Cementum annulation and age determination in Homo sapiens. II. Estimates and accuracyкод для вставкиСкачать
AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 71:321-330 (1986) Cementum Annulation and Age Determination in Homo sapiens. It. Estimates and Accuracy KEITH CONDON, DOUGLAS K. CHARLES, JAMES M. CHEVERUD, AND JANE E. BUIKSTRA Department ofAnthropology, Northwestern University, Euanston, Illinois 60201 K. C., D.K.C., J. M. C., J.E.B.);Department ofAnatomy, University of Illinois at Chicago, Chicago, Illinois 60612 K.C.):Department of Cell Biology and Anatomy, Northwestern University, Chicago, Illinois 60611 (D.K.C., J.M.C.) KEY WORDS Cementum, Aging, Teeth, Forensics ABSTRACT The cementum annulation aging technique was evaluated in a sample of 80 clinically extracted premolars (age range 11-70 years). Demineralized thin sections (7pm) stained with hematoxylin were used. The correlation (r) between age and adjusted count (number of annulations added to age of tooth eruption) was 0.78 for the entire sample (N = 73) and 0.86 for a subsample in which teeth with periodontal disease were excluded (N = 55). Standard error of the estimates ranged from 4.7 to 9.7 years depending on sex and health status of the tooth. The technique provided significantly better estimates for females than for males. The overall inaccuracy (mean absolute error) of the technique was 6.0 years, with a bias (mean error) of 0.26 years. Reduced major axis regression of adjusted count on age produced a slope of 0.797 for the entire sample and 0.889 for the nonperiodontal disease subsample. These slopes are consistent with a hypothesis of annual deposition of cementum rings given a decrease in cementogenesis with increasing age. An incremental structure of alternating light and dark bands has been reported in the dental cementum of at least 19 families of mammals, including several primates (Wada et al., 1978; Kay et al., 1984; Stott et al., 1980). Counts of these layers, or annulations, have been employed and accepted as accurate indicators of age among wildlife biologists (see reviews by Grue and Jensen, 1979; Klevezal’ and Kleinenberg, 1967; Morris, 1978; Spinage, 1973). Two recent studies (Naylor et al., 1985; Stott et al., 1982) have claimed accurate estimates of age in humans using cementum layers in thick-cut (100 pm) mineralized sections. Stott and coworkers (1982),using a sample of three cadavers (ages 57,67, and 76 years) found 4 years to be the greatest deviation between layer count (added to age of tooth eruption) and age at death. Unfortunately, the small sample size of the study precludes evaluation of the accuracy of the estimates. In the study by Naylor and coworkers (1985), which was primarily methodological, neither sample size nor accuracy of the technique is reported. In neither study are measures of in0 1986 ALAN R.LISS, INC. ter- and intraobserver error or variability between and within tooth types in layer formation considered. In a previous study, Charles and coworkers (1986) evaluated inter- and intraobserver error in cementum layer counts and inter- and intratooth variability in layer formation in a sample of human mandibular canine and first premolar pairs. Whereas canines produced counts with high interobserver error, counts based On premolars provided reasonable levels of observer error and intratooth variability. In an extended sample of 51 premolars, intraobserver variance was 2%, interobserver variance 5%, and intratooth variance (i.e., variability in counts between two sections of the same tooth) 4% of the total variance. Thus cementum layer counts in Received January 24,1986; revision accepted May 2. 1986. Address reprint requests to Keith Condon, Department of Oral Anatomy, College of Dentistry, University of Illinois at Chicago, 801 S. Paulina, Chicago, IL 60612. 322 K. CONDON, D.K. CHARLES, J.M. CHEVERUD, AND J.E. BUIKSTRA premolars offer a highly repeatable method of estimating age at death. This study evaluates the accuracy of age estimates based on cementum layer counts in a sample of 80 human premolars. Equations for age estimation and for calculating the error of these estimates will be provided. The relationship between age and layer formation will also be investigated. MATERIALS AND METHODS Unfixed clinical extractions were collected from the Oral Surgery Clinics of Northwestern University and the University of Illinois a t Chicago and from private practitioners in the Chicago area. Date of birth, date of extraction, sex, tooth identification and reason for extraction were requested for each tooth. The latter information, reason for extraction, was not always supplied. Additionally, several teeth of known age and sex were generously donated by Dr. K. Bennett of the University of Wisconsin. Canines and premolars were chosen for analysis, because examination of cadaver dentitions collected for this study (N = 112) showed a high incidence of premortem tooth loss for incisors and molars. Cadaver dentitions were not used because the cementum layers were not as clearly defined in them as in clinical extractions. In a sample of mandibular canine and first premolar pairs from 42 cadavers, consistent layer counts could not be obtained by one or both observers in 11individuals (26%; Charles et al., 1986). In contrast, in a sample of 115 clinically extracted canines and premolars, there were no instances in which counts could not be made. This suggests that some substance in the preservative used in the cadavers has a n effect on the staining quality of the layers. Demineralized root segments were sectioned longitudinally (7 pm) and stained with hematoxylin following the procedure of Charles et al. (1986). Thick-cut mineralized sections were not used because of high rates of intraobserver error reported for this technique (Charles et al., 1986).The sections were evaluated using brightfield microscopy a t magnification x 400. Areas of the cementum showing 1)clear definition of layers from the cementodentin junction to the periodontal ligament and 2) no evidence of resorption andor remodeling were photographed using Ektachrome tungsten 160 film. Two to ten such areas were photographed for each individual. The best two photographic slides based on criterion 1 above were selected for counting. The age distribution of the sample was unknown to both the photographers and observers. The photographic slides comprising the sample were randomly mixed prior to scoring. The slides were projected to approximately 22 x 28 cm, and repeated layer counts of the projection were made until a consistent count was obtained. The mean of these consistent counts was recorded as the number of annulations for that section. A cementum layer was operationally defined as a pair of adjacent dark and light bands. All slides were counted by a single observer a single time. The two counts for each individual were then averaged, and this average was used as the best estimate of layer number. Because several tooth types were used, counts were adjusted by adding the age of eruption to the mean count (Table 1).Ages of eruption for the mandibular canine and premolars were taken from Demirjian and Levesque (1980). Sex-specificeruption times were available for these teeth. Maxillary dental eruption times were taken from Thoma and Goldman (1960). The midrange of the age interval of eruption was used, and sex was not distinguished. Note that the age ranges of the maxillary dentition from Thoma and Goldman (1960)encompass those of the study by Demirjian and Levesque (1980). These studies were performed on populations from northern temperate climates. Based on the findings of Charles et al. (1986),the canines (N = 44) were eliminated from further analysis because of high interobserver error for these teeth. The age distribution of the remaining 80 individuals (represented by premolars) is shown in Figure 1. Equations for the estimation of age were generated by using least-squares regression. This method assumes no error in the measure of the independent variable (i.e., adjusted layer count; Sokal and Rohlf, 1981). Although this is clearly not the case, as demonstrated by the presence of observer error (Charles et al., 19861, least-squares regression is still the preferred technique since we are interested in estimating age from layer counts and attaching confidence intervals to these estimates (Seim and Saether, 1983). The degree of association between age and adjusted layer count was measured by correlation coefficient and measures of inaccuracy and bias. Inaccuracy is the average absolute 323 CEMENTUM ANNULATION AGING TECHNIQUE 12 > 0 z 9 W 3 c l W a LL 6 3 10 15 20 25 30 35 40 45 50 55 60 65 AGE Fig. 1. Age distribution of the premolar sample (N = 80). The numbers below each bar indicate the beginning age in years of the age class. TABLE 1. Dental eruption ages used in adjusting counts for aee estimates Demirjian and Levesque, 1980 Tooth Male Female Thoma and Goldman, 1960 (combined) 9.6 10.5 9-10 10.3 10.7 10-12 11.2 11.6 - - 11-12 11-12 ~ C P1 ~ - P2 c P1 - - - 10-11 P2 - - 10-12 error of age estimation without reference to over- or underaging, and bias is the mean over- or underprediction (Lovejoy et al., 1985). These latter measures (inaccuracy and bias) are held to be more effective indicators than the correlation coefficient, which is influenced by the age range and composition of the test sample (Lovejoy et al., 1985).Inaccuracy and bias were calculated for each decade of the sample. Because of difficulties in generating a sufficient sample, no additional teeth were available to test the equations using an independent sample. This problem was circumvented by using the jackknife technique (Sokal and Rohlf, 1981).In this test, N regressions are run on sample sizes of N - 1,leaving out a different individual in each run. Age of the excluded individual is then estimated using the derived regression. The standard deviation of the residuals or standard error of the estimate was computed using standard jackknife techniques (Sokal and Rohlf, 1981). Thus the jackknife provides an evaluation of the standard error of the regression using an outside sample. Little difference was detected between standard and jackknife estimates of parameters or errors. Finally, the structural relationship between age and layer formation was analyzed using reduced major axis regression. This model assumes measurement error in both variables and better describes the relationship between the two variables than does least-squares regression. This model assumes the relative sizes of the two error variances to be equal (Seim and Saether, 1983). RESULTS Age estimates Three of the 80 individuals (4%;two males, one female; ages 24, 37, 63 years) showed no areas of layer formation suitable for estima- 324 K. CONDON, D.K. CHARLES, J.M. CHEVERUD. AND J.E. BUIKSTRA tion. [This percentage was slightly higher in the canines, where 12% (five of 44) of the individuals showed incomplete or poorly defined incremental cementum structure.] An additional four individuals were removed from analysis following least-squares regression of the remaining 77 individuals. Examination of the residuals showed these four to be outliers having a very strong effect on the regression (Sokal and Rohlf, 1981). Examination of the four cases showed all to be possible cases of “doubling,” a phenomenon in which twice as many layers are present as would be predicted given a n annual deposition. “Doubling” has been reported in several species (Grue and Jensen, 1979; Spinage, 1973). All four cases were maxillary second premolars from males aged 30-59. The least-squares regression of age on adjusted count for the remaining 73 individuals produced a regression line with slope 0.973 and a y intercept of 2.4 years (Table 2, equation l).The slope is not significantly different from 1.0, and the y intercept is not significantly different from 0 years. The correlation coefficient is 0.78, and the standard error of the estimate is 9.6 years. A breakdown by sex of this sample produced no major differences; neither slope nor y intercept is significantly different. Similarly, the correlation coefficients for the two sexes were not statistically different, although the standard error of the estimate was smaller for females (8.2 years) than for males (10.9 years). Among the 73 individuals of the sample, there were 18 reported cases of clinical periodontal disease. Since cementogenesis involves calcification of the periodontal ligament, any pathology altering the integrity of the periodontium might be predicted to have a n effect on cementogenesis and cementum ring formation. As a partial test of this hypothesis, the 18 known cases of periodontal disease were removed. The resulting regression produced a slope of 0.965 with a y inter- cept of 0.4 years (N = 55; Table 2, equation 2). Again, the slope is not significantly different from 1.0, and the y intercept is not different from 0.0 years. The correlation coefficient is 0.86 which is not significantly different from that of equation 1. However, the standard error of the estimates has decreased to 7.4 years. Partitioning this sample by sex produced significantly different (P < .0001) correlation coefficients (males 0.733; females 0.946). However, neither the slopes nor the y intercepts of the regression lines are significantly different. Standard errors of the estimates are 4.7 years for females and 9.4 years for males. The significantly different correlation coefficients between sexes suggests using sexspecific equations for age estimation. If sex is unknown, and periodontal disease not present, then equation 2 should be used. Table 3 shows the inaccuracy (mean absolute error) and bias (mean error) by decade for the nonperiodontal disease sample. Normalized values for the entire age range show a n inaccuracy of 6.0 years, with essentially no bias. Error estimates Plots of adjusted count on residuals showed all the samples to be heteroscedastic, with the spread of the residuals increasing with increased count. This heteroscedasticity produces large standard estimates of the error for all age ranges, although examination of the scatterplot (Fig. 2) shows that age estimates for younger individuals will probably be better than estimates for older individuals. To compensate for the heteroscedasticity, two alternate error estimates can be used with equation 2. The first alternate was computed by dividing the sample into two groups, those with adjusted counts < 30 and those with adjusted counts > 30. The mean adjusted counts of the two groups were plotted against the mean of their respective residuals and the slope and intercept of the line TABLE 2. Regression equations and correlation coefficients - Eauation N Y Intercent Sloue r r2 s.e. 1 la: males lb: females 2 2a: males 2b: females 73 34 39 55 26 29 36.2 37.4 35.3 32.2 34.8 29.8 2.4 5.4 0.0 0.4 2.6 - 1.4 ,973 ,854 1.085 ,965 .775* .651* .864* .858* .733* .946* ,601 ,423 ,746 ,735 ,537 2394 9.7 10.9 8.2 7.4 9.4 4.7 ,884 1.046 325 CEMENTUM ANNULATION AGING TECHNIQUE TABLE 3. Comparison of inaccuracy and bias ofthe cementum annulation aging technique with the summary age and revised pubis techniques Age group (years) Cementum 11-17 Inaccuracy Bias 18-29 _ _ _. Inaccuracy Bias 30-39 Inaccuracy Bias 40-49 Inaccuracy Bias 50-59 Inaccuracy Bias All ages Inaccuracy Bias All aees normalized' Inaccuracy Bias Revised pubis' Summary' 3.0 3.0 - - 4.9 2.6 3.0 2.7 3.1 1.9 8.7 5.8 5.6 4.3 5.5 3.5 4.3 -3.0 5.7 1.2 6.1 0.2 8.9 -7.1 7.6 3.4 7.5 1.7 5.6 0.8 5.2 2.5 5.8 1.3 6.0 0.3 5.2 1.7 6.5 -0.4 - I 'From Lovejoy et al. (1985;Table 3). 'Normalized error measures are the unweighted averages over all decades. P . P A& A k > P P P P lo A0 ADJUST!; COUNT 10 40 60 1 Fig. 2. Scatterplot and regression line of age on adjusted count for the nonperiodontal disease sample (N = 55). 326 K. CONDON, D.K. CHARLES, J.M. CHEVERUD, AND J.E. BUIKSTRA connecting them derived. Multiplying the adjusted count by the slope and adding the y intercept provides a n error estimate that increases with increasing age. This error estimate is calculated by the following equation: Sz = 0.342 + 0.178 (adjusted count). (3) The second alternative was derived by regressing the absolute value of the residuals against adjusted count. Again, the error estimate is obtained by multiplying the adjusted count by the slope and adding this to the y intercept, i.e., S3 = 1.295 + 0.1332 (adjusted count). (4) Examination of the two alternate error estimates shows both to be a function of adjusted counts. Thus individuals with low adjusted counts will have error estimates less than the general standard error of the regression given in Table 2. For example, a n individual with an adjusted count of 24 produces error estimates of 7.4 years (sl; standard error of the regression), 4.6 years (sz), and 4.5 years ( ~ 3 )For . a n individual with a n adjusted count of 60, the error estimates are 7.4 years (sl), 11.0 years (sz), and 9.3 years ( ~ 3 ) . Weighted least-squares regression was not used to correct for heteroscedasticity because it provides a single point estimate of the standard error of the regression. We believe that the standard errors for predicted ages should increase with age, as provided by the alternatives (sz and s3) above. In summary, three error estimates are provided for use with age estimates derived from equation 2. The first is the standard error of the estimate of the regression (7.4 years; Table 2, equation 2). This estimate does not compensate for the heteroscedasticity of the sample. The other two estimates are derived using equations 3 and 4 and compensate for the heteroscedasticity, providing narrower error estimates for younger ages and wider estimates for older ages. Thejackknife The jackknife analysis yielded standard error estimates essentially identical to those produced by the regression. In the sample of 73 individuals (equation 11, the standard error was 9.7 years and the standard error of the jackknife 9.8 years. Similarly, the standard error of the estimate for the nonperio- dontal disease sample was 7.4 years and the jackknife, 7.5 years. Jackknife regression coefficients and associated standard errors were also nearly indistinguishable from those obtained by standard least-squares techniques. Reduced major axis If cementum rings are produced annually, then reduced major axis regression of adjusted count on age should produce an axis with slope 1.0 and y intercept of 0.0 counts. The reduced major axis of the full sample has a slope of 0.797 and a y intercept of 5.4 counts. Removing the cases of known periodontal disease increased the slope to 0.889 with a y intercept of 4.3 counts. Subdividing this sample by sex showed females (N = 29) to have a n axis with a slope (0.905) and y intercept (2.8 counts) close to the norm for annual deposition. The male (N = 26) axis was flatter (slope = 0.829), with a correspondingly higher intercept (7.6 counts). DISCUSSION The results of this study show a high correlation between adjusted counts of cementum layers (i.e., the number of layers added to age of tooth eruption) and age at extractiorddeath. Thus cementum layers in premolars can be used to estimate age with error estimates ranging from 4.7 to 9.7 years, depending on sex of the individual and health status of the tooth. Table 4 compares the correlation coefficients reported in this study with those reported for a variety of age indicators by Lovejoy et al. (1985). The correlation coefficient for the entire premolar sample (N = 73; r = 0.78) is equal to the best single indicator reported by Lovejoy et al. (1985), a revision of the Todd pubic symphysis technique. The correlation coefficient for the nonperiodontal disease sample (N = 55; r = 0.86) is roughly equal to the summary age technique, a multivariate composite aging method. However, whereas the summary age technique reports higher correlations for males (r = 0.90), the cementum technique produces a higher correlation for females (r = 0.95). The inaccuracy and bias of the cementum annulation technique for the nonperiodontal sample compare favorably with the \ d u e s reported for the summary age technique and are superior to those reported for the revised pubis technique (Lovejoy et al., 1985; Table CEMENTUM ANNULATION AGING TECHNIQUE TABLE 4. Comparison of correlations of age indicators with. real age Age indicator Cementum annulation Full sample Nonperio. subsample Males Females Summary age' Males Females Revised pubis' Revised auricular surface' Proximal femur' Revised suture' Dental wear' Correlation 0.78 0.86 0.73 0.95 0.85 0.90 0.79 0.78 0.71 0.53 0.53 0.71 'From Lovejoy et al. (1985;Tables 5 and 6) 3). Overall (normalizedvalues), the summary age technique provides slightly better age estimates (5.2 vs. 6.0 years) with a bias to overestimate. The accuracy of the two techniques is similar for all decades except for the decade of 30-39 years, for which the summary age is superior. The reported inaccuracy of the cementum technique in the decade of 20-29 years is heavily influenced by a single case of doubling. If this individual is removed (age 24; adjusted count 45),the inaccuracy falls to 3.2 years with a bias of 0.7 years. The cementum technique tends to overestimate age until the fourth decade, although overall the technique is unbiased. Overall, the cementum annulation aging method seems to be the most accurate single aging criterion available for adults, being comparable in accuracy to the best multivariate aging techniques (e.g., summary age; Lovejoy et al., 1985). The addition of this criterion to those already available should increase the accuracy of age estimation in forensic research. The technique is also potentially applicable to archeological material. A moderately well preserved mandibular canine from a Late Woodland site in Illinois ( - 1000 years BP) was sectioned using both mineralized and demineralized techniques (Charles et al., 1986). Although the annulations were fainter than in modern material, counts were possible. The demineralized sections appeared slightly macerated, suggesting that further experimentation with the technique (e.g., gentler acids) will be necessary for archeological specimens. 327 The technique appears to work better for females than for males, with error estimates for females being half those for males. The reason for this disparity is not clear. Although the female subsample has a slightly younger mean age (29.8 years) than the males (34.8 years), examination of the respective scatterplots shows females to form a tighter cluster of points about the regression line for all ages. It is possible that the sample of males contains cases of unreported periodontal disease. Unfortunately, it was not possible to identify positively all individuals with periodontal disease solely by examination of the cementum sections. Although the presence of periodontal disease does not appear to alter either the slope or the intercept of the regression significantly, it does substantially increase the error of the estimate (Table 2). Periodontal disease also lowers the slope of the reduced major axis of counts on age, and the males have a shallower slope than do the females. The possible existence of unreported cases of periodontal disease in the male sample illustrates the difficulty in generating a sample to evaluate the technique. In that most individuals are reluctant to have healthy teeth extracted for scientific purposes, and curators of skeletal samples of known age are reluctant to allow destructive analysis, the sample must be generated by clinical extraction. Except for the relatively rare cases of trauma and some cases of orthodontia, the majority of extracted teeth are probably diseased (decay, periodontal disease, etc.). The effect of any or all of these diseases on cementum layer formation is not clear. However, comparison of the reduced major axes of samples with (N = 73) and without (N = 55) known cases of periodontal disease shows the presence of this pathology to decrease the slope; i.e., fewer layers are present than would be expected given an annual deposition. Whether this is a direct result of the disease or the result of altered tooth function cannot be determined here. The same effect can be applied to other dental diseases. For example, although decay may not effect the root proper, the presence of decay (and pain) may result in avoidance of tooth use, thus altering the mechanical environment of the tooth and, therefore, possibly cementogenesis and/or layer formation. In other words, this sample may represent a worse case, with 328 K. CONDON, D.K. CHARLES, J.M. CHEVERUD, AND J.E. BUIKSTRA better results obtainable in a working sample of healthy teeth. The technique is not applicable to all individuals. In the premolar sample (N = BO), three individuals (4%) showed a n incomplete or poorly defined incremental layer structure in their cementum. However, this phenomenon is true of all individuals in that most areas of dental cementum in all individuals show incomplete incremental structure. Although clarity of definition does differ between individuals, there is no relationship between clarity of definition and accuracy of age estimate. The source of this variability in cementum structure both within a tooth and between individuals is unknown. Grue and Jensen (1979) note that among other species there is a broad relationship between distinctiveness in layer differentiation and the range between seasonal extremes in climate attributes. For example, in a northsouth gradient, lines are more difficult to differentiate in animals from tropical regions. Alternatively, the variability in cementum structure may not, in fact, exist. The various stains used in evaluating cementum layers are not specific for either of the bands, as is demonstrated by the general staining of all the tissues in the section. The stains are used simply to dye the essentially transparent sections and render them optically opaque. Therefore, the failure of the various stains to define the layers cannot be interpreted as a n absence of incremental structure. Another problem associated with the technique is the phenomenon of “doubling,” in which approximately twice as many layers are present as would be predicted by the age of the specimen given a n annual deposition. The doubling phenomenon has been reported previously in humans (Stott et al., 19821, marmosets (Stott et al., 1980), black bears (Sauer et al., 19661, and other species (Klevezal’ and Kleinenberg, 1967). In this sample, there were seven cases (9%)of possible doubling based on visual examination of the scores (six males, one female; four maxillary second premolars, three maxillary first premolars). Only four of these were removed from the regressions (all males, all maxillary second premolars) on the basis of high leverage. Unfortunately, examination of the suspected sections revealed no a priori way of identifying doubling except in cases involving older individuals. For example, if layer counts exceed 100, then one might logically suspect a case of doubling and divide the counts in half for a n approximate count estimate. Although the majority of doubling cases in this sample were male, and all involved maxillary premolars, the sample size is too small to determine if this is a bias. Therefore, in estimating age from layer counts in premolars, there is roughly a one in ten chance that doubling is present. The accuracy of the technique appears to diminish with age and is therefore similar to other physiological aging methods. Johanson (19711, in a n evaluation of the Gustafson aging technique, noted that the general relationship of a n increase in cementum thickness with increase in age tailed off in the higher age ranges. Gasaway and coworkers (19781, Lowe (19671, and Ransom (1969) noted among other species, a n increasing discrepancy between layer count with age in moose, red deer, and white-tail deer, respectively. The origin of the age-related discrepancy is not known simply because the underlying physiological process leading to the deposition of cementum in alternating light and dark bands is essentially unknown, although various explanatory hypotheses have been proposed (summarized in Morris, 1978; Spinage, 1973). The annual character of the layer is most commonly assumed. Grue and Jensen (1979) suggest a complex interaction of nutritional, climatic, genetic, and functional factors to be responsible, but a multifactorial model has yet to be formulated. Reduced major axis analysis shows all the samples to depart from the theoretical norm of a n annual deposition. All the slopes were less than 1.0, suggesting that age (years) accrues faster than cementum layers. If the slopes are interpreted literally, then this suggests that layer counts do not represent annual depositions and that rates of formation potentially can vary between populations, thus requiring population-specific equations for age estimates. Further, this indicates that the counts provide a measure of physiological rather than chronological age. It is possible that the alternating light and dark bands represent simply the method of cementogenesis, analogous to growth reversal lines in bone. However, a n axis with a slope of less than 1.0 does not negate a n annual deposition if cementogenesis and layer differentiation are independent processes. (Note that a n axis with a slope greater than 1.0 does negate any CEMENTUM ANNULATION AGING TECHNIQUE 329 annual hypothesis.) For example, if cementogenesis is a response to functional stimulation (i.e., occlusal forces) and the deposition of the dark (or light) layer the result of some annual environmental stimulus, then the corresponding band would be produced only if the root was undergoing cementogenesis at the time of the stimulus. As noted by Johanson (1971), the increase in cementum width with increasing age tails off in the older age ranges, suggesting decreased cementogenesis with age. It is also documented that the cross-sectional areas (and therefore probably force) of the masticatory muscles decrease with age (Weijs and Hillen, 1985). A decrease in masticatory force might alter tooth loading and therefore alter cementogenesis. If this model is correct, one would predict a decrease in cementogenesis with age and therefore a decrease in band formation, resulting in a slope of less than 1.0. Note that a decrease in layer formation in the older age ranges would not only decrease the slope but correspondingly increase the y intercept, which is the relationship seen in this study. Although we present no evidence for or against this model, it illustrates the difficulty in demonstrating the annularity of cementum rings in the absence of knowledge of band etiology. However, the fact that the axes approach the theoretical norm and do not exceed a slope of 1.0 is consistent with an annual deposition. This model is also consistent with the lower slope and higher y intercept found in the sample containing cases of periodontal disease, the presence of which one would predict to inhibit cementogenesis. In that periodontal disease is most common in older age groups, the effect would be to decrease the slope and increase the y intercept of the major axis. mates being approximately half those for males. The presence of periodontal disease also increases the error of the estimate. The technique is not applicable in all cases; a minority of individuals (4%) showed incomplete or absent incremental cementum structure. Another 9% exhibited the phenomenon of “doubling,” for which there appears to be no a priori means of identification. Reduced major axis analysis of age on adjusted counts produced slopes not inconsistent with the hypothesis of annual deposition of cementum rings given that cementogenesis decreases with age. Thus the technique may be population-independent. However, since the reduced major axes departed from the theoretical norm of slope equal to 1.0 and y intercept of 0.0, the issue of annularity of cementum ring formation needs further evaluation. CONCLUSIONS Charles, DK, Condon, K, Cheverud, JM,and Buikstra, JE (1986) Cementum annulation and age determination in Homo supiens. I. Tooth variability and observer error. Am. J. Phys. Anthropol. 71:311-320. Demirjian, A and Levesque, G-Y (1980) Sexual differences in dental development and prediction of emergence. J. Dent. Res. 59:lllO-1126. Gasaway, WC, Harkness, DB, and Rausch, RA (1978) Accuracy of moose age determination from incisor cementum layers. J. Wildlife Management 425.58-563. Grue, H, and Jensen, B (1979) Review of the formation of incremental lines in tooth cementum of terrestrial mammals. Dan. Rev. Game Biol. 11:l-48. Johanson, G (1971) Age determination from human teeth a critical evaluation with special consideration of changes after fourteen years of age. Odontologisk Revy 22 [Suppl. 211 :1-126. Kay, RF, Rasmussen, DT, and Beard, KC (1984) Cemen- Adjusted counts of cementum rings using 7pm demineralized thin sections stained with hematoxylin can provide reasonable estimates of age a t deathlextraction. Error estimates range from 4.7 to 9.7 years depending on the sex of the individual and health status of the tooth. Overall, the inaccuracy of the technique is 6.0 years, with essentially no bias. This compares favorably with the summary age technique of Lovejoy et al. (1985) and is superior to any single macroscopic technique reported to date. The technique appears to provide better age estimates for females, with error esti- ACKNOWLEDGMENTS This research was funded by NSF grant BNS-8318587 to J.E.B. and J.M.C. Kerry Knox and Rita Quinn assisted in the study. We thank the directors of the Oral Surgery Clinics of Northwestern University and the University of Illinois at Chicago, Drs. Peccar0 and Blaustein, respectively, for their cooperation. The following dentists generously provided additional specimens: Drs. Bertoglio, Borden, Bork, Brown, Cottrell, Fine, Foertsch, Gargiulo, Grotz, Huddleston, A. Martin, E. Martin, Maxson, Mayr, McGahey, Messina, Nelson, Newton, Recktenwall, Scapino, Schleifer, Schultz, Soderstrom, Sokoloff, Urban, Van Hoozen, Varland, and Wallace. Dr. Kenneth Bennett of the University of Wisconsin-Madison also generously donated teeth for this analysis. LITERATURE CITED 330 K. CONDON, D.K. CHARLES, J.M. CHEVERUD, AND J.E. BUIKSTRA tum annulus counts provide a means for age determination in Macuca nulatta (Primates, Anthropoidea). Folia Primatol. 4285-95. Klevezal’, GA, and Kleinenberg, SE (1967) Age Determination of Mammals by Layered Structure in Teeth and Bone. (In Russian, translation by Fisheries Research Board of Canada, Arctic Biological Station, Ste. Anne de Bellevue, Quebec, Canada; Publ. No. 1024, 1969). Lovejoy, CO,Meindl, RS, Mensforth, RP, and Barton, TJ (1985) Multifactorial determination of skeletal age at death: A method and blind tests of its accuracy. Am. J. Phys. Anthropol. 68:l-14. Lowe, VPW (1967)Teeth as indicators of age with special reference to red deer (Cervus elaphus) of known age from Rhum. J. 2001.152:137-153. Morris, P (1978)The use of teeth for estimating the age of wild mammals. In PM Butler and KA Joysey (eds): Development, Function and Evolution of Teeth. New York: Academic, pp. 483-484. Naylor, JW, Miller, WG, Stokes, GN, and Stott, GG (1985) Cemental annulation enhancement: a technique for age determination in man. Am J. Phys. Anthropol. 68r197-201. Ransom, AB (1969)Determining age of white-tailed deer from layers in cementum of molars. J. Wildlife Management 30:197-199. Sauer, PR, Free, S, and Browne, S (1966) Age determination in black bears from canine tooth sections. N.Y. Fish Game J. 13:125-139. Seim, E, and Saether, B (1983)On rethinking allometry: Which regression model to use? J. Theor. Biol. 104:161168. Sokal, RR, and Rohlf, FJ (1981) Biometry: The Principles and Practice of Statistics in Biology, 2nd ed. San Francisco: W.H. Freeman. Spinage, CA (1973)A review of the age determination of mammals by means of teeth, with especial reference to Africa. E. Afr. Wildlife J. I1:165-187. Stott, GG, Sis, RF, and Levy, BM (1980) Cemental annulation as a n age criterion in the common marmoset (Cullithrixjucchus). J . Med. Primatol. 9:274-285. Stott, GG, Sis, RF, and Levy, BM (1982) Cemental annulation a s an age criterion in forensic dentistry. J. Dent. Res. 61:814-817. Thoma, KH, and Goldman, R (1960) Oral Pathology. 5th ed. St. Louis: C.V. Mosby. Wada, K, Ohtaishi, N, and Machiya, N (1978) Determination of age in the Japanese monkey from growth layers in the dental cementum. Primates 19:775-784. Weijs, WA, and Hillen, B (1985) Physiological cross-section of the human jaw muscles. Acta Anat. 121:31-35.