometric measures that may be more fruitfully employed in future studies to avoid potential taphonomic biases and to increase sample sizes. We end with a series of suggestions to improve the archaeological study of oyster shell morphometrics for both archaeological interpretation and to increase the impact and relevance of these studies for broader interdisciplinary socioecological research. These suggestions include (1) devoting more attention to shell shape and the relationships between distinct measurements, (2) establishing interpretive frameworks for linking specific ecological and environmental variables to shell morphometrics and, similarly, (3) conducting modern experimental studies to better understand the impacts of small-scale anthropogenic pressures (similar to those relevant to past societies) on oyster resources and shell morphometrics. This last suggestion is critical if we are to contribute to understanding the actual mechanisms at the human-oyster-ecosystem interface that drive individual and population characteristics of eastern oysters.

Fig. 1.

Map of the Georgia Coast (top left) and Ossabaw Island (bottom left), showing locations from which oyster shell assemblages analyzed were excavated across the site of Middle Place (9CH158) (right). Maps created in Esri ArcGIS Pro Version 3.7(https://www.esri.com) using “World Imagery” Basemap with Sources: Esri, DigitalGlobe, GeoEye, i-cubed, USDA FSA, USGS, AEX, Getmapping, Aerogrid, IGN, IGP, swisstopo, and the GIS User Community. https://services.arcgisonline.com/ArcGIS/rest/services/World_Imagery/MapServer.

Results

Mean shell weight is not appropriate for assessing survivorship biases or for reconstructing mean valve sizes

Recent work has leveraged a series of complex calculations to reconstruct mean shell weights for archaeological assemblages (Wallis et al. 2025). Expanding on this work, and to better understand the relationship between oyster shell morphometrics (especially sizes) and weight, we shift the unit of analysis from the context, or whole assemblage, to individual whole left valves. Instead of relying on mean oyster weights, pooled estimated assemblage weights, and calculating these metrics from averaged weight contributions from left and right valves, we made morphometric measurements (Fig. 2) and weighed individual left valves from two archaeological assemblages (described in more detail in the Methods section). Briefly, one assemblage is from a c. 70 cm deep, extensive sheet midden likely deposited rapidly by an Indigenous community c. 500 years ago on Ossabaw Island along the Georgia coast. This assemblage, Pond SM3 (or PSM3), consists of 1,131 whole left valves exhibiting little evidence of breakage, largely absent of edge fractures. The second assemblage, a c. 50 cm deep midden referred to as Freedmen’s SM5 (or FSM5), is located at the same site on Ossabaw Island, but was deposited by a post-Emancipation, Black community between AD 1865 and 1945. FSM5 consists of 500 whole left eastern oyster valves also exhibiting little evidence of breakage, almost entirely absent of even edge fractures.

Fig. 2.

Common shell morphometric measurements made by zooarchaeologists on eastern oyster (Crassostrea virginica).

The R² values for simple regressions evaluating the relationship between left valve height and weight for the PSM3 and FSM5 assemblages were 0.48 and 0.62 respectively. The residual standard error (RSE) for PSM3 weight was 10.55 g while for FSM5 it was 9.14 g. For height, the RSE for PSM3 was 16.36 mm and was 12.72 mm for FSM5 (Table 1a; Fig. 3).

Table 1.

Results of all statistical analyses discussed in the text.

	n	Mean (Weight)	SD (Weight)	CV (Weight)	R	R 2	RSE (Weight (g))	RSE (Valve Height (mm))	p	Slope (m)	Intercept
a. Whole Valve Height v. Weight Correlations
Freedmens Shell Midden 5 (FSM5)	501	21.69	14.82	0.68	0.79	0.62	9.14	12.72	<0.001	0.57	-21
Pond SM3 (PSM3) (Full Assemblage)	1101	14.64	14.57	0.99	0.69	0.48	10.55	16.36	<0.001	0.45	-15
Pond SM3 Shape Category 1 (<1.50, Roundest)	240	16.53	16.84	1.02	0.86	0.74	8.51	7.61	<0.001	0.97	-38
Pond SM3 Shape Category 2 (1.50-1.75, Rounder)	256	16.67	18.18	1.09	0.85	0.72	9.58	9.58	<0.001	0.85	-36
Pond SM3 Shape Category 3 (1.75-2.00, Longer)	234	12.71	11.98	0.94	0.86	0.74	6.15	9.69	<0.001	0.55	-22
Pond SM3 Shape Category 4 (>2.00, Longest)	371	13.92	11.46	0.82	0.85	0.72	5.99	13.5	<0.001	0.38	-16
Multiple Regression: PSM3 Weight ~ Valve Height + Valve Length	1101	–	–	–	–	0.71	7.87	–	<0.001	–	–
Multiple Regression: FSM5 Weight ~ Valve Height + Valve Length	501	–	–	–	–	0.77	7.22	–	<0.001	–	–
b. Whole Valve Height v. Whole Hinge Length Correlations
	n	R	R2	RSE (Hinge Lengths (mm))	RSE (Valve Heights (mm))	p	Slope (m)	Intercept
Freedmens Shell Middens 3 and 5 (FSM5 and FSM3) (Full Assemblage)	987	0.7	0.49	1.53	15.18	<0.001	0.07	1.7
Pond SM3 (PSM3) (Full Assemblage)	1354	0.76	0.58	1.42	14.32	<0.001	0.075	1.1
Indigenous Households (Full Assemblage)	9786	0.78	0.61	1.23	12.59	<0.001	0.076	0.91
Households Shape Category 1 (<1.50, Roundest)	3481	0.75	0.56	1.03	8.88	<0.001	0.087	0.46
Households Shape Category 2 (1.50-1.75, Rounder)	2433	0.77	0.59	1.18	10.26	<0.001	0.089	0.24
Households Shape Category 3 (1.75-2.00, Longer)	1458	0.79	0.62	1.32	11.98	<0.001	0.087	0.19
Households Shape Category 4 (>2.00, Longest)	1627	0.81	0.66	1.39	15.43	<0.001	0.073	0.74
c. Whole and Reconstructed Valve Heights (Indigenous Households Only)
Comaprison to Whole Shell Valve Height Measurements
	n	Mean	SD	CV	Mann-Whitney U	Polyserial Correlation
Valve Heights for Whole Shells	9785	63.02	20.18	0.32	–	–
Reconstructed Valve Heights for Whole Shells from Hinge Lengths	9785	63.02	15.77	0.25	<0.001	0 (Negligible Difference)
Reconstructed Valve Heights for Broken Valves from Hinge Lengths	8500	65.89	17.81	0.27	<0.001	-0.09 (Negligible Difference)
Combined Whole Valve Heights with Reconstructed Valve Heights from Hinge Lengths of Broken Shells	18,286	64.35	19.17	0.3	<0.001	-0.04 (Negligible Difference)
d. Hinge Lengths v Taphonomic States
	n	Mean	SD	CV	Mann-Whitney U (p)	Polyserial Correlation
Freedmens (FSM5 and FSM3) Broken	1072	7.09	2.31	0.32	0.1823	n/a (No Difference)
Freedmens (FSM5 and FSM3) Whole	987	6.92	2.13	0.31
Pond (PSM3) Broken	1647	6.33	2.25	0.36	<0.05	-0.10 (Negligible Difference)
Pond (PSM3) Whole	1353	6.16	2.17	0.35
Households Broken	8502	6.06	2.22	0.37	<0.001	-0.04 (Negligible Difference)
Households Whole	9785	5.71	1.97	0.35

Fig. 3.

Plots illustrating the relationship between valve height and shell weight across all whole left valves excavated from Pond SM3 (left) and Freedmen’s SM5 (right).

Eastern oysters grow in different shapes along a continuous spectrum of valve height to length ratios. This greatly complicates and problematizes shell height studies as a whole (expanded on in the discussion) but also challenges the use of mean weights at the assemblage level as an appropriate tool for reconstructing shell size. A large round shell and a large narrow or elongated shell may weigh the same, but they could have significantly different valve height measurements. To formally evaluate the effect of shell shape on the reliability of weight as a useful measure of shell size, we segmented the PSM3 assemblage into four shape categories based on ratios of height to length from roundest to longest (roundest (< 1.50), rounder (1.50–1.75), longer (1.75-2.00), and longest (< 2.00) to capture variation between shorter/rounder and longer/skinnier informed by both Kent¹⁶ and Mizuta and Wilkfors¹⁷.

The R² values for all four shape categories ranged between 0.72 and 0.74, increasing significantly from the R² value of the pooled PSM3 assemblage (R² = 0.48) (Table 1a; Fig. 4). RSE also fell dramatically, indicating that weight has more predictive power and potential for reconstructing height when shape is accounted for. The most complicating factor for using weight, however, emerges in the evaluation of the slopes of the resulting regression equations. The slope for the total PSM3 assemblage regression is 0.45. For the roundest and rounder shape categories, slopes are 0.97 and 0.85, respectively. For the longer and longest shape categories, slopes are 0.55 and 0.38, respectively. This indicates that the relationship between valve height and weight is fundamentally different between rounder and longer oysters, and that weight increases at a different rate for rounder than for longer oysters, essentially rendering any attempt at assemblage mean weight calculations inappropriate for reconstructing mean oyster sizes from fragmented archaeological assemblages.

Fig. 4.

Plots illustrating the relationship between valve height and shell weight across whole left valves excavated from Pond SM3, subdivided into four shape categories from roundest (top left and top right) to longest (middle left and middle right). The bottom two plots depict the ranges of whole left valve weights by shape category excavated from Pond SM3.

Hinge length is a reliable alternative measurement to valve height that can be made on both whole and broken specimens

A number of studies have demonstrated the utility in using hinge measurements to estimate whole size sizes, though few have focused on eastern oyster in particular^13–15; see Rick et al. ⁷for an exception). Here we demonstrate hinge length (the width between the edges of the umbo) (Fig. 2) to be a useful and reliable alternative measurement to whole valve height for eastern oysters. One of the key benefits of hinge length is it can be measured on specimens where only the umbo has survived, making it possible to include measurements on most if not all specimens included in MNI (minimum number of individuals) calculations. The relationship between shell height and shell weight has been demonstrated (e.g.¹⁸, however, given the various drivers of shell weight (particularly those that affect shell thickness) it is near impossible to estimate shell height from the weight of a broken valve. We again use the PSM3 assemblage, the FSM5 assemblage combined with another contemporary historic shell midden (FSM3), as well as a third, large assemblage excavated from 23 discrete Indigenous household middens at the same archaeological site on Ossabaw Island (referred to as “Indigenous Households” and described in more detail in the Methods). PSM3 yielded a sample of 1,354 whole left valves for which valve height and hinge length could be measured, FSM5 and FSM3 together yielded 987 whole left valves for which both measurements could be made, while the Indigenous households yielded 9,786 (8,999 of which could yield measurements for both valve height and length to make shape determinations).

R² values for PSM3, FSM5, and the Indigenous households were 0.58, 0.49, and 0.61 respectively (Table 1b; Fig. 5). RSEs for valve height ranged from 12.59 mm to 15.18 mm, overall indicating a similar range of variation in the association of valve heights to hinge lengths as valve heights to weights. Unlike the valve height to weight regressions, however, the measured relationship between valve heights and hinge lengths remains consistent across shape categories (Table 1b; Fig. 5). The Indigenous households assemblage was split up between the same four shape categories used previously. The resulting R² values (0.56, 0.59, 0.62, and 0.66) deviated far less from the full assemblage value (0.61) than the deviations observed in the height to weight regressions which were substantively impacted when accounting for shape. Similarly, the slope associated with the valve height to hinge length regressions were consistent across shape classes, suggesting hinge length as a much more reliable and useful analog for reconstructing valve height compared to weight (Table 1b). Beyond reconstructing valve heights from hinge lengths, hinge lengths alone may represent a productive measurement across which to archaeological assemblages of eastern oysters.

Fig. 5.

Plots illustrating the relationship between valve height and valve length measurements made on whole left valves from twenty-three Indigenous households (top left), Pond SM3 (top middle), and Freedmen’s FSM5 (top right). The bottom four plots illustrate the relationship between valve height and valve length measurements made on whole left valves from twenty-three Indigenous households subdivided into four shape categories from roundest (middle two plots) to longest (bottom two plots).

In cases where height reconstructions might be appropriate (e.g., for comparing to published studies), we further evaluated how well-known valve height distributions could be reconstructed from their hinge lengths. To do so, we once again leveraged the Indigenous household assemblage (n = 9,786). Using the equation produced through the regression assessing the relationships between valve height and hinge length for the Indigenous household assemblage, we converted hinge lengths from these whole left valves into valve heights and compared the difference between the actual, measured heights of these left valves and the heights reconstructed from their hinge lengths (Table 1c; Fig. 6). The mean measured valve height for both actual measurements and reconstructed measurements was 63.02. The standard deviation shifted from 20.18 for actual measurements and 15.77 for reconstructed heights. Similarly, the CV dropped from 0.32 for actual measurements and 0.25 for reconstructed measurements. These metrics indicate that, while the mean itself may not be affected, there may be a slight shift in the overall distribution of reconstructed valve heights compared to actual measurements of valve heights. Indeed, a Mann-Whitney U test indicates a statistically significant difference between the two distributions, but a further polyserial correlation coefficient of exactly zero indicates that this difference is negligible. Further visual review of box plot and histogram comparisons suggests that, if anything, the reconstruction of valve heights from hinge lengths may result in the loss of the largest outliers (Fig. 6), but no significant difference in mean estimations or in the overall distribution. In general then, as far as gross comparative potential and reliability are concerned, hinge length may be a valuable alternative to valve height. However, we still suggest the largest value of this relationship is the ability to include measurements of hinge length to estimate valve height for broken valves of assemblages.

Fig. 6.

Plots illustrating differences whole left valve height measurements and height measurements reconstructed from hinge lengths. The top two plots compare whole valve measurements from Indigenous households with reconstructed height measurements for the same dataset based on hinge lengths. The middle two plots compare whole valve height measurements to reconstructed height measurements from the hinge lengths of broken shell. The bottom two plots compare whole valve height measurements to a combined dataset of all whole valve heights and reconstructed heights from hinge lengths of broken valves.

Given the complexity of drivers that contribute to eastern oyster size and shape, we would guess that regression formulae for reconstructing valve heights from hinge lengths is relative and should only be produced and applied for geographically (and likely temporally) distinct assemblages. That said, more modern studies are needed to better understand these relationships and to control for known environmental variables, a point that we expand upon in the Discussion section.

Archaeological eastern oyster morphometric studies are likely not substantively impacted by survivorship biases

With hinge lengths established as a useful measurement for comparing oyster sizes, we now turn to the question of survivorship or taphonomic biases. If there are substantive taphonomic biases, then we should observe substantive differences in hinge length measurements (and the distribution of those measurements) for hinge lengths measured for whole valves versus those measured for broken valves (where only hinges survived and whole valve height cannot be determined). If Wallis et al. (2025) conclusion is correct that in the existence of survivorship bias towards larger shells, then our analysis of hinges should show that measurements associated with broken valves (hinges only) should trend smaller, while hinge length measurements made on whole, unbroken valves should generally be larger. We find this is not the case.

Mann-Whitney U tests for hinge length comparisons between whole and broken valves found little evidence for substantive effects of survivorship biases (Table 1d; Fig. 7). For FSM5, there is no statistical difference between the distributions of hinge lengths for broken and whole valves. For both PSM3 and the Indigenous households assemblages, Mann-Whitney U tests did indicate statistical differences between whole and broken hinge lengths, but accompanying polyserial correlation coefficients indicate that these differences are negligible. One observable difference between hinge length distributions from whole versus broken valves is the mean hinge length for broken valves was consistently smaller than the mean length for whole valves (Table 1d). Thus, if any survivorship bias is to be cited, we offer the opposite conclusion from Wallis et al. (2025) that, in fact, smaller valves are more likely to be broken and thus under-accounted for. That said, for our cases, accounting for this potential bias does not substantively alter any interpretations or comparative potential.

Fig. 7.

Plots illustrating hinge lengths by taphonomic state (whole v. broken valves).

To more fully quantify the impact of any kind of taphonomic or survivorship bias, we conducted a series of reconstructions to assess how the inclusion of valve heights reconstructed from hinge lengths on broken shells would alter or change the interpretation of our assemblage level morphometric data. First, we compare measured whole valve heights to valve heights reconstructed from broken valve hinge lengths (Table 1c; Fig. 6). That is, we are essentially comparing the valve heights of our broken assemblage to the valve heights of our assemblage of whole shells. For these reconstructions we again relied on the Indigenous household assemblage. Recall that the mean valve height for whole valves from this assemblage is 63.02 mm. The average reconstructed valve height from broken valves is 65.89 mm. Once again the Mann-Whitney U test indicates a statistically significant difference in these distributions, but the calculated polyserial correlation coefficient of just − 0.09 suggests a negligible, even non-existent, difference.

For the next comparison, we combined the whole valve height measurements with the reconstructed valve heights from the broken valves. We then compared this combined assemblage to whole valve heights alone. This was meant to evaluate how the inclusion of reconstructed valve heights might alter our interpretations. The mean valve height for this combined assemblage was 64.35 mm, just 1.33 mm different from the whole valve assemblage alone. While the Mann-Whitney U test indicates statistically significant differences between the two, the polyserial correlation coefficient is even lower, nearing zero, at -0.04. In none of these tests have we been able to identify any substantive impacts of survivorship biases on the interpretation of shell size morphometrics, or rather these analyses are possible under various taphonomic processes that producing differential amounts of whole and broken valves.

Eastern oyster shape is a key variable that archaeologists must consider

One final result that emerged from these analyses is that oyster shape, the ratio of valve height to valve length, is a complex variable for which archaeologists must account for if they are going to produce robust interpretations (and comparisons) of eastern oyster morphometric data. We discuss the implications of this in the Discussion section but reiterate some of the key results here that illuminate the necessity of considering oyster shape.

In the analysis of weight to valve height, shape had a substantive impact on the nature of this relationship. Rounder oysters accumulate weight at a fundamentally different rate when compared to oysters that fall into the longer shape categories. This means that two left valves of the same height may have substantively different weights, while two oysters of the same weight may exhibit wildly different heights (Fig. 3).

Additionally, shape is of course tied to a complex array of ecological, environmental, and climatic variables. This reality makes using height as a sole indicator of oyster health (and stability) somewhat problematic, as height alone may be driven by a variety of factors and itself tied to key differences in shape. That is, two oysters of the exact same height may be significantly different in shape, driven by key differences in environmental stressors (including potentially humans). If we are to take this further, and begin to pose questions about human consumption, choice, and differential exploitation, shape must be accounted for. Indeed, modern artisanal oyster farmers claim a relationship between shape and taste due to biophysiological and biochemical differences¹⁹. While this may be difficult of course to test empirically, or to understand the role of “taste” in the archaeological past, what does stand to be better understood is the relationship between height, shape, and meat yield, which is potentially relevant to understanding human predation patterns, ecological drivers, and even potentially oyster health. We provide more discussion on these points below.

Discussion

The analyses presented above accomplish two goals: (1) provide justification, confirmation, and confidence in the validity and robustness of recent eastern oyster morphometric studies, in spite of the taphonomic vagaries of the archaeological record; and (2) illuminate areas that archaeologists need to devote more serious attention and consideration if we aim for our studies to continue to contribute meaningfully to addressing key modern challenges across the ecosystem and applied sciences. In this regard, we end with a discussion of what we find to be key considerations that must be made by archaeologists moving forward and some potential direction for productive future research in the area of shell morphometric studies.

The first consideration regards shape. Oyster paleobiology studies commonly use valve height measurements, but here valve length measurements were also collected to generate ratio values to evaluate overall oyster shape, which is sensitive to both environmental and human predation pressures^16,20–24. Overall shape is driven by a complex mixing of factors but in general intertidal oysters, those variably inundated, tend to be rounder while subtidal oysters, those permanently submerged, tend to be longer and skinnier^20,22–24. This is likely also related to the density of the oyster beds themselves and the substrate upon which beds rest, further complicating any straightforward interpretations. These shapes of course exist along a continuous spectrum, and disentangling the exact drivers remains difficult in the absence of explicit modern studies aimed at understanding the processes, mechanisms, drivers, and pressures of shape under different conditions. For instance, work on other mollusk species has offered evidence that tidal flow speed and currents, and human predation, can drive oyster shape^25,26. Indeed, little archaeological work has engaged with eastern oyster shape in substantive ways and archaeologists have yet to base any key interpretations of past socioecological dynamics on eastern oyster shape data.

Our second suggestion is the alignment of archaeologists’ considerations of health and stability with how these concepts are conceptualized and measured by modern practitioners and researchers. More broadly, beyond shape, we suggest that archaeologists should become more attuned towards the metrics, variables, and measures used by modern ecosystem and applied scientists to assess and compare the health and stability of oyster resources. How do ecologists, marine scientists, oyster farmers, or conservationists measure “health” within oyster populations? What analyses and observations can we provide from the archaeological record that align with indicators used by contemporary scientists to assess, for example, health, pressure, and stress?

In ecological, marine, conservation, and related studies focusing on oyster populations, health is examined at both the individual level and at the broader bed population level. At the individual level, researchers analyze growth-related metrics, including shell size²⁷, growth rates^28,29, and size-to-age relationships³⁰, informing on oyster responses to environmental conditions and energy allocation over time. Shell condition represents an important indicator for an individual’s health, acting as a proxy for experienced environmental and biological conditions³¹. Recent studies have examined shell condition through analyses of epibiont (i.e. parasitic) activity, including sponge boring and polychaete worms present on oyster shells^16,32. These species bore into oyster shell, disproportionately targeting larger individuals, and weakening structural integrity, altering growth, and increasing threat of the introduction of bacteria and secondary parasites, such as pea crabs^33,34 that can have detrimental effects to individuals and broader populations.

Beyond individual oysters, health is assessed at the bed-level through sampling strategies meant to evaluate disease prevalence, mortality rates, bed density and expansion rates, and spatial distribution across waterways³⁵. Recent work examines how diseases and parasites spread within oyster beds and how population health is determinate on their expansion. Through bed-wide samples, population-level studies consider the proportion of dead oysters, mortality rates over time, and how death occurred^36–38. These studies also consider how environmental conditions, including water temperature and salinity, influence the health of the bed and are identified through stable isotopic analyses and water sampling^28,30. Overall, individual- and population-level studies frame oyster health as an outcome of biological pressures and environmental conditions both within and external to the bed itself. This situates oyster bed compositions and health within dynamic systems, allowing research to better investigate the conditions driving empirical patterns.

To contribute in increasingly more valuable ways to interdisciplinary socioecological and paleoecological research, archaeologists should work to align our target metrics, analyses, and measurements with those that are most often employed, or leveraged, by ecologists, policymakers, and conservationists. This means that beyond size and shape alone, archaeologists have the opportunity to contribute via the collection of such datasets as epibiont activity (e.g., ³⁹, dead v. live collection determinations, and growth rates via sclerochronological methods (e.g., ⁸. The long-term temporal perspectives that archaeologists could provide on these trends have the potential to be of direct value to understand deep time patterns of oyster health, stability, and ecology.

Our final two suggestions are calls for archaeologists to work on developing stronger interpretative frameworks by which archaeological morphometric data can be evaluated and made meaningful, not just for archaeological questions, but for ecological questions, too. Both of these suggestions are a call for archaeologists to undertake critical modern observational and experimental studies to better tease apart the complex web of ecological and anthropogenic drivers that structure the morphometrics of both individual oysters and oyster populations.

Indeed, a small but key body of extant ethnographic and ethnoarchaeological work on the socioecological dynamics of shellfishing and shellfishery management among Aboriginal communities of the northern Australian coast^40,41, the Meriam islanders of the Torres Strait (e.g., ^42–45, coastal Mozambique fishers (e.g., ⁴⁶, the Solomon Islands (e.g., ⁴⁷, fishers of Kiribati (e.g., ^48–50; and the Indigenous Seri along the Gulf of California in northwestern Mexico (e.g., ⁵¹ has demonstrated that the relationships between shellfishers and shellfish resources and ecosystems are deeply complex. That is, archaeologists need to be better consider the complexities of decisions, routines, labor, timing, property regimes, management strategies, demographics, and energetics of shellfishing alongside the deeply complex ecological and environmental conditions within which particular shellfisheries are situated to more robustly, and accurately, interpret the archaeological record.

In this regard, our third suggestion is thus that archaeologists go out into these estuarine landscapes and quantitatively assess many of the assumptions that underly our archaeological interpretations but have yet to be empirically evaluated, akin to a recent study in Tampa Bay, Florida⁵². First, there is a need for very simple understandings of intra- and inter-oyster bed variation. What does the size distribution profile look like for one oyster bed versus another? How much variation do we expect within a single bed or between beds? Not just for size, but for shape? And then further, we must keenly measure and record ecological, environmental, and climatic conditions within which beds are located to better understand how this variation, both within and between beds, is driven by growth environments. What is the exact relationship between bed density and valve height, or shape? Between substrate and valve height, or shape? How does total time of inundation affect these morphometric variables? Further, and bringing into the fold human-decision making, what is the relationship between shell morphometrics and actual meat yield? We do know an allometric relationship exists between valve length and biomass for eastern oyster^53–55 which at the genus level carries genetic correlation⁵⁶, but as Powell et al. ⁵⁷ demonstrate the relationship between length and biomass can diverge significantly over salinity gradients and recognize the uncertainty in input of shape as a component of biomass. Dame (1972) further highlights the impact of intertidal vs. subtidal environments on such relationships. Finally, and again more broadly, the issue arises with an absence in archaeological studies directly and intensively engaging with modern fisheries literature. Having a handle on these relationships will facilitate more robust archaeological interpretations that can more effectively be leveraged to effectively contribute to inter- and trans-disciplinary conversations.

Our fourth and final suggestions follow in the same nature as our third but suggests explicitly the design of modern experiments to better understand the impacts of different anthropogenic pressures at small scales, scales relevant to the kinds of harvesting practices that would have predominated past human-oyster interactions. How long does it take for a community to deplete an oyster bed? At what scale can a single community realistically impact? Are these impacts at these small scales indeed reflected in temporal changes to valve heights? What kinds of impacts might different harvesting practices have, and how might we see these different practices in the archaeological record? For instance, what impacts does mass harvesting have on an oyster bed and on the temporal trends in oyster size and shape that are subsequently deposited at archaeological sites? Alternatively, what is the archaeological signature of harvesting practices that discriminate between sizes or shapes, collecting only individuals that fall into preconceived folk categories? We know that industrialization impacted oysters at scale (e.g., ^6,7, but we currently lack a formal understanding of how the smaller scale anthropogenic pressures that defined past harvesting practices may have affected oyster beds at more local scales (and thus how they may have affected the archaeological signatures of these practices and their impacts). These questions are simple, yet their impact on interpreting the archaeological record of shell morphometrics would be invaluable and can be explored via the implementation of simple experimental, longitudinal oyster collecting studies.

None of the suggestions or analyses presented here are overly complex or complicated. In fact, we would argue that the points made throughout this paper are simple. That said, we would argue that these simple points have a disproportionate impact on the value and utility of archaeological studies. Much of what we have done in this paper are simple validations, empirical exercises, or simple illuminations of considerations that are otherwise foundational to any archaeological research. We have aimed simply to encourage more scientific, empirical approaches to archaeological shell morphometric studies. In doing so, we hope to have provided some productive pathways to not just enhance our archaeological research and creating more robust narratives of the past, but to further fortify the value and role of archaeology to the broader socioecological sciences and to address some of the most critical ecological challenges facing coastal communities today.

Methods

Archaeological contexts

Three different archaeological contexts yielded oyster shell assemblages that were used in this study. These contexts are meant to span a varying range of taphonomic conditions to which archaeological oyster shell was subject, specifically Indigenous Household contexts, Indigenous non-domestic contexts, and a post-emancipation Freedmen’s household midden. The assemblage described as “Indigenous Households” was excavated in the summer of 2023. This assemblage was produced via excavations into 23 discrete shell midden deposits associated with domestic refuse from the Indigenous Guale town of Middle Place on Ossabaw Island, Georgia. Excavation locations targeted distinct household deposits from across the large town, which dates to between c. AD 1100 and 1525, with each household deposit accumulating over an estimated 30–40 year span⁵⁸.Each test excavation, placed in the thickest portion of the midden as identified through systematic probing, measured 50 cm x 50 cm and was dug in arbitrary 10-cm levels until sterile subsoil was reached. All midden matrix was screened through 1/8” screen and all artifacts and shell material were kept for later analyses. This assemblage yielded an eastern oyster MNI (determined from left valves) of 9,791 whole individuals and 8,505 broken specimens (umbos/hinges only). This assemblage was not used in weight calculations as individual shell weights were not collected at the time of measurement collection for this dataset.

The assemblage described as Pond SM3, or PSM3, is also from the Indigenous Guale town at the Middle Place site. PSM3 represents an extensive sheet midden that spans the edge of a freshwater pond that bounds the western edge of the town. PSM3 is a 50 cm x 50 cm test excavation placed into this sheet midden. Unlike the domestic middens which were c. 8–15 m in diameter and 30–40 cm in depth, PSM3 is part of a 100 m long midden with depths up to 70 cm or more. Additionally, radiocarbon dating of PSM3 indicates that this portion of the sheet midden accumulated rapidly, with a modelled range, from 4 sequential AMS radiocarbon dates from 10 cm-levels 2–5, to ca. AD 1385–1430 (68% probability)⁵⁸. This assemblage is comprised of 1,101 whole eastern oyster left valves that could be measured and weighed, and 894 broken left valves (hinges only).

The assemblage referred to as Freedmen’s SM5, or FSM5 (and FSM3), is a shell midden that is similar in shape, structure, deposition, and use to the Indigenous middens, but was produced by freed Black communities after Emancipation and accumulated between c. AD 1865 and 1945, based on historical documents referencing various occupation durations of the island post-emancipation⁵⁹. Thus, this assemblage is at least 300–800 years younger than the Indigenous Household middens and PSM3. With regard to duration, is likely these deposits represent a more similar, shortened duration of use to the Indigenous contexts, but without results from ongoing AMS radiocarbon dating and materials analyses, we use the aforementioned age range. This assemblage is comprised of 500 whole eastern oyster left valves from FSM5 that could be measured and weighed, and 487 from FSM3 that were combined with FSM5 for valve height and hinge length comparisons. There are 371 broken left valves (hinges only) from FSM5 and 701 from FSM3. This does represent differential relationship between broken and whole shells between Indigenous and freed Black deposits. Even with the more conservative range of deposition for the freed Black deposits at ca. 80 years, the rate of deposition between the three contexts is likely not a significant factor in shell survival and breakage. There could be a number of taphonomic processes, both chemical and physical, anthropogenic and non-anthropogenic, to produce this difference. Visual comparative assessments of the three assemblages suggest similarly high rates of preservation. In the analyses for each individual context, PSM3 and FSM5, we empirically assess the impact of survivorship biases.

Eastern oyster morphometrics

To assess oyster morphology through time, a series of measurements were collected on all left oyster valves. Measurements for oyster height (LVH), usually the largest dimension from the hinge to the growing edge, and length (LVL), usually the smaller dimension perpendicular to height, were recorded in millimeters following standard methods outlined in Lulewicz et al. (2017). Hinge length (LHL), which is parallel to valve length, was also measured and is the width of the umbo. All oysters < 35 mm usually represent juveniles/spat, oysters within or around their first year of growth, and were excluded from this dataset for a few compounding reasons following standard modern, historical, and archaeological methods to reduce potential preservation biases in the archaeological record, differences between modern and archaeological assemblages and collection practices, and the inconsistency in juveniles representing patterns observed in larger individuals due to the absence of ontological pressures on shape and size to ensure data more directly reflect growth environment or human selection (a target of many archaeological applications of oyster size) rather than the genetic drivers of physiological traits^{7,22,52,60–65}.

Statistical comparisons

All statistical analyses of measurements were conducted in R ⁶⁶. Since none of the datasets are normally distributed, all comparisons between broken and whole oysters were made via Mann-Whitney U tests. These tests were used to determine whether the distributions of measurements belonging to different categories of shell (e.g., whole or broke) were statistically distinct from one another. To further aid in the interpretation of these analyses, an accompanying polyserial correlation used to evaluate the actual correlation between the continuous variable (measurements) and the categorical variable (taphonomic state of whole or broken). In this way, we do not rely solely on the Mann-Whitney U tests but also employ polyserial correlations to contextualize any statistical significance suggested through the Mann-Whitney U tests. This is important, because in all cases, the polyserial correlations indicated that although some distributions were statistically distinct, this difference was always negligible. Where appropriate, statistical significance was noted at three different levels P ≤ 0.05, ≤ 0.01, and ≤ 0.001. Polyserial correlation metrics range from − 1 to 1, where a value of 0 indicates weak (or nonexistent) association between the continuous variable and the categorical variable.

Author contributions

IHL and JHL conceptualized the study and prepared the original draft of text. JHL conducted all formal analyses. MPK collected data for the study and contributed text to the original draft. All authors reviewed and edited the manuscript.

Data availability

All data is archived and available online at https://zenodo.org/records/19825312 .

Declarations

Competing interests

The authors declare no competing interests.