On the evolution of limestone-tempered pottery in the American Midwest: an experimental assessment of vessel weight and its relationship to other functional/mechanical properties

ABSTRACT

During the Middle and Late Woodland periods in the American Midwest some small-scale societies transitioned from grit to limestone as the primary clay temper. Limestone offers experimentally demonstrated benefits to vessel manufacture, including decreased wall thickness, but given the society-wide changes in mobility and exchange that also occurred, we investigated whether the use of limestone temper resulted in a different vessel weight relative to an analogous grit-tempered vessel. Our analyses demonstrated a significant difference: post-firing, limestone-tempered vessels were 4.5% lighter than grit-tempered ones. The combination of reduced weight and other benefits could help explain why limestone became the predominant temper type throughout much of the Midwest. These issues have direct analogs in the biological realm, and we use three concepts from evolutionary biology – modularity, mosaic evolution, and constraint – to investigate the role limestone temper played not only in vessel weight but also in other aspects of vessel production and use.

GRAPHICAL ABSTRACT

KEYWORDS:

• Constraint
• experimental archaeology
• hitchhiking
• limestone temper
• pottery
• mosaic evolution

Introduction

The appearance in the archaeological record of a new variant of a cultural trait – such as a different kind of temper in ceramic cooking vessels – could potentially be the result of a number of processes (Bettinger and Eerkens 1999; Cochrane, Rieth, and Dickinson 2013; Eerkens and Lipo 2005, 2007, 2014; Lycett 2015; Mesoudi and O’Brien 2008a, 2008b; O’Brien and Lyman 2002; O’Brien and Shennan 2010; O’Brien, Lyman, and Leonard 1998; Walsh, Riede, and O’Neill 2019). For example, a migrating people could introduce a variant into a region; an established group could adopt it from neighbors; or individuals on their own might invent a new variant that eventually becomes fixed in their cultural repertoire. The question then arises as to why prehistoric people adopted a new variant. Did it function better than other, established variants? Were raw materials for the new variant easier to procure? Were certain materials easier to work with than others? Did a favorite pottery mentor positively influence students such that they were biased toward that mentor’s preferred variant? Or was adoption of a variant a random occurrence, mere happenstance, an accident of history? To complicate matters, these questions are not mutually exclusive, meaning that a combination of factors could have played a role in the appearance of a new variant.

This was certainly the case in the American Midwest,¹ where new ceramic variants appeared constantly throughout the Middle Woodland (200 B.C.–A.D. 450) and early Late Woodland (A.D. 450–750) periods. Prior to that, vessels were coarse and thick walled (9–18 mm), with shapes resembling flat-bottomed tubs, barrels, or flowerpots (Farnsworth and Asch 1986). The dominant temper was a coarse, dark igneous or mafic material, with particles up to 5 mm in diameter.² Beginning around 400 B.C. and in some areas of the Midwest overlapping with the thick ware, a new, thinner, hard-paste ware became predominant in the central and upper Mississippi Valley. Vessels were tempered with large amounts of sand and/or finely crushed igneous rock, and in some areas, other minerals such chert were significant additions (Begg and Riley 1990; Farnsworth and Asch 1986; Munson 1982, 1986). In some parts of the Midwest east of the Mississippi River, crushed sherds (grog), not minerals, was the primary temper (Emerson 1986). We suspect that by 200 B.C., there were literally dozens of local ceramic industries spread across the Midwest (O’Brien et al. 1994), each producing its own wares but with considerable overlap in terms of vessel size and shape, wall thickness, and the kinds of temper used in clay pastes.

With respect to those traditions, one trend that has long been noted is a gradual shift in temper from grit to limestone that began in some places as early as ca. A.D. 200 (Braun 1983, 1987; Lepper 2005; Morgan 1985; O’Brien 1996; Prufer 1968; Rutkoski, Bebber, and Eren 2019). The effects of that change have been studied for decades (e.g. Braun 1985a, 1987; Bronitsky and Hamer 1986; Feathers 2006; Hoard et al. 1995; O’Brien et al. 1994; Rutkoski, Bebber, and Eren 2019), with most studies focusing on functional and technological aspects involving the interplay between temper and wall thickness with respect to thermal conductivity, resistance to crack propagation during production and use of pots, and vessel failure (Bebber 2017; Bebber and Eren 2019; Bebber, Spurlock, and Fisch 2018; Braun 1983; Bronitsky and Hamer 1986; Hoard et al. 1995; Rutkoski, Bebber, and Eren 2019). One aspect that has not been studied is the correlation, if any, between type of temper and vessel weight. Did limestone temper, for example, allow for production of lighter-weight pots than grit temper? We set up an experiment to examine that issue. Although the results are interesting, looked at in isolation they cannot address whether weight was a target of selection during manufacture and firing or part of a coinciding technological optima with, say, crack resistance or thermal conductivity. Alternatively, was weight entirely incidental to sought-after features involving the change from grit to limestone as a temper? These questions have direct analogs in the biological realm, and we use three concepts from evolutionary biology – modularity, mosaic evolution, and constraint – to investigate the role limestone temper played not only in vessel weight but also in other aspects of vessel production and use.

Why temper?

Temper is an additive incorporated into clay during the formation of a ceramic vessel (Rice 2015). Given that temper acquisition and preparation are energy expensive, there must be selective benefits that outweigh the utility costs of adding temper to clay (Bebber 2017). One benefit is in retarding crack propagation in unfired clays, which are characterized by weak bonds among particles. Most clays shrink an average of 15% or more during drying, and uneven drying along the surface of the vessel or between the interior and/or exterior of the vessel wall causes uneven rates of shrinkage, which usually will lead to cracking. The addition of temper increases porosity and promotes drying that is more rapid and even, which reduces stresses (Braun 1982). Temper also decreases the stickiness of wet clay, allowing the potter to produce a greater range of sizes and shapes (Herbert 2008).

With respect to grain size, there is a balancing act between crack initiation and crack propagation (Shepard 1965). Everything being equal, the use-life of a vessel can be increased by increasing the average size of temper particles. Fine-grained fired pastes have higher resistance to fracture initiation than coarse-paste pastes but lower resistance to crack propagation. Once fractures begin, they propagate rapidly, which results in the mechanical failure of the vessel wall. Conversely, particles much larger than the bonded clay particles in a fired paste have less resistance to fracture initiation but increased resistance to crack propagation. Once fractures occur in coarser pastes, they will be truncated by a coarse inclusion and, as a result, will have a lower tendency to propagate during an episode of mechanical stress (Braun 1982).

The use of temper can be thought of as a kind of specialized knowledge that demonstrates the ability of prehistoric potters to understand how it could (1) affect vessel size and shape, improve drying, prevent warping, and eliminate blowouts during firing (Bebber 2017; Bebber, Spurlock, and Fisch 2018; Bronitsky and Hamer 1986; Feathers 1989) and (2) yield performance-based benefits associated with post-firing vessel use. Midwestern potters undoubtedly knew from experience that varying the type, amount, and size of temper particles created different kinds of vessels, each of which had certain performance characteristics. Midwestern potters, like their contemporaries elsewhere, must have constantly wrestled with balancing the technological aspects of pottery manufacture with vessel performance (Braun 1982, 1985a, 1987). Could larger pots be made without thickening the vessel walls to a point where they no longer functioned effectively for cooking? Could the height of a vessel be adjusted relative to its girth to create desired changes in volume?

Limestone vs. grit as a temper

The gradual shift from grit to limestone temper in some midwestern locales began with the production of small, thin-walled (Hopewell) jars and bowls that were used for nonfood-related purposes. The coincidence in timing suggests the technology needed to produce larger, thinner-walled cooking vessels grew out of that specialized technology (Braun 1983, 1987), although it was not immediately transferred from one vessel kind to another. For several hundred years, there actually was an increase in wall thickness as potters learned to transfer the limestone-based technology from the small vessels to larger cooking containers. For example, ca. 200 B.C., the mean wall thickness of cooking vessels from three river valleys in western Illinois was approximately 7.7 mm, which increased to a maximum of approximately 8.2 mm by the beginning of the Christian era (Braun 1983, 1985b). By ca. A.D. 50, wall thickness declined monotonically at a rapid rate until ca. A.D. 300 (mean wall thickness of approximately 6.1 mm), at which point it decreased at a slower rate until ca. A.D. 550 (mean wall thickness of approximately 5.8 mm), when the rate of decrease again accelerated. By A.D. 750, mean wall thickness was approximately 5.3 mm – well over a 2 mm decrease from 200 B.C. A similar trend was found in eastern Missouri (O’Brien 1996), although wall thickness remained higher than in western Illinois. A decrease of 2 mm may not appear to be important, but if current experimental data is replicated (e.g. Bowen and Harry 2019), then this indicates that it can have a substantial impact on resisting thermal shock and thus on vessel failure.

The effect of wall thickness on heating effectiveness has been studied experimentally, but the results are mixed. Braun (1983) argued that thinner walls promote more rapid heating of the vessel’s contents compared to thicker variants, but Skibo, Schiffer, and Reid (1989) found that the contents of thicker-walled, sand-tempered vessels boiled in less time compared to the contents in thinner-walled, organic-tempered vessels. More-recent studies by Bowen and Harry (2019) found that the thickness of vessel walls does not have a significant effect on content heating rate, but it greatly affects thermal-shock resistance. Wilcox et al. (2022) found that thick-walled (18 mm) vessels could not reach the boiling point without interior surface treatment.

The addition of limestone to the repertoire of tempers in the Midwest is an interesting phenomenon in terms of the ability of prehistoric potters to work with problematic tempers, the most notorious of which are limestone and shell, both carbonate materials. Given certain firing conditions, vessels tempered with calcium carbonate are subject to failure through lime spalling (Laird and Worcester 1956), which occurs as an end process of heating calcium carbonate (CaCO₃), which is then converted to calcium oxide (CaO) and carbon dioxide (CO₂). After the gas dissipates, and upon cooling, the hygroscopic CaO hydrates to form calcium hydroxide Ca(OH)₂. As the volume of the inclusion expands, stresses cause popping and cracking.

In terms of advantages, limestone’s chemical composition greatly increases the workability of clay over that gained through using other tempers as a result of the flocculation that takes place as the calcium ions interact with the moist clay. This process causes the clay body to become more workable, which in turn facilitates the creation of larger and thinner vessels (Bebber 2017; Bebber, Spurlock, and Fisch 2018; Feathers 1990, 2006; Hoard et al. 1995; Lawrence 1972; Rutkoski, Bebber, and Eren 2019; Stimmell, Heimann, and Hancock 1982). A recent experiment by Bebber and Eren (2019), however, found just the opposite: that carbonate tempers did not make clay bodies more workable and in fact were consistently, sometimes significantly, inferior to silicate grit-tempered clay bodies. There’s a caveat here, however: All clays are not the same. Bebber and Eren used locally sourced glacial clay from a gravel mine in northeastern Ohio to assess the interactions “between the illite clays and temper types that are documented in the archaeological record” (243). As opposed to illite clays, mineralogical analysis of bottomland clays from the Mississippi River and its major tributaries such as the Illinois shows that they are high in montmorillonite (Holmes and Hearn 1942), which, as Stimmell, Heimann, and Hancock (1982) point out, is typically unsuitable for use in low-fired vessels because of its extreme plasticity and high shrink/swell ratio. Temper can counter this effect, but only if used in suitable quantities (Stimmell, Heimann, and Hancock 1982).

A series of studies on thermal properties of grit and limestone has demonstrated that the improved performance of limestone-tempered pottery both pre- and post-firing likely played a significant role in the transition from grit as a temper (Bebber 2017; Bebber, Spurlock, and Fisch 2018; Hoard et al. 1995; Lepper 2005; O’Brien et al. 1994; Pollack, Henderson, and Raymer 2008; Rutkoski, Bebber, and Eren 2019). For example, experimental results on post-firing vessel strength show that limestone-tempered samples are more resistant to mechanical stress compared to grit-tempered samples (Bebber 2017; Bebber, Spurlock, and Fisch 2018; Hoard et al. 1995). Experiments have also suggested that limestone-tempered pottery is more resistant to thermal stress than grit-tempered pottery, as the carbonate additive has a thermal expansion rate similar to that of clay (Hoard et al. 1995; O’Brien et al. 1994).

Despite the technological advantages that have been documented for limestone temper, several proposed performance-related advantages still need testing. One example is Braun’s (1983, 118) proposal that thinner vessel walls expedited cooking – a proposal based on principles from materials science (Van Vlack 1964):

The thermal conductivity of a vessel wall is its ability to conduct heat from one face to the other, determining the rate at which a change in temperature at one face will be transmitted to the other. Other things being equal, the thinner a wall, the higher its heat conductivity.

As Bowen and Harry (2019) point out, after Braun’s work the association of thinner-walled vessels with both greater heating efficiency and thermal-shock resistance became gospel (e.g. Eerkens 2003; Frink and Harry 2008; Hart 2012; Hoard et al. 1995; Kooiman 2016; O’Brien et al. 1994; Schiffer and Skibo 1987, 1997; Skibo 2013; Skibo and Blinman 1999; Skibo, Malainey, and Kooiman 2016).

O’Brien (1987, 1996) advanced the issue further, proposing that the coincidence of thinner walls with the rapid increase in the economic importance of seeds of native annuals such as chenopodium, sumpweed, and maygrass suggested that vessel walls were being made thinner to allow more rapid heat conduction. The advance in technology in turn may have affected population growth rates. If early Late Woodland groups prepared food, especially starchy gruels and porridges, more efficiently, perhaps infants could have been weaned earlier by substituting these foods for mother’s milk (Buikstra, Koningsberg, and Bullington 1986). Any resulting decrease in the lactation period might have allowed the birth rate to rise slightly, if only by one additional offspring per childbearing woman. Buikstra, Koningsberg, and Bullington (1986) did, in fact, note an increase in fertility in the western Illinois River valley during the period A.D. 450–750. Given the decrease in wall thickness and the concomitant rise not only in the importance of native annuals but also in fertility (Asch, Farnsworth, and Asch 1979; Braun 1977, 1985b; Buikstra, Koningsberg, and Bullington 1986; O’Brien 1996; O’Brien et al. 1994), the proposal for expedited cooking is reasonable. Further evidence of the role played by vessels in cooking the seeds of native annuals is provided by hundreds of sherds from a site in eastern Missouri, which contain thick (often up to 1 mm) carbonized residues on the interiors. Similar residues occur on Middle Woodland and early Late Woodland sherds from western Illinois.

The problem is, until very recently, the proposal had never been adequately tested (Bebber and Eren 2019), and the one exception produced equivocal results (Bowen and Harry 2019). In that experiment, modern untempered vessel replicas, some with thin walls and some with thick walls, were fired at 900°C. Despite their equivocal results – there was, at best, a 3% difference in heat conductivity – Bowen and Harry made an excellent point:

While heat may be transmitted into the pot more quickly, it may also be transmitted more quickly out of the pot. Thus, while we might expect the contents of thinner vessels to heat up more quickly, they may also require greater amounts of fuel over the long run to maintain a simmer or boil. (Bowen and Harry 2019, 261)

Vessel weight: experimental evidence

Mobility and exchange are important elements in the lives of all small-scale societies (e.g. Barnard and Wendrich 2008; Sellet, Greaves, and Yu 2006), and society-wide changes in these elements across much of the Midwest are well documented (e.g. Bareis and Porter 1984; Braun 1977, 1983, 1985b, 1986, 1987, 1991; Braun and Plog 1982; Chapman 1980; Charles 1992; Emerson, McElrath, and Fortier 2000; Lepper 2005; Mehrer 1995; Morse and Morse 1983; Muller 1986; O’Brien 1987, 1996; Roper 1979). As such, decreasing tool-transport costs by reducing vessel weight could have increased the efficiency of mobility and exchange at multiple scales. Human energy expenditure would have decreased as a result of reduced weight during either long-distance travel or repeated short trips measured at the scale of meters (e.g. Surovell and O’Brien 2016). Or, in terms of exchange, more vessels could potentially be transported if each weighed less. Finally, by reducing the weight of containers, more carrying weight could be dedicated to vessel contents (e.g. seeds and water) rather than to the vessel itself, increasing task efficiency (Surovell and O’Brien 2016).

Ideally, examination of selective pressures relative to changing technological and functional characteristics of prehistoric vessels would employ the actual archaeological materials, but prehistoric pottery is a relatively low-fired, porous ware that no longer retains its original properties. Although sherds can serve as accurate measures of some dimensions, such as vessel-wall thickness and vessel size (e.g. Braun 1977, 1985b; O’Brien et al. 1994), and tempering agents can be identified, postdepositional processes (Schiffer 1983, 1987) such as leaching of temper particles by groundwater and rehydration of clay minerals have led to compositional change and therefore reduction in strength. Thus, strength measurements conducted on sherds or pots from archaeological contexts are residual measurements (Braun 1983; Mabry et al. 1988; Steponaitis 1983) and thus are not good proxies for original properties.

To circumvent this problem, the experimental work discussed below was carried out using modern ceramic replicas manufactured to resemble in several important respects prehistoric materials. By using replicas of past technologies, or aspects of past technologies, experimental archaeologists can determine how one variant performs relative to another in a variety of tasks (Bronitsky 1986; Eren et al. 2016; Herbert 2008; Johnson et al. 1988; Schiffer and Skibo 1987, 1989; Skibo, Schiffer, and Reid 1989; Vaz Pinto et al. 1987). If a functional advantage can be demonstrated experimentally, then function can be invoked as a possible contributing factor to a variant’s occurrence in the archaeological record. Sometimes multiple advantages, or coinciding optima, can be experimentally demonstrated.

Materials and methods

For the weight experiment, we processed locally sourced glacial clay via levigation, which entailed soaking the clay in water for three days and then sieving to remove small pieces of gravel and sand using geological bucket screens ranging in mesh size from 1.0 to 6.0 mm. After the clay was screened, it was stirred and levigated for 48 h. We removed the top layer of water, and the clay mixture levigated for another 48 h. After the final round, a clay slurry remained within the bucket, which we poured onto a fabric-covered wooden frame, where the clay slowly dried over the course of 72 h. Once a workable consistency was achieved, the clay was enclosed in sealed buckets to maintain a consistent moisture level.

For temper, we purchased grit (quartz) and carbonate (limestone), crushed both, and sorted them into three groups by size (< 1 mm, 1–2 mm, and 2–4 mm), which are the size classes and percentages of each class found via petrographic analysis of Late Woodland sherds from the Scioto River valley, Ohio (Bebber 2017). The tempers were then mixed and added to the wet clay at 28% by volume, which is the amount present in archaeological samples from southern Ohio (Bebber 2017). We added and wedged the temper into the clay body to ensure it was evenly distributed throughout the matrix and that air bubbles were removed.

Once mixed, the tempered clay was measured into six different weight (in kilograms) sets per temper type: 0.0625, 0.125, 0.250, 0.500, 1.000, and 2.000. There were five samples in each set (n = 60). We condensed the clay into spherical shapes, then wrapped and bagged them in plastic to prevent premature drying. A.M. produced the clay pots (Figure 1) using the pinch technique. We left the pots to gradually dry over the course of several weeks. The slow drying process prevented uneven drying and crack formation. The pots were then fired in a Skutt electric kiln at a firing temperature of 773°C and held at that temperature for three hours, followed by a natural cool down. Ethnoarchaeological studies show that the temperature range for low-fired vessels fired in an open “bonfire” setup can range from 550° to 900° C (Gosselain and Livingstone-Smith 1995; Skibo, Schiffer, and Reid 1989). The pots for this study were fired in a Skutt electric kiln to a mid-range temperature of 773^o C (cone 016) and held at that temperature for 30 min. After firing, the pots were immediately weighed upon removal from the kiln, and their measurements recorded (see Supplementary Online Materials).

Figure 1. Examples of pottery vessels produced for the experiment: (a) grit-tempered bowls pre-firing; (b) limestone-tempered bowls pre-firing; (c) grit-tempered bowls post-firing; (d) limestone-tempered bowls post-firing.

Sixty clay pots, each with one of two tempers, were measured on three scales: temper (binary variable, 0 = grit; 1 = limestone), mass (continuous variable, pre-fired mass of sample, in kg), and loss (continuous variable, proportion of total mass lost in firing). Variation in loss due to temper was controlled (n = m = 30 of each temper type). Mass represents a potential covariate, ranging from 0.0625 kg to 2.000 kg. It is also tightly controlled; that is, the distributions of mass within each of the two temper categories are identical.

Two regression models to predict loss of mass with “temper” as a binary variable (zero-one) were estimated: $\mathit{M}odel1(\mathrm{two}\text{-}\mathrm{group}\mathrm{comparison}):\mathrm{Expected}(\mathrm{loss})=b_0+b_1(\mathrm{temper})$ $\mathit{M}odel2(\mathrm{potential}\mathrm{covariate}\mathrm{added}):\mathrm{Expected}(\mathrm{loss})=b_0+b_1(\mathrm{temper})+b_2(\mathrm{mass})$

The Model 1 regression is equivalent to the pooled t-test in every way. That is, Model 1 reproduces exactly the two-sample, homoscedasticity-assumed, normal-theory test of hypothesis and parameter estimation. Thus, b₁ is the difference between the means, and the standard error of this slope (s_b) would be exactly the standard error of the difference between the means. Also, in either format, the estimated parameters (either the estimation of β₁ or µ_G – µ_L), the confidence intervals for either, or the two calculated t-values would be identical. However, the regression format was chosen to conform conceptually with Model 2, which adds a continuous regressor as a potential covariate. Model 2 is the simple analysis of covariance, without an interaction term (which was nonsignificant here, and not shown).

Results

Model 2 is no improvement over Model 1. The adjusted R-square values and therefore the standard deviations from regression in both models are virtually identical. The absolute value of the t-statistic for inclusion of the covariate is much less than 1.00, and the confidence interval for β₂ is approximately symmetric about zero (Table 1). Therefore, Model 1, the simple two-group comparison, is necessary and sufficient, and this is the chosen model.

Table 1. Estimates of coefficients.

Model 1: (adjR^2 = .944; sy:Model = .0055)
Variable	bj	sb	t58	p<	95% confidence interval
Intercept	.2111				pr (.2091 < β1 < .2131) = .95
Temper	.0447	.0014	31.68	.000	pr (.0419 < β1 < .0475) = .95
Model 2: (adjR^2 = .944; sy:Model = .0055)
Variable	bj	sb	t57	p<	95% confidence interval
Intercept	.2109				pr (.2085 < β1 < .2134) = .95
Temper	.0447	.0014	31.43	.000	pr (.0419 < β1 < .0476) = .95
Mass	.000028	.000107	0.26	.794	pr (–.0018 < β1 < .0024) = .95

There are two inferences, and they are unambiguous. The primary result is that the estimate for the difference between means (sample estimate of µ_L – µ_G, or of β₁) is highly significant (t = 31.68, Table 1) and exceptionally strong (Cohen’s d effect size = 8.19), with the limestone and the grit sample means more than eight common standard deviations apart. The second inference is that there is no measurable effect of variation of pre-fired mass on proportion of weight loss. In other words, with the mass covariate forced into the model (Model 2), the adjusted mean difference (b₁ = 0.0447) is identical to three digits to the unadjusted 2-group experimental difference.

In short, our results support the hypothesis that midwestern potters’ use of limestone temper would have reduced the post-firing weight of a ceramic vessel relative to that of a silicate grit-tempered one. On average, the difference in weight is ∼4.5%. We suggest three corollaries emerge from this result. First, differences in energy acquisition and expenditure were likely key catalysts of past cultural diversity, innovation, and artifact design (Kuhn and Miller 2015; Mika et al. 2022; Stiner and Kuhn 2016; Surovell 2009; Torrence 1989). To paraphrase Mika et al. (2022), when considered over the entire use-life of a ceramic vessel, differences in vessel weight could potentially benefit energy or time stores. Over hundreds or thousands of carrying events, minor but significant differences in vessel weight could save individuals many thousands of kilojoules or hours of time. With respect to the 4.5% vessel-weight difference reported here, either the absolute weight of the Woodland vessel would have been reduced or more transport weight could have been dedicated to vessel contents by using limestone over grit.

Second, we suspect that the 4.5% reduction is a minimum estimate during the grit- to limestone-temper transition because limestone vessels also tended to have thinner walls, which would have further reduced vessel weight. Thinner walls perhaps emerged from increases in society-wide vessel-construction skill. If so, then these two independent production strategies – the use of limestone as a temper and production of thinner walls – might have functioned in concert to minimize vessel carrying weight, further augmenting energy or time benefits.

Third, bearing in mind that the use of limestone temper imparts other experimentally demonstrated benefits to ceramic vessels in terms of resistance to mechanical and thermal stresses, it is likely that weight reduction was not the sole target of selection. In this case, breakage resistance and weight reduction were coinciding optima (Brantingham and Kuhn 2001; Lycett and Eren 2013), although in some cases the weight-reduction benefits stemming from limestone might have been entirely incidental to the sought-after breakage-resistance features (Bebber, Spurlock, and Fisch 2018). In this regard, we note that the level of sedentism varied widely throughout the Middle and early Late Woodland periods, and researchers are not in agreement as to how frequently pottery was moved. Indeed, research focused on pottery use by residentially mobile groups argues that Woodland pots were used with relatively expedient hearths (Beck, Hill, and Khandelwal 2022) and that residentially mobile groups prioritize weight reduction in ceramic manufacture (Beck 2009). In contrast, other scholars (Gibbs 2012) assert that cooking and storage vessels were rarely if ever moved and thus there would be no benefit in weight reduction. Thus, while “coinciding optima” may potentially help explain the widespread occurrence, innovation, or adoption of limestone temper across the Midwest, in particular contexts its weight-reduction benefits need not have automatically applied. We examine this critical issue below in a wider perspective.

Discussion

Our view of the evolution of ceramic vessels parallels that of Neff (1992, 173):

The ceramic attributes that are the most likely targets of selection are those that affect a pot’s ability to perform the basic cooking, carrying, and storage needs of the social unit within which it is made and used. Selective pressures on pottery-making may stem from changes in the uses to which vessels are put or from key technological innovations that alter the relative efficiencies of a whole complex of ceramic technological practices.

As important as formal properties of a vessel are, in that they provide windows into processes of technological change, by themselves they do not tell us much about human behavior. To address “why-type” questions requires that those mechanical properties be translated into performance characteristics, such as ease of manufacture, heating effectiveness, impact resistance, thermal-shock resistance, abrasion resistance, and the focus here, portability (Schiffer and Skibo 1987).

The traits that together created a Woodland pot – type of clay, type of temper, and firing temperature – can be seen as parts of recipes, which include the materials (ingredients) required to make a vessel and the behavioral rules (instructions) required to construct and use it (Lyman and O’Brien 2003; Mesoudi and O’Brien 2008a; O’Brien et al. 2010). Mesoudi and O’Brien (2008a) suggest that recipes can best be thought of as being hierarchically structured, with the finished pot comprising several behavioral subroutines – for example, preparation of material, production, and use – each of which in turn can be subdivided into a sequence of constituent lower-level actions required to complete each subroutine. Hierarchical organization is advantageous when subunits are repeated in one or more recipes, as repeated subunits must be learned only once and so reduce the overall costs of learning. This is especially the case when learning is associated with some degree of both error and cost. If errors are localized, they are far less likely to lead to failure. Given the prevalence of hierarchical learning, we would expect to find evidence of repeated subunits in the archaeological record – for example, where the same technological component is repeated in a single tool (Mesoudi and O’Brien 2008a).

Modularity

Modularity, a concept that has a long history in biology (e.g. Bolker 2000; Bonner 1988; Lewontin 1978; Riedl 1977; Von Dassow and Munro 1999; Wagner 2001; Winther 2001), refers to an organization of an entity, whether an organism or a clay pot, that favors evolvability by allowing one module to change without interfering with the rest of the entity:

Simultaneous random changes in many parts of a highly integrated structure are not likely to improve its function, as the chance improvement of one part will almost always be swamped by deleterious effects in many other parts. But if the parts are variationally independent, selection gets the chance to tune them one at a time, thereby improving the probability of finding improvements. (Hansen 2003, 84)

Modularity has played a small but growing role in archaeology, typically with respect to projectile points (e.g. Buchanan et al. 2015; González-José and Charlin 2012; O’Brien et al. 2014, 2016; Shott and Otárola-Castillo 2022; Smith and Goebel 2018; Smith, Jennings, and Smallwood 2021). Most archaeologists still regard points as integral wholes, which in one sense they are, but the concept of “integral whole” obscures the fact that

even the smallest hafted point is made up of parts. To its users, the different segments or ‘modules’ of a point served different purposes and were treated differently. From the perspective of hunters, minimally, tip modules initiated wounds in animal prey, blade modules deepened them, and stems articulated the exposed blade to the composite tool. (Shott and Otárola-Castillo 2022, 80–81)

But if we measure, say, only the length or width of points – thus treating them as undifferentiated wholes – we are obscuring the segmentation – the modularity – that might be there. As we will see, ceramic vessels also exhibit modularity.

Mosaic evolution

In mosaic evolution, some traits (modules) evolve independently (Carroll 1997; Foley 2016). Some might coevolve with others because of pleiotropic effects, and the timing of the evolution of one trait can be independent from the evolution of another (Smallwood et al. 2019). “Thus, different traits appear and change at different times, and the rates of evolution vary not just between periods but also between elements of the … extended phenotype” (Foley 2016, 12). In this case, ceramic vessels are parts of the extended phenotype of midwestern potters (O’Brien and Holland 1992, 1995).

Some traits hitchhike, which occurs when two or more traits are mechanically linked but only one is being sorted by selection. As a result of the linkage, the “hitchhiker” is also sorted. Those two processes – mosaic evolution and hitchhiking – both of which have long played significant roles in biological evolutionary theory (e.g. Stanley 1979) and paleoanthropology (e.g. Brace 1995; Foley 2016; McHenry 1994; Skelton and McHenry 1998) – are beginning to play similar roles in archaeology (Darwent and O’Brien 2006; Lyman and O’Brien 2003; Mesoudi and O’Brien 2008a; O’Brien, Darwent, and Lyman 2001, 2010; Prentiss et al. 2011, 2016; Smallwood et al. 2019; Smith and Goebel 2018), although, like with modularity, research has focused primarily on projectile points.

Constraint

One of the most misunderstood, and hence misused, concepts in evolutionary biology is constraint (Antonovics and Van Tienderen 1991). Here are two useful definitions:

Evolutionary constraints are restrictions, limitations, or biases on the course or outcome of adaptive evolution. The term usually describes factors that limit or channel the action of natural selection. … Constraints occur when a trait is precluded from reaching, shifted away from, or slowed down in its approach to a (defined) selective optimum. (Hansen 2015, 1)

Similarly, a constraint is a “property of a trait that, although possibly adaptive in the environment in which it originally evolved, acts to place limits on the production of new phenotypic variants” (Blomberg and Garland 2002, 901). As Blomberg and Garland (2002) point out, Futuyma (1998) listed several consequences of constraint: (1) the absence of adaptive characters – organisms can be constrained from evolving adaptive traits because of a lack of variation in the required direction; (2) the presence of directional trends as a result of developmental pathways making some variants more likely than others; and (3) low rates of evolution because of limited genetic variation. Although he was referring to organisms, the same evolutionary constraints apply to ceramic vessels. As with modularity and mosaic evolution, most archaeological discussions of constraint have focused on projectile points (e.g. Charlin and Cardillo 2018) or other stone tools (e.g. Brantingham and Kuhn 2001).

Evolutionary processes and limestone-tempered pottery

All three evolutionary processes contribute to our understanding of the shift from grit- to limestone-tempered pots. Everything starts with modularity and the ability of midwestern potters to tune out any deleterious effects of decisions they made. Recall our earlier statement that if errors are localized, they are far less likely to lead to failure:

Simultaneous random changes in many parts of a highly integrated structure are not likely to improve its function, as the chance improvement of one part will almost always be swamped by deleterious effects in many other parts. But if the parts are variationally independent, selection gets the chance to tune them one at a time, thereby improving the probability of finding improvements. (Hansen 2003, 84)

The modularity in Woodland pots is evident in such things as variation in the amount of temper, either limestone or grit, that could be added without leading to failure (Braun 1983). It also is evident in such features as the use of mixed tempers in the same vessel (O’Brien 1996; O’Brien et al. 1994) and the presence of limestone-tempered vessels alongside contemporary grit-tempered vessels on the same site (e.g. Bebber, Spurlock, and Fisch 2018). The same comments apply to pottery from Southeast Missouri tempered with grit and shell (Dunnell and Feathers 1991; Feathers and Scott 1989). Although Braun (1982, 185) is talking specifically about temper-particle size, his statements apply to all other aspects of modularity and to constraints as well:

The selection of particles for inclusion in a ceramic paste … not only will follow the requirements of the process of manufacture, but minimally must not interfere with the desired mechanical performance of the resulting vessels. Appropriate selection can, in fact, improve vessel performance under particular circumstances. The potter must establish a compromise in selecting particles for inclusion, balancing the choices according to the risks and benefits of particular materials under particular conditions of manufacture and use. This balancing also involves the consideration of other factors, including the labor and material costs of different choices, intended wall shape and thickness, and whole vessel size and shape. The last four of these also directly relate to vessel intended use and mechanical response characteristics.

There is every reason to suspect that midwestern potters had excellent control over recipes for creating successful pots but were always making adjustments based on their previous successes and failures as well as those of potters around them. Note that this cultural selection is analogous to Jacob’s (1977, 1163) comments on natural selection, which works not like an engineer but like a tinkerer, “who does not know exactly what he is going to produce but uses whatever he finds around him whether it be pieces of string, fragments of wood, or old cardboards … to produce some kind of workable object.” The archaeological record indicates that tinkering was a slow process during much of the Middle and early Late Woodland periods. For example, in eastern Missouri, limestone began to be used as a temper for cooking vessels by the start of the Christian era, but it was sporadic, and vessels were often tempered with a mix of grit and limestone. It rose in prevalence by ca. A.D. 350, when limestone was used as the only temper in some vessels, but it did not become the dominant temper type until ca. A.D. 750.

Some archaeologists speculate on whether prehistoric potters recognized differences in their wares. For example, Bowen and Harry (2019) pose the question that even if using thinner pots is more fuel efficient, was the difference substantial enough to have been recognized by prehistoric cooks and to have influenced indigenous potters’ production technology? We could ask the same question about vessel weight. These are interesting questions but ones that probably are unanswerable because we can never get inside the heads of prehistoric potters to understand why they made the decisions they did. In one sense we can speak, rather trivially, of intent or adaptation being proximate causes, but such a statement has little or no analytical value. Proximate causes, in any scientific framework, are functional causes – how things work – and to invoke intent and/or adaptation as explanations (as opposed to perhaps parts of an explanation) robs valid functional questions of their interesting parts and replaces them with vitalistic, directional components that lead to what Gould and Lewontin (1979) termed “just so” stories: A trait or behavior that an organism exhibits must be useful to that organism; otherwise, it wouldn’t have it. In archaeology, where adaptation and intent are often invoked as “explanations,” they often are viewed as products of vaguely referenced “selective agents” – elements of the natural or cultural environment that force human groups to change or face decline or extinction (O’Brien and Holland 1992). Of course, such stories are untestable.

The way out of this dilemma is to start with evolutionary theories of adaptive function: “By starting with a crisp definition of each adaptive problem the animal needs to solve … researchers quickly bec[o]me aware that different adaptive problems have different computational requirements” (Klein 2014, 12). To understand those computational requirements – here of vessel manufacture – archaeologists use recipes that match those of prehistoric potters to create replicates (analogous modules) to evaluate the applicability to archaeology of principles derived from material-science fields, keeping in mind that “although the principles may be universally valid, in practice they may yield very different effects under different conditions” (Bowen and Harry 2019, 272). Together with studying the computational requirements referred to by Klein (see above), there has to be a focus on the adaptive problems that vessels were built to overcome.

Based on our reading of the evidence and applying the principle of parsimony – that is, all things being equal, the most acceptable explanation of an occurrence, an event, or a phenomenon is the simplest one, meaning the explanation that involves the fewest entities – wall thickness appears to have been the major target of selection throughout the Middle and early Late Woodland periods in the Midwest, with clay workability and vessel weight as hitchhikers. How thick to make the walls of a pot is of basic concern to any potter interested in performance characteristics of the finished vessel, since that consideration affects not only the cost of manufacture but also at least three aspects of mechanical performance: thermal conductivity, flexural strength (breakage load), and resistance to thermal shock. It is the relationship among those aspects that is important: Thermal conductivity is correlated negatively with increased wall thickness, although the findings of Bowen and Harry (2019) discussed earlier showed a minimal increase in conductivity in thin-walled vessels. Thermal shock – the ability of a vessel to withstand sudden and extreme changes in temperature – is also correlated negatively with wall thickness: Everything else being equal, the thicker the walls, the more poorly a vessel withstands thermal stress. However, flexural strength – the ability of a vessel to withstand mechanical stresses without distorting or breaking – correlates positively with increased thickness, given that variables such as temper type and size are held constant. Therefore, potters moved toward the production of vessels that had greater thermal-stress resistance – a move accompanied by a technology – centered around limestone as a temper – that allowed for production of an increasingly thin-walled vessel that would not collapse during manufacture and firing (O’Brien et al. 1994).

Although as we mentioned, there are several unanswered questions surrounding the precise use of thinner-walled vessels in the Midwest – for example, the effects of thermal conductivity in quickly heating foods – a plethora of evidence suggests that the correlation between thinner vessels walls and an increase in the importance of starchy and oily seeds in midwestern diets was not incidental (Asch, Farnsworth, and Asch 1979; Braun 1977, 1983, 1985b; O’Brien 1996; O’Brien et al. 1994). In other words, the adaptive problem that Woodland vessels were built to overcome was the ability to better incorporate starchy and oily seeds into the diet.

Acknowledgements

We thank editor Alan Outram and two anonymous reviewers for their comments, which improved the manuscript.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

Mika, Eren, Meindl, and Bebber were supported by the Kent State University College of Arts and Sciences.

Notes on contributors

Anna Mika

Anna Mika, M.A., is a doctoral student of archaeology at the University of Cambridge.

Metin I. Eren

Metin I. Eren, PhD, is Associate Professor of Anthropology at Kent State University and a Research Associate in Archaeology at the Cleveland Museum of Natural History.

Richard S. Meindl

Richard S. Meindl, PhD, is Professor of Anthropology and Professor of Human Evolutionary Biology, School of Biomedical Sciences, Kent State University.

Michael J. O’Brien

Michael J. O'Brien, PhD, is Professor of History and Professor of Life Sciences, College of Arts and Sciences, Texas A&M University–San Antonio.

Michelle R. Bebber

Michelle R. Bebber, PhD, is Assistant Professor of Anthropology, College of Arts and Sciences, Kent State University.

Notes

1 The American Midwest includes areas drained by the upper and central Mississippi River and the lower Ohio and Missouri rivers.

2 We’re bypassing discussion of early fiber-tempered wares. Whereas they are widespread over portions of the Southeast, in the Midwest their occurrence appears to be restricted to the Late Archaic-period Nebo Hill site near Kansas City, Missouri (Reid 1984a, 1984b), with dates that fall in the 2600–1500 B.C. range. Nebo Hill sherds are extremely soft (< 2.5 on the Mohs’ scale), and pastes contain 5–15% silt-sized particles and no sand. A mixture of two types of temper was used: (1) crushed potsherds or fired clay and organic material, including switchgrass, big bluestem, and sedge. This early ware is part of what Rice (1999) referred to as the “software horizon” – a “period characterized by sparse ceramics that are often thick, low fired, with limited surface treatment and either organic inclusions (or temper) or no deliberate addition of nonplastic materials at all” (Rocek 2013, 231).