James K. Feathers (Reference Feathers2023) offers thoughtful comments and constructive criticism, not polemic, that more archaeologists might emulate. He raises important questions that deserve reflection and seeks to “make the problem of use life more tractable.” Nevertheless, Feathers's argument is not persuasive.
The larger study (Shott Reference Shott2018) from which my article derived (1) argued the importance of use life (“L” for reasons given there) in generating ceramic assemblages, (2) acquired longitudinal Michoacán ethnoarchaeological data that replaced informant estimates with computed L values, (3) identified robust correlations between vessel size and L in Michoacán and a substantial cross-cultural longitudinal dataset, and (4) conducted survivorship analysis that identified different failure causes (chiefly chance versus attrition) by vessel type and size that affect inferences from assemblages—failure by chance occurring at more regular rates. Secondarily, the study advocated vessels, not sherds, as valid units of observation, following Orton's (Reference Orton1993) theoretical arguments. The comparative merits of units of observation seem to be Feathers's chief concern.
Feathers highlights an issue noted in my article but explored at length in the larger study (Shott Reference Shott2018:118–137): distributions of values of L. His comment is a salutary clarification that distributions, observable in ethnoarchaeological data, are practically impossible to compile in archaeological context. Yet properties of ethnoarchaeological distributions can be gauged (e.g., in Weibull failure analysis) across cross-cultural datasets, those properties then projected to comparable archaeological types. Nevertheless, the important issue of L distributions is tangential to Feathers's concern.
Units of Observation
Feathers proposes three units of observation: sherds, original vessels, and assemblages. He agrees that fragmentation rate—which affects sherd count dramatically, vessel count not at all—is highly variable, therefore an uncontrolled source of variation. More equivocally, he assumes that vessel-type weights vary little between assemblages. In Michoacán (Shott Reference Shott2018:45–48) and cross-cultural datasets (Shott Reference Shott2018:Table 7.3), they vary considerably within and between assemblages, casting doubt on Feathers's assumption.
Advocating assemblages as observation units, practically Feathers means assemblages of sherds. Yet he evidently agrees that sherd count is invalid and that vessel counts are valid. Feathers then argues, somehow, that sherd-count assemblages are valid. The argument puzzles, engaging the severe deficiencies of sherd counts in quantification (Orton Reference Orton1993): uncontrolled variation in fragmentation rates, size, and L among originating vessels. Variation in vessel size alone overrepresents larger vessels in ways that sherd counts cannot control. Vessels are fundamental units of original use (ignoring, obviously, occasional use of sherds as tools), whereas sherds are merely a convenient unit. Assemblages are contexts of accumulation, not observation units. However defined, they are sets of sherds or original vessels (the latter as wholes or, usually, parts) derived, differently, from those observation units.
Feathers questions vessels as units of quantification only on the pragmatics of (1) converting fragmentary sherds to original vessels and (2) inference to L. Sherd-to-vessel conversion is challenging, not insurmountable. Refitting is not always unfeasible. As Feathers notes, intact vessels are rare in the record and, potentially, unrepresentative. But quantification options from Orton's (Reference Orton1993) units to rim diameter supply valid units. Estimated vessel equivalents (EVEs), for instance, express quantity in proportion to measures of vessel size such as volume, weight, and surface area (Shott Reference Shott2018:164; Waagen Reference Waagen2022:531) and features such as rims and orifice diameters.
Feathers notes correctly that vessel L is a product of size and durability, supply, function, and replacement cost. Michoacán data were analyzed by functional class, thereby controlling function, and tested for but revealed no supply effects (Shott Reference Shott2018:105–109). Vessel size correlated significantly with L in Michoacán and in several cross-cultural datasets (Shott Reference Shott2018:147–149). A robust connection emerges between vessel size and L, but of course further research is needed on its conceivable complication by durability and replacement effects. Until then, size emerges as a key variable controlling for L's effect on archaeological assemblages, further validating vessels as observation units.
On Seriation
Feathers agrees that better seriation candidates are short-lived vessels that fail by chance, not attrition. In Michoacán, small cooking vessels best fit this description. Short L minimizes the risk that vessel life exceeds occupation span; chance failure minimizes variation in discard rate (Shott Reference Shott2018:17–21, 173), both thereby reducing extraneous variation. Pottery not experiencing isotopic decay, vessel “half-life” is perhaps a misnomer.
Acknowledging sherd-count problems, in application Feathers finesses them. He summarizes Orton's discussion of life versus death assemblages (see also Shott Reference Shott2018:163–165), which pertains only to vessels; there are no life assemblages of sherds. In Feathers's hypothetical example, assemblage composition is “affected by . . . sample size”; actually, joint size-composition is explained not by span alone but also L and vessel—not sherd—systemic number (Shott Reference Shott2010). To Feathers, L “is similar for vessels of similar kind.” If by “kind” Feathers means function, the assumption is refuted in Michoacán (Shott Reference Shott2018:111–118) and elsewhere by considerable variation in L across size classes of similar function. As above, the key property that captures L variation is vessel size, variation that Feathers and I agree biases the sherd counts that he considers valid to quantify seriation assemblages. Treating sherds as some natural counting unit, to justify assemblages as observation units, ignores the problems, enumerated above, that invalidate sherds for that purpose. Feathers also does not explain how seriations “identify differential use life”—a property of vessels, not sherds. “Differential use life” is a problem baked into sherd counts, which cannot control for it.
Some Pragmatics of Ceramic Analysis
Feathers and I both value pragmatics. He argues that the refitting and calculation of Orton's units are time consuming and expensive. But all research shares those qualities, including collection of large sherd assemblages of taphonomically ravaged original vessels (which vary widely in Orton's “brokenness” or fragmentation rate and his “completeness” or inclusion of only some sherds from vessels) from secondary or tertiary contexts that are integral to the approach he advocates.
Yes, it is easier to count and weigh sherds than to estimate EVEs. But narrow focus on cost confines options to the cheapest and easiest, a sort of neoliberal approach better suited to making T-shirts than to analyzing ceramics. First, we should identify valid units of observation and quantification—only then the legitimate pragmatics of their acquisition.
An irony of Feathers's preference for sherds over vessels is his suggestion that sherd thickness and curvature correlate with original-vessel size (see also Zvietcovich et al. Reference Zvietcovich, Navarro, Saldana, Castillo and Castaneda2016). If testing bears it out, this is another pragmatic inference to original-vessel size, making the need for sherd counts in the first place even more moot. (Admittedly, it does not address the completeness problem.) The matter is for ceramic analysts to address, which Feathers probably joins me in encouraging.
Conclusion
Feathers and I agree that not all inferences from ceramic assemblages require knowing number of original vessels or L by type (for example, sourcing paste or temper, technical studies of fracture resistance; e.g., Shott Reference Shott2018:5). Yet L affects inferences—from dating to estimating population and occupation span, even to studies of pots as the tools they were (Braun Reference Braun, Moore and Keene1983)—that he believes can be drawn while ignoring it. There is no doubting the difficulties that Feathers emphasizes. But, confronting them, the Michoacán and earlier studies already offer partial, provisional solutions that deserve serious consideration, against Feathers's sherd counts to form assemblages somehow considered immune to the severe deficiencies of sherds as counting units.
L's inference is a “problem,” not an insurmountable obstacle. Unless we are content that uncontrolled sources of variation complicate, in unmeasurable directions and degrees, inferences drawn from ceramic assemblages, the only choice is to confront the challenge. Refitting, for instance, is time consuming, but it usefully corrected Hill's Broken K Pueblo conclusions (Shott Reference Shott2018:163) because “patterning . . . attributed to social processes . . . resulted from redundant entries of the same vessel” (Skibo et al. Reference Skibo, Schiffer and Kowalski1989:395)—a completeness effect that original emphasis on vessels, not sherds, would have revealed. Similarly, the Michoacán study was an effort to improve methods, focused more on inference to L than to numbers of original vessels. Throwing up our hands instead is a counsel of despair that Feathers probably agrees is unworthy of serious regard. Such responses, as Waagen ruefully noted, value economy of effort over validity and “make weighting [sic] and counting the batches of sherds . . . the modus operandi” (Reference Waagen2022:532). This is tantamount to conceding that valid quantification is difficult, so instead do what is easy and then hope for the best. This is uncritical economizing, not pragmatism, and a poor foundation for valid inference from ceramic assemblages.
Respectfully, Feathers's argument endorses the status quo's deeply flawed modus operandi: count and weigh sherds in the blithe faith that those measures somehow suffice. No good reasons support this belief, and several—as the Michoacán and other studies identified—cast grave doubt on it.
You have
Access
James K. Feathers (Reference Feathers2023) offers thoughtful comments and constructive criticism, not polemic, that more archaeologists might emulate. He raises important questions that deserve reflection and seeks to “make the problem of use life more tractable.” Nevertheless, Feathers's argument is not persuasive.
The larger study (Shott Reference Shott2018) from which my article derived (1) argued the importance of use life (“L” for reasons given there) in generating ceramic assemblages, (2) acquired longitudinal Michoacán ethnoarchaeological data that replaced informant estimates with computed L values, (3) identified robust correlations between vessel size and L in Michoacán and a substantial cross-cultural longitudinal dataset, and (4) conducted survivorship analysis that identified different failure causes (chiefly chance versus attrition) by vessel type and size that affect inferences from assemblages—failure by chance occurring at more regular rates. Secondarily, the study advocated vessels, not sherds, as valid units of observation, following Orton's (Reference Orton1993) theoretical arguments. The comparative merits of units of observation seem to be Feathers's chief concern.
Feathers highlights an issue noted in my article but explored at length in the larger study (Shott Reference Shott2018:118–137): distributions of values of L. His comment is a salutary clarification that distributions, observable in ethnoarchaeological data, are practically impossible to compile in archaeological context. Yet properties of ethnoarchaeological distributions can be gauged (e.g., in Weibull failure analysis) across cross-cultural datasets, those properties then projected to comparable archaeological types. Nevertheless, the important issue of L distributions is tangential to Feathers's concern.
Units of Observation
Feathers proposes three units of observation: sherds, original vessels, and assemblages. He agrees that fragmentation rate—which affects sherd count dramatically, vessel count not at all—is highly variable, therefore an uncontrolled source of variation. More equivocally, he assumes that vessel-type weights vary little between assemblages. In Michoacán (Shott Reference Shott2018:45–48) and cross-cultural datasets (Shott Reference Shott2018:Table 7.3), they vary considerably within and between assemblages, casting doubt on Feathers's assumption.
Advocating assemblages as observation units, practically Feathers means assemblages of sherds. Yet he evidently agrees that sherd count is invalid and that vessel counts are valid. Feathers then argues, somehow, that sherd-count assemblages are valid. The argument puzzles, engaging the severe deficiencies of sherd counts in quantification (Orton Reference Orton1993): uncontrolled variation in fragmentation rates, size, and L among originating vessels. Variation in vessel size alone overrepresents larger vessels in ways that sherd counts cannot control. Vessels are fundamental units of original use (ignoring, obviously, occasional use of sherds as tools), whereas sherds are merely a convenient unit. Assemblages are contexts of accumulation, not observation units. However defined, they are sets of sherds or original vessels (the latter as wholes or, usually, parts) derived, differently, from those observation units.
Feathers questions vessels as units of quantification only on the pragmatics of (1) converting fragmentary sherds to original vessels and (2) inference to L. Sherd-to-vessel conversion is challenging, not insurmountable. Refitting is not always unfeasible. As Feathers notes, intact vessels are rare in the record and, potentially, unrepresentative. But quantification options from Orton's (Reference Orton1993) units to rim diameter supply valid units. Estimated vessel equivalents (EVEs), for instance, express quantity in proportion to measures of vessel size such as volume, weight, and surface area (Shott Reference Shott2018:164; Waagen Reference Waagen2022:531) and features such as rims and orifice diameters.
Feathers notes correctly that vessel L is a product of size and durability, supply, function, and replacement cost. Michoacán data were analyzed by functional class, thereby controlling function, and tested for but revealed no supply effects (Shott Reference Shott2018:105–109). Vessel size correlated significantly with L in Michoacán and in several cross-cultural datasets (Shott Reference Shott2018:147–149). A robust connection emerges between vessel size and L, but of course further research is needed on its conceivable complication by durability and replacement effects. Until then, size emerges as a key variable controlling for L's effect on archaeological assemblages, further validating vessels as observation units.
On Seriation
Feathers agrees that better seriation candidates are short-lived vessels that fail by chance, not attrition. In Michoacán, small cooking vessels best fit this description. Short L minimizes the risk that vessel life exceeds occupation span; chance failure minimizes variation in discard rate (Shott Reference Shott2018:17–21, 173), both thereby reducing extraneous variation. Pottery not experiencing isotopic decay, vessel “half-life” is perhaps a misnomer.
Acknowledging sherd-count problems, in application Feathers finesses them. He summarizes Orton's discussion of life versus death assemblages (see also Shott Reference Shott2018:163–165), which pertains only to vessels; there are no life assemblages of sherds. In Feathers's hypothetical example, assemblage composition is “affected by . . . sample size”; actually, joint size-composition is explained not by span alone but also L and vessel—not sherd—systemic number (Shott Reference Shott2010). To Feathers, L “is similar for vessels of similar kind.” If by “kind” Feathers means function, the assumption is refuted in Michoacán (Shott Reference Shott2018:111–118) and elsewhere by considerable variation in L across size classes of similar function. As above, the key property that captures L variation is vessel size, variation that Feathers and I agree biases the sherd counts that he considers valid to quantify seriation assemblages. Treating sherds as some natural counting unit, to justify assemblages as observation units, ignores the problems, enumerated above, that invalidate sherds for that purpose. Feathers also does not explain how seriations “identify differential use life”—a property of vessels, not sherds. “Differential use life” is a problem baked into sherd counts, which cannot control for it.
Some Pragmatics of Ceramic Analysis
Feathers and I both value pragmatics. He argues that the refitting and calculation of Orton's units are time consuming and expensive. But all research shares those qualities, including collection of large sherd assemblages of taphonomically ravaged original vessels (which vary widely in Orton's “brokenness” or fragmentation rate and his “completeness” or inclusion of only some sherds from vessels) from secondary or tertiary contexts that are integral to the approach he advocates.
Yes, it is easier to count and weigh sherds than to estimate EVEs. But narrow focus on cost confines options to the cheapest and easiest, a sort of neoliberal approach better suited to making T-shirts than to analyzing ceramics. First, we should identify valid units of observation and quantification—only then the legitimate pragmatics of their acquisition.
An irony of Feathers's preference for sherds over vessels is his suggestion that sherd thickness and curvature correlate with original-vessel size (see also Zvietcovich et al. Reference Zvietcovich, Navarro, Saldana, Castillo and Castaneda2016). If testing bears it out, this is another pragmatic inference to original-vessel size, making the need for sherd counts in the first place even more moot. (Admittedly, it does not address the completeness problem.) The matter is for ceramic analysts to address, which Feathers probably joins me in encouraging.
Conclusion
Feathers and I agree that not all inferences from ceramic assemblages require knowing number of original vessels or L by type (for example, sourcing paste or temper, technical studies of fracture resistance; e.g., Shott Reference Shott2018:5). Yet L affects inferences—from dating to estimating population and occupation span, even to studies of pots as the tools they were (Braun Reference Braun, Moore and Keene1983)—that he believes can be drawn while ignoring it. There is no doubting the difficulties that Feathers emphasizes. But, confronting them, the Michoacán and earlier studies already offer partial, provisional solutions that deserve serious consideration, against Feathers's sherd counts to form assemblages somehow considered immune to the severe deficiencies of sherds as counting units.
L's inference is a “problem,” not an insurmountable obstacle. Unless we are content that uncontrolled sources of variation complicate, in unmeasurable directions and degrees, inferences drawn from ceramic assemblages, the only choice is to confront the challenge. Refitting, for instance, is time consuming, but it usefully corrected Hill's Broken K Pueblo conclusions (Shott Reference Shott2018:163) because “patterning . . . attributed to social processes . . . resulted from redundant entries of the same vessel” (Skibo et al. Reference Skibo, Schiffer and Kowalski1989:395)—a completeness effect that original emphasis on vessels, not sherds, would have revealed. Similarly, the Michoacán study was an effort to improve methods, focused more on inference to L than to numbers of original vessels. Throwing up our hands instead is a counsel of despair that Feathers probably agrees is unworthy of serious regard. Such responses, as Waagen ruefully noted, value economy of effort over validity and “make weighting [sic] and counting the batches of sherds . . . the modus operandi” (Reference Waagen2022:532). This is tantamount to conceding that valid quantification is difficult, so instead do what is easy and then hope for the best. This is uncritical economizing, not pragmatism, and a poor foundation for valid inference from ceramic assemblages.
Respectfully, Feathers's argument endorses the status quo's deeply flawed modus operandi: count and weigh sherds in the blithe faith that those measures somehow suffice. No good reasons support this belief, and several—as the Michoacán and other studies identified—cast grave doubt on it.