AGE, SEX, AND SEASON OF DEATH OF THE GRANDVILLE MASTODONT

Daniel C. Fisher


ABSTRACT

The Grandville mastodont, from Kent County, Michigan, has been radiocarbon dated at 11,320 ± 140 B.R Longitudinal sectioning of its right tusk provides access to samples of tusk dentin formed throughout life and exposes incremental features that allow direct correlation of annual dentin increments among samples. Enumeration of annual increments indicates that 33 years are recorded in the tusk, providing an estimate of age at death. Mammut americanum typically shows well developed sexual dimorphism in tusk and body size, and the Grandville mastodont was clearly a male. A short-term decline in the rate of increase of tusk length around age 15 may reflect expulsion from a matriarchal family unit at the onset of sexual maturity. Microscopic analysis of subannual incremental laminae near the tusk pulp cavity indicates that this animal died in mid-autumn, at the end of the fifteenth fortnight following the most recent winter-spring boundary. The age and life history data obtained from this individual will be useful in characterizing the population dynamics of late Pleistocene mastodonts of the Great Lakes region and in evaluating hypotheses for their extinction.

INTRODUCTION

In October of 1985, two young boys, Jeremie and Kenneth Butts, while playing on a large pile of marly clay on a residential construction site in Grandville, Michigan, discovered a rib of what came to be known as "Smitty," the Grandville mastodont. Upon report of the discovery to Grand Valley State University, the site (Kent County, Grandville Township, Section 18, T6NR12W; Shoshani 1989) was investigated by Professor Richard E. Flanders and an enthusiastic group of volunteers. The landowner and contractor responsible for construction, Mr. Mark E. Smith, gave permission for a recovery effort, and additional skeletal elements were subsequently found in the spoil pile that had produced the initial rib. All skeletal material occurred in marly clay that had been dug out for the foundation and basement of a house. The advanced stage of construction precluded excavation of immediately adjacent, undisturbed areas, but cores near the house encountered only sandy sediment, suggesting that the clay had been of limited extent. Recovery operations continued for about three months and consisted of hand excavation of the entire mass of sediment removed during construction. It is a credit to all involved that recovery of the skeleton was as complete as it was, but detailed information regarding the original context of this material is not available.

The lack of data on in situ bone distribution, state of articulation, and site stratigraphy precludes certain types of taphonomic and paleoenvironmental studies of the Grandville mastodont. Nevertheless, there is a great deal of paleobiological information recoverable from the fossil material itself. This paper represents an initial treatment of some of this information, focusing on absolute age at death, sex, season of death, and preliminary indications of tusk growth rates and the timing of molar eruption and wear. A subsequent study will consider other aspects of tooth formation and skeletal growth and maturation. Although only limited implications can be drawn from analysis of a single individual, comparative treatment of such data from a larger sample will play an important role in characterizing the population dynamics of mastodonts around the time of their extinction and in evaluating hypotheses for the cause of their extinction.

GEOLOGIC AGE

Following recovery of the mastodont, Professor Flanders submitted a sample of tusk dentin and a sample of bone (fragments of the sinus region of the cranial vault) for radiocarbon dating of the whole-coliagen component. The tusk was dated at 10,920 ± 190 B.P. (Beta-15265), and the bone at 11,320 ± 140 B.P. (Beta-15266).

Most bone preservation for this individual is very good, with cortical bone retaining much of its original physical character (Class 11 of Stafford et al. 1988). However, the samples submitted for dating appear unlikey to have been as pristine as the best material available from the site. Tusk dentin is generally more porous than dense, cortical bone and thus more subject to contamination by younger carbon; and judging from remaining skull material and mastodont skulls from other sites, the dated bone had a rather finely lamellar to cancerous structure. Since younger humates may have been present, the whole-collagen dates obtained here are probably minimum estimates of geologic age (Thomas W Stafford, Jr., personal communication 1990). Rather than average the two dates, it may thus be best to treat the bone date (ie., the older of the two) as the best available estimate of geologic age for the Grandville mastodont. In any event, it seems likely that this individual is well within the general age range of late glacial mastodont populations of the south-central Great Lakes region (Holman et al. 1986).

ANNUAL AND SEASONAL CYCLES IN PROBOSCIDEAN DENTIN

Recent work on patterns of incremental lamination in proboscidean tusk and molar dentin has demonstrated repetitive variation in structure and composition on at least three different scales. These patterns have been described in greater detail elsewhere (Fisher 1987, 1988; Koch et al. 1989) and so will be summarized only briefly here, focusing on tusk dentin. The largest scale of incremental lamination has been shown to be annual, allowing an entire tusk to be interpreted as a multi-year record of growth and life history. The permanent tusks of extant elephants begin to form near birth and continue growing throughout life (Sikes 1971). Continuous growth seems also to characterize the permanent tusks of mastodonts and mammoths, and although initiation near birth has not been confirmed (e.g., by a neonatal line in a tusk or another tooth whose annual record could be correlated with that of a tusk), it is a plausible working hypothesis. New dentin is added to the conical surface demarcating the pulp cavity, at the proximal end of the tusk, and is displaced distally through tusk eruption. Except for loss of material from the distal end of the tusk through fracture or gradual attrition, the permanent tusk may thus represent a complete record of an individual's life.

In some cases the most obvious expression of annual (first-order) incremental features in proboscidean tusks is a series of circumferential (or annular, ie., ring-like) swellings and constrictions along the outer surface. The apparent cause of these features is a seasonally shifting balance (one cycle per year) between the rate of extension of the proximal dentin margin and the rate of dentin apposition (measured normal to incremental surfaces). This produces undulations of the outer surface of the dentin which are, in places, observable on the outer surface of the tusk. These tend to be most pronounced proximally and least evident distally, although this is not simply an ontogenetic trend in the primary amplitude of undulations. It results in part from an interaction involving progressive coverage of the dentin-cementum interface by cementum, while still within the alveolus, and abrasion of both cementum and dentin, after eruption. Since cementum has just begun to be laid down at the proximal end of the tusk, undulations are usually clearest in this region. Where the tusk emerges from the alveolus, the cementum is thickest and tends to obscure the annual, annular swellings. Distal to this point, the cementum often thins due to abrasion, and in some cases, the resulting attritional surface truncates dentin on the high shoulders of annulae, leaving cementum in intervening areas, again highlighting the annular pattern. Still nearer the tusk tip, cementum may be lost entirely to abrasion, leaving no topographic trace of annulation.

Another feature that sometimes marks the annual incremental pattern in proboscidean dentin is color banding, with a single, major, dark-light cycle per year. This may be absent or developed only in shades of white or pale yellow in fresh material and in some of the best preserved fossil material, from clay. It is better developed, presumably through diagenetic enhancement of original structural differences, in fossil material preserved in peat or peaty clay. Color banding is usually most evident in transverse cross sections of dentin, though it may also be seen in longitudinal sections. On an intact tusk, color bands are most likely to be seen in the narrow zone where surface attrition has most recently exposed dentin from under its mantle of cementum. Nearer the tusk tip, where dentin has been exposed longer, surface discoloration usually obscures the bands. If cementum fractures away from the underlying dentin, color banding may be quite obvious along the freshly exposed dentin-cementum interface.

An annual cycle is also represented in the spacing of higher-order incremental features. This may only be evident after quantitative analysis of those features, making it difficult to use as a means for initially tracing tusk increments, but it is still important in correlating between samples of dentin and in evaluating subannual intervals in life history. There appear to be two additional orders of incremental features: second-order laminae, with either fortnightly or lunar monthly periodicity, depending on the age and sex of the individual; and third-order laminae, with daily periodicity. The biological basis of the former is poorly understood, though a relation to reproductive cycles has been suggested (Fisher 1987, 1988); the latter may reflect circadian variation in metabolic rate. Third-order laminae will not figure further in the present analysis, but the number and spacing of second-order laminae will be used to identify winter-spring boundaries and, relative to this temporal landmark, determine the season of death.

Individual seasons within the annual cycle of deposition of tusk dentin have been identified in three ways. Many extant mammals show annual color banding of dentin, and the dark bands almost always represent winter depostion (Klevezal and Kleinenberg 1969). In addition, this portion of the year is characterized by thinner second-order increments, implying slower growth, which in turn seems most likely to be associated with winter. Finally, oxygen isotopic analyses of the carbonate fraction of tusk hydroxyapatite show regular patterns with expected winter values in the dark bands and values more characteristic of spring-autumn in the light bands (Koch et al. 1989).

Another feature that bears a relation to incremental lamination is desiccation-related fracturing of tusk dentin. As tusk dentin dries out, it experiences substantial volume reduction, and because the outer portions tend to dry before the axial region, stresses are induced that lead to fracturing. The orientations of major stress gradients and the resulting fracture systems are controlled by the conical to cylindrical geometry of incremental features and of the tusk as a whole. Most fracturing is either parallel to incremental features, accentuating their cone-within-cone geometry, or approximately perpendicular to incremental features, in a radial pattern. Of these two fracture systems, the conical fractures are especially important. Although their genesis is entirely secondary to incremental features, in a cross section of tusk dentin they are often the most easily traced indicator of incremental features, making them useful in correlating from one area to another within a single tusk. Color bands can also be used in this way, but they are often too subtle to be followed unambiguously over great distances, and even when well developed, their boundaries are less discrete than a single fracture. Conical fractures sometimes even have seasonal significance in that they often occur at the approximate position of the winter-spring boundary. This is probably because the steepest spatial gradients in physical properties are located in this position, where the slow growth of winter gives way to the rapid growth of spring. This association should be used cautiously, however, as fracture position can be influenced by other factors as well.

MATERIAL AND METHODS

The Grandville mastodont is reposited in the Archaeology Laboratory of the Department of Anthropology and Sociology, Grand Valley State University, Allendale, Michigan. A few skeletal elements are displayed in the Grandville Town Hall. Dental material consists of both tusks leach nearly complete), the left M2/,3/ and right M2/,3/. Identification of the specimen as Mammut americanum was first based on the molars, but it is confirmed by cranial morphology and by numerous diagnostic features on postcranial elements (cf. Olsen 1972). I have completed only a cursory analysis of the nondental material, but there are no evident duplications of elements or incompatibilities in size or state of epiphysis fusion to suggest that more than one individual is present.

Most of the determinations reported here are based on the right tusk. This was found in multiple segments, bounded by a combination of transverse and longitudinal fractures. Most of the transverse fractures show staining and localized diagenetic modification along their surfaces, indicating that they formed prior to excavation. Others, including damage in the vicinity of the tusk tip, are fresh and apparently formed during mechanical excavation. Because cementum undergoes less volume change with desiccation than does dentin, most of the cementum spalled off soon after excavation, exposing the dentin-cementum interface. Preservation of tusk dentin varies from excellent to very good (Class 11-111, Stafford et al. 1988). At its worst, the dentin is white, chalky, and brittle, though even much of this material is intact and retains traces of incremental features. This condition is distributed somewhat sporadically, but mainly along the tusk axis, in the distal two thirds of its length. At its best, the dentin is white to pale yellow and retains much of the hardness and resiliency of fresh ivory.

In most respects, dentinal lamination in tusks is best studied in transverse cross section, but because of the conical geometry and imbricate nature of tusk dentin increments, no single transverse cross section (other than in a tusk of a very young animal) exposes the entire interval recorded by a tusk. The only section that does display the full sequence of increments is a longitudinal section passing along the tusk axis (figure 1). Although the helical shape of proboscidean tusks makes this a complicated cut, it provides the only opportunity to trace increments directly from one part of the tusk to another. The sampling plan applied to the Grandville tusk was designed to combine the strengths of transverse and longitudinal sections by obtaining a series of samples that: (1) collectively represented the entire sequence of dentin increments, (2) could be studied in transverse section, and 13) showed annual increments that could be correlated from one sample to the next by tracing directly along a longitudinal section. The procedure involved assembling the tusk, sectioning it longitudinally, and removing the necessary samples (though logistic considerations sometimes dictated revisions of this order of events). The two longitudinal halves were left separate as a convenient teaching tool displaying tusk structure and as documentation of the temporal relationships of thin sections made from the samples.

Since each half of the reassembled tusk would display strong, helical curvature, it was essential to use adhesives that would resist the large bending moments that could be anticipated. Most of the reassembly was therefore done using epoxy. In a few instances where adhesion was less of a problem than space-filling (e.g., in radial, longitudinal fractures, or where small fragments of tusk were missing), polyester resin was used.

Large-scale sectioning of the tusk was accomplished with a specialty adapted, Delta 14-inch (35.6 cm) metal-cutting bandsaw with a3/4 inch (19.1 mm), 10 tpi, bimetal blade, operated at a blade speed of 220 fpm. Parts of the tusk were sawn freehand, with support provided by the cutting table of the bandsaw, but most of the cutting required additional support for adequate control. This was provided by a large, rolling scaffold that arched over the saw and moved along with the tusk during cutting. Explanation of the details of this arrangement and the sequence of steps in tusk restoration and sampling will be presented elsewhere, but the general pattern of cutting can be summarized briefly. The path of longitudinal sectioning followed the fine, longitudinal ridges that develop through accretionary propagation of crenulations of the proximal margin of dentin deposition. This meant that the cut displayed a gentle twist as it followed the helix of the tusk. The cut was also displaced approximately 1 cm from the tusk axis so that the axis itself, and the last-formed dentin at a given position along the tusk, would be retained consistently in one of the two tusk halves. In this instance, the cut was directed medial to the tusk axis, and all samples were removed from the lateral half. After sectioning, the cut surface of the medial half was sanded and polished to reveal more clearly the internal structure of the tusk.

Most of the removal of samples for microscopic examination was done with the same bandsaw used for large-scale cutting. Where feasible, sampling sites were chosen to avoid areas of chalky preservation. Although thin sections can be made from such material, there is greater potential for fracturing and loss of material during section preparation. In cases where only a relatively thin zone of well preserved dentin remained near the outer surface of the tusk, the procedure for sample removal was modified in order to minimize reduction of the tusk cross section. In such instances, the targeted block of dentin was isolated by blind channels ground transversely and longitudinally using a carbide bur on a flexible-shaft grinder. Methods of thin section preparation and analysis are outlined by Fisher (1988).

AGE AT DEATH

Cementum spalling exposed the dentin-cementum interface over most of the proximal half of the tusk. In about the proximal third, there are conspicuous annulations of the cementum-dentin interface and pronounced color bands cropping out on this surface. If annual increments had been manifested over the whole tusk in this fashion, it would have been relatively straightforward to estimate age at death. However, surficial annulations and color bands on earlier-formed portions of the tusk were so subtle that they could not be distinguished and counted with certainty. In the distal third of the tusk, the attritional dentin surface bore scarcely any trace of these features.

After longitudinal sectioning, annual increments could in principle be traced within the interior of the tusk. In most of the distal two thirds of this specimen, however, enumeration of annual increments, even given the longitudinal cross section, was difficult; on this part of the cross section, the contrast between summer and winter color bands was simply not great enough to trace increments far enough to provide unequivocal counts. Indications of conical geometry were evident throughout the tusk, but most of these were sets of approximately parallel fracture surfaces, developed en eschelon, flanking the tusk axis. Despite some tendency (noted above) for such fractures to occur near winter-spring boundaries, fractures are not consistently enough associated with the annual cycle to use them as a direct indicator of age.

The strategy adopted to deal with this difficulty was to start sampling proximally (G in Figure 1), where annual increments were most evident, and then proceed distally (toward A in Figure 1). On the transverse cuts made for sample removal, annual color banding was usually clear. Fractures paralleling incremental features were traced from the transverse surface, "around the corner" onto the longitudinal surface, and then toward a more distal part of the tusk. In this direction, more recently formed increments dropped out (distal to the location of the apex of the pulp cavity at the time of their formation), and earlier-formed increments were encountered. A site for the next sample was chosen to insure, where possible, an overlap of several years between samples. In other words, samples were located so that the outermost (earliest-formed) well preserved years of one sample were duplicated as the innermost (latest-formed) well preserved years of the next distal sample.

Although simple in concept, this strategy was somewhat complicated in practice. In general, well preserved dentin did not extend all the way to the tusk axis, so samples had to be more closely spaced than would have been necessary in a better preserved tusk. There were sometimes multiple routes by which the sample series might have been extended, with the choice among them influenced by multiple factors (e.g., quality of preservation, clarity of dentinostratigraphic relationships, and minimization of the amount of material being removed). The final choice involved some shifting between ventral and dorsal positions along the longitudinal surface; this was possible because fractures following incremental features frequently extend onto the opposite side of the tusk through a v-shaped apex at the tusk axis. Although sample sites were chosen in proximal-distal sequence, they are identified (A-GI in a distal-proxirnal sequence as a reminder of the order of dentin apposition. Only about a half year was duplicated by samples A and B, but generous overlap was obtained between all other samples.

Dimensions of samples were influenced by several factors. Preliminary checks revealed that the full extent of fractures and areas of chalky preservation was not always obvious from their expression on the primary longitudinal section. In order to enhance the likelihood of having adequately preserved material for microscopic analysis, more material was removed in areas of problematic preservation. Where samples had a radial dimension greater than would fit on a single petrographic microscope slide, thickness in the transverse direction was augmented (up to 2.5 cm) to allow subdivision oblique to incremental lamination, to provide overlap between slides. In addition, samples were cut large enough in the longitudinal direction (5 cm or more) to allow for isotopic analysis at a later date.

The approximate location and dentinostratigraphic relationships of samples are illustrated in Figures 1 and 2. Figure 1 is based on a photograph of the reassembled and longitudinally sectioned tusk and is thus a realistic depiction of tusk shape. It is idealized, however, in the clarity with which the incremental pattern is portrayed. On the actual specimen, viewed at this scale, only faint color bands and discontinuous fractures would be seen, but to demonstrate the pattern that emerges from integration of information throughout the tusk, the figure shows the approximate reconstructed course, in longitudinal section, of every third winter-spring boundary (showing each such boundary would make the drawing too complex). While this gives a clear sense of the organization of the tusk, its scale does not allow explicit portrayal of the distribution of well preserved portions of annual increments throughout the series of samples. This information is presented in Figure 2, a schematic diagram in which tusk morphology is greatly foreshortened and incremental features are regularized in order to show more clearly the temporal relations between samples. The straightened tusk axis, the constant land large) apical angle of dentin increments, the constant thickness of dentin increments, and inexact longitudinal spacing of samples are all departures from reality that simplify graphical presentation. In addition, for consistency with Figure 1, only every third winter-spring boundary is shown. Annual increments are numbered from first to last. Columns of numbers indicate individual samples, with numbers shown only where a year is represented by well-preserved material within a given sample. One feature that may appear inadvertent, but that is in fact realistic, is the emergence of the tusk axis (shown as a solid line) from the ventral surface of the tip. This is a common pattern resulting from use wear on the tusk tip during life; as a result of this asymmetrical removal of some of the earliest-formed dentin, the first year represented in the tusk can be sampled only dorsally, at some distance from the tip.

The correlation and enumeration of annual increments presented in Figure 2 has not yet been confirmed by microscopic analysis, but previous experience suggests that it is probably not off by more than a year or so. The last half of the lifespan seems unproblematic; if there is an error, it is probably in the earlier portion of the tusk. It is especially hard to recognize winter-spring boundaries, without thin sections, in the vicinity of the first recorded year. For purposes of enumeration, these boundaries have been placed assuming comparable annual increment thickness to that seen in years 2-4.

Given this approximation, the first-formed dentin present in the tusk may have formed during the spring, such that the first year is essentially complete, from one winter-spring boundary to the next. The last year recorded in the tusk is clearly not complete (see discussion of season of death), but it may still be tallied as one year for purposes of age estimation. The pulp cavity is in excellent condition, leaving, as far as can be determined, no possibility of unrepresented time at the end of life. In all, it appears that 33 years are recorded in the right tusk of the Grandville mastodont. Although the tip of the left tusk is in better physical condition than the tip of the, right the tips have similar profiles and the whole tusks, similar lengths. It therefore seems unlikely that the left tusk would provide any significant early extension of the record of dentin deposition. It is likely that at least some of the earliest-formed dentin has been abraded away, but the amount is unknown. Considered by itself, this would make the measured age, 33 years old, a minimum age. However, it is also possible, though probably less likely, that dentin deposition began in the permanent tusk somewhat prior to birth. I suspect that the number of years represented in the tusk is a slight (1-2 year) underestimate of the actual age at death, but at present, the most objective figure for comparative purposes is the measured age.

SEX AND LIFE HISTORY

As has been noted for some time (Osborn 1936), Mammut americanum shows marked sexual dimorphism, most clearly in tusk dimensions and body size. This dimorphism has never been adequately documented, but it parallels the dimorphism known for African elephants (Elder 1970) and has been applied in the reporting of sex ratios characteristic of different taphonomic categories of late Pleistocene mastodont occurrence (Fisher 1987). The attributes that may be used at this time to sex the Grandville mastodont include the basal circumference of the tusk (52.5 cm, measured normal to the tusk axis; some cementum missing), tusk length 1257 cm, measured along the outside curvature; some reconstruction of tip re- quired), and these dimensions considered relative to age (33) or state of molar eruption and wear (right M2/ worn past point of dentin confluence on all lophs; left M2/ deeply worn and roots mostly resorbed; right and left M3/ fully erupted and dentin exposed on first three lophs; approxi- mately age group XXIII using Laws' system for relative aging of African elephants; Laws 1966). Tusks as large as this are not seen on any individual recognized as female, and certainly not on females of comparable age or stage of molar eruption and wear. For comparison, the Powers mastodont (Western Michigan University; Garland and Cogswell 1985) shows 30 years in its left tusk and an only slightly less advanced molar stage (Laws' age group XXII); its tusk basal circumference is 22.0 cm (cementum intact), and its tusk length is 140 cm. To demonstrate that the difference in measured age between these two individuals cannot account for much of the disparity in tusk size, the dimensions for the Grandville mastodont can be "backed up," or taken at the thirtieth winter-spring boundary, yielding a circumference of 50.9 cm (some cementum missing) and a length of 242 cm. Although more thorough documentation of sexual dimorphism would be useful, and is intended for a later analysis, this is a striking difference in tusk development. By analogy with extant elephants, it seems clear that the Grandville mastodont is a male.

Although terminal measurements of tusk length and basal circumference suffice for some purposes, comparison with the Powers mastodont illustrates the usefulness of a more complete representation of tusk shape. Tusks, like invertebrate skeletons that show accretionary growth, record ontogenetic information "longitudinally" (ie., as a succession of stages of a single individual) and thus provide an alternative to "cross sectional" studies of multiple individuals assessed terminally, at different ages. In addition to factoring out effects of genetic variation, this provides more information from a smaller sample. Figure 3 shows an ontogenetic trajectory of tusk length and basal circumference measurements for the Grandville mastodont. For this plot, data are given at arbitrary 110 cm) intervals, initially measured from the proximal end of the tusk, because of excavation damage to the tip, but reported relative to the reconstructed distal end of the tusk, for consistency with ontogenetic order. Proximallywinter-spring boundaries show up clearly on the dentin-cementum surface, their position is indicated by tick marks on a secondary abscissa (age). In earlier ontogeny, where the position of these boundaries is less evident without microscopic analysis, only an estimate of boundary position at three-year intervals is given (without tick marks). Another factor urging caution in the reporting of winter-spring boundary positions in early ontogeny is that tusk abrasion shifts outcrop position distally, although the magnitude of this effect probably diminishes significantly after the first few years.

Tusk data presented as in Figure 3 may be used directly in studies of sexual dimorphism but are also a source of information on growth rates that may have other implications for life history and reproductive biology. For instance, Figure 4 shows annual tusk length increments throughout life. Where successive winter-spring boundaries are clear, increments are plotted as solid dots; where they are not clear, they are estimated and averaged over three-year intervals and plotted as open circles at the midpoint of the relevant interval. Broad features of the resulting pattern are nonetheless clear throughout and appear not to be dependent on the necessary approximations. The graph is essentially a coarse record of rates of increase in tusk length throughout ontogeny. As with many growth rates, we see high values early in ontogeny, declining into adulthood. During adulthood are minor fluctuations that are probably best interpreted as representing normal variations in health status and/or food availability due to climate. There may be some question as to how closely rates of increase in tusk length reflect nutritional status, but a strong relationship is likely. The conical geometry of tusks suggests that length increments should be highly correlated with appositional thickness, and this is supported in the years that have been analyzed by thin section. Appositional thickness of dentin in turn shows seasonal variation (discussed below) that follows a pattern suggestive of nutritional control.

Against this background, an interesting feature of Figure 4 is the relatively short-term but pronounced drop in growth rate around age 15. A possible interpretation of this is that it reflects a decline in nutritional status associated with expulsion from a matriarchal family group, following puberty. The implied social structure and age of attainment of sexual maturity are similar to what is seen in African elephants (Laws 1966; Sikes 1971); and although directly comparable tusk data have not been gathered for extant elephants, an inflection in Hanks' (1972:265) graph of tusk circumference versus age can be interpreted in these terms. Studies of additional individuals (both male and female) are clearly necessary, and tests involving extant elephants would be helpful, but there is a possibility that information on age of sexual maturity for male mastodonts is much more readily available than might have been anticipated.

SEASON OF DEATH

For determination of the season of death, a small block of dentin was cut from sample G, extending from the pulp cavity to a desiccation fracture that paralleled incremental lamination. This block had an appositional thickness of about 2.3 cm and served as the source of a transverse thin section, perpendicular to incremental lamination. Annual color banding, still subtle on the unpolished transverse surface, became progressively more evident with polishing. Examination of the polished surface, even without magnification and in reflected light, left little doubt that the desiccation fracture approximated a winter-spring boundary, that four more winters followed, and that death occurred in the autumn, before development of the next dark, winter band. Completion and analysis of the thin section confirmed this initial impression and provided a quantitative record of rates of dentin deposition that could be used to characterize the last several years of life and evaluate the timing of death relative to the last winter-spring boundary.

Figure 5 shows a graph of trends in second-order increment thickness for about the last three years examined by thin section. Winter-spring boundaries are indicated by arrows and are placed, by convention (Fisher 1987), at the end of the thinnest second-order increment, within a sequence of thin increments, associated with a dark, winter band. Identifications of winter-spring boundaries can also be evaluated for consistency with the partially independent judgments made for adjoining years and with the more fully independent judgments made on different individuals. The three boundaries shown in Figure 5 occur within regions with comparable increment profiles, are separated by an appropriate number of second-order (fortnightly) increments, and show first and second-order increment thicknesses that are comparable to those measured on other individuals (Fisher 1987, 1988; Koch et al. 1989). The first full year in Figure 5 is a little unusual in failing to show the high rates of growth characteristic of most summers, but the following years are more normal. Relative to the last winter-spring boundary, the Grandville mastodont died at the end of the fifteenth fortnight, or about mid-autumn.

External tusk topography also provides some indication of season of death. The proximal margin of the tusk is broken, but most of this damage follows a path just proximal to a "shoulder" representing one of the topographic annuli of the dentin-cementum interface. This breakage is thus at the approximate position of a winter-spring boundary. Along the medial portion of the proximal dentin margin, dentin extended further than elsewhere, at least half way into the following summer. The thickness of dentin at this point, projected along its angle of incidence with the dentin-cementum interface, is compatible with an autumn death.

DISCUSSION

The information obtained through these analyses goes beyond answering questions about the identity of a single individual. Studies involving age profiles, the distribution of age at death in a sample of individuals, are playing an increasing role in paleobiological and archaeological investigations (Klein and Cruz-Uribe 1984). Until recently, the best available approach to studying age profiles of mastodonts (e.g., Saunders 1977) has been rather qualitative: individuals were placed in relative age categories defined on the basis of molar eruption and wear patterns of African elephants (Laws 1966). In general, age assignments were possible only by analogy with African elephants, in units of "African-equivalent years" (AEY, Saunders 1980). However, there are several problems with such an approach. One is that the dentitions of mastodonts and elephants, though similar in some respects, differ in details of the molar eruption and wear pattern. As a result, molar stages are only broadly comparable between these taxa. Second, even the African elephant age categories have been criticized for irregularities in the amount of time represented by each, manifested as consistent patterns of uneven frequency distribution of individuals among age categories (Fatti et al. 1980). This poses problems for statistical analyses of elephant age profiles that may well carry over into studies of mastodonts. Third, relative age categories allow only limited studies of ecology and population biology. Absolute age at death, combined with molar eruption and wear stage, for the Grandville mastodont provides one control point for the development and calibration of absolute age scales for American mastodonts. Data on more individuals will be required to meet this goal, but in the beginning, the contribution of each case is critical.

At present, there is little published data to compare with that reported here. Osborn (1936:181) recognized on the Warren mastodont (AMNH 9951) the features now referred to as first-order annuli and suspected that they might represent annual increments. On the best-preserved tusk of this individual, he could discern only the proximal-most 18 annuli. However, he estimated that there might be 28 such increments in the en- tire tusk. He does not specify the assumptions (about magnitude of tusk increments in early ontogeny) on which this estimate was based, but his figure may be provisionally accepted. The molar stage of this individual is elsewhere described sufficiently (Warren 1852:21) to assign a Laws age category of XXII (implied age: 39 AEY). An additional case is the Powers mastodont, mentioned above estimated age: 30 yr; molar stage XXII; implied age: 39 AEY). Although detailed evaluation of these data would be premature, it is intriguing that all three determinations are in close accord and that they all suggest an absolute age significantly less than the attributed age in "African-equivalent years." It seems unlikely that loss of material from the tusk tip could account for all of the 9-11 year difference. This suggests that late Pleistocene American mastodonts took less time than do African elephants to reach comparable molar stages. A more relevant comparison would be with earlier populations of American mastodonts. Although this information is not yet in hand, it would offer an interesting opportunity to evaluate whether mastodonts, near the time of their extinction, showed growth trajectories indicative of nutritional stress. This in turn would provide a test of some climatic hypotheses for their extinction (Graham and Lundelius 1984; King and Saunders 1984).

Identification of the sex of the Grandville mastodont will probably have few repercussions on its own, but the life history information that may emerge from the combination of sex and growth data could provide important insights into population dynamics. We have at present little information on mastodont (or mammoth) life history and reproductive biology. Most of what we think we know about these animals comes from analogies with extant elephants. Our understanding of the paleobiology and eventual extinction of late Pleistocene megafauna would be improved if methods can be developed for direct assessment of relevant attributes from fossil material. The suggestion that a drop in tusk growth rate in male mastodonts signals the onset of sexual maturity needs a great deal of additional work, but may be a step in a useful direction.

The season of death determination reported here is interesting mainly in comparison to the bimodal pattern of seasonal mortality (mid to late autumn and late winter to early spring) identified for other late Pleistocene mastodonts of the Great Lakes region (Fisher 1987; Fisher and Koch 1983). It fits at the margin of, but still within, the autumn mode in mortality. Where taphonomic analysis of skeletal condition has been possible, winter-spring deaths appear to represent natural deaths, with no human association, while autumn deaths consistently show signs of carcass processing by humans (Fisher 1987). This has been interpreted as supporting the hypothesis that the autumn deaths were caused by human hunting, and it raises the question of whether the Grandville mastodont might have been hunted and butchered. Unfortunately, little can be said at present about the cause of death and early postmortem history of this individual. No unambiguous indications of human association were recognized during excavation or cleaning of the bones, but taphonomic analysis is not yet complete (and will in any event be limited by the lack of data on in situ bone distribution). The general geologic context of the bones, in pond sediments, is comparable to cases that have been interpreted as representing meat caching by Paleoindians (Fisher 1989), but there is too little information to evaluate this possibility further at this time. I am confident that additional information will be gleaned from subsequent analyses, but at the moment we know more about the history of this animal's life than about the circumstances surrounding its death.

The shortcomings of available information should not, however, overshadow the value of what is in hand. Time is an ecologically critial parameter that is rarely available in studies in studies of fossil material, but that is recorded by the cyclical character of proboscidean dentin deposition. This constitutes an unusual source of biological information that awaits interesting ecological and evolutionary applications.