at SciVerse ScienceDirect

Journal of Archaeological Science 40 (2013) 3903e3913

Contents lists available

Journal of Archaeological Science

journal homepage: http : / /www.elsevier .com/locate/ jas

Nursing mothers and feeding bottles: reconstructing breastfeedingand weaning patterns in Greek Byzantine populations (6the15thcenturies AD) using carbon and nitrogen stable isotope ratios

Chryssi Bourbou a,*, Benjamin T. Fuller b,c,d, Sandra J. Garvie-Lok e, Michael P. Richards b,f

aHellenic Ministry of Culture, 28th Ephorate of Byzantine Antiquities, Sourmeli 24, 73110 Chania, Crete, GreecebDepartment of Human Evolution, Max Planck Institute for Evolutionary Anthropology, 6 Deutscher Platz, D-04103 Leipzig, Germanyc Laboratory of Biodiversity and Evolutionary Genomics, Centre for Archaeological Sciences, University of Leuven, Ch. Debériotstraat 32,B-3000 Leuven, BelgiumdDepartment of Earth System Science, University of California, Irvine, Keck CCAMS Group, B321 Croul Hall, Irvine, CA 92697, USAeDepartment of Anthropology, University of Alberta, Edmonton, Alberta T6G 2H4, CanadafDepartment of Anthropology, University of British Columbia, Vancouver, British Columbia V6T 1Z1, Canada

a r t i c l e i n f o

Article history:Received 11 July 2012Received in revised form11 March 2013Accepted 22 April 2013

Keywords:GreeceByzantine periodStable isotopesBreastfeeding and weaning patterns

* Corresponding author. Almyridos 10, 73133 Cha69735 54028; fax: þ30 28210 44114.

E-mail address: chryssab@gmail.com (C. Bourbou)

0305-4403/$ e see front matter � 2013 Elsevier Ltd.http://dx.doi.org/10.1016/j.jas.2013.04.020

a b s t r a c t

Traditionally, documentary evidence (mainly medical works and the vitae and miracles of saints) servedas the primary source of information for perinatal nutrition in the Byzantine era. In the last decade,however, stable carbon and nitrogen isotope ratio analysis has also been applied for the reconstruction ofByzantine breastfeeding and weaning practices. This paper reviews the documentary evidence forByzantine weaning and compares it to isotopic data for eight Greek Byzantine skeletal samples from thesites of Eleutherna, Kastella, Messene, Sourtara, Stylos, Nemea, Petras and Servia (6the15th centuriesAD). The documentary evidence suggests that Byzantine children were weaned at a relatively late age.The age patterning of the stable isotope data is not as clear as normally seen at single sites. However, thepresence of a higher proportion of elevated values in juveniles aged three years or less suggests thatweaning was completed by the fourth yearda pattern consistent with the written sources. The dataavailable from this study allow us to make some tentative suggestions about cultural and temporaldifferences in weaning, and to recommend directions for further research. A comparison of the Byzantinedata presented here to data published for Roman-era sites from the Mediterranean and Western Europesuggests that the Byzantines maintained a Roman-era practice of relatively late weaning. In contrast,medieval data for Western Europe and the few measurements made on post-Byzantine Greek materialsuggest more variation, with some groups weaning late and others weaning earlier, by two years of age.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Stable isotope ratio analysis of bone collagen has been widelyapplied toGreek skeletal populations to reconstruct adult diets fromprehistoric times through to the Byzantine period (for reviews seeBourbou, 2010; Bourbou et al., 2011; Papathanasiou et al., in press).However, this method has been used less frequently to addressquestions pertaining to juvenile diets. Previous work has beenconducted on Greek populations of the Byzantine era (Bourbouand Garvie-Lok, 2009, in press; Bourbou, 2010) and Bronze Age

nia, Crete, Greece. Tel.: þ30

.

All rights reserved.

(Triantaphyllou et al., 2008), as well as on the Greek colonists ofApollonia on the Black Sea (Kwok, 2007; Holt, 2009). One reason forthe relative paucity of stable isotope studies of weaning in Greekarchaeological populations is the problem of sample size: for manyeras the typical cemetery size is small and the preservation is oftenpoor, with the result that sites often contain insufficient juvenileremains for a confident reconstruction to be done.

This study undertakes a large-scale survey of juvenile humanremains from the early (4the9th centuries AD) and middleByzantine (10the15th centuries AD) periods by integrating theisotopic results of 61 non-adult individuals from eight sites, fromvarious geographical locations in Greece. This work is an expansionof a smaller pilot study that used the same approach (Bourbou andGarvie-Lok, 2009), and acts as a follow up to a previous discussion

C. Bourbou et al. / Journal of Archaeological Science 40 (2013) 3903e39133904

of the Byzantine adult diet based on stable isotope values from thesame eight sites (Bourbou et al., 2011).

In the context of these juvenile isotopic values, the possibleeffects of weaning stress on non-adult mortality in Byzantinepopulations as well as the development of specific pathologicalconditions (i.e., metabolic disorders) at the onset of weaning arediscussed. In addition, the infant feeding habits of the ByzantineEmpire are compared and contrasted to the practices of WesternEurope. Breastfeeding and weaning patterns of Roman and medi-eval populations from the Mediterranean region and WesternEurope have been investigated in a number of relevant isotopicstudies (Dittman and Grupe, 2000; Dupras et al., 2001; Mays et al.,2002; Richards et al., 2002; Herrscher, 2003; Prowse et al., 2004;Turner et al., 2007). However, no attempt has been made tocompare isotopic results between Greek and Western Mediterra-nean populations and it is believed that such a comparison canprovide useful information on differing attitudes toward infantfeeding practices in the two regions.

2. Reconstructing breastfeeding and weaning patterns bystable isotope ratio analysis

Stable isotope studies of breastfeeding andweaning patterns arebased on the shifts in infant tissue d15N and d13C values that typi-cally occur at the onset and termination of breastfeeding. Obser-vations made on fingernail clippings of modern human mothereinfant pairs (Fogel et al., 1989; Fuller et al., 2006a) show that infantd15N values rise rapidly with the onset of breastfeeding, reaching aplateau roughly one trophic level (2e3&) above themother’s tissuevalue. As supplementation with other foods begins the values startto drop, falling to a level similar to the mother’s after nursing hashalted completely. A similar, though subtler, effect (w1&) is pre-sent for d13C and this can be used to better understand the timing ofthe introduction of solid foods (Fuller et al., 2006a).

Since the initial documentation of the effects of nursing andweaning on human d15N values, nitrogen isotope ratio analysis hasbeen used to study nursing and weaning in many archaeologicalhuman populations (e.g., Katzenberg and Pfeiffer, 1995; Katzenberget al., 1996; Schurr, 1997; Herring et al., 1998; Dupras et al., 2001;Richards et al., 2002, 2006; Prowse et al., 2004; Schurr and Powell,2005; Clayton et al., 2006; Fuller et al., 2006b; Dupras and Tocheri,2007; Turner et al., 2007; Jay et al., 2008; Nitsch et al., 2011;Redfern et al., 2012; Howcroft et al., 2012). This application ispossible because bone collagen, like other body proteins, reflectsthe changes in the d15N value of an infant’s diet that occur duringthe initiation and termination of breastfeeding. The most commonapproach to studying weaning through archaeological remains isthe study of bone samples from a number of infants and children ina burial group. These samples are analyzed and the d15N values areplotted by age at death and compared to the mean value for adultfemales in the population. Typically, such studies show d15N valuesfor neonates that are similar to those of adult females. The valuesthen rise fairly rapidly to roughly one trophic level above the meanadult female value as nursing incorporates 15N-enriched collageninto the bones. After a given age, the d15N values start to drop again;the point at which this drop begins is taken to be the age by whichsignificant consumption of protein-rich foods other than breastmilk typically began. Because bone is slower to turnover than mosttissues, this age will lag behind the ‘true’ age at which weaningactually began in the population. However, although the bones ofadults may retain the isotopic signal of an earlier diet for decades(Stenhouse and Baxter, 1979; Hedges et al., 2007), bone turnover ina rapidly growing infant is considerably faster, especially in smalland porous bones such as the ribs (Richards et al., 2002). Obser-vations on large samples of infants from archaeological skeletal

populations have led researchers such as Katzenberg et al. (1996)and Richards et al. (2002) to suggest that the lag time betweenweaning and a shift in infant rib d15N value is quite brief, on theorder of a few months or less.

Although powerful, this application of nitrogen isotopes has itslimitations and requires caution. Because of the delay caused byturnover, the onset of a drop in d15N values must be taken as thetime by which weaning had typically begun for a population ratherthan a precise estimate of weaning age. Problems may also arisefrom any differences between a culture’s approach to weaninghealthy and sick infants, as the weaning pattern reflected in thedeceased infants from a burial population is likely to be biasedtoward the approach preferred for an unhealthy child (Katzenberget al., 1996; Fuller et al., 2003). The delay in the weaning signalcaused by bone turnover can be circumvented to a degree bysampling the growing edge of bones and teeth in order to accesscollagen that was being formed at the time of death (Waters-Ristand Katzenberg, 2010; Waters-Rist et al., 2011). To compensate inpart for the bias resulting from the study of non-survivors the teethof older individuals can be analyzed, using the d15N and d13C valuesof dentine that formed at given ages to provide a record of shiftingdietary d15N and d13C during the childhood years of those whosurvived the weaning process (Fuller et al., 2003; Dupras andTocheri, 2007; Eerkens et al., 2011). However, these approachesrepresent only partial solutions to the biases and inaccuraciesinherent to the method.

As a result of the bias introduced by studying non-survivors, andbecause of a certain normal variation in weaning age expected forany population, studies that reconstruct weaning behavior usingbone collagen values should ideally sample the remains of a largenumber of infants, and where large burial groups are available, thisis commonly done (e.g., Herring et al., 1998; Dupras et al., 2001;Richards et al., 2002; Prowse et al., 2004; Schurr and Powell, 2005).However, this is a problem for researchers studying cultures forwhich large infant burial groups are unavailable for reasons such asa preference for smaller, less centralized burial sites, a tendency tobury infants and adults separately, or small social group size. Insuch cases researchers have chosen to tentatively reconstructweaning patterns based on a small series of juvenile remains,cautioning that the resulting weaning estimates are very general(e.g., Richards et al., 2003). Here we take a different approach,combining juvenile remains from several sites of the same cultureto produce an aggregate sample. The juvenile values from each siteare converted to d15N and d13C departures from that site’s adultfemale mean, allowing them to be directly compared to oneanother. At least one other study has successfully used pooledstable isotope data from different sites to study weaning behavior(Williams et al., 2005). In doing so here, we hope to contribute tothe understanding of weaning practices from the Greek Byzantineperiod and also to illustrate the value of this approach for recon-structing weaning in past human cultures for which large series ofinfant remains are unavailable.

2.1. Feeding the infant in Byzantine Greece: documentary evidence

The Byzantine Empire was the predominantly Greek-speakingcontinuation of the eastern Roman Empire. Although the westernportion of the Roman Empire fell in the 5th century AD its easternportion, including the area known today as Greece, remained underimperial control. As the Byzantine Empire it would persist throughperiods of expansion and decline until its fall to the Ottoman Em-pire in the 15th century AD. The documentary and zooarchaeo-logical evidence for adult diets in Byzantine Greece has beenpresented elsewhere (Bourbou, 2010; Bourbou et al., 2011; Bourbouand Garvie-Lok, in press). To summarize, the adult diet was based

Table 1List of sites for which human remains are studied. For details on site locations anddates, see text.

Site Date (AD) n Reference

Eleutherna 6the7th c. 16 Bourbou, Fuller and Richards 2009a

Kastella 11th c. 7 Bourbou and Richards, 2007;Bourbou and Garvie-Lok 2009

Messene 6the7th c. 5 Bourbou, Fuller and Richards 2009a

Nemea 12the13th c. 2 Garvie-Lok 2001; Bourbou andGarvie-Lok 2009

Petras 12the13th c. 4 Garvie-Lok 2001; Bourbou andGarvie-Lok 2009

Servia 11the15th c. 3 Garvie-Lok 2001; Bourbou andGarvie-Lok 2009

Sourtara 6the7th c. 16 Bourbou, Fuller and Richards 2008a

Stylos 11the12th c. 8 Bourbou, Fuller and Richards 2008a

a Project on the investigation of Byzantine dietary patterns conducted at the MaxPlank Institute (2007e2009).

C. Bourbou et al. / Journal of Archaeological Science 40 (2013) 3903e3913 3905

on grain, olive oil and wine. The main grains used were wheat andbarley, both C3 cultivars; some use of the C4 grain millet is alsodocumented. To these staples were added legumes, fruits, vegeta-bles and nuts, eggs and dairy products, the meat of domesticatedanimals, game, and marine resources. The Orthodox fastingregimen required abstinence from meat on many days of the yearand should have encouragedmarine resource use at all social levels.Stable isotope ratio analyses of archaeological human remains(Garvie-Lok, 2001, in press; Bourbou and Richards, 2007; Bourbouet al., 2008, 2011; Bourbou, 2010; Bourbou and Garvie-Lok, inpress) confirm this general pattern while adding interesting detail.The values suggest a dependence on C3 grains at most sites, withvariable subsidiary use of millet. Generally high d15N values andlow d13C values suggest that dietary protein came primarily frommeat or dairy products rather than legumes or marine resources,although coastal communities do show some evidence for marineresource use.

Information on the foods considered appropriate for infants andchildren comes in part from Byzantine medical treatises discussingperinatal nutrition and the proper selection of a wet-nurse. Theseworks combine the eclectic knowledge of physicians of the RomanImperial period, such as Soranus of Ephesus (1st century AD), withadditional remarks and clarifications (Rey, 2004; Bourbou andGarvie-Lok, 2009; Bourbou, 2010). Especially valuable is Oribasius(c. AD 325e400), who gave detailed information concerning achild’s nutrition from birth through the age of 14 focusing mainlyon the first two years (Raeder, 1964; Laskaratos and Poulakou-Rebelakou, 2003). Additional useful information about the age ofweaning can be found in the lives and miracles of saints, whichcontain passing references to many aspects of daily life includingchild-rearing (Beaucamp, 1982; Bourbou and Garvie-Lok, 2009;Bourbou, 2010).

Breast milk, provided by the mother or a wet-nurse, was seenas the infant’s ideal food (Beaucamp, 1982; Laskaratos andPoulakou-Rebelakou, 2003; Bourbou and Garvie-Lok, 2009).Because most physicians of the time felt that colostrum was un-healthy, they advised a few days’ wait between birth and the onsetof nursing (Temkin, 1991; Laskaratos and Poulakou-Rebelakou,2003). During this period, the newborn was to be fed with clearhoney of high quality, followed by drops of a lukewarm solution ofhoney and water (hydromel) or by a mixture of honey and goat’smilk. After about four days, nursing could commence and for thenext few months the infant was to be exclusively breast-fed.Women were enjoined to eat properly during lactation to ensurethe quality of their milk. The suggested diet was diverse andincluded plentiful animal protein in the form of meats and fish. Itdiffered from the general adult diet in its prohibition of somefoods thought to be damaging to the quality or flavor of the milkdue to their taste, odor or other properties, including bitter oracidic foods, cheese, legumes, nuts, some river fish, and pungentherbs such as garlic (Laskaratos and Poulakou-Rebelakou, 2003).The degree to which a lactating woman’s diet differed from herdiet at other times would have depended on how closely shefollowed these proscriptions and on the importance of the pro-scribed foods in her community. In general, Roman and Byzantinephysicians considered weaning a fairly serious step to beapproached with caution, and suggested a gradual weaning pro-cess. Soranus recommended the introduction of solid food byroughly six months of age. This could consist at first of crumbledbread softened with hydromel (water with honey), milk (fromsheep or goats, as cow’s milk was not used), sweet wine orhoneyed wine, moving on later to spelt soup, very moist porridge,or an egg cooked soft enough to be sipped. However, the child wasnot to be weaned abruptly at this time, since a diet of breast milksupplemented by these soft solid foods was to be continued for at

least another year. Adult foods were to be introduced graduallyand the child would not have been completely weaned until theage of two, if not later (Heiberg, 1921; Olivieri, 1935; Raeder, 1964;Burguière et al., 1990; Temkin, 1991). Descriptions in the lives andmiracles of saints suggest that weaning was sometimes notcompleted until the age of three or four (Bourbou and Garvie-Lok,2009; Bourbou, 2010).

3. Materials and methods

The study includes the isotopic analysis of bone samples takenfrom 61 non-adult human skeletons from eight sites in Greece, bothinland and coastal (Table 1; Fig. 1). All individuals were aged lessthan 13 years at death based on their dental and/or skeletaldevelopment.

Samples from three of the sites (Servia, Petras and Nemea)were analyzed by Garvie-Lok (2001). The Servia (n¼ 3) material isderived from sub-floor graves in the episcopal church of Kastron.This series of burials likely dates to the initial use of the churchafter its construction in the 11th century AD. Samples for Petras(n ¼ 4) come from a rural cemetery directly on the coast dating tothe late 12th or early 13th century AD. The Nemea (n ¼ 2) samplesare from a Byzantine farming community occupying the area inand around ancient Nemea, and date to the 12th through 13thcenturies. The five remaining sites were prepared and analyzed byBourbou and Richards (2007), Bourbou et al. (2008) and Bourbou(2010). The rural site of Eleutherna (n ¼ 16) is located 10 km fromthe north coast of Crete, dating to the 6the7th centuries AD, andthe burials came to light during the excavation of a basilica.Burials from ancient Messene (n ¼ 5) represent an urban popu-lation occupying a fertile valley approximately 25 km from thesouth coast of the Peloponnese. Burials from the rural cemetery ofSourtara (n ¼ 16) also date to the 6the7th centuries AD; the site islocated in inland Northern Greece, close to the site of Servia. Theburials excavated at the church of Ag. Ioannis Theologos at Stylos(n ¼ 8) date to the 11the12th centuries AD. Finally, the cemeteryof Kastella (n ¼ 7) dates to the 11th century AD, a few hundredmeters from the north coast of Crete. These eight cemetery pop-ulations represent communities mainly depending on agricultureand animal husbandry, or involved in other subsidiary activities(e.g., household manufacture, trade, and fishing). Most settle-ments were of rural character, with the exception of Messenewhere a more urban organization of the settlement can beidentified.

Because the sites were originally analyzed in a number ofdifferent studies, preparation methods varied. Bone samples from

Fig. 1. Map showing locations of sites discussed.

C. Bourbou et al. / Journal of Archaeological Science 40 (2013) 3903e39133906

Nemea, Petras and Servia represent ribs and long bones. Collagenwas isolated from the samples by soaking in repeated changes of1% HCl solution (Sealy, 1986), then treated with 0.125 M NaOH for20 h to remove humic and fulvic acid contaminants (Katzenbergand Weber, 1999). The collagen was analyzed at the StableIsotope Laboratory of the University of Calgary (Canada),Department of Physics and Astronomy, using an NA 1500elemental analyzer coupled to a Finnigan Mat TracerMat massspectrometer.

The Eleutherna, Kastella, Messene, Sourtara, and Stylos sampleswere also collected mainly from ribs and long bones. Collagen wasextracted following a modified Longin procedure (Longin, 1971), asoutlined in detail elsewhere (Richards and Hedges, 1999), with theaddition of an ultrafiltration step (Brown et al., 1988). Isotopicvalues for the Kastella samples were measured using a Carlo Erbaelemental analyzer coupled to a ThermoFinnigan Delta Plus XLmass spectrometer at the Isotope Laboratory, Department ofArchaeological Sciences, University of Bradford. Isotopic values forthe samples from Eleutherna, Messene, Sourtara and Stylos weremeasured using a ThermoFinnigan Flash elemental analyzercoupled to a ThermoFinnigan Delta Plus XL mass spectrometer atthe Max Planck Institute for Evolutionary Anthropology, Depart-ment of Human Evolution, Leipzig, Germany. The extractedcollagen from all samples was generally verywell preserved, and allthe samples reported here had collagen yields of over 1% and C:Nratios between 2.9 and 3.6, which is indicative of collagen suitablefor isotopic analysis (DeNiro, 1985).

4. Results

The stable carbon and nitrogen isotope data for the non-adultsare presented in Table 2. The average adult female d15N and d13Cvalues vary between the eight sites, reflecting local variation in diet(Bourbou et al., 2011). In order to allow the data to be interpreted asa group, each non-adult value is considered in terms of its rela-tionship to the mean adult female value for its site (Table S1).Because this format standardizes the non-adult values in terms oftheir elevation relative to adult female values from their own sites,it is useful for comparisons between sites and eras (e.g., Schurr,1997; Schurr and Powell, 2005; Williams et al., 2005) and in thiscase provides a composite impression of weaning patterns atByzantine sites. Figs. 2 and 3 present the data for the eight sites inthis format.

The d15N results do not present the completely straightforwardweaning curves seen in many single-site archaeological samples,with all or almost all young infants elevated substantially above theadult female mean and a clear descent to adult d15N at a particularage (Fig. 2). Although many of the infants show elevated d15Nvalues, many are also scattered within �1& of the adult femalemean. However, a closer look at the data shows that age-specificpatterning has indeed been recovered. Of the 23 non-adults agedthree years or less at death, 12 (52%) are elevated by more than 1&over the mean adult female value for the same site; eight of theseindividuals (35%) are elevated above the adult female mean bymore than 2&. Of the 38 non-adults aged over three years at death,

Table 2Non-adult d15N and d13C values, collagen quality indicators and osteological information for all non-adult samples along with adult female site means (data from Garvie-Lok,2001; Bourbou and Richards, 2007; Bourbou and Garvie-Lok, 2009; Bourbou, 2010).

Sample no. Age d13C (&) d15N (&) C:N %C %N % Yield

Eleutherna#8 Birth-3 years �19.3 7.7 3.3 41.5 14.7 4.4#9 Birth-3 years �19.4 6.9 3.2 40.2 14.6 3.2#12 18 months � 6 months �19.2 10.8 3.5 39.5 13.4 1.1#10 2 years � 8 months �19.9 7.0 3.3 38.6 13.7 1.0#13 2e4 years �18.9 11.4 3.6 39.9 12.9 1.5#22 5 years � 16 months �19.1 8.1 3.6 40.8 13.3 1.7#19 4e8 years �19.1 11.7 3.5 37.6 12.5 2.4#20 4e8 years �19.0 9.2 3.5 40.2 13.6 4.1#34 4e8 years �19.1 8.1 3.2 39.8 14.4 6.7#1 4e12 years �19.4 7.0 3.2 41.2 14.9 3.3#2 4e12 years �20.3 7.5 3.3 36.2 12.9 0.9#35 4e12 years �19.0 7.9 3.5 24.3 8.2 2.9#44 4e12 years �18.2 8.0 3.3 39.7 14.1 1.9#26 6e10 years �19.0 8.3 3.4 34.2 11.7 1.7#4 7e11 years �19.3 6.7 3.3 44.4 15.8 1.3#27 7e11 years �18.7 8.7 3.3 40.4 14.5 3.2Adult female d15N mean ¼ 7.9 � 1.4Adult female d13C mean ¼ �18.9 � 0.6Kastella#S1 5e9 years �19.0 8.9 3.3 40.3 14.4 5.6#S2 2 years � 8 months �18.0 12.1 3.2 40.8 14.7 6.3#S3 0.5 monthse1.5 years �18.3 11.6 3.3 33.0 11.7 3.0#S4 10 years � 30 months �18.8 7.9 3.2 37.2 13.4 4.0#S8 4.5e5.5 years �18.7 8.3 3.2 26.5 9.8 1.9#S10 7.5e8.5 years �18.7 8.5 3.3 38.2 13.7 2.6#S11 4 years � 12 months �19.4 7.5 3.4 46.5 15.8 1.4Adult female d15N mean ¼ 8.7 � 0.6Adult female d13C mean ¼ �19.0 � 0.2Messene#4 1e3 years �18.9 8.5 3.4 39.7 13.8 3.3#11 1e3 years �19.6 8.7 3.3 39.2 13.7 1.3#17 1e3 years �20.4 7.6 3.5 34.6 11.5 1.2#24 1e3 years �18.9 7.2 3.4 29.6 10.2 4.7#12 9.5e10.5 years �18.9 7.1 3.3 41.3 14.5 1.1Adult female d15N mean ¼ 8.3 � 0.6Adult female d13C mean ¼ �19.2 � 0.3NemeaSNE-32 16e32 months �19.0 8.6 3.3 44.3 15.6 7.1SNE-31 5e10 years �19.2 7.5 3.3 44.1 15.4 4.9Adult female d15N mean ¼ 8.5 � 1.1Adult female d13C mean ¼ �19.1 � 0.1PetrasSPE-16 30e42 months �18.6 12.7 3.3 47.1 16.7 18.6SPE-02 4e5 years �20.2 8.6 3.2 45.2 16.8 19.4SPE-14 5.5e6.5 years �18.9 9.6 3.3 43.5 15.4 4.2SPE-07 6e7 years �19.7 8.9 3.3 47.0 16.7 13.6Adult female d15N mean ¼ 9.2 � 0.3Adult female d13C mean ¼ �19.3 � 0.2ServiaSSE-37 Ca. 12 months �17.6 11.3 3.3 44.9 15.9 14.0SSE-03 3e7 years �18.6 9.5 3.2 43.6 15.7 13.5SSE-02 8e11 years �19.0 9.2 3.4 43.9 14.9 4.5Adult female d15N mean ¼ 8.5 � 0.2Adult female d13C mean ¼ �18.9 � 0.1Sourtara#19 Birth-3 years �17.6 11.0 3.2 41.3 15.1 3.8#20 Birth-3 years �16.9 12.1 3.2 42.5 15.5 4.5#21 Birth-3 years �17.2 10.2 3.2 42.6 15.5 5.1#11 18 months � 6 months �18.0 12.0 3.3 42.1 14.8 3.6#43 2e3 years �17.5 11.0 3.3 32.5 11.5 0.7#29 3e4 years �18.4 10.6 3.2 40.2 14.5 2.2#25 2e4 years �18.4 10.2 3.4 37.3 12.7 2.6#30 4e5 years �18.1 9.6 3.2 42.2 15.4 1.8#23 5e6 years �18.5 9.5 3.2 39.9 14.4 3.3#14 4e8 years �18.4 9.5 3.2 41.9 15.1 2.1#16 4e8 years �17.7 9.5 3.2 42.4 15.4 5.5#28 8e9 years �18.7 9.7 3.3 37.8 13.5 1.7#34 Ca. 9 years �18.5 9.6 3.2 40.7 14.1 1.5#8 10 years � 30 months �18.5 9.9 3.3 41.6 14.9 1.9#10 10 years � 30 months �18.1 10.2 3.2 42.4 14.9 5.0#12 10 years � 30 months �18.6 9.6 3.3 41.7 14.8 2.0

(continued on next page)

C. Bourbou et al. / Journal of Archaeological Science 40 (2013) 3903e3913 3907

Table 2 (continued )

Sample no. Age d13C (&) d15N (&) C:N %C %N % Yield

Adult female d15N mean ¼ 9.5 � 0.4Adult female d13C mean ¼ �18.2 � 0.2Stylos#8 Fetus �19.2 10.1 3.5 39.3 13.1 2.0#9 Newborn �18.3 9.3 3.2 26.4 9.7 3.0#4 Newborn �19.4 10.5 3.5 34.8 11.7 2.1#2 6.5e7.5 years �19.7 8.8 3.3 27.4 9.7 3.0#12 7e8 years �19.6 10.1 3.4 42.5 14.4 1.3#3 7.5e8.5 years �19.7 9.1 3.3 43.1 15.0 1.8#7 7.5e8.5 years �19.5 9.1 3.4 27.6 9.5 1.4#10 7.5e8.5 years �18.8 9.7 3.3 44.3 15.6 4.0Adult female d15N mean ¼ 8.8 � 1.2Adult female d13C mean ¼ �18.9 � 0.8

2.0

C. Bourbou et al. / Journal of Archaeological Science 40 (2013) 3903e39133908

four (10%) are elevated by more than 1& over the mean adult fe-male value for the same site; only one of these individuals (3%) iselevated above the adult female mean by more than 2&.

The carbon isotopes show similar overall patterning by age atdeath, with most non-adult d13C values falling into a �0.5& fluc-tuation around the adult female mean (Fig. 3). As with the d15Nvalues, age-dependent patterning can be seen. Of the 23 non-adultsaged three years or less at death, nine (39%) show relatively highd13C values elevated 0.6&e1.3& above the adult female mean fortheir site. Of the children above three years of age, only one indi-vidual (3%) shows a d13C value in this range. The pattern here is notas definite, but this would be expected given that the smaller tro-phic effect for d13C makes nursing elevation more difficult todistinguish from background d13C variation in a population.

Considering d15N and d13C together (Fig. 4) shows their rela-tionship more clearly. In this figure, sample d15N and d13C elevationabove the adult female mean is considered by age groupings ratherthan site. Juveniles older than about three years of age plot aroundthe intersection of the axes (0& elevation for d15N and d13C),reflecting their position within the adult d15N and d13C ranges fortheir respective sites. Some juveniles younger than three also fallinto this area. However, age-related patterning is clearly seen in thearea of high d15N and d13C elevation at the upper right of the graph(marked by an ellipse); here, almost all individuals are aged lessthan about three years.

5. Discussion

5.1. Interpreting the site data

The d15N and d13C data recovered for the eight sites consideredhere do not show the clear weaning curves usual for single-sitesamples, but they do show age-related patterning. This patterning

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0

5.0

0 1 2 3 4 5 6 7 8 9 10 11 12δN

val

ues

rela

tive

to

adul

t fem

ale

mea

n (‰

)

EleuthernaKastellaMesseneNemeaPetrasServiaSourtaraStylos

age at death (years)

Fig. 2. Non-adult d15N values expressed in terms of per mil elevation relative to theirsites’ adult female means.

separates children under three, who are more likely to showelevated d15N and d13C values, from those over three, who are lesslikely to show elevated relative to adult women. This generalpattern is consistent with our expectations for Byzantine-erabreastfeeding based on the documentary sources reviewed above.However, the separation between younger and older children is notas clear as was expected, and a large proportion of infants fail toshow an elevation above the female d15N and d13C means. Thereasons for this patterning bear discussion.

It is possible that the lack of a clear general shift from elevationto adult-like values is an artifact of the way in which the data arecollated, with individuals from several different sites standardizedto those sites’ adult female data. If there were, in fact, differences inweaning age between the groups in question, collating the data inthis way could produce this pattern. In this scenario, the samplesanalyzed for a given age could include both juveniles from earlier-weaning groups who had already reached the adult female d15Nand d13C means and juveniles from later-weaning groups who stillshowed considerable d15N and d13C elevation. If this was the case,we might expect to see differential patterning of non-adults fromthe different sites. Some instances of this might be seen in the data.For example, a group of two-year-olds fromMessene seems to forma cluster with uniformly low d15N (Fig. 2). This could be taken asevidence for an early weaning age at Messene. However, other in-terpretations are possible, as the one other non-adult from that site(aged roughly ten years) also shows a much lower d15N value thanchildren its age from the other sites. This could suggest a moregeneral difference, with diet at Messene different from the othersites for children of all ages. Stable isotope analyses of adults fromthe eight sites (Bourbou et al., 2011) did not suggest anything un-usual about adult diet at Messene. However, it is conceivable that

1.0

1.5 (‰

)

0 5

0.0

0.5

adul

t fe

mal

e m

ean

EleuthernaKastellaMesseneNemeaPetrasServia

-1.5

-1.0

- .

val

ues

rela

tive

to

SourtaraStylos

-2.00 1 2 3 4 5 6 7 8 9 10 11 12

C

age at death (years)

Fig. 3. Non-adult d13C values expressed in terms of per mil elevation relative to theirsites’ adult female means.

Fig. 4. Non-adult d15N and d13C values by age group, expressed in terms of per mil elevation relative to their sites’ adult female means. The ellipse marks the area of elevated d15Nand d13C values likely associated with nursing.

C. Bourbou et al. / Journal of Archaeological Science 40 (2013) 3903e3913 3909

local practices could produce juvenile diets at the site with un-usually low d15N relative to adult diets. The possibility that signif-icant differences inweaning practices or more general juvenile dietexisted between communities is one that should be investigated infuture research.

While inter-site variation may be part of the explanation for theage distribution of the grouped data, it cannot fully explain the factthat at several of the sites, young juveniles with no noticeable d15Nor d13C elevation above the adult female means were recovered. Ford15N, almost half of the non-adults aged less than three years fallwithin�1& of the adult female mean. A cluster of younger childrenfrom Messene and Eleutherna with low d15N values is particularlynotable (Fig. 2). Young children with no apparent nursing elevationare not unusual in archaeological samples, though they are usuallyrecovered at a lower percentage than seen here. There are manypotential explanations for such individuals involving maternalmortality (loss of the mother at a young age, with subsequentforced weaning), maternal morbidity (illness of the mother severeenough to cause loss of her milk), and short birth spacing (withforced early weaning to accommodate the needs of the youngersibling). What could be causing the higher frequency at some ofthese Byzantine sites is unclear. Although Soranus and Oribasiusboth stated firmly that breastfeeding was best for the child(Bourbou and Garvie-Lok, 2009), was there a factor in operation e

particularly at Eleutherna and Messene, perhaps e that made earlyweaning unavoidable for some?

Some interesting patterning can also be noted in the older non-adults aged three years or more. The general �1& fluctuation ind15N around the adult female mean for these older children pre-sumably reflects normal individual variation within each of thepopulations, and is seen in adults from these sites as well. Anexception to this pattern is one child from Eleutherna (#19) agedapproximately four to eight years whose d15N value is 3.8& abovethe female mean for that site. This value could reflect unusualtreatment, such as prolonged breastfeeding or a distinctive diet; itmight also be linked to nutritional stress (Fuller et al., 2005). Mostolder juveniles have d13C values within �0.5& of the adult femalemean. However, a number of older children have d13C values fallingnoticeably below this �0.5& range. This may simply reflect

idiosyncratic or inter-site variation, but could also conceivablyreflect dietary differences between children and adults. It is alsointeresting to note that despite evidence for some millet use in thenorthern Greek communities of Sourtara and Servia (Bourbou et al.,2011), juveniles from those sites do not show any sign of relativelyhigher d13C value that could reflect selective use of millet or dairyproducts from millet-fed animals in weaning foods. This placesthem in contrast to late Roman Egypt, where millet appears to haveplayed this role (Dupras et al., 2001).

5.2. Isotopes and Byzantine weaning practices

There are few data to compare to our results because very littleisotopic research has been conducted on the dietary patterns ofinfants and children in Greek skeletal samples. Studying a muchearlier era, Triantaphyllou et al. (2008) investigated the breast-feeding and weaning patterns from the Bronze Age population ofLerna. Their results suggest that weaning generally started at orbefore the age of 2.5 years to three years. Work by Kwok (2007) andHolt (2009) on the Greek colonists of Classical-Hellenistic Apolloniasuggest that consumption of weaning foods started around sixmonths to one year of age and weaning was generally complete bytwo to four years of age. The data available from our study do notallow firm general conclusions to be drawn about weaningbehavior in Byzantine Greece. The contrast between the higherproportion of elevated d15N and d13C values seen in non-adults agedthree years or less and the lower proportion seen in older childrenis consistent with the documentary evidence that weaning wasusually complete by the fourth year (Bourbou and Garvie-Lok,2009).

Arriving at a more precise age of weaning than “before fouryears of age” is not possible given our data, largely because of thescatter seen in d15N and d13C values at all ages. The sample includesa number of juveniles aged approximately two to three years atdeath with substantial d15N and d13C elevations above the adultfemale mean for their sites, suggesting that mother’s milk stillformed a significant portion of their diets. However, there are alsosome older children showing substantial elevation, and severalyounger children show little if any d15N or d13C elevation above

C. Bourbou et al. / Journal of Archaeological Science 40 (2013) 3903e39133910

adult females. Some potential reasons for this variation have beendiscussed above, including variation in weaning practices or ingeneral juvenile feeding practices between communities. The po-tential for idiosyncratic variation also should not be overlooked. Asreviewed in Bourbou and Garvie-Lok (2009), accounts of Byzantinechildhood in the lives of the saints provide indications of individualvariation inweaning history, with some childrenweaned by the ageof three but at least one still breastfeeding at four. There is alsosome evidence that adjustment of weaning practices to individualneeds was expected; for example, Soranus advocated that a child’sweaning schedule should consider the season and the child’shealth, and that weaning should be halted if the child became ill(Temkin, 1991). Given these indications, some of the scatter in thevalues found for this study may reflect simple individual choice onthe part of the parents. It will be interesting, as more sites arestudied, to seewhether a similar pattern of high variability is found.

5.3. Byzantine weaning and health

After birth, breast milk provides the newborn with immuno-logical protection and a supply of balanced nutrients, as it containsover 100 different constituents that contribute to the health,growth and development of the newborn (Lewis, 2007, 97e9). Theonset of weaning represents a crucial period in a child’s life. Theintroduction of solid foods can be related to the development ofspecific pathological conditions (for example, susceptibility to in-fections), and to an increased risk of mortality (e.g., Wheeler, 2010).This offers a plausible explanation for the frequent clustering ofdeaths around the weaning age in non-adult mortality patterns(Katzenberg et al., 1996). Bourbou (2010) examined a total of 174non-adult individuals from six sites in Crete (Greece) during theByzantine period (6the12th centuries AD), and found that in themajority of the sites, childmortality exceeded infant mortality, with56.8% (n¼ 99) dying in this age range (Fig. 5). This predominance ofchild deaths could be rooted largely in environmental factors suchas poor sanitation and cultural factors influencing weaning prac-tices. Furthermore, specific pathological (hematopoietic: cribraorbitalia and porotic hyperostosis) and metabolic (scurvy) condi-tions observed in that sample tend to develop in the years followingthe weaning age suggested by our stable isotope data. Similar re-sults were reported for the Queenford Farm cemetery in Britain(Fuller et al., 2006b). There, despite the few individuals betweenbirth and 1.5 years available for isotopic analysis and the range ofuncertainty in the age estimation, the authors observed a peak innon-adult mortality associated with a decrease in d13C and d15N

Age categories adapted from Scheuer and Black (2004, 468) and Lewis (2007, 2)

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Eleutherna Gortyn Kastella Kefali Knossos Stylos

Adolescent

Child

Infant

Neonate

Fetus

Fig. 5. Non-adult mortality in six Byzantine populations (6th/7the12th centuries AD)from Crete (data after Bourbou, 2010).

values between two and three years, suggesting possible nutri-tionally inadequate and/or contaminated weaning foods.

The particular choice of honey and goat’s milk in infant feeding,documented in the written sources, could have had serious adverseconsequences on infant health, as evidenced by the high rates ofcribra orbitalia observed in the Dakhleh Oasis samples (Fairgrieveand Molto, 2000). The use of goat’s milk as a dietary staple for in-fants can result in serious complications. Although goat’s milk ispopularly seen as similar to human milk, it is relatively low in bothcobalamine (0.1 mg/L) and folic acid (6 mg/L) compared to humanmilk (4 mg/L and 52 mg/L, respectively). Infants who are fed goat’smilk rather than human milk starting shortly after birth developsevere megaloblastic anemia by the age of about three to fivemonths due to folic acid deficiency (Chanarin, 1990).

Similarly, honey is often contaminated with the spores of Clos-tridium botulinum, the bacterium that secretes the toxin that causesbotulism, a severe and often fatal form of food poisoning (Nevaset al., 2005). Botulinum toxin blocks the transmission of chemicalsignals at neuromuscular junctions; left untreated, it can causedeath through paralysis of the muscles of respiration (Passmoreand Eastwood, 1986). In the infant, C. botulinum spores can colo-nize the intestinal tract, leading to infant botulism, an illness ofvarying severity whose symptoms include reduced muscle tone,difficulty in suckling, and sometimes respiratory problems (Arnonet al., 1979; Merenstein et al., 1991; Shapiro et al., 1998). In mod-ern populations, most infants who develop infant botulism recoverafter an illness lasting several weeks or several months; in antiq-uity, however, recovery may have been less certain. The possiblepresence of honey and goat’s milk in Byzantine infant diets shouldbe considered a potential health stress for the period.

5.4. Comparison with western medieval feeding practices

As with the Byzantine physicians, the works of earlier writerssuch as Soranus influenced the medical tradition of western phy-sicians on infant feeding practices. It is only in the 15th century ADthat western authors first introduce useful comments and notesbased on observations and direct experience (Fildes, 1986). Medi-eval writers in Western Europe suggest the gradual cessation ofbreastfeeding by about two years of age (Shahar, 1990). Liquid orsemi-liquid foods, such as animal milk, gruel, or pap (a mixture offlour and bread cooked in water) might be given prior to theemergence of the first teeth, before about six to ninemonths of age;afterward other foods might be introduced (Fildes, 1986, 1995).

Isotopic studies on the duration of breastfeeding and the timingof weaning include a small number of relevant publications fromthe Mediterranean region and Western Europe, dating from theRoman to the late medieval period (Dittman and Grupe, 2000;Dupras et al., 2001; Mays et al., 2002; Richards et al., 2002;Herrscher, 2003; Prowse et al., 2004; Turner et al., 2007). The studyby Prowse et al. (2004) on samples from Isola Sacra near Romedating to the 1ste3rd centuries AD found that infants were fullyweaned by 2.5e3 years of age, and by 4e5 years of age d13C andd15N values were similar to adult values. The isotopic analysis byDupras et al. (2001) of samples from the Kellis 2 cemetery at theDakhleh Oasis of Egypt, dating to the Romano-Christian period (ca.AD 250e450), revealed a gradual weaning process lasting untilapproximately three years of age. They also suggest that supple-mentary foods, likely including the milk of goats and cows that hadbeen fed with millet, were introduced into the diet at around sixmonths of age. Dittman and Grupe (2000) analyzed non-adultsamples from the early medieval population (AD 500e700) atWenigumstadt in southern Germany, finding the highest d15Nvalues (around 12&) at one year of age. By the age of three theinfants were nearly completely weaned as the d15N values drop to

C. Bourbou et al. / Journal of Archaeological Science 40 (2013) 3903e3913 3911

approximately 9&. A gradual and prolonged weaning process (twoto four years) is also suggested from the isotopic analysis of samplesfromQueenford Farm in Britain, dating to the late 4th throughmid-6th centuries AD (Fuller et al., 2006b). The study by Herrscher(2003) on non-adult samples from the necropolis of Saint-Laurent de Grenoble (Isère, France), dating to the 13the15th cen-turies AD, also indicates the onset of weaning between 2.6 and 3.3years of age. In contrast to these studies, the d13C and d15N valuesfor the British site of Wharram Percy (10the16th centuries AD)indicate that weaning must have occurred earlier, at or shortlybefore the age of two years (Mays et al., 2002; Richards et al., 2002).

The isotopic data from Isola Sacra, the Dakhleh Oasis, Weni-gumstadt and Queenford Farm demonstrate relatively late weaningages. Despite the wide geographic separation of these populationsand presumed local cultural differences, their stable isotope valuessuggest a similar weaning pattern, with some individuals nursingto the age of three years. Fuller et al. (2006b, 50) have pointed out acontrast between this apparent “Roman weaning pattern” and thatseen at the later medieval site of Wharram Percy, whose stableisotope values suggest that children were weaned at an earlier age,between 1.5 and 2 years. This contrast could reflect a stronger in-fluence by the physician Soranus and other medical scholars amongthe Roman and early medieval populations, who appear to havefollowed the physicians’ advice for a gradual weaning process,while later medieval populations chose to wean their children at ayounger age. However, early weaning is not observed in the Saint-Laurent de Grenoble sample, although it dates to the late medievalperiod. As with the Byzantine sites considered in this paper, thepotential for different practices by different communities must beconsidered. Given the distances between Wharram Percy andSaint-Laurent de Grenoble, it would not be surprising for differentlocal cultural practices applied toward infant feeding. Perhaps suchlocal variation increased in the later period.

Insofar as they can be associated with a rough weaning age, theGreek Byzantine data presented here are consistent with the laterweaning ages found among Roman and earlymedieval populations.This may indicate that Greek Byzantine populations persisted in a“Roman weaning pattern” influenced by the early medical advice.Although the available data are sparse, Bourbou and Garvie-Lok(2009) have attempted to determine whether a difference inweaning practices existed between the varied populations of latemedieval Greece. In this era, invasions and migrations broughtcommunities of Western European (Frankish) origin to Greece(Lock,1995).While few skeletal remains of those communities havebeen examined, the d15N values of the small number of infant re-mains analyzed are consistent with an earlier weaning age similar

3

4

5

ean

(‰)

0

1

2

to a

dult

fem

ale

me

Byzantine

Frankish

-2

-1

0

0 1 2 3 4 5 6 7 8 9 10 11 12

N v

alue

s re

lati

ve t

0 1 2 3 4 5 6 7 8 9 10 11 12age at death (years)

Fig. 6. Non-adult d15N values for Byzantine Greece (this study) and later medievalWestern European (Frankish) cemeteries in Greece (data after Bourbou and Garvie-Lok, 2009).

to that of Wharram Percy (Fig. 6). The bioarchaeological study ofFrankish communities in Greece is in its early stages, and willbenefit from techniques such as dentine stable isotope analysis thatmaximize the information from small burial groups (Garvie-Lok, inpress). Byzantine Greece, while more thoroughly studied to date,will also benefit from data provided by additional sites and newtechniques. Thus it is hoped that future research will provide moreevidence on the subject of comparative Greek and Western wean-ing traditions.

6. Conclusions

Written sources have been traditionally used for the recon-struction of infant feeding practices in the past, but to what extentthe recommendations cited in these documents were actually fol-lowed cannot be determined by their study alone. Although iso-topic signatures can only be used as indicators of weaning age,rather than an absolute measurement, the application of stableisotope ratios is a powerful tool that can add considerably to ourknowledge of infant feeding in the past. This technique is difficult toapply to past cultures that tended to bury their dead in smallgroups because many sites do not contain enough juvenile remainsfor a reliable statistical analysis of infant values to bemade. Herewehave taken an alternative approach, combining juvenile valuesfrom an assortment of sites and standardizing them to adult femalevalues to produce an aggregate sample. While this process mustnecessarily introduce more error than using the remains from asingle site, here it has produced useful results where few otherapproaches were possible.

In this study a general contrast between a higher proportion ofelevated d13C and d15N values in non-adults aged three years or lessand a lower proportion seen in older children suggests thatweaning during the Byzantine period in Greece (6the15th cen-turies AD) was complete by the fourth year, a pattern consistentwith the documentary evidence. Although arriving at a more pre-cise age of weaning than “before four years of age” is not possible,the elevation of d13C and d15N values in some children aged around2.5e3 years indicates that breast milk formed a substantial part ofsome infants’ diets well into the third year of life. However, there isalso a substantial number of children under three years of age thatshow little or no elevation over their site’s adult female mean. Thiscould reflect a number of factors, including local conditions at twoof the communities that contribute a large proportion of the chil-dren with unexpectedly low values.

Theweaning process is often linked to an increased risk of infantmortality and potentially exposes non-adults to newpathogens andnutritional stress. A clustering of non-adult deaths at around thetime of weaning and the development of pathological conditionsassociated with weaning stress have been frequently attested in thebioarchaeological analysis of Byzantine populations. Byzantinepopulations seem to have maintained traditions carried over fromthe Roman period, following later weaning ages as did Roman andearly medieval populations from the Mediterranean region andWestern Europe. It seems, however, that this so far uniform patternis in contrast with the isotopic data obtained from the later medi-eval period in Western Europe. During that period, the two avail-able studies suggest variation with one site indicating thatcessation of breastfeeding was more abrupt and occurred at anearlier age, at or shortly before two years of age (Wharram Percy),while the other indicates that later weaning was still the preferablepractice (Saint-Laurent de Grenoble). Further isotopic studies andrelevant comparative analysis should provide additional evidenceto determine whether a difference in weaning practices existedbetween Roman/Byzantine and later medieval western Europeanpopulations.

C. Bourbou et al. / Journal of Archaeological Science 40 (2013) 3903e39133912

Acknowledgments

Funding for isotopic analyses was given by the British School atAthens (Kastella), Dumbarton Oaks Project Grant (Sourtara), Ioan-nis Costopoulos Foundation (Eleutherna and Messene), the SocialSciences and Humanities Research Council of Canada (Nemea,Petras and Servia), the University of Calgary (Nemea, Petras andServia), and the Max Planck Society. The initial phases of Garvie-Lok’s workwere donewhile on fellowship at theWiener Laboratoryof the American School of Classical Studies at Athens, and shethanks the laboratory for its financial and logistical support.

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.jas.2013.04.020.

References

Arnon, S.S., Madura, T.F., Damus, K., Thompson, B., Wood, R.M., Chin, J., 1979. Honeyand other environmental risk factors for infant botulism. J. Pediatr. 94, 331e336.

Beaucamp, J., 1982. L’allaitement: mère ou nourrice? JÖB 32.2, 549e558.Bourbou, C., 2010. Health and Disease in Byzantine Crete (7the12th Centuries A.D.).

Ashgate Publishing Ltd, Surrey.Bourbou, C., Garvie-Lok, S., 2009. Breast-feeding and weaning patterns in Byzantine

times: evidence from human remains and written sources. In:Papaconstantinou, A., Talbot, A.-M. (Eds.), Becoming Byzantine: Children andChildhood in Byzantium. Dumbarton Oaks Research Library and Collection,Washington DC, pp. 65e83.

Bourbou, C., Richards, M.P., 2007. The middle-Byzantine menu: stable carbon andnitrogen isotope values from the Greek site of Kastella, Crete. Int. J. Osteo-archaeol. 17, 63e72.

Bourbou, C., Garvie-Lok, S. Bread, oil, wine e and milk: feeding infants and adults inByzantine Greece. In: Papathanasiou, A., Richards, M.P., Fox, S. (Eds.), StableIsotope Studies of Prehistoric and Historic Greek Populations. OccasionalWiener Laboratory Series (Hesperia Supplement). Princeton, in press.

Bourbou, C., Fuller, B.T., Garvie-Lok, S.J., Richards, M.P., 2011. Reconstructing thediets of Greek Byzantine populations (6the15th centuries AD) using carbon andnitrogen stable isotope ratios. Am. J. Phys. Anthropol. 146, 569e581.

Bourbou, C., Fuller, B.T., Richards, M.P., 2008. The application of stable isotope ratioanalysis for detecting dietary patterns in Greek Byzantine populations (7the13th centuries AD). In: Paper Presented at the 5th Symposium of the HellenicSociety of Archaeometry, Athens 8e10 October 2008.

Brown, T.A., Nelson, D.E., Southon, J.R., 1988. Improved collagen extraction bymodified Longin method. Radiocarbon 30, 171e177.

Burguière, P., Gourevitch, D., Malinas, Y., 1990. Soranos d’ Ephèse. In: Maladies desFemmes, tome II, livre II. Les Belles Lettres, Paris.

Chanarin, I., 1990. The Megaloblastic Anemias. Blackwell Scientific, Oxford.Clayton, F., Sealy, J., Pfeiffer, S., 2006. Weaning age among foragers at Matjes river

rock shelter, South Africa, from stable nitrogen and carbon isotope analyses.Am. J. Phys. Anthropol. 129, 311e317.

DeNiro, M.J., 1985. Postmortem preservation and alteration of in vivo bone collagenisotope ratios in relation to palaeodietary reconstruction. Nature 317, 806e809.

Dittman, K., Grupe, G., 2000. Biochemical and paleopathological investigations onweaning and infant mortality in the early Middle Ages. Anthropol. Anz. 58,345e355.

Dupras, T.L., Tocheri, M.W., 2007. Reconstructing infant weaning histories at Romanperiod Kellis, Egypt using stable isotope analysis of dentition. Am. J. Phys.Anthropol. 134, 63e74.

Dupras, T.L., Schwarcz, H.P., Fairgrieve, S.I., 2001. Infant feeding and weaningpractices in Roman Egypt. Am. J. Phys. Anthropol. 115, 204e212.

Eerkens, J.W., Berget, A.G., Bartelink, E.J., 2011. Estimating weaning and earlychildhood diet from serial micro-samples of dentine collagen. J. Archaeol. Sci.38, 3101e3111.

Fairgrieve, S.I., Molto, J.E., 2000. Cribra orbitalia in two temporally disjunctpopulation samples from the Dakhleh Oasis, Egypt. Am. J. Phys. Anthropol.111, 319e331.

Fildes, V.A., 1986. Breasts, Bottles, and Babies: a History of Infant Feeding. EdinburghUniversity Press, Edinburgh.

Fildes, V.A., 1995. The culture and biology of breastfeeding: an historical review ofWestern Europe. In: Stuart-Macadam, P., Dettwyler, K.A. (Eds.), Breastfeeding:Biocultural Perspectives. Transaction Publishers, New York, pp. 101e126.

Fogel, M.L., Tuross, N., Owsley, D., 1989. Nitrogen isotope tracers of human lactationin modern and archaeological populations. In: Yearb. Carnegie Inst. Wash., vol.88, pp. 111e117.

Fuller, B.T., Richards, M.P., Mays, S.A., 2003. Stable carbon and nitrogen isotopevariations in tooth dentine serial sections from Wharram Percy. J. Archaeol. Sci.30, 1673e1684.

Fuller, B.T., Fuller, J.L., Sage, N.E., Harris, D.A., O’Connell, T.C., Hedges, R.E.M., 2005.Nitrogen balance and d15N: why you are not what you eat during nutritionalstress. Rapid Commun. Mass Spectrom. 19, 2497e2506.

Fuller, B.T., Fuller, J.L., Harris, D.A., Hedges, R.E.M., 2006a. Detection of breastfeedingand weaning in modern human infants with carbon and nitrogen stable isotoperatios. Am. J. Phys. Anthropol. 129, 279e293.

Fuller, B.T., Molleson, T.I., Harris, D.A., Gilmour, L.T., Hedges, R.E.M., 2006b. Isotopicevidence for breastfeeding and possible adult dietary differences from late/sub-Roman Britain. Am. J. Phys. Anthropol. 129, 45e54.

Garvie-Lok, S.J., 2001. Loaves and Fishes: a Stable Isotope Reconstruction of Diet inMedieval Greece. Ph. D. dissertation. University of Calgary.

Garvie-Lok, S. Greek, Frank, other: differentiating cultural and ancestral groups inthe Frankish Morea using human remains analysis. In: Gerstel, S.E.J. (Ed.),Viewing the Morea: Land and People in the Late Medieval Peloponnese.Dumbarton Oaks Research Library and Collection, in press.

Hedges, R.E.M., Clement, J.G., Thomas, C.D.L., O’Connell, T.C., 2007. Collagen turn-over in the adult femoral mid-shaft: modeled from anthropogenic radiocarbontracer measurements. Am. J. Phys. Anthropol. 133, 808e816.

Herring, D.A., Saunders, S.R., Katzenberg, M.A., 1998. Investigating the weaningprocess in past populations. Am. J. Phys. Anthropol. 105, 425e439.

Heiberg, J. (Ed.), 1921. Paulus Aeginita. Corpus Medicorum Graecorum, 2 vols..Teubner, Leipzig.

Herrscher, E., 2003. Alimentation d’une population historique: Analyse des donnéesisotopiques de la nécropole Saint-Laurent de Grenoble (XIIIeeXVe siècle,France). Bull. Mem. Soc. Anthropol. Paris 3e4, 145e320.

Holt, S., 2009. Individuals and Variation: Stable Isotope Analysis of Weaning UsingDental Serial Sections. M.A. thesis. McMaster University.

Howcroft, R., Eriksson, G., Lidén, K., 2012. Conformity in diversity? Isotopic inves-tigation of infant feeding practices in two iron age populations from SouthernÖland, Sweden. Am. J. Phys. Anthropol. 149, 217e230.

Jay, M., Fuller, B.T., Richards, M.P., Knüsel, C.J., King, S.S., 2008. Iron age breastfeedingpractices in Britain: isotopic evidence from Wetwang Slack, East Yorkshire. Am.J. Phys. Anthropol. 136, 327e337.

Katzenberg, M.A., Pfeiffer, S., 1995. Nitrogen isotope evidence for weaning age in anineteenth century Canadian skeletal sample. In: Grauer, A. (Ed.), Bodies ofEvidence. Wiley-Liss, New York, pp. 221e235.

Katzenberg, M.A., Weber, A., 1999. Stable isotope ecology and palaeodiet in the LakeBaikal region of Siberia. J. Archaeol. Sci., 651e659.

Katzenberg, M.A., Herring, D.A., Saunders, S.R., 1996. Weaning and Infant Mor-tality: Evaluating the Skeletal Evidence. In: Yearb. Phys. Anthropol., vol. 39,pp. 177e199.

Kwok, C., 2007. Baby Bones, Food, and Health: Stable Isotopic Evidence for InfantFeeding Practices in the Greek Colony of Apollonia (5the2nd Centuries B.C.).M.A. thesis. Trent University.

Laskaratos, J., Poulakou-Rebelakou, E., 2003. Oribasius (fourth century) andearly Byzantine perinatal nutrition. J. Pediatr. Gastroenterol. Nutr. 36,186e189.

Lewis, M.E., 2007. The Bioarchaeology of Children. Perspectives from Biological andForensic Anthropology. Cambridge University Press, Cambridge.

Lock, P., 1995. The Franks in the Aegean, 1204e1500. Longman, London.Longin, R., 1971. New method of collagen extraction for radiocarbon dating. Nature,

241e242.Mays, S.A., Richards, M.P., Fuller, B.T., 2002. Bone stable isotope evidence for infant

feeding in Medieval England. Antiquity, 654e656.Merenstein, G.B., Kaplan, D.W., Rosenberg, A.A., 1991. Silver, Kempe Bruyn and

Fulginiti’s Handbook of Pediatrics, 16th ed. Appleton & Lange, Norwalk.Nevas, M., Lindström, M., Hautamäki, K., Puoskari, S., Korkeala, H., 2005. Prevalence

and diversity of Clostridium botulinum types A, B, E and F in honey produced inthe Nordic countries. Int. J. Food Microbiol. 105, 145e151.

Nitsch, E.K., Humphrey, L.T., Hedges, R.E.M., 2011. Using stable isotope analysis toexamine the effect of economic change on breastfeeding practices in Spital-fields, London, UK. Am. J. Phys. Anthropol. 146, 619e628.

Olivieri, A., 1935. Aetii Amideni libri medicinales IeIV, Corpus Medicorum Grae-corum. Teubner, Leipzig.

Papathanasiou, A., Richards, M.P., Fox, S. (Eds.) Stable Isotope Studies of Prehistoricand Historic Greek Populations. Occasional Wiener Laboratory Series, (HesperiaSupplement). Princeton, in press.

Passmore, R., Eastwood, M.A., 1986. Human Nutrition and Dietetics. Churchill Liv-ingstone, London.

Prowse, T., Schwarcz, H.P., Saunders, S., Macchiarelli, R., Bondioli, L., 2004. Isotopicpaleodiet studies of skeletons from the Imperial Roman-age cemetery of IsolaSacra, Rome, Italy. J. Archaeol. Sci. 31, 259e272.

Raeder, I., 1964. Oribasii collectionum medicarum reliquiae. A.M. Hakkert,Amsterdam.

Redfern, R.C., Millard, A.R., Hamlin, C., 2012. A regional investigation of subadultdietary patterns and health in late Iron Age and Roman Dorset, England.J. Archaeol. Sci. 39, 1249e1259.

Rey, A.L., 2004. Autour des nourrissons byzantins et de leur régime. In: Dasen, V.(Ed.), Naissance et petite enfance dans l’Antiquité, Actes du colloque deFribourg, 28 novembre-1er décembre 2001. Academic Press, Fribourg,pp. 363e375.

Richards, M.P., Hedges, R.E.M., 1999. How chemical analysis of human bones can tellus the diets of people who lived in the past. In: Tzedakis, Y., Martlew, H. (Eds.),Flavours of Their Times, Food and Drink in Minoan and Mycenaean Times.Kapon Publishers, Athens.

C. Bourbou et al. / Journal of Archaeological Science 40 (2013) 3903e3913 3913

Richards, M.P., Pearson, J.A., Molleson, T.I., Russell, N., Martin, L., 2003. Stableisotope evidence of diet at Neolithic Çatalhöyük, Turkey. J. Archaeol. Sci. 30,67e76.

Richards, M.P., Fuller, B.T., Molleson, T.I., 2006. Stable isotope palaeodietary study ofhumans and fauna from the multi-period (Iron Age, Viking and Late Medieval)site of Newark Bay, Orkney. J. Archaeol. Sci. 33, 122e131.

Richards, M.P., Mays, S., Fuller, B.T., 2002. Stable carbon and nitrogen isotope valuesof bone and teeth reflect weaning age at the Medieval Wharram Percy site,Yorkshire, U.K. Am. J. Phys. Anthropol. 119, 205e210.

Schurr, M.R., 1997. Stable nitrogen isotopes as evidence for the age of weaning at theAngel site: a comparison of isotopic and demographic measures of weaning age.J. Archaeol. Sci. 24, 919e927.

Schurr, M.R., Powell, M.L., 2005. The role of changing childhood diets in the pre-historic evolution of food production: an isotopic assessment. Am. J. Phys.Anthropol. 128, 278e294.

Sealy, J., 1986. Stable Carbon Isotopes and Prehistoric Diets in the South-westernCape Province, South Africa. In: BAR International Series 293. Archaeopress,Oxford.

Shahar, S., 1990. Childhood in the Middle Ages. Taylor & Francis, London.Shapiro, R.L., Hatheway, C., Swerdlow, D.L., 1998. Botulism in the United States: a

clinical and epidemiologic review. Ann. Intern. Med. 129, 221e228.

Stenhouse, M.J., Baxter, M.S., 1979. The uptake of bomb 14C in humans in radio-carbon dating. In: Berger, R., Suess, H.E. (Eds.), Radiocarbon Dating. University ofCalifornia Press, Berkeley, pp. 324e341.

Temkin, O., 1991. Soranus’s Gynecology. The Johns Hopkins University Press,Baltimore.

Triantaphyllou, S., Richards, M.P., Zerner, C., Voutsaki, S., 2008. Isotopic dietaryreconstruction of humans from Middle Bronze Age Lerna, Argolid, Greece.J. Archaeol. Sci. 35, 3028e3034.

Turner, B.L., Edwards, J.L., Quinn, E.A., Kingston, J.D., van Gerven, D.P., 2007. Age-related variation in isotopic indicators of diet at Medieval Kulubnarti, SudaneseNubia. Int. J. Osteoarchaeol. 17, 1e25.

Waters-Rist, A.L., Katzenberg, M.A., 2010. The effect of growth on stable nitrogenisotope ratios in subadult bone collagen. Int. J. Osteoarchaeol. 20, 172e191.

Waters-Rist, A.L., Bazaliiskii, V.I.,Weber, A.W., Katzenberg,M.A., 2011. Infant and childdiet in Neolithic hunter-fisher-gatherers fromCis-Baikal, Siberia: Intra-long bonestable nitrogen and carbon isotope ratios. Am. J. Phys. Anthropol. 146, 225e241.

Wheeler, S.M., 2010. Nutritional and disease stress of juveniles from the DakhlehOasis, Egypt. Int. J. Osteoarchaeol. 22, 219e234.

Williams, J.S., White, C.D., Longstaffe, F.J., 2005. Trophic level and macronutrientshift effects associated with the weaning process in the Postclassic Maya. Am. J.Phys. Anthropol. 128, 781e790.