Analysis of dental attrition and mortality in the Medieval village of Tirup Denmark.код для вставкиСкачать
AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 126:169 –176 (2005) Analysis of Dental Attrition and Mortality in the Medieval Village of Tirup, Denmark Jesper L. Boldsen* Institute of Forensic Medicine, Department of Anthropology (ADBOU), University of Southern Denmark, DK 5000 Odense C, Denmark KEY WORDS dental attrition; whole mouth analysis; heterogeneity; selective mortality; nutrition; Tirup ABSTRACT New directions and new questions raised in the study of health in the past justify this reanalysis of the pattern of dental attrition in the Medieval Danish population of Tirup. Dental attrition was scored on all permanent molars from the Tirup skeletal sample. Scores were analyzed by means of logistic regression of the probability of having entered a given stage of wear for a given tooth in a way that is very similar to transition analysis. The primary determinant of dental attrition was age at death. In addition to age, the effects of sex, side, and dating were analyzed. In order to assess the homogeneity of the process of wearing teeth down, a third-order polynomial in age-at-death was also ﬁtted to the transition probabilities. It was found that age is the single most Dental attrition is a more direct consequence of the action of the environment on human hard tissue than any other marker. Teeth wear in the process of chewing. Contrary to what happens in the skeleton, there is no buildup of new mineralized tissue on the crowns of worn teeth. These facts open unique possibilities for studying the processes of wear and tear in the past. An analysis based on such individually seen unidirectional series of transitions has a much stronger possibility for revealing the effect of selective mortality and other aspects of heterogeneity of health and living conditions than other types of osteological analyses. It is a natural consequence of the unidirectionality of dental attrition that it is frequently used as a means of individual age estimation. In the absence of selective mortality, dental attrition can provide a population-speciﬁc means of age estimation; but archaeologically relevant reference samples, i.e., samples of skeletons of known age and sex coming from a population with the same way of living as any given archaeological sample of skeletons, are virtually nonexistent. Therefore, age estimation of ancient individuals from dental attrition is difﬁcult (cf. Brothwell, 1989; Lovejoy, 1985). Skeletal samples are, at best, random samples of the dead members of a once-living community (Wood et al., 1992). At all ages, it was usually the most frail in the population who died; possible exceptions would be high-status warriors, but probably not the © 2004 WILEY-LISS, INC. important determinant of dental attrition, and that sex or side did not differentiate the rate of attrition. In several transitions, there was evidence of heterogeneity, indicating both random and systematic interpersonal differences in the rate of attrition and an association between the rate of attrition and age-at-death. It was found that attrition proceeded more quickly after AD 1300 than prior to that date. It is suggested that this was due to a possible general deterioration of living conditions in Northern Europe and an increased reliance on grain for food during the ﬁrst half of the 14th century. The temporal effect on attrition rate accounts for some but not all the observed heterogeneity wear. Am J Phys Anthropol 126:169 –176, 2005. © 2004 Wiley-Liss, Inc. kind of peasant soldiers or members of rebellions who would have found their last resting place in Tirup. This means that skeletal samples are, in comparison to the age matched living population, samples of the most frail. One of the challenges is to utilize such systematically skewed samples to draw conclusions about living conditions in the equally but oppositely selected once-living population. Since the publication of Wood et al. (1992), some work was carried out to bridge the information gap between the living and the dead and promote our understanding of health in the past. A few case studies, within limited frames of reference, provided speciﬁc solutions to the problems raised by the osteological paradox (e.g., Boldsen, 1997; Poulsen et al., 2001); most of the progress (be it from a slightly different perspective) was synthesized by Hoppa and Vaupel (2002). The purpose of this paper is to follow up on *Correspondence to: Jesper L. Boldsen, Institute of Forensic Medicine, Department of Anthropology (ADBOU), SDU-Odense University, Campusvej 55, DK 5230 Odense M, Denmark. E-mail: Jboldsen@health.sdu.dk Received 4 September 2003; accepted 2 March 2004. DOI 10.1002/ajpa.20057 Published online 26 July 2004 in Wiley InterScience (www. interscience.wiley.com). 170 J.L. BOLDSEN earlier analyses of dental attrition in Medieval Denmark (Helm and Prydsø, 1979; Boldsen, 1991), and to reach an understanding of the processes in the living that created the observed patterns in the Tirup sample of Medieval Danish skeletons. The attrition scores used in this paper were published before (Boldsen, 1991), but they have not been analyzed in relationship to age estimates based on transition analysis. As there were some shortcomings in the ﬁrst paper on these data, it is justiﬁable to reanalyze and republish the Tirup dental attrition data. MATERIALS AND METHODS In total, 620 burials were identiﬁed and excavated from the Tirup cemetery (Kieffer-Olsen et al., 1986). Due to relatively few infant burials, this ﬁgure is probably a little smaller than the total number people interred in the cemetery. During the data collection for another recent research project, two additional individuals were identiﬁed among the scattered remains of bones from disturbed burials (Boldsen, 2003). The cemetery appears to have been in use from the middle of the 12th to the middle of the 14th century AD (Kieffer-Olsen et al., 1986). Based on the position of the forearms in relation to the rest of the body, it is possible to divide the burials into two only slightly overlapping temporal groups. Burials with the arms stretched along the side of the body are older than burials where the forearms were placed over the body. According to Jantzen et al. (1994), the transition between the two types of arrangements was archaeologically instantaneous, i.e., it took only a few decades for one custom to replace the other. The transition took place around AD 1300. A ditch surrounded the cemetery. There were no graves outside of the demarcated area, and an area approximately 2 m wide just inside the ditch was also empty of graves. Only a small area in the southern part of the cemetery was excavated prior to the recognition of the site as a Medieval cemetery. The excavator driver who discovered the site was very thorough in collecting the human remains that led him to the discovery. It seems likely that no adult skeletons were lost due to his nonarchaeological excavation. The skeletons in the analyses in this paper are those of adults (age 18 years or more, as determined by fusion of the spheno-occipital synchondrosis) giving information on dental attrition of at least one of the 12 permanent molars, and sufﬁciently well-preserved to yield data for transition analysis age estimation.1 A total of 149 skeletons (58% of 258 skele1 Transition analysis age estimation consists of the collection of standardized osteological observations (as described by Boldsen et al., 2002) from the pubic symphysis, the iliac auricular surface, and ﬁve different suture segments. These observations (scores) are transformed to estimates of age at death through a computer program available by request from the author. The resulting age estimates TABLE 1. Definition of dental attrition scores Score 0 1 2 3 4 5 6 7 8 9 Description No information; crown is missing due to reasons other than attrition Unworn tooth Dental attrition affects only enamel Dental attrition has exposed one “dentine island” Dental attrition has exposed two “dentine islands” Dental attrition has exposed three “dentine islands” Dental attrition has exposed four “dentine islands” Dental attrition has exposed conﬂuent “dentine island” Dental attrition has removed all enamel from occlusal surface of tooth Dental attrition has removed crown altogether tons aged 18 years or more) fulﬁlled the inclusion criteria, giving attrition information on 1–12 permanent molars. On average, each skeleton provided information about 7.9 teeth (SD ⫽ 3.5). Only being able to utilize around half of the adult burials in the analyses was a potential source of bias in the analyses. However, as lack of sufﬁcient preservation was the only exclusion criterion, this possible bias is probably a minor source of systematic error in the estimates. The registration of stages of dental attrition was carried out on each of the up to 12 permanent molars separately. Scores were given based on naked eye inspection of the teeth, and followed an old tradition of concentrating on enamel exposure on the occlusal surface (Murphy, 1959; but also used on Danish Medieval skeletons by Helm and Prydsø, 1979; Boldsen 1991). Data on the angle of attrition were not collected. The scores used in this paper are summarized in Table 1. The Murphy dental attrition scoring system is quite simple and thus facilitates a relatively quick collection of attrition data on a large number of individuals. This makes Murphy scores more suited for analyses of the kind carried out here. Each of the 12 permanent molars has an opposing tooth, i.e., the molar with the same number from the same side but from the other jaw. It follows from this that there are up to six sites where molars can occlude in the process of mastication. In order to get a measure of the area available for functional chewing, the number of such intact positions (functional chewing positions) was counted. The hypothesis behind this data collection scheme is that attrition rates go up as the number of surviving functional chewing positions go down. This number is a measure of total oral function. On an individual tooth level, it is also important to know if the tooth opposing the tooth under study is functionally present or not. The hypothesis behind this registration is that the attrition rate declines if the opposing tooth is lost. consist of a point estimate (a maximum likelihood estimate of age at death; MLA), a standard error for this estimate, and a 95% conﬁdence interval. 171 DENTAL ATTRITION AND MORTALITY IN MIDDLE AGES TABLE 2. Numbers of molars from Tirup partitioned by degree of occlusal wear Upper molar Score 1 2 3 4 5 6 7 8 9 Lower molar Description 1 2 3 1 2 3 Unworn Enamel only 1 dentine island exposed 2 dentine islands exposed 3 dentine islands exposed 4 dentine islands exposed Conﬂuent dentine islands No enamel left No crown left 0 3 6 8 21 18 65 60 21 0 58 33 24 13 3 45 15 5 3 93 20 12 10 1 11 4 1 0 2 3 3 10 38 75 51 28 0 23 16 16 27 30 61 29 10 0 79 15 19 9 15 27 21 4 In addition to data on dental conditions, age at death was estimated by means of transition analysis (Boldsen et al., 2002). Usher (2000) made the scorings of the pubic symphysis, the iliac auricular surface, and the cranial suture segments that were used for age estimation. The analyses of attrition are carried out as logistic regression analyses of the probability of having reached a given stage of attrition in the same way that Boldsen et al. (2002) did in deﬁning the curves for transition analysis age estimation. Logistic regression is regression analysis carried out on the probability parameter of a dichotomous, i.e., a binomially distributed random variable. Logistic regression is particularly well-suited to analyze individually age-progressive traits, like dental occlusal attrition in samples where individuals can only be observed once and at a random age, namely, age at death. Other types of general linear (GLM) models (logistic regression analysis is a member of the GLM family) might statistically be as suited for the analysis of data like those analyzed here, but the interpretation of results of a logistic regression analysis is much more straightforward than the interpretation of results of other types of GLM analyses. In the process of dichotomizing data for logistic regression analysis, the focus is ﬁxed on one particular event in the dental attrition history of each individual. It is the probability that the individual has reached (and/or passed through) this stage of attrition that is the object of the analysis. Here let p̂i(kPc) denote the estimated probability that tooth “i” has reached attrition score “k” or higher for any given combination of covariates (age, side, sex, presence of opposing tooth, number of functional chewing positions, or dating) “c.” The statistic p̂i(kPc) is limited to values from 0 –1, so it does not lend itself directly to regression analysis. Consequently, it is the logit-transformation of p̂i(kPc) , i.e., l̂i(kPc)that is analyzed. 冉 l̂ i (kPc) ⫽ ln 冊 p̂ i (kPc) 1 ⫺ p̂ i (kPc) ˆ p̂ i (kPc) ⫽ e li (kPc) ˆ 1 ⫹ e li (kPc) (1) l̂i(kPc) is analyzed using the following model: l̂ i (kPc) ⫽ ␣ˆ ⫹ ␤ˆ c ⫹ ε ⫽ l̂ i (kPc) ⫹ ε (2) The ␣ and ␤ parameters are estimated through (logistic) regression, and ε is an independent error term with mean 0. Here ␤ is a vector of parameters, and not necessarily a single number. The estimated ␣ and ␤ values are used to reconstruct the attrition transition curves in order to illustrate the different effect that acted on dental wear in the Tirup population. In order to capture a possible effect of logistic nonlinearity on the transition to higher stages of dental attrition, and thus ﬁnd evidence for selective mortality, the age effect includes a third-order polynomial of the transition analysis maximum likelihood estimate of age at death. The analyses are carried out as single tooth attrition stage transition estimation in two rounds. First, each tooth position is analyzed on an individual basis, with no concern given to the conditions of other teeth. In the second round of analyses, data on the number of functional molar chewing positions in the mouth and on the presence of the opposing molar are included. In the ﬁrst round of analyses, all teeth can be utilized; in the second round, only the teeth coming from individuals with information on the presence of all molars are analyzed. RESULTS Table 2 gives the number of molars classiﬁed by number position and attrition stage. Table 2 clearly reveals a gradient of attrition with, on average, higher scores on the ﬁrst molar and the lowest on the third molar. Further, it is clear that attrition stage 6 (dentine exposed in four islands) is rarely seen in second and third upper molars. This is not a reﬂection of a strange pattern of attrition, but is rather caused by the fact that many of these molars only have three cusps. To avoid any consequences of such anatomical conditions and to reduce the correlations among the transition curve estimates, numerical restrictions were applied in choosing which transitions to analyze. In the analyses, the observations were dichotomized into those who had not (the “⫺” group) and those who had passed the transition analyzed (the “⫹” group). To carry out the analysis of a transition, it was required that both groups 172 J.L. BOLDSEN TABLE 3. Test statistics (2 tests with 1 degree of freedom; significant values, P ⬍ 0.05, in bold) for transition curves describing entry into attrition scores for six different types of molars Entry into score number First upper molar 5 6 7 8 9 First lower molar 6 7 8 9 Second upper molar 3 4 7 8 Second lower molar 3 4 5 6 7 8 Third upper molar 3 4 Third lower molar 3 5 7 8 Side Sex Age Age2 Age3 Sex-age interaction Median age at transition 0.14 0.02 1.17 1.96 3.53 0.03 2.69 0.14 0.06 1.55 6.49 14.78 21.01 25.95 12.06 1.31 0.25 1.23 0.81 1.53 1.18 0.33 1.29 0.44 1.66 0.06 1.89 0.21 0.10 1.18 Age Age Age Age Age 15.7 21.1 24.2 45.6 95.0 0.57 0.28 0.58 1.33 2.91 0.76 0.51 3.09 5.46 15.20 16.56 1.22 0.06 2.21 1.47 0.34 0.06 1.87 1.24 0.26 0.11 0.33 1.29 2.59 Age Age Age None 5.4 19.8 51.1 0.42 5.07 10.80 3.54 3.17 1.90 Age ⫹ sex Model M: 7.4 F: 25.0 26.9 63.1 109.8 0.22 1.04 0.46 0.02 0.06 1.59 9.12 9.25 7.10 6.51 1.31 0.01 4.97 0.90 0.00 0.72 0.25 0.98 Age ⫹ age ⫹ age Age Age 0.87 0.12 0.34 0.30 0.23 0.20 0.30 0.01 1.53 0.03 0.72 0.00 7.37 4.73 7.26 12.14 7.30 10.20 1.08 4.06 5.57 2.02 0.01 0.01 1.10 3.75 4.63 1.52 0.01 0.08 0.33 0.02 1.43 0.14 0.07 0.08 Age Age ⫹ age2 ⫹ age3 Age ⫹ age2 ⫹ age3 Age Age Age 1.0 20.2 22.8 23.8 41.7 91.4 1.14 0.94 1.90 0.21 5.10 4.54 0.15 0.12 0.06 0.17 2.95 0.14 Age Age 64.5 90.2 0.37 0.54 0.06 0.00 1.03 0.78 0.45 0.10 6.94 4.71 5.10 0.01 5.67 4.22 0.10 0.02 5.04 4.08 0.01 0.09 0.03 1.18 0.71 0.25 Age ⫹ age2 ⫹ age3 Age ⫹ age2 ⫹ age3 Age None 26.7 66.1 86.6 - 2 3 consisted of more than 15 observations, and that more that 15 observations had moved from the “⫹” to the “⫺” group compared to the previous transition analyzed. This means, for example, that only the entry to stages 3, 5, 7, and 8 was analyzed for the third lower molar. Individual tooth analysis Table 3 summarizes all tests for both discrete and continuous effects acting on the probability of having entered a given attritional stage for each of the six molar positions. Table 3 clearly reveals the general ﬁnding that age in all but two transitions plays a signiﬁcant part in the transition into a given attrition stage. This fact is illustrated in Figure 1, which shows the transition curves for the ﬁrst upper molar. On the other hand, there is never a side differentiation in the transition probability. This ﬁnding is not unexpected, but it shows that molars were not used extensively to perform other tasks than chewing in Tirup. Sex and sex-age interaction is only signiﬁcant for one transition out of 25 tests, so this can probably be considered an example of mass signiﬁcance. There are ﬁve transitions that did not proceed in a logistic-linear way. These could also be examples of mass signiﬁcance, but they all show the same pattern of logistic nonlinearity. Figure 2 illustrates this effect by showing all the logistic nonlinear transition curves in the same plot. It is clear that the transition proceeded quickly in ages be- Fig. 1. Transition among stages of attrition of ﬁrst upper molar. This is an example of an uncomplicated logistic-linear progression of dental attrition. tween 20 and around 40, followed by an age interval with apparently no attrition, from around 40 to around 60 years, after which age, attrition appears to have started again. Whole dentition analysis In Table 4, the effects of the presence of the opposing molar and the number of functional chewing DENTAL ATTRITION AND MORTALITY IN MIDDLE AGES Fig. 2. Examples of logistic-nonlinear transition curves. Flat and even declining parts of these curves can only be interpreted as evidence that selective mortality as dental attrition on an individual level is an absolutely irreversible process, so lack of increasing attrition with increasing age on population level can only be brought about by a positive association between rate of wear and risk of dying. Curve 1, transition into attrition stage 5 in lower third molar; curve 2, transition into attrition stage 3 in lower third molar; curve 3, transition into attrition stage 4 in upper second molar; curve 4, transition into attrition stage 5 in lower second molar. positions between permanent molars are analyzed. The picture created by Table 4 is much more confusing than the one created by Table 3. In general, there are fewer signiﬁcant associations—probably at least partly a consequence of having fewer teeth to analyze, but the pattern of associations is also less consistent among teeth. The clearest image is created for the higher attrition scores of the ﬁrst upper molar. As generally expected, this tooth shows higher levels of wear when the opposing tooth was in position in the mouth. Further, this tooth clearly shows the attrition effect of the number of functional chewing positions in the mouth. The ﬁrst lower molar shows a very similar pattern of associations, but with fewer statistically signiﬁcant associations. Figure 3 illustrates the effect of the number of chewing positions on the transition into stage 8 for the ﬁrst lower molar. For the second and third molars, the effect of the presence of the opposing tooth seems to be reverse of the pattern seen for the ﬁrst molars. This might be a reﬂection of a higher level of mobility of the teeth further back in the mouth, so that attrition becomes less dependent on the immediately opposing tooth and more a reﬂection of the generally increased load brought about by missing at least one functional chewing position. Temporal aspects of dental attrition Chronology is the only potential source of heterogeneity that is possible to analyze with the present 173 level of registration of the skeletons. The ﬁrst step in this analysis is to estimate the effect of late vs. early dating of the graves on the transition probabilities for the different stages in all six molars on the level of the individual tooth. The results of this analysis are summarized in Table 5. The majority of odds ratios are over 1 (20 out of 25 estimates, sign test 2 ⫽ 17.64, df ⫽ 1, P ⬍ 0.001). This indicates that attrition for most teeth and into most stages proceeded more quickly late (ca. AD 1250 –1350) than early (ca. AD 1150 –1300) in the history of the Tirup population. Nearly half of the odds ratios (11 out of 25) are statistically signiﬁcant (P ⬍ 0.05), and all signiﬁcant odds ratios are larger than 1. However, this might partly be due to intercorrelation among the odds ratios, as the analysis of six different types of molars (upper and lower M1, M2, and M3) does not provide six times as much information as a analysis of a single tooth. In spite of this, the results reveal a temporal heterogeneity in dental attrition in the Tirup population. Five transition curves show evidence for signiﬁcant deviations from logistic linear (Table 3) progression with age, and thus for an association between individual-level risk of dying (frailty) and dental attrition. Four of these ﬁve transitions show a statistically signiﬁcant dating effect (Table 5). The models describing the relative importance of heterogeneity independent of a dating effect and a dating effect independent of heterogeneity (logistic nonlinearity of the transition curves) are not nested in each other. This means that it is not possible directly to test them against each other. However, it is possible to test them against a comprehensive model containing both effects at the same time. This comprehensive model describes the transition probability (or rather, logistic transformation of this probability) as a function of the following factors: age, age2, age3, dating, age * dating, age2 * dating, and age3 * dating. In this list, “*” indicates interaction effects. The two models nested in this comprehensive model can be denoted the heterogeneity model (with the factors age, age2, and age3) and the dating model (with the factors age, dating, and age * dating). Table 6 summarizes the analysis of the two last models in relationship to the ﬁrst, comprehensive model. Dating of the burial is clearly an aspect of the heterogeneity that can be seen in the rate of attrition of the permanent molars in the Tirup population. However, it is also clear that dating is not the only factor affecting interpersonal differences in the rate of attrition. Figure 4 illustrates the transition curve for entry to stage 4 in the second upper molar. Figure 4 illustrates the quicker transition in the late than in the early period, and it shows how substantial the heterogeneity of individual transition rates remains after removing the dating effect. It appears that heterogeneity played a much bigger part in this transition in the early than in the late period. 174 J.L. BOLDSEN TABLE 4. Estimates of effect (odds ratio) of opposing tooth being present and of having one additional functional chewing position along with 2 test statistics for hypothesis that these odds ratios are 11 Whole mouth effects Opposing tooth present Stage First upper molar 5 6 7 8 9 First lower molar 6 7 8 9 Second upper molar 3 4 7 8 Second lower molar 3 4 5 6 7 8 Third upper molar 3 4 Third lower molar 3 5 7 8 1 One extra chewing position 2 test Odds ratio 2 test 0.00 1.23 36.59 43.02 4.12 0.03 0.03 7.78 8.60 0.84 0.97 0.75 0.53 0.50 0.56 0.01 1.29 5.50 11.98 4.40 Age Age Age ⫹ OP ⫹ PO Age ⫹ OP ⫹ PO Age ⫹ PO Age Age Age None 0.21 3.40 3.21 1.28 0.96 1.20 1.46 0.06 0.76 0.70 0.49 0.58 0.81 2.98 13.91 6.77 Age Age Age ⫹ PO PO Age ⫹ sex Age ⫹ age2 ⫹ age3 Age Age 0.00 0.37 0.44 8.95 0.09 0.54 0.75 2.04 1.04 0.89 0.78 0.48 0.04 0.45 2.66 6.77 Age ⫹ sex Age ⫹ age2 Age Age ⫹ PO Age Age ⫹ age2 ⫹ age3 Age ⫹ age2 ⫹ age3 Age Age Age 0.57 0.12 0.25 0.19 0.54 0.15 0.42 4.06 2.22 4.03 0.53 1.22 0.85 0.98 0.85 0.85 0.65 0.95 0.67 0.01 1.05 1.14 14.43 Age Age 0.98 0.62 0.00 0.11 1.23 1.18 1.60 0.92 0–36 0.30 0.49 0.27 1.67 2.90 1.06 2.58 0.66 0.69 0.58 0.56 7.09 8.68 17.63 17.93 Single tooth model Age Age Age Age Age Age ⫹ age2 ⫹ age3 Age ⫹ age2 ⫹ age3 Age Odds ratio Whole mouth model Age Age Age ⫹ age2 ⫹ age3 ⫹ OP Age Age ⫹ OP Age ⫹ PO Age Age Age ⫹ PO Age ⫹ PO PO PO df ⫽ 1 ⫺ signiﬁcant values, p ⬍ 0.05, in bold. OP, opposing tooth present; PO, one additional functional chewing position. DISCUSSION Fig. 3. Estimated transition probabilities for entry to stage 8 in ﬁrst lower molar, shown as function of age at death and number of functional chewing positions in mouth. Lowest curve is one for dentitions with six surviving functional chewing positions, and others are ordered in sequence, down to zero functional chewing positions (top curve). For stage 8 of ﬁrst lower molar, number of chewing positions is a stronger determinant of transition probability than age, but both are highly signiﬁcant. Dental attrition is a nonpathological outcome of wear and tear. Teeth are the only hard parts of the body that do not react to external forces by both breaking down and building up of tissue. Therefore, dental attrition is a truly unidirectional process. On an individual level, dental attrition can only proceed from one stage to the next at whatever pace is set by the strength of the individual teeth and by the hardness and quantity of the material chewed. It is obvious that the time a tooth has been exposed to attrition is a generally important aspect of the amount of wear the tooth has experienced. This fact is reﬂected in Figure 1, showing the smooth succession of transition curves for the ﬁrst molar lower. However, the relationship between time of use of a tooth (i.e., age) and level of attrition is confounded by several factors that make it impossible to use attrition scores for age estimation, even within a relatively homogeneous population such as Tirup. Clearly, some of the heterogeneity might be a consequence of the inaccuracy of the age estimates used in the analyses. However, inaccuracy is not enough to create the image of diverse attrition rates; a direct association between dental attrition and other aspects of skeletal aging would be required to create the image illustrated in Figure 2. DENTAL ATTRITION AND MORTALITY IN MIDDLE AGES 175 TABLE 5. Estimates of the effect (odds ratio) of late vs. early dating of grave on transition probability along with 2 test statistics for hypothesis that these odds ratios are 11 Stage First upper molar 5 6 7 8 9 First lower molar 6 7 8 9 Second upper molar 3 4 7 8 Second lower molar 3 4 5 6 7 8 Third upper molar 3 4 Third lower molar 3 5 7 8 1 Odds ratio 2 Age Age Age Age Age 4433.5 5.1 2.2 1.4 0.3 9.51 7.92 1.60 0.77 2.75 Age Age Age 5.4 3.4 1.4 0.8 3.68 5.45 0.61 0.00 Age ⫹ sex Age ⫹ age2 ⫹ age3 Age Age 0.4 1.5 1.2 1.2 0.00 3.97 2.75 1.11 Age Age ⫹ age2 ⫹ age3 Age ⫹ age2 ⫹ age3 Age Age Age 2.8 1.2 1.3 1.7 1.4 0.7 2.57 2.46 3.92 8.31 6.21 1.81 Age Age 1.7 1.8 5.08 3.94 Age ⫹ age2 ⫹ age3 Age ⫹ age2 ⫹ age3 Age 1.6 1.8 0.7 0.8 9.56 5.43 1.75 0.16 Single tooth model df ⫽ 1 ⫺ signiﬁcant values, p ⬍ 0.05, in bold. TABLE 6. Tests for reduction of model containing both dating and heterogeneity effects to nested, simpler models containing only one of two factors1 Model 2 test for Tooth and stage Second upper molar, 4 Second lower molar, 4 Second lower molar, 5 Third lower molar, 3 Third lower molar, 5 Dating Heterogeneity 15.22 4.38 6.77 8.34 5.43 10.06 2.59 4.20 15.11 6.94 Model accepted/ best model Comprehensive Heterogeneity Heterogeneity Dating Dating 1 A signiﬁcant test indicates a signiﬁcant loss of ﬁt when moving from comprehensive to reduced model. Signiﬁcant test statistics are given in bold. The Tirup skeletal sample is (like all skeletal samples) a cross section of the population at death. It is only the oral condition at time of death that it is possible to examine. The lack of time depth in the registrations is probably the reason why the presence of the opposing tooth plays so minor a part for the level of attrition. The total number of functional chewing positions is more stable, as it is like an average that contains information from 12 different teeth. This is probably the reason why this measure is more closely correlated to the level of attrition than is the presence of the opposing tooth. It is, under all circumstances, important to evaluate the whole mouth when analyzing attrition. In the present study, only the molars were analyzed. This means that it was not possible to analyze the poten- Fig. 4. Estimated transition probabilities for entry into attrition stage 4 of second upper molar, shown separately for early (AD 1150 –1310) and late (AD 1290 –1350) graves. tial consequences of tooth loss/compensatory mastication area in the dental arcs in front of the molars. In future studies of dental attrition and oral function of past populations, it will be important to include whole mouth information in the registrations. Probably the distribution of strength of teeth has remained more or less unchanged over the centuries, so the primary difference between patterns of attrition is brought about by the hardness and quantity of the consumed food, chewed by different individual human beings and in general by people in different historical periods. Most skeletal samples are (like the Tirup sample) accumulated over prolonged periods of time, comprising both relatively afﬂuent and relatively poor phases. Such ﬂuctuations will inevitably create heterogeneity among individuals, both in the rate of dental attrition and in the risk of dying. But this is not the only source of systematic differences among the people forming a skeletal sample. In virtually all societies, social status differences are reﬂected in access to and quality of food; such differences will usually also be reﬂected by the risk of dying at different ages. On top of systematic temporal and social differences among individual human beings, more random differences of the genetically determined strength of teeth create heterogeneity. The random component of attritional heterogeneity is unlikely to be associated with the risk of dying; but the systematic, temporally, or socially associated heterogeneity is likely to be reﬂected in mortality. The general uniformity of Christian Medieval burials and the lack of strong and uniform indicators of social differences in the pattern of burials (as seen in Löddeköpinge; see Cinthio and Boldsen, 1984) make it virtually impossible to analyze the effect of social inequality on the rate of dental attrition. This makes temporal differences the only source of sys- 176 J.L. BOLDSEN tematic interpersonal heterogeneity in the rate of dental attrition that can be analyzed in the present context. The period from around 1150 –1350 saw some dramatic changes in living conditions in Northern Europe. It appears that the population had a maximum size by the end of the 13th century. Prior to 1300, it appears that the population was generally and slowly growing, leading to the formation of many rural village communities such as Tirup. This came to an end in the 14th century. In the second and third decades of that century, most of Northern Europe was affected by several consecutive years of famine (e.g., Campbell, 1991; Dyer, 1989; Hybel, 1997, 2002). In the middle of the century, the Black Death took away a large portion of the population. It is possible that a downturn of living conditions in the ﬁrst half of the 14th century is reﬂected in the higher rate of dental attrition among people buried in late arm position B (forearms placed over rather than along the body) graves. However, it is more likely that a change in subsistence pattern with increasing reliance on grain for food from the 12th to the 14th century is the main reason for the increased level of dental attrition. On the other hand, it is clear that this temporal shift in the rate of attrition cannot account for all the heterogeneity in the rate of dental attrition observed in the Tirup skeletal sample. The analyses in this paper clearly show that there is no such thing as a preindustrial pattern of dental attrition that can be used for age estimation. In general, teeth of older individuals were more worn than teeth of younger people, as expected. However, heterogeneity of the rate of attrition (on an individual level, on a social level, and on a temporal level) changed attrition scores from indicators of age to indicators of health and well-being. Other aspects of oral health and general survival were tested and found negative for a dating/arm position effect (Boldsen, 1988, 1997, 2002). The fact that teeth preserve well and that they, through chewing, interact directly with the environment is probably the reason it has been possible to ﬁnd a dating effect on dental attrition and not on other indicators of public health in the Tirup population. Probably the lack of significant dating effects for other variables in a statistical sense is a type 2 error. More information may be extracted from the Tirup skeletons by the application of more elaborate registration schemes. However, it is likely that the lower number of complete observations, which such schemes would lead to, are sufﬁcient to remove statistical signiﬁcance from all or most of the test. The planning of osteological research into the health of past populations is a delicate balance between quantity and quality of data. ACKNOWLEDGMENTS This research would not have been possible without the highly skilled technical assistance from lab- leader Ulla H. Freund. Bethany M. Usher is gratefully acknowledged for giving access to her scorings for transitions analysis age estimation. Historian Nils Hybel of the University of Copenhagen read the manuscript, which beneﬁted considerably from his comments. LITERATURE CITED Boldsen JL. 1988. Two methods for reconstructing the empirical mortality proﬁle. Hum Evol 3:335–342. Boldsen JL. 1991. Ageing and dental attrition in a medieval rural Danish population. Int J Anthropol 6:217–224. Boldsen JL. 1997. Estimating patterns of disease and mortality in a Medieval Danish village. In: Paine RR, editor. Integrating archaeological demography: multi-disciplinary approaches to prehistoric population. Carbondale: University of Southern Illinois Press. p 229 –241. Boldsen JL. 1999. A child, a tooth—the cost of reproduction in the Middle Ages. Research report, Danish Center for Demographic Research 9. Odense: University of Southern Denmark. Boldsen JL. 2002. Demograﬁsk struktur i landsbyen Tirup. Hikuin 27:233–244. Boldsen JL. 2004. Leprosy and mortality in the medieval Danish village of Tirup. Am J Phys Anthropol (in press). Boldsen JL, Milner GR, Konigsberg LW, Wood JW. 2002. Transition analysis: a new method for estimating age from skeletons. In: Hoppa RD, Vaupel JW, editors. Paleodemography. Age distributions from skeletal samples. Cambridge: Cambridge University Press. p 73–106. Brothwell DR. 1989. The relationship of tooth wear to ageing. In: Iscan MY, editor. Age markers in the human skeleton. Springﬁeld, IL: Charles C. Thomas, Ltd. p 303–316. Campbell BMS, editor. 1991. Before the Black Death. Studies in the “crisis” of the early fourteenth century. Manchester: Manchester University Press. Cinthio H, Boldsen JL. 1984. Patterns of distribution in the early medieval cemetery at Löddeköpinge. The Löddeköpinge investigation IV. Meddelanden Lunds Univ Hist Mus 1983–1984: 116 –127. Dyer C. 1989. Standards of living in the late Middle Ages. Social change in England c. 1200 –1520. Cambridge: Cambridge University Press. Helm S, Prydsø U. 1979. Assesment of age at death from mandibular molar attrition in medieval Danes. Scand J Dent Res 87:79 –90. Hoppa RD, Vaupel JW, editors. 2002. Paleodemography. Age distributions from skeletal samples. Cambridge: Cambridge University Press. Hybel N. 2002. Klima og hungersnød i middelalderen. Hist Tidsskr 102:265–281. Jantzen C, Kieffer-Olsen J, Madsen PK. 1994. De små brødres hus i Ribe. Mark Montre 1994:26 –36. Kieffer-Olsen J, Boldsen JL, Pentz P. 1986. En nyfunden kirke ved Bygholm. Vejle Amts Arbog 1986:24 –51. Lovejoy CO. 1985. Dental wear in the Libben population: its functional pattern and role in the determination of adult skeletal age at death. Am J Phys Anthropol 68:15–28. Murphy T. 1959. Gradients of dentin exposure in human molar tooth attrition. Am J Phys Amthropol 17:179 –186. Poulsen LW, Qvesel D, Brixen K, Vesterby A, Boldsen JL. 2001. Low bone mineral density in the femoral neck of Medieval women: a result of multiparity? Bone 26:454 – 456. Usher BM. 2000. A multistate model for health and mortality in paleodemography: Tirup cemetery. Ph.D. dissertation, Pennsylvania State University. Wood JW, Milner GR, Harpending HC, Weis KM. 1992. The osteological paradox. Curr Anthropol 33:343–370.