Accurately Estimating Vessel Volume from Profile Illustrations
Author(s): Louise M. Senior and Dunbar P. Birnie, III Reviewed work(s): Source: American Antiquity, Vol. 60, No. 2 (Apr., 1995), pp. 319-334 Published by: Society for American Archaeology Stable URL: http://www.jstor.org/stable/282143 . Accessed: 23/07/2012 10:58 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. . Society for American Archaeology is collaborating with JSTOR to digitize, preserve and extend access to American Antiquity. http://www.jstor.org ACCURATELY ESTIMATING VESSEL VOLUME FROM PROFILE ILLUSTRATIONS Louise M. Senior and Dunbar P. Birnie III The highly fragmented nature of most archaeological ceramic assemblages makes whole or reconstructible vessels valuable and rare finds. Vessel volume has rarely been systematically quantified because convenient reconstruction methods dealing with sherds and partial vessels have been lacking. Now, with the method presented in this paper, highly accurate volumetric capacities of fragmented vessels can be calculated from carefully prepared vessel profile illustrations. The profile is digitized using a small number of points per vessel (20 to 30 points are usually sufficient). These data are then converted to a volumetric measure using a computerized algorithm based on the geometry of stacked bevel-walled cylinders. This method of determining vessel volumes was tested and shown to be highly repeatable and accurate. Quantifiable sources of error are generally limited to less than one percent per vessel, with the final accuracy limited chiefly by the quality of illustration. With this computerized technique, fragmented vessels no longer need to be fully reconstructed in order to obtain volumetric information. Los hallazgos de vasijas completas o reconstruibles son muy raros debido a la gran fragmentacion de la mayoria de los conjuntos cerdmicos arqueologicos. El volumen de las vasijas no ha sido cuantificado sistemdticamente debido a la falta de mdetodos adecuados que utilicen vasijas parciales o fragmentos de estas mismas. Ahora, con el metodo presentado en este reporte, capacidades volumetricas exactas pueden ser calculadas para vasijas parciales o frag- mentadas utilizando ilustraciones cuidadosas de los perfiles de las vasijas. El perfil es digitalizado usando pocos puntos para cada perfil (20 o 30 puntos son usualmente suficientes). Los datos son convertidos a un estimado de capacidad volumetrica usando un algoritmo computarizado que se basa en la geometria de cilindros con paredes inclinadas, montados uno encima de otro. Este metodo de determinacion de volumenes de vasijas ha sido puesto a prueba; el metodo es preciso y exacto. Fuentes de error cuantificable son limitadas a menos del uno por ciento por vasija, con la exactitud final dependiente, en su mayor parte, de la calidad de la ilustracion del perfil de la vasija. Con esta tecnica computarizada, la vasijas fragmentadas no necesitan ser completamente reconstruidas con el proposito de obtener informacion volumetrica. W hy should we examine pot volumes? Prehistoric vessels were made to put things in, yet most archaeological analyses study only what is on the vessels. Prehistoric vessels were manufactured as containers; whether for cooking, storage, or serving, ves- sel volume (capacity) is an essential attribute that is tied directly to use and function of pottery. Most methods used to determine vessel capacity depend on the use of whole, or reconstructed, pots. Because of this, most analyses of prehistoric pottery, worldwide, are restricted to surface decoration techniques ("style") rather than pottery function since such studies can be made from the large frag- mented assemblages generally encountered archaeologically.1 Techniques of quick, ac- curate volumetric determination from vessel profiles now exist; thus, this attribute (ca- pacity) should no longer be neglected in ce- ramic analyses. Vessel volume generally relates to size, weight, and transportability of ceramic ves- sels; thus it is significant in exchange and use- life models based on ceramics (Deal 1983: 155-6; DeBoer 1985; Graves 1985:23; Lon- gacre 1981:63-64, 1985; Rice 1987:293-306; inter alia). Design variability can also be re- lated to variability in vessel size, as has been documented in the Kalinga ethnoarchaeolog- Louise M. Senior * Department of Anthropology, University of Arizona, Tucson, AZ 85721 Dunbar P. Birnie III * Department of Materials Science and Engineering, University of Arizona, Tucson, AZ 85721 American Antiquity, 60(2), 1995, pp. 319-334. Copyright ? 1995 by the Society for American Archaeology 319 AMERICAN ANTIQUITY ical assemblage by Graves (1981, 1985). De- sign variability within an assemblage may also be partitioned by vessel size because of the varied length of time that vessels were used; thus small pots may have more up-to-date design motifs than large vessels if sampled synchronically as was observed by DeBoer (1984:557) among the Shipibo-Conibo. Vessel volume is also an attribute from which artifact function can be deduced, al- though multiple lines of inference are needed to adequately assess systemic function from vessel shape (e.g., Hally 1986; Smith 1988). Vessel capacity measurements have some- times been interpreted as reflecting house- hold size and thus been incorporated into models of regional settlement, subsistence, and paleodemography (e.g., Blinman 1986a: 86ff.; Blinman 1986b:602-607; Nelson 1981; Smith 1988; Turner and Lofgren 1966). Systematic vessel volume studies may re- veal emic categorization of vessels by size and/or capacity (e.g., suggestions by Longa- cre et al. 1988; Miller 1982). Relative stan- dard units of volume or multiples of such units may be expressed and disclosed through vessel capacity measurements, and thus a portion of the metrological scheme of a past culture can be deduced (e.g., Gelb 1982; Mainkar 1984; Nelson 1985; Powell 1989; Rottliinder 1967; Turner and Lofgren 1966). In addition, the standardization hypothesis (Balfet 1965:163; Costin 1991; Feinman et al. 1984:299; Rice 1981:220-221; Rice 1991; Sinopoli 1988:586) frequently invoked as a gauge of relative craft specialization in ce- ramic studies may be more appropriately ap- plied to vessel capacities than to other metric attributes of pottery. Most studies of stan- dardization implicitly assume that produc- tion efficiency is a byproduct of craft routin- ization; further examination of why stan- dardization is important or linked to more complex levels of sociopolitical organization is frequently left unstated.2 The "reason" that vessel sizes are so stan- dardized in complex societies may be more rooted in concern for specific, market-related volumetric measures than in motor skill rou- tinization alone. The interpretation of Uruk Bevel Rim bowls as ration vessels, although controversial, is partly based on their stan- dardized capacities (Alden 1973; Beale 1978; Chazan and Lehner 1990; Johnson 1973; Le Brun 1980; Miller 1981; and Surenhagen 1975). Ration vessels are noted in Assyriol- ogical literature (e.g., Gelb 1965; Powell 1989) and the Akkadian concern for accurate men- suration undoubtedly led to standardization of vessel capacities by potters (Senior and Weiss 1992). As suggested by the brief overview above, study of ceramic vessel capacity will contrib- ute very significant data to our knowledge of ancient material culture. Application of the technique described in this paper will greatly increase the ability of researchers to obtain volumetric information from their frag- mented ceramic assemblages. In turn, this will begin to redirect the study of ceramics from what is "on" pottery to how specific pots functioned. Previous Estimations of Vessel Capacity Even though the attribute "vessel capacity" has been relatively neglected in archaeolog- ical analyses of ceramics, varied methods that determine vessel volume exist in the litera- ture. Most of these techniques rely on the use of whole or reconstructed pots; thus they have been useful only for those relatively rare finds. Other techniques, if based on specific key measurements of vessels (e.g., height, diam- eter, etc.), are relevant only when the assem- blage analyzed is highly standardized in terms of vessel form (e.g., Ericson and DeAtley 1976; Fitting 1970:174-175; Fitting and Hal- sey 1966; Hagstrum and Hildebrand 1990; Michael, Grantz and Maslowski 1974; Mounier 1987). The primary techniques that have been applied to archaeological finds are discussed below. Researchers should realize that vessel ca- pacity can be defined in at least two ways. It can be measured as "total possible capaci- ty"-that is, capacity up to the meniscus of the vessel rim. Obviously, this is probably not a practical capacity for the vessel, but it 320 [Vol. 60, No. 2, 1995] REPORTS is easily replicable between researchers. The "effective volume" of a vessel is frequently cited in archaeological literature (e.g., Hally 1986); this is a measure of capacity up to the point of greatest constriction, or to some point at which the investigator believes the vessel was considered full. Though perhaps more realistic than the total capacity, estimations of effective volume are subject to observer error. There is also danger inherent in Hally's point that vessel orifice is related to vessel capacity; though somewhat true in a broad sense, this factor varies by specific assem- blages. Either the total possible capacity or the effective volume can be measured with the methods listed below. Fluid Volume Method Measurements of the water contained in a vessel is one of the most obvious and expe- dient ways to ascertain comparative volumes of vessels. It is also advantageous because it does not require any special equipment or programs for calculation. There are four pri- mary limitations to the accuracy of this meth- od. First, it is possible for water to be ab- sorbed by the vessel walls, making a slight change in the apparent volume; this can be remedied by lining the vessel with plastic be- fore water is added, or by measuring the ca- pacity after the vessel walls are fully saturat- ed. Second, it is limited by the true accuracy of the measuring container. For example, most common laboratory glassware is rated only to within five percent of its nominal volume. Third, use of water to measure ca- pacity limits the analysis exclusively to whole and reconstructed vessels; fragments are not useful in any way. Finally, many museums and other institutions will not allow use of water after the objects are accessioned into collections for fear of damage. Water, or oth- er materials placed inside vessels, may phys- ically damage and contaminate fragile paint- ed decorations or residues on pots. Dry Volume Method This method can be performed with any kind of free-flowing solid (lentils, rice, bird, or mustard seeds, etc.), but many museums al- low only the use of lightweight polystyrene packing material ("styritos") since these are believed to be stable, inert substances that will not alter the vessel in any way. The meth- od is essentially similar in all respects to the water method detailed above; vessels are filled with a solid substance, and then the quantity of this substance is carefully measured. This method may not be as precise or re- peatable as the water method: Styritos tend to pack with a lot of air space between them. When styritos are measured with beakers or other measuring devices, only a fairly rough estimate of quantity is calculated. Also, the point at which the volume is "leveled off," such as at the rim of the vessel (or measuring device), is somewhat arbitrary and subject to the investigator's eye (whereas water would run out of the vessel if it exceeded the ca- pacity, even slightly). This method also re- quires complete or reconstructed vessels; profiles of vessels cannot be measured. Should free-flowing solids be used in volumetric measurements, researchers should try to use materials with very small radii (such as mus- tard seed), since these are not as subject to errors of heaping and packing. Heaping can easily be remedied by leveling the mounded solid with a straight rod drawn across the mouth of the vessel (see Mainkar 1984:148). Density Methods Vessel capacity can also be estimated by mea- suring the mass of material (of known den- sity) filling the vessel. This may eliminate additive errors accumulated when measuring devices are repeatedly used (such as measur- ing a 30-liter vessel with a one-liter contain- er). Mainkar (1984:148) summarizes this method: Take well-sifted sand. . . allow the sand to flow under gravity into the pot and when a heap of sand has been formed on the mouth of the pot, stop the flow. Level the heap with a straight rod or plate so that the sand is filled up to the top of the mouth of the pot. Weigh the pot and the sand. Remove the sand and weigh the empty pot. Determine the density of sand in a suitable manner. The volume of the pot can be calculated from the weight of the sand and its density. 321 AMERICAN ANTIQUITY Note that vessel capacity may also be mea- sured this way using water, which has a more easily standardized density than sand. This method, though eliminating some error due to repeated use of laboratory glassware, is dependent on both the accuracy of the scale used and the utilization of complete vessels. Methods Using the Assessment of Geometric Solids Several investigators have employed for- mulae derived from geometry to assess the general shape of pottery studied, and thus derive their respective volumes using key measurements from the pots in these for- mulae. Primary examples of these methods have been discussed by Castillo Tejero, and Litvak (1968), Ericson and Stickel (1973) and Johnson (1973). Although these methods can generally render a useful value for vessel ca- pacity, they are usually not very accurate if the assemblage is varied in form and little standardized. Care must be taken to com- pensate for vessel wall thickness. Because ex- terior pot dimensions are most commonly tabulated, the volume could be overestimat- ed, even if the geometry is well-suited to the vessel form. As discussed by Rice (1987:220-222), the Castillo Tejero-Litvak method assigns nu- merical values to vessel silhouettes and to reference points on the silhouettes. Vessels are thus described as sections or combina- tions of the geometric forms described by these points (e.g., a rectangular solid minus an ovaloid of specific dimensions). This method can be used to calculate capacity. It also has the benefit of not disturbing the in- terior contents of the vessel (since it is based entirely on observation of vessel dimen- sions), and thus will not disturb prospects for residue analyses. One drawback to this method is that it yields only an idealized representation of the vessel volume. The accuracy of the geometric representation depends on how smoothly the inside surfaces of the vessel mimic the set of shapes used to render the volume estimate. Thus it may only be useful for vessels that have convenient geometric outlines. A mod- ification of this method was applied by Senior (1990). The method employed by Ericson and Stickel (1973) incorporates both the geomet- ric form(s) representative of the vessel and the appropriate measurements (in millime- ters) of the vessel in volumetric formulae. Advantages and drawbacks to this method are essentially the same as those stated above for the Castillo-Tejero-Litvak method. Johnson (1973:135) used measured attri- butes of Uruk straight-sided bevel-rim bowls to ascertain volumes geometrically. These quickly made vessels were hypothesized to be ration bowls; thus their capacity (and its range of variation) was deemed interesting. Because the bowls all have the same general shape, a single formula was used. The ad- vantages of this method are that it can be applied to fragmentary as well as whole ves- sels; all that is needed is a complete vessel profile so that all the pertinent attributes could be measured. Disadvantages are that the method is specific only to straight-sided, open vessel forms. Calculus Method-Stacked Cylinders Nelson (1985:312-131) estimates vessel vol- ume through the calculus method of "stacked cylinders," which envisions the vessel as di- vided horizontally into a series of slices (see Figure 1). The interior diameter of the vessel is measured for each slice, from mouth to base; such slices represent the diameters of very thin cylinders (greatly exaggerated in Figure 1). Stacked one on top of another, these cylinders represent the entire vessel's volume (Rice 1987:221-222). Several researchers in Japan have used this method for making measurements from scale drawings (e.g., Fu- jimura 1981; Kobayashi 1992). Smith (1983; 1985:262) also determines vessel capacity with calculus. Pot capacity is calculated "by integration as the volume of the solid of revolution formed by revolving the profile curve around the x axis" (Smith 1985:262). Smith first derives a polynomial expression to fit each vessel profile, but this 322 [Vol. 60, No. 2, 1995] REPORTS Figure 1. Schematic representation of the stacked cyl- inder technique for calculating the volume of a vessel. is essentially the stacked cylinder calculus method. These polynomial expressions can easily be used to assess vessel volume from integration of its revolution (Smith 1985:291- 293). Smith's polynomials are derived so that he can assess other vessel variables besides capacity from the expression with his com- puter-simulation technique (SHAPE). Smith's integration of vessel profile polynomials is highly accurate if the polynomial truly ren- ders the pot shape; however, it is still only as good as the theoretical limit of the calculus integration, wherein the "true" value of the vessel volume is calculated from addition of the areas of an infinite number of "slices" or stacked cylinders. Data from a vessel profile, if accurately rendered, can be utilized when working with calculus-based capacity measures; thus, re- construction or dependence on whole vessels is not necessary. This opens up a much larger sample for analysis since the majority of ar- chaeological finds are highly fragmented and even if whole vessels are found, full recon- struction is very time consuming and thus costly. This method can also yield fairly ac- curate volumetric data if enough points are measured (i.e., if enough "cylinders" are cal- culated). The limitation of this method lies in the rough boundary formed by the cylinders of different diameters stacked on each other (Figure 1). Also, accurate measurement of the various cylinders is somewhat difficult and time consuming to procure; this could be fa- b R1 H1 R2 H2 R3 / Figure 2. Schematic representation of the use of bevel- walled cylinders for integrating the volume of a vessel. The illustration uses the same vessel but fills it with edge- matching beveled cylinders. The small outlined box (top right) defines the enlarged area, which shows exactly how different radii and heights are defined. R1, R2, and R3 are radii measured at sequential positions along the ves- sel interior. H, and H2 are the heights of successive slices. cilitated by taking measurements from well- executed scale drawings of the vessel profiles (see new method, below). Because the cal- culations required in Nelson's "stacked cyl- inder method" are fairly time consuming, this method would be greatly expedited through use of a computer spread-sheet program, even though this is not mentioned in the published work (Nelson 1985:312-313). Such a pro- gram would quickly process the computa- tions and store the estimated capacities so that they could later be statistically com- pared. 323 AMERICAN ANTIQUITY New Method of Vessel Capacity Estimation Vessel volumes can now be quickly measured from scale drawings.3 The cross sections (pro- files) are digitized and fed into a computer where an integration is performed to yield several characteristic features: volume, max- imum diameter, rim diameter, and height. This method, like others using calculus, as- sumes vessel symmetry around the axis of the illustrated profile; it is thus not suitable for dramatically asymmetric pottery. In-depth descriptions of the algorithm, its implemen- tation, and utilization are presented in fol- lowing sections. Algorithm The new method uses an improved stacked- cylinder method. The algorithm, rather than being restricted to vertical-walled cylinders, incorporates the calculus for cylinders with slanted edges.4 The edges match as they are stacked, and thus provide a smoother, more continuous, rendition of the vessel profile. Figure 2a shows the use of slanted cylinders for modeling the vessel capacity (compare with Figure 1 drawn with the same vessel). When using the stacked cylinder method above, each cylinder has a radius measured at half of its height, so that the overlapping comers will yield some canceling of errors and approximate the true volume. Figure 2a shows how the same vessel would be modeled by slanted cylinders. This gives a shape that is much smoother than that given by the ver- tical-walled stack represented in Figure 1. The vessel is first represented by a list of coordinates that are recorded directly from the outline of the vessel (to get the vessel's capacity the outline must be taken on the interior wall as presented in the projection of the cross section). This defines a sequence of points where the radius is known. These known radius positions then define the tops and bottoms of beveled cylinders, which, when summed, determine the volume of the vessel. Each beveled cylinder has a volume that is determined by the height of the cylinder and the top and bottom radii (defined as H, R1, and R2, respectively). The individual beveled cylinder volume is: V= (R2+ RR2 + R2) 3 2 2 (1) This formula was used by Johnson (1973) to estimate volume of straight sided vessels; for his application, R1 is the rim diameter, R2 is the base diameter, and H is the total height. In the present algorithm, the total vessel volume is simply the sum of the volumes of all N cylinders that have been defined by trac- ing the outline of the vessel interior: V= - '(R2 + RRi+RI + R2 1) (2) i=1 When writing a computer program to im- plement this summation, care must be taken to ensure that the outline is defined in a par- ticular order around the vessel. It is usually most expedient to start at one rim and work directly around the vessel. Each cylinder is defined by the points used to define the input outline. Adjacent outline points define suc- cessive radii (Ri and Ri+ ) by calculating the horizontal distance from the center line. And, the vertical separation between the two points defines the height (Hi). Figure 2b illustrates how the successive radii and heights are se- lected. This close-up segment from Figure 2a shows two beveled cylinders being used to model part of the volume. The upper beveled cylinder has R, as the top radius, R2 as the bottom radius, and H1 as the height. After Equation 1 has been evaluated for this slice, then the summation moves to the next slice in sequence. To force the beveled cylinders to match smoothly the bottom radius from the previous cylinder is used as the top for the next one; thus R2, R3, and H2 will be used for the next calculation, and so on. Since the height calculation is based on a sequentially traced contour, any reversals in direction will yield a negative height that will result in a negative volume for that slice. This assures that concave-bottomed vessels are accurately rendered using this method, be- cause the dimpled base region actually re- duces the capacity of the vessel. 324 [Vol. 60, No. 2, 1995] REPORTS Implementation This technique has been implemented using an IBM AT-compatible personal computer and a Summagraphics "Summasketch" dig- itizing tablet. A powerful computer is not re- quired to run this program efficiently and quickly; it has also been used with a dual- floppy portable PC computer connected to the same digitizer. The computer and digi- tizer are attached through the parallel com- munications port (using RS232 protocol). The present program for communicating with the tablet and for later integrating the volume is written in PASCAL, although any language would suffice. The digitizing tablet is the key element be- cause it enables very easy data collection from scale illustrations. The "mouse" (or "puck") has a cross-hair that can accurately point at sequential positions along contours, and their location is specified simply by pressing a but- ton on the mouse. This arrangement allows a large number of points to be accurately measured from a scale drawing of the vessel profile. Two versions of the program have been implemented. The first works with full-pot illustrations; the second version uses half-pot illustrations but requires the specification of the location of the midline of the pot (since there are not two sides to average). The half- pot program is favored by these authors for three reasons: (1) interior vessel profiles are generally only illustrated on one side of a vessel following common illustration con- ventions; (2) it is faster to digitize only half of the vessel; and (3) there is no noted dif- ference in accuracy between the two methods so the slightly faster version is preferred. Procedures As with all techniques, reasonable care must be taken when measuring the volume. The illustration must be carefully sized and ori- ented to guarantee accurate data collection. As an example, we describe the measurement of a half-pot cross section illustration since this is slightly more complex and, as noted above, it is our preferred method. The scale drawing is first taped to the dig- itizing tablet. The entire drawing must be dig- itized without removing it; thus, the illustra- tion must be proportioned to fit within the active area of the digitizer (usually about one foot square). This may sometimes require photocopy reductions to make the drawing fit. As long as the illustration's scale is re- duced by the same ratio, then the potential distortion of the photocopying process will have little impact on the final measured vol- ume.5 Care must be taken when affixing the draw- ing to the tablet; it must be directly vertical since the X-Y position of the cross-hairs is used to define the sequential positions along the contour. To assure this, the illustration's centerline should be aligned such that it is perpendicular to the edge of the digitizer's active area. This can easily be done with a drafting triangle or T-square. After attach- ment, the illustration scale is digitized by pointing to each end of the scale marker and then entering the measured length value through the computer keyboard. Next, point after point along the vessel profile is entered into the program. Each point is entered by carefully moving the cross-hairs so that it is directly on the contour. Once it is in position, one of the mouse buttons is pressed to register its location. Then the mouse is moved to the next position along the contour, thus tracing the outline. Measured points can be taken with arbitrary spacing along the contour; the program does not require that they be evenly spaced to produce a capacity measurement. There is no specific guideline on how close- ly the data points should be spaced along the contour. It is usually better to gather as many data points as patience will allow, but the points can usually be spaced somewhere be- tween a millimeter and a centimeter apart and still yield excellent volumes, as discussed below. The spacing of the data points can also be adjusted by the user when entering the data; portions of the cross section that have sharp curvature, carinations, or other odd shapes will require more closely spaced input points. With practice, most vessel profiles can 325 AMERICAN ANTIQUITY be digitized in less than two minutes using this program. After the half-contour is completely spec- ified, then the centerline is specified. When these data have been collected, the program quickly goes about integrating the sequential volume elements and presenting the final ves- sel characteristics. These calculations are completed within only a few seconds. Discussion This procedure offers distinct advantages over many other methods for determining volume because whole vessels are not required. The level of accuracy afforded by this technique is examined in close detail below. Quanti- tative comparisons with other techniques are also made. Many Computer-Aided Design (CAD) programs offer software that can per- form similar functions to the procedure de- scribed in this article; however, such pro- grams are usually very expensive and may not be feasible for all archaeological projects. The sources of error described below will be roughly equivalent whether a CAD program or the "Senior-Birnie" method is applied to estimate vessel capacity from scale drawings. Over the last 30 years, efforts to identify and reduce various sources of sampling, ex- planation, and measurement errors in ar- chaeological investigations have increased. Assessment of measurement error is especial- ly important when developing new methods such as that discussed in this article because error in measurement can be magnified when using highly sensitive digitizing equipment and thus be a significant source of variation in archaeological observations. Such contrib- uted variability must not, however, be con- fused with the variation present in the ar- chaeological record.6 Before making any in-depth analysis of the sources and magnitudes of error found for different techniques, it is useful to define var- ious terms that relate to measurement error in general. The most important distinction to make is between accuracy and precision. Accuracy defines how close the measured volume is to the "true" volume. Accuracy is the fundamental evaluation of measurements that have been made; we desire our mea- surements to be as accurate as possible. Pre- cision, on the other hand, is an assessment of the quantitative quality of the measure- ment itself and is not a relative comparison with the "true" value. Two of the key ele- ments that play a part in precision are (1) the number of significant digits that a measure- ment has, and (2) the repeatability of a par- ticular measurement.7 The number of signif- icant digits is usually the number of decimal places that can be obtained from a particular measuring instrument (ruler, tape measure, beaker, etc.). Note that most measuring de- vices have relatively low instrumental pre- cision.8 For instance, most metric rulers have graduated scales at increments of 1 mm, forc- ing measurements to have only about one significant digit at the centimeter level. Repeatability is related to the number of significant digits, but it also incorporates some level of random error that might cause vari- ation in repeated measurements using the particular instrument. The repeatability can never be better than the precision of the mea- surement, but it could be far worse if the random error contribution is large. Random errors are usually assumed to be normally distributed, which is the basis for using the mean and standard deviation as quantifica- tion of the variability of measured values. Even after a set of repeated measurements has been made (yielding a mean and standard deviation), there may not be certainty that the resultant values represent an accurate an- swer. In fact, many measurement techniques have systematic error contributions to each measurement. An example of systematic er- ror would be a ruler that had changed length through thermal expansion or had its end abraded through use, thus reducing its length. These errors give an average bias in one di- rection or another, causing the result to be repeatable, but inaccurate. These errors are usually much more difficult to identify and quantify than random measurement error. In summary, accuracy is limited by all sources of error, including instrumental precision, random measurement error, and systematic error in measurement. Because of these def- 326 [Vol. 60, No. 2, 1995] REPORTS initions, accuracy can never be better than precision, although precision may easily be better than accuracy for any given measure- ment. The sources of error found in this method fall into three basic categories: those inherent to creating a digitized illustration, those from the digitization process, and those that are part of the mathematical integration of vol- ume elements. These various contributions will be discussed here; wherever possible, at- tempts have been made to quantify these contributions. Error in Illustration Preparation The image digitized is usually a photocopy of a higher quality illustration. This illustra- tion (and its photocopy) have built-in error contributions that come from several factors: 1) Vessel profile reconstruction from the existing sherds may be inaccurate; 2) Illustration of the vessel profile may be inaccurate;9 3) Rendering of the scale marker may be inaccurate; and 4) Photocopying may stretch or distort the illustration. 10 Whenever fragmented vessels are recon- structed (especially for partial reconstruc- tions), the possibility exists that the sherds will not be glued into exactly their original place, shape, or configuration. If only vessel profiles are being reconstructed, this could contribute error to the production of the pro- file sketch. Great care should be taken in ex- actly matching the existing sherds during re- construction; best results will probably be ob- tained by experienced curators. There will be more error with large vessels than small ones simply because more sherds are being refit, and potentially ill-fit, into the vessel profile. Ironically, it is generally easier to produce an accurate interior profile rendering from a fragmented pot than an unbroken one, so this method is probably best when used with frag- mented remains, despite possible distortions due to reconstruction. Because the present method relies heavily on the use of an accurately rendered scale illustration of the vessel profile after recon- struction, production of this illustration is probably the greatest potential source of error in the method. In many regions of the world, highly accurate illustrations of all vessel pro- files are routinely prepared by professional illustrators; because these illustrators pro- duce so many sherd drawings, their work is assumed to be very consistent. It is very dif- ficult, however, to assess the real accuracy of this step. Vessel profile renderings are rarely esti- mated from smaller fragments; when they are, illustrators routinely signify such guessti- mates with dotted lines. For obvious reasons, use of such illustrations should be avoided since any error in the estimation will be dra- matically compounded when the metrics are converted to volumetric units. When applying this (or any) method to a large comparative sample, it is ideal that all drawings used be prepared by the same il- lustrator.1' This will reduce random varia- tions between illustrators, and any systematic error will not interfere with comparisons made between various vessel assemblages. Any possible error in rendering the scale on a drawing is particularly insidious because it will magnify or compress the entire illus- tration by the amount of error in the scale. Since we are calculating volumes directly based on this linear scale marker, such error will be amplified by the mathematical oper- ations used. This will typically triple the ef- fective error from scale marker to final mea- sured capacity because length is taken to the third power to find volume.12 Photocopying of published works or illus- trations (which might be necessary to allow attachment to the surface of the digitizing tablet) can also contribute error. To minimize such error, it is critical that the illustration's scale be included on the image copied. In this way, the scale will be subjected to the same possible distortions as the rest of the image. It is assumed that photocopying error is a relatively uniform stretching or reduction specific to each machine used. For most cop- ying machines this effect is rather minor and would only become apparent if copies were 327 AMERICAN ANTIQUITY made sequentially from earlier copies, am- plifying the effect. Mechanical Errors in Digitization The digitization process itself also has limi- tations. These limitations have two sources: mechanical and human. Mechanical limita- tions can best be overcome if the illustration digitized is at least a 1:1 rendering of the pot; however, this may not always be feasible be- cause of the expense of extra-large digitizers needed for over-sized drawings. The more an illustration is reduced, the greater the amount of error introduced by slight differences in location of digitized points. It is desirable to use as much of the active area of the digitizing tablet as possible. This increases the resolu- tion of the measurements being made. Even though there are limits to the preci- sion of the numbers obtained from the dig- itizing tablet, no constraint is placed nor- mally on the accuracy based on this effect. Digitizing tablets usually measure X-Y po- sitions with an absolute accuracy of about 1/20 of a millimeter; this significantly exceeds the accuracy of most illustrations or photocopies. Human Error in Digitization The manual dexterity, "eye," and general pa- tience of the individual digitizing an illustra- tion can also contribute to, or influence, the human error in implementation of this meth- od. The actual choice of digitized points is a crucial factor in this method. The subjective choice of digitized points impacts three areas: (1) defining the outline, (2) defining a centerline for a half-vessel il- lustration, and (3) defining the length of the illustration scale. These three categories are difficult to separate from each other. Instead, the aggregate error for these contributions can be estimated by having a single user measure the same drawing several times and calcu- lating the coefficient of variation in the re- sultant volumes. A measurement of these human error ef- fects was conducted using the geometric ves- sel profile shown in Figure 3. This illustration has an idealized shape that allows the exact volume to be calculated directly from the known dimensions used to plot it (again using Equation 2). Thus, comparison between the measurement and the true value is allowed. Figure 3 was measured 10 times; each new measurement was performed after removing and reattaching the illustration to the active area of the digitizing tablet. Any human error involved in accuracy of illustration orienta- tion, placement, and attachment was there- fore included in this process. The vessel volume from these 10 mea- surements is 24,690 ? 105 cc. This compares very favorably with the true calculated value of 24,609 cc. The coefficient of variation in volume measurements was .004. Since these measurements included all steps of the pro- cess, the coefficient of variation directly quantifies the aggregate contribution of aim- ing error because of variations in manual dex- terity when "aiming" the digitizer puck. This extremely small error contribution is be- lieved to be significantly smaller than errors derived from illustration or reconstruction processes. Good results also depend to a large extent on whether the user is familiar with the dig- itizing equipment, the program, and the pro- cess of performing the measurement. The quality of the technique, even when applied by beginners, is illustrated by another com- parison: the variation in volumes as mea- sured by many different novice users. In sampling 24 individuals, many of whom had never touched any digitizing equipment, we found a coefficient of variation of .029.13 A real pot illustration, rather than an ideal- ized one, was used in this study. This indi- cates that even novice users can achieve a precision of closer than three percent! We believe that any of these users could reach precision levels closer to .004 with only lim- ited practice. Error in Mathematical Algorithm Finally there may be some error in the math- ematical method for estimating volume. This is consistently true for calculus-based tech- niques. All of these techniques become more 328 [Vol. 60, No. 2, 1995] REPORTS I I I I I I 0 10 Figure 3. Cross section of a hypothetical geometrical object for testing the human error accuracy of the digi- tizer volume-measurement technique. The true volume can be found by summing two beveled disk volumes using Equation 2. accurate as the size of the increment becomes smaller. In the limit of infinitely small incre- ments, then the answer reaches the true val- ue. Such a mathematical limit is never ac- tually achieved. The mathematical errors accumulated when integrating the vessel pro- file illustrations can be simulated by digitiz- ing a curved shape as shown in Figure 4. This perfect hemisphere also has a volume that can be geometrically calculated. When it was measured 10 separate times (with careful re- moval and reattachment to the digitizing tab- let), the volume was 16,470 + 86 cc, giving a coefficient of variation of .005. The true volume was calculated to be 16,755 cc, based on the geometry of the drawing. The digitized volume is approximately 1.7 percent smaller than the true volume, significantly beyond the random error of the technique. This is a result of the concave hemispherical shape: I I I I I I 0 10 Figure 4. Cross section of a hemisphere for quantifi- cation of summation errors resulting from the use of bev- eled disk volume elements in calculating capacity. each digitized layer is approximated as hav- ing a straight wall, which erroneously trun- cates a tiny bit of the inside of the volume. Still, this error contribution is extremely small. Comparison with Other Methods Finally, we have compared all of the above techniques with our computer-based meth- od. For this comparison we have chosen two vessels that were available to us both as whole vessels and as cross-sectional illustrations. Figure 5 shows the cross-sectional views of the two vessels. The left and right vessels were drawn to the same scale and will be referred to as "A" and "B," respectively. For both of these pots we measured their volume using the liquid capacity (with water) and their dry capacity (using rice and "styritos"). Each of these measures was made three times for each substance in each vessel; these values were averaged and are presented in Table 1. The volume was then calculated using a simple ellipsoid model using: DrD2H 6 (3) where D is the largest diameter and H is the height. Next, the volume was calculated by the method of stacked cylinders, using incre- ments of 1 cm and measuring the radius at the center of each slab. Finally, the illustra- 329 AMERICAN ANTIQUITY tion was digitized using the new method. The resultant volumes (in cc) are given in Table 1. The last row of the table gives our best estimate of the absolute accuracy of each technique. If capacity measured with rice grains or water is considered the "true" capacity of vessels, then the Senior-Birnie method de- scribed in this paper compares favorably as shown in Table 1. Rice and water are shown to yield approximately the same figures for vessel capacity. Rice could give slightly larger values than water because it can be "heaped," whereas water will drain out.'4 These mea- surements are our most accurate direct mea- surements of volume. "Styritos" are less ac- curate and are not recommended for further use in volumetric studies. The inaccuracy of polystyrene packing material is caused by (1) the inability to pack them tightly and (2) the inability to "level" off the measuring device. They are also difficult to work with since stat- ic electricity makes them cling to vessel walls. The geometric solid method may be good for these vessels since their form is nearly ellip- soidal. The Senior-Bimie method is mainly lim- ited by the accuracy of the illustration in this comparison, since the method was proven to be highly repeatable (precise) in the tests on Figures 3 and 4. This method has been dem- onstrated to be highly accurate and fast; once the illustration is created, this method is the fastest of all methods tested. Future Directions The technique described in this paper relies very heavily on an accurate illustration; in fact, it can be said to be only as good as the illustrator involved. It may thus seem more fruitful to digitize vessel photographs to elim- inate the rendering error. However, this will only partially alleviate errors in volumetric estimations for at least four reasons. First, great care must be exercised during the photo shoot so that the vessel is photographed as a true profile without any included distortion. Second, even if the vessel is in the exact pro- file position, parallax distortion around the edge may hinder exact rendering of the profile with the digitizer. Third, care must be taken to place the scale for the photograph in ex- actly the same plane as the pot to avoid ex- aggeration of errors based on the scale. Fi- nally, most vessels photographed this way will be whole pots; therefore the capacity es- timated will be based only on the exterior profile of the vessel, and will equal the vessel capacity plus the clay used in vessel produc- tion. Were these factors addressed, then use of photographs rather than scale drawings could be expeditious. The digitization of photographs is an im- portant future direction for applications of the Senior-Birnie method since many ar- chaeologists routinely use photographs rather than illustrator's renderings. Parallax distor- tion may be an easily resolvable problem; the photogrammetry and remote sensing litera- tures should be consulted for algorithms that compensate for the distortions of different kinds of lenses. Vessels should be photo- graphed with the longest possible lenses and with the largest format cameras practicable so as to minimize distortion. Future research directions should also include estimates that compensate for vessel wall thickness when using photographs of exterior vessel profiles. Conclusions A new technique for measuring accurate ves- sel capacities from profile illustrations has been developed. It is based on the use of a digitizing tablet to quantify the shape of a vessel cross-sectional illustration and a com- puter to integrate the volume numerically. Whole or reconstructed vessels are not need- ed, although this method will also work on complete artifacts. Additionally, capacity of symmetrical archaeological features can be calculated using this method if accurate fea- ture sections are digitized. Future investigations might also use pho- tographs, rather than vessel renderings, for production of profile information. New il- lustration techniques developed by Harrison Eiteljorg and Nancy Wilkie may also greatly facilitate use of this method of capacity es- 330 [Vol. 60, No. 2, 1995] REPORTS Figure 5. Ethnographic vessels from Paradijon, Philippines, collected by M. Neupert (illustrated by Masashi Kobayashi). Note that the ridges shown on the interior of the lower vessel are rills remaining from manufacture; they are not illustration flaws or products of computer-scanning technologies. timation: pottery profiles can now be accu- rately drawn using AutoCAD 12 and a dig- itizing light pen (Eiteljorg 1994). Advantages of this technique are that this program runs very quickly on a standard per- sonal computer (PC). Most vessels can be accurately measured from scale drawings in only a couple of minutes. The brief time re- quired for this method allows investigators to average several readings for the best pos- sible assessment of the vessel capacity. Most archaeological projects are now equipped with a computer adequate for this computation; additional costs of obtaining a digitizing tab- let currently run approximately $250 to $300. The method described herein provides a low budget alternative that supplies very fast, ac- curate, and repeatable measurement of vessel or feature capacities using scale illustrations. Public Access to Pot-Volume Program The authors will provide the compiled PAS- CAL program on diskette as "shareware." No Macintosh version is available. Send your ad- dress, disk density, and size preference plus U.S. $5.00 (to cover postage and disk pur- chase) to the authors. Please do not request this material through American Antiquity. Acknowledgments. The comments on this manuscript by the American Antiquity editorial staff, the anonymous Table 1. Comparison of Volumes Measured Using Several Different Techniques. Geometric Stacked Stacked Measurement Solid Vertical Bevel-walled Method Water Rice Styritos Ellipsoids Cylinders Cylinders Vessel A 1350 1400 1650 1640 1580 1515 Vessel B 2400 2400 2550 3385 2280 2440 Estimated Accuracy 5% 5% 5% + ?% . % . ?% + Ren% 2% + Ren% Note: The estimated accuracies have been assigned based on quantifiable sources of error discussed in the text. All three of our direct volume measurements used measuring devices with about 5% basic accuracy. When using styritos, however, an unknown amount of additional inaccuracy occurs due to problems with packing and measuring the larger pieces. Note that the measurement of Vessel A is about 20% larger than that found using water or rice. For the geometric shape representation method we have not quantified the error contribution, but again Vessel A yields a much larger value, suggesting error in the range of 15 to 20%. Of course, this depends on how well the pot matches the geometric representation. Finally, both stacked cylinder methods derive their measurements from illustrations of the vessels rather than directly from the vessels. Rendering error from the illustration process has not been quantified but is included in these two columns as "Ren%." We believe that experienced illustrators will have accuracies at about the 5% level. (As explained in the text, this level of error would arise from linear measurement and rendering error at a level of less than 2%.) For the vertical-walled cylinder method we have not quantified the error contributions because of the mathematical summation; however, Vessel A gave a value about 15% larger than the direct measurement. In the case of the bevel-walled cylinder method, the 2% estimated error contribution is based on the 1.7% difference between calculated and measured values found when testing Figure 4. This includes all contributions due to illustration attachment, pointing error, and repeatability. 331 AMERICAN ANTIQUITY reviewers, and Barbara Mills were greatly appreciated by the authors, and we thank them for significantly im- proving the final product. Terry Majewski and Janet Walker greatly facilitated submission of this article with their fine editorial skills: thank you very much! We also wish to thank the "EGGS" of the University of Arizona Anthropology Department (particularly Andrea Free- man, Charles Tompkins, and Helga W6cherl) for initial inspiration and challenge in developing this method. Nino Aimo kindly translated our abstract. Many thanks to Carol Kramer and William Longacre, as well as to Mark Neupert, for the use of their ethnographically collected vessels in testing this method. Kramer graciously al- lowed use of her excellent illustrations, as well as field data collected on vessel capacities. Masashi Kobayashi allowed use of his vessel illustrations produced for his Ph.D. dissertation on Philippine ceramic ethnoarchaeol- ogy. Diane Dittemore allowed Senior access to Kalinga vessels stored in the Arizona State Museum and gra- ciously supplied a seemingly endless volume of "styri- tos" for use in vessel capacity measurements. We heartily thank you all for your help and support. A preliminary report and test of this method appeared in Senior and Gann (1992). References Cited Alden, J. R. 1973 The Question of Trade in Proto-Elamite Iran. Unpublished Master's thesis, Department of An- thropology, University of Pennsylvania, Philadel- phia. Arnold, P. J. 1991 Domestic Ceramic Production and Spatial Or- ganization: A Mexican Case Study in Ethnoar- chaeology. Cambridge University Press, Cam- bridge. Balfet, H. 1965 Ethnographical Observations in North Africa and Archaeological Interpretation: The Pottery of the Maghreb. In Ceramics and Man, edited by F. Matson, pp. 161-177. Viking Fund Publications in Anthropology No. 41, New York. Beale, T. W. 1978 Bevel Rim Bowls and Their Implications for Change and Economic Organization in the Later Fourth Millennium B.C. Journal of Near Eastern Studies 37:289-313. Blinman, E. 1986a Additive Technologies Group Final Report. Chapter 2 in Dolores Archaeological Program: Final Synthetic Report, compiled by David A. Breternitz, C. K. Robinson, and G. T. Gross, pp. 633-661. U.S. Department of the Interior, Bureau of Reclamation, Engineering and Research Center, Denver. 1986b Technology: Ceramic Containers. In Dolores Archaeological Program: Final Synthetic Report, compiled by David A. Breternitz, C. K. Robinson, and G. T. Gross, pp. 595-609. U. S. Department of the Interior, Bureau of Reclamation, Engineering and Research Center, Denver. Castillo Tejero, N., and J. Litvak 1968 Un Sistema de Estudio Para Formas de Vasi- jas. Technologia 2. Departamento de Prehist6oria, Instituto Nacional de Antropologia e Historia, Mex- ico City. Chazan, M., and M. Lehner 1990 An Ancient Analogy: Pot Baked Bread in An- cient Egypt and Mesopotamia. Paleorient 16:21-35. Coombes, G. B. 1979 The Archaeology of the Northeast Mojave Des- ert. Bureau of Land Management, Cultural Re- sources Publications, Archaeology. Costin, C. 1991 Craft Specialization: Issues in Defining, Doc- umenting, and Explaining the Organization of Pro- duction. In Archaeological Method and Theory, vol. 3, edited by M. B. Schiffer, pp. 1-55. University of Arizona Press, Tucson. Daniels, F., J. H. Mathews, J. W. Williams, P. Bender, and R. A. Alberty 1956 Experimental Physical Chemistry. McGraw- Hill, New York. Deal, M. 1983 Pottery Ethnoarchaeology Among the Tzeltal Maya. Ph.D. dissertation, Simon Fraser University, Burnaby, British Columbia. DeBoer, W. R. 1980 Vessel Shape from Rim Sherds: An Experi- ment on the Effect of the Individual Illustrator. Jour- nal of Field Archaeology 7:131-135. 1984 The Last Pottery Show: System and Sense in Ceramic Studies. In The Many Dimensions of Pot- tery, edited by S. Van der Leeuw and A. C. Prit- chard, pp. 527-571. University of Amsterdam, Al- bert Egges van Giffen Instituut voor Prae-en Pro- tohistorie, Amsterdam. 1985 Pots and Pans Do Not Speak, Nor Do They Lie. In Decoding Prehistoric Ceramics, edited by B. Nelson, 347-357. Southern Illinois University Press, Carbondale. Dibble, H. L., and M. C. Bernard 1980 A Comparative Study of Basic Edge Angle Measurements. American Antiquity 45:857-865. Eiteljorg, H. II 1994 AutoCAD for Pottery Profiles. CSA Newsletter 6(4):5-9. Center for the Study of Architecture, Bryn Mawr, Pennsylvania. Ericson, J. E., and S. P. DeAtley 1976 Reconstructing Ceramic Assemblages: An Ex- periment to Derive the Morphology and Capacity of Parent Vessels from Sherds. American Antiquity 41:484-489. Ericson, J. E., and E. G. Stickel 1973 A Proposed Classification System for Ceram- ics. World Archaeology 4:357-367. Feinman, G., S. Kowalewski, and R. Blanton 1984 Modelling Ceramic Production and Organi- zational Change in the Pre-Hispanic Valley of Oa- xaca, Mexico. In The Many Dimensions of Pottery, edited by S. Van der Leeuw and A. C. Pritchard, pp. 297-337. University of Amsterdam, Albert Egges van Giffen Instituut voor Prae-en Protohistorie, Amsterdam. Fish, P. R. 1978 Consistency in Archaeological Measurement and Classification: A Pilot Study. American Antiq- uity 43:86-89. 332 [Vol. 60, No. 2, 1995] REPORTS Fitting, J. E. 1970 The Archaeology of Michigan. Natural History Press, Garden City, New York. Fitting, J. E., and J. P. Halsey 1966 Rim Diameter and Vessel Size in Wayne Ware Vessels. The Wisconsin Archaeologist 47:208-211. Fujimura, H. 1981 Measuring Vessel Volume. Archaeological Re- search (Koukogaku Kenkyu) 28:106-117. Gelb, I. J. 1982 Measures of Dry and Liquid Capacity. Journal of the American Oriental Society 103:585-590. Graves, M. W. 1981 Ethnoarchaeology of Kalinga Ceramic Design. Ph.D. dissertation, Department of Anthropology, University of Arizona, Tucson. 1985 Ceramic Design Variability within a Kalinga Village: Temporal and Spatial Processes. In Decod- ing Prehistoric Ceramics, edited by B. Nelson, pp. 9-34. Southern Illinois University Press, Carbon- dale. Hagstrum, M., and J. Hildebrand 1990 The Two-curvature Method for Reconstruct- ing Ceramic Morphology. American Antiquity 55: 388-403. Hally, D. J. 1986 The Identification of Vessel Function: A Case Study from Northwest Georgia. American Antiquity 51:267-295. Johnson, G. A. 1973 Local Exchange and Early State Development in Southwestern Iran. Anthropology Paper No. 51. Museum of Anthropology, University of Michigan, Ann Arbor. Kobayashi, M. 1992 Changes in a Formal Assemblage from Jomon Pottery to Yayoi Pottery. Hokuetsu Archaeology (Hokuetsu Koukogaku) 5:1-34. LeBrun, A. 1980 Les Ecuelles Grossieres: Etat de la Question. In L 'Archeologie de L 'Iraq, edited by M. T. Barrelet, pp. 59-70. Centre Nationale Recherche Scientifique (CNRS), Paris. Longacre, W. A. 1981 Kalinga Pottery, and Ethnoarchaeological Study. In Pattern of the Past, edited by I. Hodder, G. Isaac, and N. Hammond, pp. 49-66. Cambridge University Press, Cambridge. 1985 Pottery Use-life Among the Kalinga, Northern Luzon, the Philippines. In Decoding Prehistoric Ce- ramics, edited by B. Nelson, pp. 334-346. Southern Illinois University Press, Carbondale. Longacre, W. A., K. Kvamme, and M. Kobayashi 1988 Southwestern Pottery Standardization: An Ethnoarchaeological View from the Philippines. The Kiva 53:101-112. Mainkar, V. B. 1984 Metrology in the Indus Civilization. In Fron- tiers of the Indus Civilization, edited by B. B. Lal and S. P. Gupta, pp. 141-151. Indian Archaeolog- ical Society, New Delhi. Michael, R. L., D. Grantz, and R. Maslowski 1974 Vessel Diameters from Sherds: A Mathemat- ical Approach. Pennsylvania Archaeologist 44(4):42- 43. Miller, A. 1981 Straw Tempered Ware. In An Early Town on the Deh Luran Plain: Excavations at Tepe Faru- khabad, edited by H. T. Wright. Memoirs No. 13, pp. 126-130. Museum of Anthropology, University of Michigan, Ann Arbor. Miller, D. 1982 Artefacts as Products of Human Categorisation Processes. In Symbolic and Structural Archaeology, edited by I. Hodder, pp. 17-25. Cambridge Uni- versity Press, Cambridge. Mounier, R. A. 1987 Estimation of Capacity in Aboriginal Conoidal Vessels. Journal of Middle Atlantic Archaeology 3: 95-102. Nelson, B. 1981 Ethnoarchaeology and Paleodemography: A Test of Turner and Lofgren's Hypothesis. Journal of Anthropological Research 37:107-129. 1985 Reconstructing Ceramic Vessels and Their Systemic Contexts. In Decoding Prehistoric Ceram- ics, edited by B. Nelson, pp. 310-329. Southern Il- linois University Press, Carbondale. Plog, S. 1985 Estimating Vessel Orifice Diameters: Mea- surement Methods and Measurement Error. In De- coding Prehistoric Ceramics, edited by B. Nelson, pp. 243-253. Southern Illinois University Press, Carbondale. Plog, S., F. Plog, and W. Wait 1978 Decision Making in Modern Surveys. In Ad- vances in Archaeological Method and Theory, vol. 1, edited by M. B. Schiffer, pp. 383-421. Academic Press, New York. Powell, M. A. 1989 MaBe und Gewichte. Reallexikon der Assy- riologie und Vorderasiatischen Archiologie 7:457- 517. Rice, P. 1981 Evolution of Specialized Pottery Production: A Trial Model. Current Anthropology 22:219-240. 1987 Pottery Analysis: A Sourcebook. University of Chicago Press, Chicago. 1991 Specialization, Standardization and Diversity: A Retrospective. In The Ceramic Legacy of Anna 0. Shepard, edited by R. Bishop and F. Lange, pp. 27-279. University Press of Colorado, Niwot. Rottliinder, R. C. A. 1967 Is Provincial Roman Pottery Standardized? Archaeometry 9:76-91. Senior, L. M. 1990 Cracked Pot Ideas in Archaeology: An Esti- mation of Prehistoric Value Systems. Paper pre- sented at the 57th Annual Meeting of the Society for American Archaeology, Pittsburgh. Senior, L. M., and D. Gann 1992 Measuring Ceramic Vessel Capacities. Poster presented at the Third Southwest Symposium, Tuc- son, Arizona. Senior, L. M., and H. Weiss 1992 Tell Leilan "Sila Bowls" and the Akkadian Re- organization of Subarian Agricultural Production. Orient-Express 2:16-24. Sinopoli, C. 1988 The Organization of Production at Vijayana- 333 AMERICAN ANTIQUITY gara, South India. American Anthropologist 90:580- 597. Smith, M. F., Jr. 1983 The Study of Ceramic Function from Artifact Size and Shape, Ph.D. dissertation, University of Oregon, University Microfilms, Ann Arbor. 1985 Toward an Economic Interpretation of Ceram- ics: Relating Vessel Size and Shape to Use. In De- coding Prehistoric Ceramics, edited by B. Nelson, pp. 254-309. Southern Illinois University Press, Carbondale. 1988 Function from Whole Vessel Shape: A Method and an Application to Anasazi Black Mesa, Arizona. American Anthropologist 90:912-923. Surenhagen, D. 1975 Untersuchungen zur Keramikproduktion in- nerhalb der Spat-Urukzeitlichen Siedlung Habuba Kabira-Sud in Nordsyrien. Acta Praehistorica et Ar- chaeologica 5/6:43-164. Tuggle, H. David 1970 Prehistoric Community Relations in East-Cen- tral Arizona. Ph.D. dissertation, University of Ar- izona. University Microfilms, Ann Arbor. Turner, C. G. II, and L. Lofgren 1966 Household Size of Prehistoric Western Pueblo Indians. Southwestern Journal of Anthropology 22: 117-132. Notes ' We do not imply that style is easier to infer from sherds than function; rather, style is commonly examined in order to establish the chronological and ethnic affiliation of archaeological assemblages. One could argue that style from fragmented assemblages is just as difficult to de- termine, if not more so, than function. 2 Exceptions to this are the connections made between political economy and standardization explicitly dis- cussed by Feinman et al. 1984 and Costin 1991. Note also Philip Arnold's recent work wherein a high degree of standardization is evident in the absence of craft spe- cialization in Tuxtla, Mexico (Arnold 1991). 3 Use of photographs, rather than scaled drawings, is discussed at the end of this article. 4 In geometry this shape is called afrustrum (pl.frustra); "truncated cone" is also sometimes used. 5 Photocopy distortion is often greatest around the edges of the image and drawings should therefore be scruti- nized in these regions for error. 6 Various archaeologists have already discussed this problem (Coombs 1979:68-70; DeBoer 1980; Dibble and Bernard 1980; Fish 1978; S. Plog, F. Plog and Wait 1978: 414; S. Plog 1985; Tuggle 1970:86; inter alia). 7 Repeatability is frequently called replicability in the archaeological literature. 8 Instrumental precision is also called "sensitivity" in archaeological, and other, literatures. 9 Use of appropriate photographs could render this prob- lem moot; however, most photographs would not yield interior capacity measurements since the outside of the vessel wall would usually be digitized. Use of photo- graphs, and errors inherent to their use, is discussed later in this article. 10 Many photocopy machines are constructed to delib- erately slightly reduce the image copied to approximately 99-98 percent of the original size. I If photographs are used, it is also best that they be prepared by the same photographer, or at least one using the same criteria for aligning the vessel in true profile position. 12 The effective error is not cubed. During the mathe- matical operation of multiplication, the fractional error contributions from the two factors are added. So, when finding the volume, each of the three dimensions adds to the error (see Daniels et al. 1956:324ff. for discussion of experimental error calculation). 13 These individuals participated in this project at the Third Southwest Symposium, Tucson, Arizona, during the poster session presentation by Senior and Gann (1992). 14 See Mainkar (1984:148) for method to avoid heaping of small solids. Received February 24, 1993; accepted July 20, 1994. 334 [Vol. 60, No. 2, 1995]