Journal of Anthropological Archaeology 33 (2014) 99–107

Contents lists available at ScienceDirect

Journal of Anthropological Archaeology

journal homepage: www.elsevier .com/ locate/ jaa

Copying errors of potters from three cultures: Predictable directionsfor a so-called random phenomenon

0278-4165/$ - see front matter � 2013 Elsevier Inc. All rights reserved.http://dx.doi.org/10.1016/j.jaa.2013.12.003

⇑ Corresponding author.E-mail address: gandon.enora@gmail.com (E. Gandon).

Enora Gandon a,⇑, Valentine Roux b, Thelma Coyle c

a Computerized Archaeology Laboratory, Institute of Archaeology, Mt. Scopus, 91905 Jerusalem, Israelb CNRS, UMR 7055, Préhistoire et Technologie, Maison de l’Archéologie et de l’Ethnologie, 21 Allée de l’Université, 92023 Nanterre cedex, Francec Aix-Marseille University, CNRS, ISM UMR 7287, 13288 Marseille cedex 09, France

a r t i c l e i n f o

Article history:Received 27 June 2013Revision received 15 December 2013

Keywords:Copying errorMotor skillCultural learning nicheRandom evolutionCeramic vesselsMorpho-metric attributes

a b s t r a c t

The impact of copying error on change in artifact morphology is studied through a field experiment withthree groups of potters, each with a distinct potting tradition (one from France and two from India). Thenine French potters and the 12 Indian potters had to reproduce – in five specimens – four different modelshapes with two different weights of clay (in total, each potter threw 40 pots). Results show that the var-iability generated while copying depends on both the difficulty of the task and the cultural learningniches of the potters. We conclude that, even though unintended, the copying error is culturally con-strained and therefore its amplitude and directions predictable. This is attributed to the cultural selectionof motor skills during apprenticeship.

� 2013 Elsevier Inc. All rights reserved.

Introduction

Within the framework of the dual-inheritance theory, culturaltransmission is seen as the primary mechanism for the evolutionof cultural features (Boyd and Richerson, 1985; Cavalli-Sforzaand Feldman, 1981; Durham, 1991; Henrich and McElreath,2003; Mesoudi and O’Brien, 2008; Richerson and Boyd, 2005;Shennan, 2008). According to this theory, in the course of transmis-sion, cultural features are reproduced with variations on whichselective forces of different kinds are exerted. These variationscould be partially generated by unintended copying errors occur-ring during learning (e.g. Eerkens, 2000; Eerkens and Bettinger,2001; Eerkens and Lipo, 2005, 2007; Hamilton and Buchanan,2009; Kempe et al., 2012; Steele et al., 2010). These copying errorsresult from the physiological limitation of the human perceptionwhich prevents individuals from accurately reproducing modelartifacts. According to previous studies, such copying errors wouldbe random with no ‘‘predefined or predictable direction’’ (Eerkensand Lipo, 2005: 319).

From an archaeological point of view, ‘‘models’’ are traditionallydefined in terms of morpho-metric classes or types once related tofunctional or chrono-cultural attributes (e.g. Gardin, 1980; Karasikand Smilansky, 2011; Orton et al., 1993; Read, 2009; Whallon,1972). The implicit assumption is that these classes or types corre-

spond to ‘‘cultural models’’ which craftsmen have copied to fulfillspecific cultural demands. They are characterized by intra-class(or type) variability which defines their degree of standardization.This degree of standardization has been shown to increase with thefrequency at which the artifacts are made, highly standardizedartifacts being characterized by a coefficient of variation of around3% and even less in the case of mass production (Roux, 2003). Thislow variability, verified empirically on craft products, correspondsto the magnitude of values considered to be inherent in humanperceptual-motor capacities (Eerkens, 2000; Eerkens and Bettinger,2001; Eerkens and Lipo, 2005).

It is generally accepted that copying errors lead to random vari-ations which will amplify overtime, giving rise to different mor-phological types and possibly to distinct lineages of objects, ifmultiple chains of transmission evolve independently and if noother cultural evolutionary forces are at work (e.g. Eerkens andLipo, 2005; Kempe et al., 2012; Neiman, 1995). Our aim here isto provide an explicit experimental study to analyze the amplitudeand the nature of the copying errors. Even though previous exper-imental studies have been conducted (Eerkens, 2000; Kempe et al.,2012), these studies have mainly involved perceptive (shape andlength perception) and cognitive skills (e.g. Mesoudi and Whiten,2004; Mesoudi et al., 2006). However, the replication of a modelartifact not only implies the capacity to faithfully perceive the mod-el shape but also the capacity to faithfully produce forms (i.e. a 3Dshape) from the model shape (a 2D shape). It is therefore still un-clear (i) whether the variation created in the course of transmission

100 E. Gandon et al. / Journal of Anthropological Archaeology 33 (2014) 99–107

relates to motor skills, perceptive skills or both; (ii) whether thevariations created by motor skill and/or perception are of the sameamplitude; (iii) what role, if any, the level of task difficulty and thecultural context play in the copying errors.

To address these issues, we conducted a field experiment wherewe asked French and Indian potters to replicate models of vesselsfrom visual observation. Experimenting in France and in Indiaallows us to take into account the role of the cultural context inartifact variability. As Washburn’s study has showed with Artstudents, cultural knowledge influences the replication of imagesand objects (Washburn, 2001). Hence it is likely that the culturalsetting influences the way traditional craftsmen reproduce specificmodels. The method adopted enables us to measure the intra-potter and inter-group variability and to assess the mechanismsunderlying the copying error phenomenon. As we shall see, thevariability in replicating objects is inherent to both motor and per-ceptive skills and the distance between the replicas and the modelsdepends on the difficulty of the task and on the cultural learningniches. In the discussion, we ask therefore whether copying errors,an unintended phenomenon considered as universal, nonethelessfollow predictable directions determined by the task difficultyand the learning traditions from which they stem.

Materials and method

A standardized experiment in three cultural settings

Three cultural groups of expert potters participated in thestudy: nine French potters (group F), six Indian potters from theHindu community (group Prajapati, abbreviated Pr), and six Indianpotters from the Muslim community (group Multani, abbreviatedMu). All the participants had practiced the craft for more thanten years. Potters from the two Indian groups (Pr and Mu) origi-nated from the same region in Northern India. However, theybelong to distinct cultural groups and have different technologicaltraditions (Roux, 2013). In India, pottery production is a specializedactivity that is carried out on a domestic scale. It is learned withinendogamous castes that produce traditional objects in mass pro-duction (Kramer, 1997; Roux and Corbetta, 1989; Saraswati et al.,1966). By contrast in France, the social organization of pottery as ahandicraft has been modernized since the industrial revolution:the apprenticeship has been transferred from family workshopsto public schools and the mass production of standardized usualobjects has since been replaced by an artistic personal production.As a result, the training conditions of the nine French potters weremuch more heterogeneous than those of the Indian potters. Inaddition to these distinctive social organizations, the instruments(wheels) vary in the three cultural contexts. French potters use

Fig. 1. Wheels used in the three cultural settings. From left to right: the electrical (pedalPrajapati potters, and the low-inertia kick-wheel used by the Multani potters.

electrical wheels activated by a pedal (Fig. 1, left panel), whilethe Prajapati use high-inertia fly wheels launched with a woodenstick (Fig. 1, middle panel), and Multani potters use foot-operated,low-inertia kick-wheels (Fig. 1, right panel).

A standardized experiment was set up in one pottery workshoplocated in France (Bourgogne) and in two pottery workshops (onePr and one Mu) located in the same Indian village (Bulandshar dis-trict, Uttar Pradesh). Working in their usual conditions of practice,potters were asked to reproduce four different model shapes usingtwo different quantities of clay, giving a total of eight experimentalconditions (Table 1). The shapes (referred to as cylinder, bowl,sphere and vase, respectively) were presented as 2D drawingswithout providing any indication of absolute dimensions to be pro-duced. The quantity of clay provided for each test corresponded toa mass of 0.75 kg or 2.25 kg. The four shapes were not part of any ofthe potters’ daily repertoire. The vase was the most difficult shapeto produce, due to the very small height of the maximum diameterand therefore the high risk of collapse (Gandon et al., 2011). Theeight experimental conditions were produced in five specimens,thus each participant produced a total of 40 pots.

The participants were instructed to faithfully reproduce themodel shapes, to throw vessels with the thinnest walls possible,and to refrain from embellishment operations once the vesselwas thrown. Participants practiced the task the day before theexperiment, producing at least one vessel under each of the eightexperimental conditions. During the experiment proper, the orderof the different conditions was randomised within each block ofeight trials so as to avoid systematic learning effects.

Data recording and analysis

The experimental sessions were videotaped (Panasonic NV-GS320) and the image of each finished vessel was extracted fromthe films. The 2D coordinates of the cross-sectional profiles ofthe vessel were obtained from each image by tracing each profileout on a Cintiq 21UX Wacom� tablet with integrated screen. Thecoordinates were converted from pixels to centimeters using a pre-determined calibration factor. The coordinates were re-sampled togenerate an equal number of points (256 in total) at regular heightintervals along the y-axis and were finally smoothed with a lowpass filter.

The absolute dimensions of each vessel were then measured.For the two open shapes [cylinder and bowl], we measured thebase, the aperture and the height, and for the two closed shapes[sphere and vase], we added the maximal diameter and the heightof the maximal diameter (Fig. 2). For each potter, we also calcu-lated, for each of the eight experimental conditions, the averageprofile which corresponds to the average of the five specimens pro-duced for each experimental condition.

operated) wheel used by the French potters, the high inertia stick-wheel used by the

Table 1The eight experimental conditions. Potters were asked to throw on their usual wheels four shapes (cylinder, bowl, sphere and vase); each shape being produced with two differentclay masses (0.75 and 2.25 kg).

Shape Cylinder Bowl Sphere Vase

Model (2D drawing)

Mass of clay (kg) 0.75 2.25 0.75 2.25 0.75 2.25 0.75 2.25Experimental condition 1 2 3 4 5 6 7 8

Fig. 2. Absolute dimensions measured for the production. For the open shapes [cylinder and bowl]: base, aperture and height; for the closed shapes [sphere and vase]: base,aperture, height, maximal diameter and height of the maximal diameter.

E. Gandon et al. / Journal of Anthropological Archaeology 33 (2014) 99–107 101

The data analysis provides two kinds of results, the intra-individual variability (i.e. between-specimens variability per pot-ter) and the copying error from the model. The intra-individualvariability has been represented by the coefficient of variation(CV) calculated for the absolute dimensions of the five specimensproduced in each condition. For the two open shapes [cylinderand bowl], CV was calculated for the base, height and aperture;for the two closed shapes [sphere and vase], CV was also calculatedfor the maximal diameter and the height of the maximal diameter.

For the analysis of the copying error we calculated both a globaland a partial error of reproduction. The global error gives an overallmeasure of the copying error while the partial error is specific to aparticular dimension. Both measures involve a comparison with amodel which is initially provided as a geometric shape, withoutan intrinsic scale. For the purposes of the global error calculations,the models were scaled using the height of the model profile as areference such that the model height was set to unity and all othercoordinates became ratios of this maximum height. Similarly, eachaverage profile was re-scaled such that each x and y coordinate wasexpressed as a ratio of the average profile maximum height. Theglobal error was then estimated using a RMS calculation of thepoint-by-point differences between the 256 points of the scaledmodel profile and the corresponding points in the re-scaled aver-age profile produced by each potter, for the eight experimentalconditions. As the RMS was performed on ratios, a simple �100multiplication meant the global error could be finally expressedas a percentage (the maximum height of the reproduced vesselsbeing the reference value). The partial error was defined as the dif-ference between the dimensions of the model and those of theaverage profile produced by each potter for each condition. In or-der to maintain the original units, the models were each scaledto the maximum height of the corresponding average profiles.The rescaled model dimensions were subtracted from the corre-sponding dimensions of the average profiles to give the partialerror estimates. Hence, partial error was calculated solely for thebase and aperture of the open shapes and for the base, aperture,maximal diameter and height of the maximal diameter of theclosed shapes; it could not be calculated for the height.

Repeated measures analyses of variance (ANOVA) on both theCVs and the global error of reproduction were used to test theeffects of Group, Shape and Mass. Significant (a = .05) main effectsand interactions were further explored using Newman–Keuls post-hoc tests. These statistical analyses, as well as the absolute valuesof the partial errors, enable us to assess the inter-potter variabilitywithin each group and from one group to the other.

Results

Variability of the experimental production

Fig. 3 shows the cross-sectional profiles of the complete assem-blage (40 vessels) thrown by one representative potter of each cul-tural group. By representative potter, we mean a potter whoseexperimental production displayed metric dimensions close tothe overall group average. Visual inspection of this figure com-bined with the low CVs (Table 2) indicate a low intra-individualvariability (i.e. between-specimens variability) for all the threegroups, which in turn indicates a fair degree of standardizationas would be expected among highly skilled potters.

The analysis of variance performed on the CVs averaged overdifferent absolute dimensions reveals a non-significant effect ofthe factor Mass (F(1,18) = 3.89, ns) but a significant effect of thefactor Group (F(2,18) = 3.56, p < .05), a clear-cut effect of the factorShape (F(3,54) = 20.15, p < .001), and an interaction between thesetwo factors Group and Shape (F(6,54) = 2.75, p < .02). For theFrench assemblage, the CVs of the closed shapes (6.7% and 7.8%,respectively for the sphere and vase) are higher than those of theopen shapes (4.3% and 5.0%, respectively for the cylinder andbowl). For the Prajapati assemblage the CV of the vase (8.0%) ishigher than those of the three other shapes which are undifferen-tiated (5.9%, 4.5%, and 5.9%, respectively for the cylinder, bowl andsphere). And for the Multani assemblage, the CV of the vase (6.1%)is higher than that of the bowl (4.1%) and undifferentiated fromthose of the cylinder (5.4%) and sphere (5.3%). Interestingly, forthe three groups, the CV of the vase – which was the most difficult

Fig. 3. Complete assemblages (40 pots) of a representative potter for each group. The cross-sectional profiles of the eight experimental conditions are shown: the four shapes(cylinder, bowl, sphere, and vase) and the two masses of clay: 0.75 kg in blue and 2.25 kg in red. (For interpretation of the references to color in this figure legend, the reader isreferred to the web version of this article.)

Table 2Coefficients of variation of the absolute dimensions of the production. The values are presented for the three groups (French, Prajapati and Multani) and the eight experimentalconditions (four shapes * two masses). C: Cylinder, B: Bowl, S: Sphere, V: Vase, H: Height, B: Base, A: Aperture, MD: Maximal Diameter, HMD: Height of Maximal Diameter.

Shape Mass F Pr Mu

(kg) H B A MD HMD H B A MD HMD H B A MD HMD(%) (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) (%)

C 0.75 4.5 4.2 4.7 3.7 8.9 6.3 6.4 5.1 4.82.25 4.6 3.6 4.2 3.4 6.5 6.9 5.8 5.8 4.4

B 0.75 5.0 7.2 3.4 5.4 6.5 3.3 5.4 4.6 2.82.25 5.5 5.6 3.3 4.1 4.4 3.5 4.5 4.5 2.5

S 0.75 5.6 7.6 9.9 3.5 8.5 4.0 7.8 8.2 3.3 5.7 4.0 5.1 6.1 4.0 6.62.25 6.8 5.5 9.6 3.2 6.6 4.0 6.9 9.7 3.7 6.2 6.3 4.7 6.3 1.9 7.5

V 0.75 7.6 6.9 12.6 3.9 9.8 7.4 7.6 8.6 3.3 12.5 4.7 6.2 6.9 4.3 8.72.25 6.5 6.4 11.7 2.8 9.3 6.0 8.8 10.1 3.4 12.0 4.7 6.4 7.8 3.1 7.9

102 E. Gandon et al. / Journal of Anthropological Archaeology 33 (2014) 99–107

shape – is higher than those of the other three shapes. Conversely,the shape reproduced with the lowest CV – which would indicatethe more familiar shape – is different for the French group (cylin-der) and the two Indian groups (bowl).

The global copying error

Figs. 4 and 5 show results related to the global copying errors.Two main results can be ascertained. On the one hand, the strongermorphological changes are found with the vase (for both masses)

and with the 2.25 kg sphere. Within all three groups, the pottershave transformed the vase shape – an atypical and difficult ceramicshape – to a sphere shape, which is an easier and more classicalceramic shape. On the other hand, the morphological changesobserved with the cylinder, the bowl and the 0.75 kg spheredepend on the groups.

Fig. 4 illustrates the average (i.e. between-specimens) profiles ofthe vessels thrown by one potter representative of each group inthe eight experimental conditions; they are presented alongsidethe profiles of the corresponding models. Except for the 0.75 kg

Fig. 4. Productions and corresponding models for a representative potter in each group. In blue, average (i.e. between-specimens) cross-sectional profiles of the reproducedvessels; in red, the profile of the models. The two masses (0.75 and 2.25 kg) are in small and big sizes. (For interpretation of the references to color in this figure legend, thereader is referred to the web version of this article.)

Fig. 5. Global copying error. The eight experimental conditions (four shapes * twomasses) are represented from left to right. Blue: French, red: Prajapati, and green:Multani. (For interpretation of the references to color in this figure legend, thereader is referred to the web version of this article.)

E. Gandon et al. / Journal of Anthropological Archaeology 33 (2014) 99–107 103

cylinder thrown by the French and Multani potters and for the0.75 kg sphere thrown by the Multani potters, all the replicasdeviate noticeably from the models. The morphological changesare especially high for the vase (in both masses) and – to a lesserextent – for the 2.25 kg sphere and for the 2.25 kg French bowl.

Fig. 5 presents the average (i.e. between-participant) globalcopying error, for the three groups of potters and for all eightexperimental conditions. Except for the bowl shape (in bothmasses), the higher global copying error is obtained by the Praja-pati potters. This result appears in a statistical main effect of thefactor Group (F(2,18) = 8.07, p < .01) and in post-hoc analysis(p < .05). While the effect of Mass is not significant(F(1,18) = 1.76, ns), Shape has a clear-cut effect (F(3,54) = 13.77,

p < .001) and also a clear-cut interaction with the factor Group(F(3,54) = 5.62, p < .001). For the French and Multani groups, theglobal copying error of the vase (17.6% and 15.7%, respectivelyfor the F and Mu groups) and that of the bowl (15.4% and 18.9%)are undifferentiated and superior to the global copying error ofthe sphere (6.4% and 7.4%) which is itself superior to that of thecylinder (2.8% and 3.4%). For the Prajapati group, the global copy-ing error of the vase (22.7%) is superior to that of the three othersshapes, the global copying error of cylinder (5.8%) is inferior tothose of the three others shapes and the global copying error ofthe sphere and bowl are undifferentiated (10.7% and 9.9%, respec-tively for the sphere and bowl).

The partial copying error

Results concerning the partial copying error provide a deeperunderstanding of the morphological changes applied to the modelshapes during their reproduction. Tables 3 and 4 show – for each ofthe 21 potters – the partial copying error, for the open and theclosed shapes respectively. We observe that – in the three groups– (a) none of the models has been faithfully reproduced as alreadyunderlined by the global copying error; (b) inter-individual varia-tion is strong in the sense that the amplitude and sign (+ or �) ofthe partial copying error vary from one potter to the other, evenwhen they belong to the same cultural group. However, this indi-vidual ‘blue print’ in the models reproduction is itself enclosed in-side the overall influence of the group. Indeed, in each of the threegroups, most of the partial copying errors that are superior to 1 cmare observable for the productions of all the group members.

Table 3Partial copying error for the cylinder and bowl. Potters 1–9 constitute the F group,potters 10–15 constitute the Pr group and potters 16–21 constitute the Mu group. Ineach group, the average (between-potter) partial error is also presented (lines in boldfont). To give an indication of significance, we have underlined the average partialerrors superior to 1 cm and whose sign (� or +) is common to more than 70% of themembers of the group. B: Base, A: Aperture.

Cylinder Bowl

0.75 kg 2.25 kg 0.75 kg 2.25 kg

B A B A B A B A(cm) (cm) (cm) (cm) (cm) (cm) (cm) (cm)

P1 �0.1 0.2 �2.3 �2.4 2.3 �6.4 1.7 �11.5P2 �0.2 0.5 �6.1 �6.0 1.6 �4.9 1.1 �9.6P3 1.0 1.2 �0.1 �0.4 �1.4 �6.6 �2.4 �8.3P4 1.5 1.7 �1.4 �0.4 2.7 �0.8 2.9 0.4P5 2.8 2.4 1.3 1.0 2.3 �5.3 3.7 �6.8P6 0.8 1.4 �0.9 0.3 1.7 �7.9 2.2 �11.0P7 �0.8 0.0 �1.2 �1.1 2.0 �2.3 4.5 0.9P8 0.0 0.6 �1.8 �1.5 1.3 �6.1 2.1 �4.0P9 �0.9 �0.6 �2.0 �2.2 �0.5 �8.7 �1.1 �11.6m 0.4 0.8 �1.6 �1.4 1.3 �5.4 1.6 �6.8P10 �2.3 �0.5 �4.7 �4.7 1.6 �0.7 2.5 �1.0P11 �2.6 �2.0 �7.7 �7.6 0.8 �4.2 2.0 �6.6P12 �2.2 �1.0 �3.6 �3.4 3.0 0.2 4.5 �3.4P13 �2.0 �0.8 �3.7 �4.4 2.2 1.0 5.3 3.1P14 �2.5 �2.2 �4.0 �5.1 2.3 1.1 3.8 1.0P15 �1.5 �0.1 �2.8 �3.6 2.1 �4.0 5.2 �3.8m �2.2 �1.1 �4.4 �4.8 2.0 �1.1 3.9 �1.8

P16 �0.7 0.5 �1.9 �1.3 3.1 �0.8 3.8 �2.5P17 0.8 1.5 �1.5 �0.6 4.5 5.0 6.7 8.6P18 �1.2 1.3 �4.3 �0.5 3.4 1.9 4.2 �0.9P19 �1.4 �0.8 �2.9 �2.6 3.6 4.3 4.7 3.0P20 2.0 1.3 0.9 �1.6 3.6 3.7 5.7 2.1P21 1.1 1.5 �2.7 �2.8 5.3 3.9 5.9 0.9m 0.1 0.9 �2.1 �1.6 3.9 3.0 5.2 1.9

104 E. Gandon et al. / Journal of Anthropological Archaeology 33 (2014) 99–107

More precisely, the 0.75 kg cylinder has been reproduced with asmall copying error in the two groups French and Multani: thus thebase and the aperture have increased by less than 1 cm. But in the

Table 4Partial copying error for the sphere and vase. Potters 1–9 constitute the F group, potters 10the average (between-potter) partial error is also presented (lines in bold font). To give an1 cm and whose sign (� or +) is common to more than 70% of the members of the group

Sphere

0.75 kg 2.25 kg

B A DM HDM B A DM HDM(cm) (cm) (cm) (cm) (cm) (cm) (cm) (cm)

P1 1.9 1.3 �0.9 0.1 0.8 �3.2 �4.4 �0.1P2 �0.3 0.4 �1.5 0.6 �2.2 �2.6 �4.9 0.7P3 �0.5 �0.3 �2.3 0.6 �0.2 �0.8 �4.6 0.8P4 2.4 0.4 0.1 0.2 3.0 �0.3 �0.9 0.4P5 0.8 0.7 �1.6 0.1 0.5 �0.5 �4.6 0.3P6 3.0 0.9 1.8 0.2 4.5 �0.3 2.0 0.3P7 1.5 0.2 �1.2 �0.1 3.3 �0.4 �1.0 �0.5P8 �0.4 �1.6 �2.7 0.6 �0.7 �3.0 �4.3 0.2P9 3.8 2.8 1.3 0.4 5.5 0.8 �0.1 0.0m 1.3 0.5 �0.8 0.3 1.6 �1.1 �2.5 0.2

P10 �1.4 �1.4 �2.5 0.4 �2.1 �4.2 �5.2 0.6P11 �2.3 �2.5 �4.9 1.3 �3.0 �5.5 �7.9 1.4P12 0.5 0.8 �1.4 1.0 �1.5 �2.1 �5.9 1.2P13 0.3 �1.1 �0.9 0.8 0.0 �4.0 �2.9 0.4P14 �0.2 �1.1 �2.0 1.1 �1.4 �4.4 �4.2 0.8P15 0.8 �0.2 �3.0 0.9 0.4 �2.7 �4.7 1.0m �0.4 �0.9 �2.4 0.9 �1.3 �3.8 �5.1 0.9

P16 0.0 �2.1 �2.7 0.6 �0.1 �2.9 �4.5 1.0P17 2.0 �0.5 �1.3 0.1 1.2 �2.3 �3.5 0.3P18 �0.8 �1.5 �3.3 0.6 �0.3 �2.7 �4.4 0.1P19 1.6 0.3 0.0 0.2 1.4 �2.3 �2.6 1.1P20 3.1 2.0 1.0 0.5 4.1 �0.1 �0.2 0.9P21 3.2 �0.6 0.9 0.2 4.1 �1.7 1.6 0.4m 1.5 �0.4 �0.9 0.4 1.7 �2.0 �2.3 0.6

Prajapati group, these same dimensions have significantly de-creased (�2.2 cm and �1.1 cm, for the base and aperture respec-tively). The 2.25 kg experimental cylinder is narrower than themodel in each of the three groups. This is visible in the decreaseof both the base and aperture. However, this decrease is significantonly in the Prajapati group (�4.4 cm and �4.8 cm, for the base andaperture respectively). Concerning the bowl, the geometric devia-tions are larger than those of the cylinder and specific to eachgroup. The French potters have significantly reduced the aperture(0.75 kg: �5.4 cm; 2.25 kg: �6.8 cm) and slightly increased thebase (0.75 kg: +1.3 cm; 2.25 kg: +1.6 cm); the Prajapati pottershave slightly reduced the aperture (0.75 kg: �1.1 cm; 2.25 kg:�1.8 cm) and increased the base to a larger extent (0.75 kg:+2.0 cm; 2.25 kg: +3.9 cm); and finally the Multani potters have in-creased both the aperture (0.75 kg: +3.0 cm; 2.25 kg: +1.9 cm) and,to a larger extent, the base (0.75 kg: +3.9 cm; 2.25 kg: +5.2 cm).

Considering the closed shapes, Table 4 indicates – in line withthe previous observations (Fig. 4) – that the main morphologicalchanges are related to the maximal diameter and are much morepronounced for the vase than for the sphere. In the three groups,these two shapes (sphere and vase) in both 0.75 and 2.25 kgmasses were reproduced with a smaller maximal diameter thanthat of the model. Maximal diameter decreases by more than5 cm for the 0.75 kg vase and by more than 7 cm for the 2.25 kgvase. This decrease of the maximal diameter is combined – in bothmasses – with an increase of the height of the maximal diameter.This increase is of more than 1 cm in the Prajapati and Multanigroups and less than 1 cm in the French group. For the sphere,the decrease of the maximal diameter is significant (>2 cm) forthe 2.25 kg vessels in the three groups and for the 0.75 kg vesselsin the Prajapati group. As for the vase, this decrease of the maximaldiameter is accompanied by an increase of the height of the max-imal diameter. However, this increase is limited (<1 cm) for thetwo masses and for the three groups. Some significant deviationsare also observed in the base and aperture of the closed shapes.

–15 constitute the Pr group and potters 16–21 constitute the Mu group. In each group,indication of significance, we have underlined the average partial errors superior to

. B: Base, A: Aperture, MD: Maximal Diameter, HMD: Height of Maximal Diameter.

Vase

0.75 kg 2.25 kg

B A DM HDM B A DM HDM(cm) (cm) (cm) (cm) (cm) (cm) (cm) (cm)

0.0 0.1 �11.0 1.2 2.0 �3.0 �13.6 0.6�1.0 0.7 �6.5 1.4 �3.6 �2.4 �15.3 2.1

6.5 1.5 0.4 0.2 11.9 1.6 2.5 �0.43.0 0.7 �4.6 0.7 5.6 1.1 �5.0 0.62.6 2.0 �4.3 0.5 2.4 0.9 �10.8 0.52.1 �0.1 �3.5 1.2 2.3 �1.2 �6.4 2.31.4 0.7 �7.6 0.4 5.1 0.7 �9.5 �0.12.0 0.5 �5.2 0.6 2.3 �2.3 �9.5 0.34.7 1.9 �4.0 0.5 5.7 �1.0 �8.1 0.3

2.4 0.9 �5.2 0.7 3.7 �0.6 �8.4 0.7

�2.7 �0.8 �7.6 2.0 �4.7 �2.6 �13.2 2.7�3.9 �1.0 �11.1 2.7 �5.2 �3.4 �18.0 3.4�2.0 2.0 �8.7 2.3 �4.2 �0.1 �15.6 3.9�0.3 �0.1 �5.2 1.4 0.3 �0.8 �7.7 1.8�0.3 1.3 �4.7 2.3 �2.0 �0.1 �6.9 2.4

1.1 2.1 �5.5 2.8 0.6 �0.3 �8.9 2.7

�1.3 0.6 �7.1 2.2 �2.5 �1.2 �11.7 2.80.1 1.0 �6.7 1.0 �1.4 0.4 �10.6 1.81.5 1.3 �5.1 1.2 0.9 0.6 �8.6 1.4�0.3 0.9 �6.7 1.5 �0.2 0.7 �7.9 1.2

0.6 1.9 �3.9 1.3 �0.3 0.1 �8.6 3.04.0 1.2 �2.7 0.8 4.8 0.5 �5.0 1.61.2 0.9 �5.2 1.2 1.5 �0.2 �6.8 1.7

1.2 1.2 �5.0 1.2 0.9 0.3 �7.9 1.8

E. Gandon et al. / Journal of Anthropological Archaeology 33 (2014) 99–107 105

The aperture of the 2.25 kg sphere has decreased in the threegroups (�1.1, �3.8, and �2.0 cm, for the French, Prajapati andMultani groups respectively). The aperture of the 2.25 kg vasehas decreased in the Pradjapati group (�1.2 cm); and the apertureof the 0.75 kg vase has increased in the Multani group (+1.2 cm).The base of the vase – in both masses – has increased in the Frenchand Multani groups and decreased in the Prajapati group.

Fig. 6. The transmission of motor skill in the ecological perspective. Motor skillemerges from the interplay of three factors that together define the perceptual-motor workspace: the task, the (physical) environment, and the agent (Newell,1986; Reed and Bril, 1996). In order to successfully perform the task, the agent (A)interacts with his environment (E) through informational (E to A) and mechanical(A to E) couplings (Bootsma, 1998; Gibson, 1979; Kugler and Turvey, 1987; Warren,2006). As a result of the joint operation of these factors, the agent’s neuromuscularsystem is harnessed into functional (i.e. task-appropriate) behavior. The body’sneuromuscular system is flexible and a specific task can be achieved by severalequivalent motor skills (Bernstein, 1967). The set of motor skills corresponding to agiven task is called the motor space. In the course of motor learning, the agentexplores the workspace in order to discover the sensory information relevant toachieve the task (Handford et al., 1997; Jacobs and Michaels, 2007; Newell, 1991).Several studies have shown that social pointers—resulting from interaction withindividuals already mastering the task—channel the learner’s attention so that hisexploratory activity occurs over an optimal area of the workspace (Al-abood et al.,2001; Boyer and Bril, 2001; Bril, 1986, 2002a; Janelle et al., 2003; Neuman and Gray,2013; Scully and Newell, 1985).

Discussion

The aim of this study was to provide experimental tests in orderto explore the copying error phenomenon, and more precisely toinvestigate how it depends on the potter’s skills and cultural con-texts. For this purpose, a field experiment was conducted with 21craft experts distributed between three different cultural groups.They were asked to replicate five specimens of four shapes intwo sizes. Each shape represented different degrees of throwingdifficulties.

Two main results were obtained. The first result is that none ofthe models were reproduced precisely by the potters whatever thecultural group they belong to as shown by the CVs of the dimen-sions of the replicas and the distances between the dimensionsof the replicas and the models. In this regard, the CVs obtainedare very much coherent with the ones proposed by theoretical re-searches on the physiological limitation of the perceptual system.However, our results show that not only perception of the artifactshapes but also motor skill play in the copying error. This is firstsuggested by the analysis of the CVs which shows a clear-cut effectof the factor shape, knowing that shapes are a good proxy forassessing throwing difficulty and therefore motor skills (Gandonet al., 2011). This is also confirmed by the values obtained for thevase which presents the highest mechanical stresses and thereforethe highest throwing difficulty. In the three groups, this shape ob-tains the highest CVs and presents the highest distances from themodel as expressed by the global and partial error scores. On thecontrary, the cylinder whose shape is the least mechanicallystressed claims the best scores for global copying error. The roleof the difficulty of the task in the copying error is also visible inthe nature of the copying errors. Among the three groups, the vaseis thus systematically transformed into a sphere with a smallermaximum diameter localized close to the mid-height of the vessel.This deformation enables the potters to reduce the risks of collapseof the vessel. In summary, there are both perceptual and motorlimitations which lead to variations between the model artifactand the reproduced artifact, and the copying error can to be seenas emerging from both motor skill and perception of the artifactshapes.

The second main result is that the direction of the copying errordepends on the learning niches. Potters reproduced the modelswith culturally distinctive traits. This applies even to the Frenchpotters despite this group being more heterogeneous than the oth-ers. This result is suggested by the analysis of the CVs which showsthat not only the factor Shape has an effect, but also the factorGroup. The morphological type produced by the French pottersproduced with the lowest CV is the cylinder; the one producedby the Indian potters with the lowest CV is the bowl. This differ-ence can be explained in terms of the familiarity of the potters withthose shapes. In the case of the French potters, cylinders are com-monly learned as ‘‘roughout’’ before learning how to shape a‘‘closed’’ shape (Colbeck, 1979). In the case of the Indian potters,bowls are the first morphological shape to be learned and are massproduced at least once a year at the occasion of religious festivals(Roux and Corbetta, 1989). The role of the learning niches in thecopying error is also suggested by the scores of the partial copyingerror. The distances between the model and the replicas are of a

same order within each group, and of a different order from onegroup to the other. This is especially visible for the open shapeswhich are less subject to throwing difficulties. Their aperture andbase are increased or decreased by the same amplitude within eachgroup, but this amplitude differs between groups. The same resultswere obtained for the base and aperture of the closed shapes.

The importance of the learning niche in the process of copyingerrors can be well explained by the mechanisms underlying thetransmission of motor skills. According to the approach developedby the ecological psychology (Bernstein, 1967; Gibson, 1979), allour daily motor skills – and their environmental effects – resultfrom (i) the constraints of the task (the goal of the skill), (ii) theconstraints of the surrounding physical environment and (iii) theconstraints of the agent (body constitution and experience)(Fig. 6). This set of constraints is called the workspace (Handfordet al., 1997; Newell, 1986; Reed and Bril, 1996). Due to the flexibil-ity of the neuromuscular system, achieving a task is not confined toa unique motor skill (Bernstein, 1967). A specific task can beachieved by several equivalent motor skills, a set of motor solu-tions called the motor space. The ‘selection’ by societies of subsetof motor solutions corresponds to the traditional (i.e. cultural)manner to achieve the task in that society (Bril, 2002a). This cul-tural selection occurs during the apprenticeship of the skill. Indeed,in the course of motor learning, the individual explores the work-space in order to discover the sensory information relevant toachieve the task (Jacobs and Michaels, 2007; Newell, 1991; Reedand Bril, 1996). Previous studies have shown that this explorationis guided by the observation of a tutor’s model (Al-abood et al.,2001; Bandura, 1969; Boyer and Bril, 2001; Bril, 1986, 2002b;Handford et al., 1997; Janelle et al., 2003; Neuman and Gray,

106 E. Gandon et al. / Journal of Anthropological Archaeology 33 (2014) 99–107

2013; Scully and Newell, 1985; Shea et al., 2000). The role of thetutor model is to educate the learner attention in order to orienthis exploratory activity over an optimal area of the workspace (Bril,1986, 2002a; Ingold, 2001). This social guidance – whichcorresponds to the cultural transmission of motor skill – not onlyfacilitates the learning process but also paves the way to the devel-opment of a specific, culturally situated, motor skill (Bril, 2002a).

This cultural transmission constrains the process of copyingerrors as shown by the deformations of the replicas in relationshipwith the cultural groups. It explains that the process of copyingerror is shaped by the learning niches. It suggests that the morethe learning niches will be specific to morphological types, the lessthe copying errors should be a source of significant deviations.

Conclusion

In conclusion, we have provided novel experimental data, inwhich the copying error phenomenon is explored not only fromthe perception perspective, but also from the motor skill perspec-tive. The results obtained show that copying error is an unintendedand universal phenomenon in the sense that it results from the hu-man perceptual-motor limitations that prevent the craftsman froma perfect replication of a given model, even if the craftsman is anexpert. However, it depends both on throwing difficulties andlearning niches. It follows that its amplitude and directions arepredictable.

These results have implications for better understanding theevolution of artifacts whose manufacture involves necessarily bothperception and motor skills. As a main consequence, copying erroris expected to be constrained by the cultural context into whichlearning takes place and therefore to be a source of predictable var-iation. In other words, variability of artifacts produced by copyingerrors should be orientated, inscribing these artifacts within cul-tural lineages of objects. In this regard, contrary to theoreticalhypotheses, it is highly improbable that solely the copying errorphenomenon could produce new types of objects and/or new lin-eages of objects. On the contrary, it is expected that drift, andtherefore unintended evolution, will present a cultural orientation,or a cultural ‘‘signature’’. In other words, the copying error shouldlead to artifact shapes close to the usual shapes of the group. Also,the magnitude of the copying error should vary in proportion tothe level of difficulty of the artifacts. In this regard, objects whichhave evolved from the same cultural context, but independentlyfrom multiple chains of transmissions, should present comparablecopying errors unless other evolutionary forces were at work. Evo-lutionary models should be able to take into account these specific-ities on the basis of the types produced in the original culturalcontext, their difficulty level and their frequency of production.These would be key factors used in the design of mathematicalmodels of such predictable drifts.

Acknowledgments

We are grateful to Avshalom Karasik for his assistance regard-ing the data processing and to Leore Grosman for her advices dur-ing the writing process. This study was funded by Aix-MarseilleUniversity and the Fyssen foundation.

References

Al-abood, S.A., Davids, K., Bennett, S.J., 2001. Specificity of task constraints andeffects of visual demonstrations and verbal instructions in directing learners’search during skill acquisition. J. Mot. Behav. 33, 295–305.

Bandura, A., 1969. Principles of Behavior Modification. Holt Rinehart and Winston,New York.

Bernstein, N.A., 1967. The Coordination and Regulation of Movements. PergamonPress, London.

Bootsma, R.J., 1998. Ecological movement principles and how much informationmatters. In: Post, A.A., Pijpers, J.R., Bosch, P., Boschker, M.S.J. (Eds.), Models inHuman Movement Science. PrintPartners Ipskamp, Enschede, pp. 51–63.

Boyd, R., Richerson, P.J., 1985. Culture and the Evolutionary Process. University ofChicago Press, Chicago.

Boyer, I., Bril, B., 2001. Structuring a child activity: a comparative study of mother–child interaction in a complex bimanual task. In: Van Der Kamp, J., Ledebt, A.,Savelsbergh, G., Thelen, E. (Eds.), Proceedings of the International Conference onMotor Development and Learning in Infancy. IFKB, Amsterdam, pp. 55–58.

Bril, B., 1986. The acquisition of an everyday technical motor skill: the pounding ofcereals in Mali (Africa). In: Wade, M.G., Whiting, H.T.A. (Eds.), Themes in MotorDevelopment. Martinus Nijhoff, Dordrecht, pp. 315–326.

Bril, B., 2002a. L’apprentissage de gestes techniques: ordre de contraintes etvariations culturelles. In: Bril, B., Roux, V. (Eds.), Le Geste Technique. RéflexionsMéthodologiques et Anthropologiques. Erès, Ramonville Saint-Agne, pp. 113–151.

Bril, B., 2002b. Apprentissage et contexte. Intellectica 35, 251–268.Cavalli-Sforza, L.L., Feldman, M., 1981. Cultural Transmission and Evolution: A

Quantitative Approach. Princeton University Press, Princeton.Colbeck, J., 1979. Pottery the Technique of Throwing. Batsford B.T. Ltd., London.Durham, W.H., 1991. Co-evolution: Genes, Culture and Human Diversity. Standford

University Press, Standford.Eerkens, J.W., 2000. Practice makes within 5% of perfect: visual perception, motor

skills, and memory in artifact variation. Curr. Anthropol. 41, 663–668.Eerkens, J.W., Bettinger, R.L., 2001. Techniques for assessing standardization in

artifact assemblages: can we scale material variability? Am. Antiq. 66, 493–504.Eerkens, J.W., Lipo, C.P., 2005. Cultural transmission, copying errors, and the

generation of variation in material culture and the archaeological record. J.Anthropol. Archaeol. 24, 316–334.

Eerkens, J.W., Lipo, C.P., 2007. Cultural transmission theory and the archaeologicalrecord: providing context to understanding variation and temporal changes inmaterial culture. J. Archaeol. Res. 15, 239–274.

Gandon, E., Casanova, R., Sainton, P., Coyle, T., Roux, V., Bril, B., Bootsma, R.J., 2011. Aproxy of potters’ throwing skill: ceramic vessels considered in terms ofmechanical stress. J. Archaeol. Sci. 38, 1080–1089.

Gardin, J.C., 1980. Archaeological Constructs: An Aspect of Theoretical Archaeology.Cambridge University Press, Cambridge.

Gibson, J.J., 1979. The Ecological Approach to Visual Perception. Houghton-Mifflin,Boston.

Hamilton, M.J., Buchanan, B., 2009. The accumulation of stochastic copying errorscauses drift in culturally transmitted technologies: quantifying Clovisevolutionary dynamics. J. Anthropol. Archaeol. 28, 55–69.

Handford, C., Davids, K., Bennett, S., Button, C., 1997. Skill acquisition in sport: someapplications of an evolving practice ecology. J. Sports Sci. 15, 621–640.

Henrich, J., McElreath, R., 2003. The evolution of cultural evolution. Evol.Anthropol.: Issues, News, Rev. 12, 123–135.

Ingold, T., 2001. From the transmission of representations to the education ofattention. In: Whithehouse, H. (Ed.), The Debated Mind: EvolutionaryPsychology Versus Ethnography. Berg, Oxford, pp. 113–153.

Jacobs, D.M., Michaels, C.F., 2007. Direct learning. Ecol. Psychol. 19, 321–349.Janelle, C.M., Champenoy, J.D., Coombes, S.A., Mousseau, M.B., 2003. Mechanisms of

attentional cueing during observational learning to facilitate motor skillacquisition. J. Sports Sci. 21, 825–838.

Karasik, A., Smilansky, U., 2011. Computerized morphological classification ofceramics. J. Archaeol. Sci. 38, 2644–2657.

Kempe, M., Lycett, S., Mesoudi, A., 2012. An experimental test of the accumulatedcopying error model of cultural mutation for acheulean handaxe size. PLoS One7, e48333.

Kramer, C., 1997. Pottery in Rajasthan: Ethnoarchaeology in Two Indian Cities.Smithsonian Institution Press, Washington.

Kugler, P.N., Turvey, M.T., 1987. Information, Natural Law, and the Self-Assembly ofRhythmic Movement: Theoretical and Experimental Investigation. LawrenceErlbaum Associates, Hillsdale, NJ.

Mesoudi, A., O’Brien, M.J., 2008. The cultural transmission of Great Basin projectile-point technology I: an experimental simulation. Am. Antiq., 3–28.

Mesoudi, A., Whiten, A., 2004. The hierarchical transformation of event knowledgein human cultural transmission. J. Cogn. Cult. 4, 1–24.

Mesoudi, A., Whiten, A., Dunbar, R., 2006. A bias for social information in humancultural transmission. Br. J. Psychol. 97, 405–423.

Neiman, F.D., 1995. Stylistic variation in evolutionary perspective. Am. Antiq. 60, 7–36.

Neuman, B., Gray, R., 2013. A direct comparison of the effects of imagery and actionobservation on hitting performance. Movement Sport Sci. 1, 11–21.

Newell, K.M., 1986. Constraints on the development of coordination. In: Wade,M.G., Whiting, H.T.A. (Eds.), Motor Development in Children: Aspects ofCoordination and Control. Martinus Nijhoff, Boston, pp. 341–360.

Newell, K.M., 1991. Motor skill acquisition. Annu. Rev. Psychol. 42, 213–237.Orton, C., Tyers, P., Vince, A., 1993. Pottery in Archaeology. Cambridge University

Press, Cambridge.Read, D.W., 2009. Artifact Classification: A Conceptual and Methodological

Approach. Left Coast Press.Reed, E.S., Bril, B., 1996. The primacy of action in development. A commentary of N.

Bernstein. In: Latash, M.L., Turvey, M.T., Bernstein, N.A. (Eds.), Dexterity and itsDevelopment. Lawrence Erlbaum Associates, Mahwah, pp. 431–451.

Richerson, P.J., Boyd, R., 2005. Not By Genes Alone: How Culture TransformedHuman Evolution. University of Chicago Press, Chicago.

E. Gandon et al. / Journal of Anthropological Archaeology 33 (2014) 99–107 107

Roux, V., 2003. Ceramic standardization and intensity of production: quantifyingdegrees of specialization. Am. Antiq. 68, 768–782.

Roux, V., 2013. Spreading of innovative technical traits and cumulative technicalevolution: continuity or discontinuity? J. Archaeol. Method Theory 20, 312–330.

Roux, V., Corbetta, D., 1989. The Potter’s Wheel. Craft Specialization and TechnicalCompetence. Oxford and IBH Publishing, New Delhi. <http://www.irit.fr/arkeotek/ebooks/potterswheel.pdf>.

Saraswati, B., Behura, N.K., Bose, N.K., 1966. Pottery techniques in peasant India.Anthropol. Surv. India 13.

Scully, D., Newell, K., 1985. Observational learning and the acquisition of motorskills: toward a visual perception perspective. J. Human Movement Stud. 11,169–186.

Shea, C.H., Wright, D.L., Wulf, G., Whitacre, C., 2000. Physical and observationalpractice afford unique learning opportunities. J. Mot. Behav. 32, 27–36.

Shennan, S., 2008. Evolution in archaeology. Annu. Rev. Anthropol. 37, 75–91.Steele, J., Glatz, C., Kandler, A., 2010. Ceramic diversity, random copying, and tests

for selectivity in ceramic production. J. Archaeol. Sci. 37, 1348–1358.Warren, W.H., 2006. The dynamics of perception and action. Psychol. Rev. 113, 358–

389.Washburn, D.K., 2001. Remembering things seen: experimental approaches to the

process of information transmittal. J. Archaeol. Method Theory 8, 67–99.Whallon, R., 1972. A new approach to pottery typology. Am. Antiq. 37, 13–33.