TAPHONOMY OF THE LARGE VERTEBRATE
FAUNA FROM THE THOMAS FARM
LOCALITY (MIOCENE, HEMINGFORDIAN),
GILCHRIST COUNTY, FLORIDA
Ann E. Pratt
Biological Sciences, Volume 35, Number 2, pp. 35-130 1990
UNIVERSITY OF FLORIDA
Numbers of the BULLETIN OF THE FLORIDA MUSEUM OF NATURAL HISTORY
BIOLOGICAL SCIENCES, are published at irregular intervals. Volumes contain about
pages and are not necessarily completed in any one calendar year.
S. DAVID WEBB, Editor
RHODA J. BRYANT, Managing Editor
Communications concerning purchase or exchange of the publications and all manuscripts shou
be addressed to: Managing Editor, Bulletin; Florida Museum of Natural History; University
Florida; Gainesville FL 32611-2035; U.S.A.
This public document was promulgated at an annual cost of $6136.00 OR
$6.136 per copy. It makes available to libraries, scholars, and all interested
persons the results of researches in the natural sciences, emphasizing the
Publication date: 6/18/90
TAPHONOMY OF THE LARGE VERTEBRATE FAUNA
FROM THE THOMAS FARM LOCALITY
GILCHRIST COUNTY, FLORIDA
Ann E. Pratt*
The results of a taphonomic investigation of the Thomas Farm locality are presented.
Scdimentological evidence indicates that deposition took place in a -large (about 35 m in
diameter), high-walled (30 m) sinkhole. Fossiliferous sediments most likely represent the
remains of a large debris cone that formed in the bottom of the sink. Laminar water-lain clays
indicate that the sinkhole was water-filled at least periodically throughout the course of
deposition. Two massive layers of limestone rubble represent two separate sequences of roof and
wall collapse of the sinkhole. A major change in the nature of the deposit occurred following
these events, as indicated by the presence of upper massive layers of calcareous sands rich in
microvertebrates. Taphonomic evidence suggests that bones of megafaunal vertebrates
accumulated attritionally. In the lower clay and sandy clay units of the site, the megavertebrate
bones show an orientation pattern consistent with that indicative of moving water. Movement of
water through the sink proceeded from northeast to southwest through underground joints in the
limestone forming the sinkhole. Current velocity was of sufficient strength to winnow out less
dense elements, primarily those belonging to Voorhies Dispersal Group I. Lack of severely
weathered, water-worn, or carnivore-chewed bone suggests that bones did not lie exposed on the
surface for long periods before being deposited on the debris cone, nor was there a strong fluvial
or carnivore-mediated influence on the bone assemblage. The faunal composition of the
mcgavcrtcbrate taxa in all but the uppermost layers is similar throughout the site, suggesting that
the fauna available for fossilization did not change greatly throughout the course of deposition.
For the major part of the depositional history of Thomas Farm, the taphonomic factors
responsible for this sampling did not change appreciably. The upper layers contain very little
mcgafauna and are not considered extensively in this study. Faunal evidence indicates that the
sink was located in a wooded area, and the climate at the time of deposition was tropical to
* Th author is an Assistant Professor of Biology at Georgia Southern University, Statesboro GA 30460. A large part of this
paper was submitted in partial fulfillment of her doctoral degree to the Department of Zoology, University of Florida.
Pratt, A. E. 1990. Taphonomy of the Large Vertebrate Fauna from the Thomas Farm Locality
(Miocene, Hemingfordian), Gilchrist County, Florida. Bull. Florida Mus. Nat. Hist., Biol. Sci.
36 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
Se presentan los resultados de una investigaci6n tafon6mica en la localidad de Thomas
Farm. La evidencia sedimentol6gica indica que la deposici6n ocurri6 en un cenote grande
(alrcdedor de 35 m de diametro), y profundo (30 m). Los sedimentos fosiliferos estAn formados
en su mayoria por los restos de un gran cono de recolecci6n, formado en el fondo del pozo. Las
arcillas laminares depositadas por el agua indican que el cenote contenia agua al menos
peri6dicamente a lo largo del curso de deposici6n. Dos grandes capas de fragments de rocas
calcarcas represcntan secuencias de colapso del techo y los lados del cenote. Ocurri6 un gran
cambio en la naturaleza de los dep6sitos despuds de estos events, lo cual es indicado por la
prescncia de grades capas superiores de arenas calcireas ricas en microvertebrados. La
evidencia tafon6mica sugiere que los huesos de megavertebrados se acumularon a lo largo del
tiempo debido a causes naturales. En las unidades de arcilla profunda y arcilla arenosa del sitio,
los huesos de megavertebrados muestran un patron de orientaci6n consistent con el causado por
agua en movimiento. El movimiento del agua en el cenote era de noreste a suroeste a travds de
grietas subterrAneas en la roca caliza que formaba el cenote. La velocidad de la corriente tenia
suificient fuerza para separar los elements menos densos, principalmente aquellos
pertenecientes al Grupo I de Dispersi6n de Voorhies. La ausencia de huesos muy
intemperizados, gastados por el agua, o mascados por carnivores, sugiere que los huesos no
permanecieron expuestos en la superficie por periods largos antes de ser depositados en el cono
de rccolecci6n, que no habia una corriente fluvial fuerte, y que los carnivores tampoco
modificaron el arreglo de los huesos. La composici6n faunistica de los taxa de megavertebrados
cs similar en todo el sitio, except en las capas m6s superficiales, lo que sugiere que la fauna
disponible para la fosilizaci6n no cambi6 much durante el curso de deposici6n. Los factors
tafon6micos responsables de esta muestra no cambiaron apreciablemente durante la mayor part
de la historic de deposici6n en Thomas Farm. Las capas superiores contienen muy poca
megafauna y no son consideradas extensamente en este studio. La evidencia faunistica indica
que el pozo estaba situado en una zona boscosa, y el clima durante la deposici6n era tropical o
TABLE OF CONTENTS
Intro d uctio n .................................................................................................................. ..................... 37
A know ledgem ents............................................................. ................................... ....................... 38
Materials and Methods................. ...................... 39
A abbreviations and A acronym s.................................................................... .............. ......... 39
F field E xcavation................................................................................................... ..................... 39
Laboratory Procedures ............................................................................... ............................ 41
R results and D iscussion............................................................................... ..................................... 46
Previous Excavations and Interpretations...................................................... ....................... 46
Environm ent of D eposition....... ............................................................................................... 52
M egafaunal T aphonom y..................................................................................... ..................... 70
Sum m ary and C onclusions.................................................................... .......................................... 124
D epositional E nvironm ent ...................................................................... ......................... 124
Source of the Bone A ccum ulation .......................................................... ........................ 125
T he T terrestrial E nvironm ent................................................................... .............. .......... 125
literature C ited ................................................................................................ ......................... 126
A ppend ix.......................................... ................................................................................. .............. 130
I'PTI': TAPIIONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 37
The Thomas Farm locality is among the best-known Miocene land
vertebrate fossil sites in North America. Specimens from this deposit are
housed in several major museums, and the vertebrate fauna has been described
in numerous publications (see Webb 1981 for a complete list). Thomas Farm
is distinctive among Florida sites for a number of reasons. The vertebrate
remains are incredibly numerous and consist of both large (megafaunal) and
small (microfaunal) vertebrates. Localities in which all size ranges of
vertebrates are found in great abundance are very rare. Although the site has
been known and worked for over half a century, fossiliferous sediments are still
The Thomas Farm locality was discovered in the spring of 1931 by state
geologist Clarence Simpson of the Florida Geological Survey, who found
fragments of fossil bone on the spoil pile from a well on the abandoned farm of
Racford Thomas. These bones, and others found in a test pit dug to the west
of the old well, were sent to George Gaylord Simpson, then at the American
Museum of Natural History. G. Simpson identified the bones and in 1932
published the first description of the site and the fossils.
No further interest was taken in Thomas Farm until 1938, when Thomas
Barbour of the Museum of Comparative Zoology, Harvard University, chanced
to see the specimens in storage at the Florida Geological Survey and realized
the great importance of the locality (Barbour 1944). He arranged for the
purchase of the abandoned 40-acre farm, and formal excavation began. In
1942 the land was deeded to Archie Carr of the University of Florida
Department of Biology, who in turn donated it to the Florida State Board of
Education. In the years from 1940-1958, the site was cooperatively excavated
by crews from the Florida Geological Survey, the Museum of Comparative
Zoology, and the University of Florida (White 1942; Olsen 1962; Puri and
Vernon 1964). It was during that time that the large collections were
accumulated, both by Harvard and the Florida Geological Survey.
Excavation was conducted sporadically throughout the mid 1960s into the
early 1970s under the direction of Walter Auffenberg and Thomas Patton of
the Florida State Museum (now called the Florida Museum of Natural
History). In 1973, field work at the site ceased and was not resumed until this
study began in the fall of 1981.
Although Thomas Farm has been known for over 50 years, prior to this
study only the excavations during the late 1940s and early 1950s were extensive.
The large number of vertebrates removed during those years attests to the
great richness of the site. In the past, the major field efforts were geared
toward recovery of fauna rather than toward analysis of this isolated
assemblage of sediment and fossils. The geology of the Thomas Farm locality
38 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
is very complex and has never been completely understood. The mode of
formation of the site and deposition of the bones has been a subject of
disagreement since the locality's discovery. The still-impressive productivity of
the site, the diverse fauna, and the fact that the locality is protected by the state
make Thomas Farm an ideal subject for taphonomic and paleoecological
analysis. This study represents the first attempt at taphonomic analysis of an
early Miocene Florida terrestrial vertebrate locality. Although other early
Miocene faunas are known from Florida (Simpson 1932; Frailey 1978, 1979),
most of the sites have long since been destroyed by mining or land
The main goals of this study are to determine, by means of controlled
excavation and analysis, the important geological features of the Thomas Farm
locality and the mode of deposition of the megafaunal vertebrate remains. In a
study of this type, a distinction is usually made between megafaunal and
microfaunal analysis. This distinction is primarily methodological; most
megafaunal taxa are those whose living weight was 1 kg or greater and whose
bones are collected by excavation in the field, whereas most microvertebrates
(weight of living animal less than 1 kg) are found by screenwashing and
processing bulk matrix. Because the collecting procedures for the two size
ranges of vertebrate fauna are different, the types of taphonomic data that are
obtained also provide different types of information. In this study, only the
results of the megafaunal taphonomy are considered. Taphonomic analysis of
the rich microvertebrate fauna is discussed in detail elsewhere (Pratt 1986,
S. David Webb and Gary Morgan kindly allowed use of the collections and facilities of the
Florida Museum of Natural History. Members of the Thomas Farm Committee granted me
permission to work at the site. Richard Hulbert, Gary Morgan, Steven Emslie, and Arthur Poyer
aided in field excavations and also made many helpful suggestions. Russell McCarty provided
valuable expertise in surveying and establishing the grid and prepared many of the specimens.
Earlier versions of this manuscript were reviewed by Gary Morgan, Richard Hulbert, Ronald
Wolff, S. David Webb, Elizabeth Wing, and Richard Kiltie. Graig Shaak provided valuable
insights concerning deposition in sinkholes. Daryl Harrison drafted Figures 19 and 20 and
assisted with other figures. Financial support for this project was provided by Sigma Xi, the
Society of Vertebrate Paleontology Bryan Patterson award, the UF Department of Zoology, and
the Florida Museum of Natural History. This paper is based on research done as part of a
doctoral dissertation in the Zoology Department, University of Florida. This is University of
Florida Contribution to Paleobiology Number 354.
PRATT': TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 39
MATERIALS AND METHODS
Abbreviations and Acronyms
The following abbreviations are used for the following museum collections that house
Thomas Farm fossils: FGS, Florida Geological Survey, Tallahassee, Florida (vertebrate fossils
from FGS are now housed at FLMNH); FLMNH, Florida Museum of Natural History (formerly
the Florida State Museum), Gainesville, Florida; MCZ, Museum of Comparative Zoology,
SIarvard University, Cambridge, Massachusetts; UF, University of Florida, Gainesville, Florida.
Other abbreviations are as follows: Directional: E, east; N, north; NE, northeast; NW,
northwest: S, south; SE, southeast; SW, southwest; W, west. Quantitative; cm, centimeter; m,
meter: g, gram; kg, kilogram; NISP, number of identifiable specimens, MNI, minimum numbers
of individuals; P, probability, R X C, row by columns.
An accurate map of a fossil locality and a grid delimiting the fossiliferous sediments are
prerequisites to a taphonomic analysis, as both enable the paleontologist to determine the
general location of a fossil or sediment within the deposit. A topographic map constructed by the
FGS in 1956 (Puri and Vernon 1964; Puri, Yon, and Oglesby 1967) eliminated the need for a
completely new map. However, as no permanent measuring grid had ever been established at the
Thomas Farm locality, it was necessary to do so prior to any excavation. In September of 1981,
Russell McCarty and I used a surveyor's theodelite to determine the locations of the grid stakes.
The stakes were placed at 5 m intervals to the north and east, respectively, of a selected 0 m
North by 0 m East point in the southwestern corner of the site. The location of this point was
marked by installation of a permanent concrete surveyor's post. The major axes of the grid run
25 m to the north and 20 m to the east from this ON, OE datum marker (Fig. 1A).
The location of each 5 m by 5 m square was identified by the coordinates, in meters north
and east of the ON,OE marker, of the northeast corner stake of each square. Each 5 m by 5 m
square was subdivided into 25 1 m by 1 m squares, starting from the square designated M1
located in the southeast corner. Using this system, it was possible to record the location, to the
nearest square meter, of any region of the site (Fig. 1B). For example, square meter 5 in the
southwestern edge of the 15N by 10E square is designated 15N,10E MS.
The position of the grid on the FGS topographic map of the site (Puri and Vernon 1964)
was determined by sighting from four relatively permanent landmarks outside the excavation
area. These reference points include: (1) the eastern edge of the limestone pinnacle located on
the northwestern side of the drainage ditch; (2) a cement marker near the northeast entry gate;
(3) the westernmost edge of the enclosed south wall of the new pole barn; (4) a large spike driven
into the large live oak tree in the southwestern corner of the site (Fig. 1A and Table 1).
Establishing a Depth Datum
To facilitate recording depth of sediment layers within the site, a depth datum point was
established at 50 cm above the ground surface on the ON,OE marker. The sediment depth at this
point was arbitrarily set at 0 cm, and all other depth measurements within the site were recorded
in centimeters below this 0 cm depth level. Using this point for reference, it was possible to
determine the relative depth of any sediment level within the site, with the exception of the steep
northern and eastern walls where no digging was planned.
40 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
5E 10E 15E 20E
Figure 1. The measuring grid constructed at Thomas Farm in fall of 1981. (A) Outline map
showing position of entire grid, composed of 5 m x 5 m grid squares. Numbered points mark
locations of reference landmarks sighted from ON x OE datum (see Table 1). (B) Close-up of a
5 m x 5 m square showing numerical designation of square meters.
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 41
At the end of the study, elevations of excavated areas were determined. The elevation of
the ON,OE marker was established as 16.85 m (55.3 feet) above sea level. During the final stages
of this study, the site was cleared and a trench was excavated (by backhoe) around the edges of
the grid to aid drainage and to provide an estimate of the depth and extent of remaining
fossiliferous layers (Fig. 2A). A backhoe was also used to remove large quantities of sandy
matrix rich in microfauna from the northern section of the 10N x 10E square. An estimated 16
tons of this matrix were transported by truck to the grounds of UF (FLMNH Behavior
I.aboratory) for storage, processing and washing.
Techniques in Fossil Collecting
All workers recorded bone orientations and other relevant data in field notebooks. On any
given day, a worker was assigned a square meter in which to excavate. The sedimentary unit
(from 1 to 17) from which a fossil was taken was also recorded. Bones were exposed in the
unconsolidated sediments with the aid of dental picks and brushes. When a fossil suitable for
mapping (any identifiable fossil at least 2 cm in any dimension) was uncovered, both a field
number and a reference point were marked on the bone in indelible ink using a rapidograph pen.
The reference point was indicated by a small "x," and its location and orientation on the bone
were described in the field notes. Bearing and plunge were also recorded. The location of the
fossil's reference point was measured in centimeters north and east, respectively, from the
southwestern edge of the square, and at some depth (in centimeters) below the 0 cm datum.
The bones were mapped using a system designed and constructed by Russell McCarty,
Arthur Poyer, and myself. This apparatus is pictured in Figures 3 and 6, and a description is
included in Appendix 1. Using this system, it was possible to record orientation data for any
bone within a 10 m2 area. Several of these measuring devices were set up in the most
fossiliferous regions of the site.
All bones encountered were collected. Bones removed without complete orientation data
were placed in bags labelled with the collection date, the collector's name, the 5 m by 5 m square,
the square meter, and the sediment level.
Each sediment layer excavated was described and measured. The major beds are sands
and clays, and the bones are concentrated in layers. The layers were named with reference to the
major bone-bearing sediments. The lowest level encountered was designated as unit 1, the
highest, unit 17. The total thickness of exposed sediment currently exceeds 4 m; however, the
majority of the bones collected were removed from units 5 through 15, in particular units 5, 6, 7,
8, 11, and 15, through a thickness of approximately 3 m. Sediment samples for screenwashing,
sediment analysis, and pollen analysis were collected from specific depths within randomly
selected square meters. Strikes and dips of beds were taken throughout the course of the study,
as were determinations of depths below the datum and thicknesses of particular sedimentary
units. This information was used to construct a composite stratigraphic section for the site.
All bones collected in the field were returned to the laboratory for cleaning, repair,
identification, and cataloguing. Fossils were washed, air-dried, and identified as completely as
possible. All relevant field data associated with the elements were recorded. Plaster jackets were
prepared in the fossil preparation lab by technicians and by myself.
42 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
SP I I
Figure 2. Areas from which fossils and sediment were removed during the course of this study.
Solid black lines show the position of the grid; each square is 5 m x 5 m.
Figure 3. Permanent measuring system used to determine bone locations. (A) 5 m east-west
pipe (cm ruled); (B) stationary T-joints and vertical supports; (C) moveable T-joints (PVC) and
1 m north-south measuring pipe (cm ruled); (D) plumb bob.
Table 1. Locations of references points used to mark ON,OE datum point. Numbers correspond
to those listed in Figure 1.
Angle (east of N) Distance (m)
Reference Point from ON, PE from ON, OE
1. Limestone pinnacle, N trench 3530 21.95 (72 ft)
2. Cement marker by N gatepost 37 38.71 (127 ft)
3. Southwest edge of pole barn siding 390 43.83 (144 ft)
4. Live-oak, SW edge of site 2060 7.15 (23.5 ft)
PRATT: TAPIIONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 43
All information recorded for fossils greater than 2 cm in length was converted to numerical
codes. Fossils were sorted into three major groups: those with field numbers and complete
locality and positional data; those collected in the field with square meter and level data only; and
those removed from screenwashed concentrate but belonging to taxa whose remains for the most
part are large enough to be identified in the field (small carnivores and small artiodactyls). The
coded data were entered and stored on an IBM mainframe computer at the Northeast Regional
Data Center (NERDC). Data categories were as follows:
1. Identification data: bone identification number, family, genus, species, type
of element, side (right or left), portion of element (complete, proximal,
distal, and so on), greatest length of element.
2. Provenience data: sedimentary unit, 5 m by 5 m square, square meter,
location within meter (in cm), depth below datum.
3. Orientation data: side of element facing up in field (medial, lateral and so
on), uncorrected bearing, uncorrected plunge, corrected bearing, corrected
plunge, compass orientation of reference point.
A total of over 5,000 megafaunal entries was made. Data were analyzed and sorted using
Statistical Analysis System programs (SAS User's Guide 1982). Sorting procedures were used to
arrange elements by data categories such as taxon, element type, or sedimentary unit. Graphics
programs were used to construct histograms of preferred orientations, to plot bone positions and
to calculate numbers of individuals and numbers of specimens per taxon, densities, and relative
representations of bones.
Positions of bones possessing field numbers and specific locality measurements were
plotted on two-dimensional maps. Upon completion of a square meter map (all bones and
sediment of a given layer removed from that square), the map was photocopied on a clear acetate
sheet. The see-through map could be placed as an overlay on maps of lower sediment levels from
the same square meter to show overall bone concentrations and orientations for a given
excavated region. The reference points of bones were also plotted on north/south
microstratigraphic sections of the site to assess relative bone concentrations within different
sedimentary levels. This procedure proved to be of use in providing a check on the field
identification of the sedimentary unit.
In an effort to observe and document the sequence of events that occur in the deposition
of a sediment cone, I constructed a small-scale replica of a sinkhole in a 10-gallon aquarium tank
filled with water. The top of the tank was covered, except for an opening 5 cm in diameter
through which sediment was introduced. The opening was located near one side of the tank so
that only half of a cone formed, and the thickness and dip of the layers could be viewed through
the glass. The initial shape of the cone was determined by introduction of a limestone "rubble
layer." I subsequently dropped sand, clay, or pebble sediments through the opening periodically
to form a cone approximately 30 cm in diameter.
44 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
Modern Taphonomic Studies
Although several experimental studies dealing the effects of running water on bone
orientation and dispersal have been undertaken by previous workers (Voorhies 1969;
Behrensmeyer 1975; Korth 1979; Pratt 1979; Hanson 1980), additional studies on skeletal element
behavior in standing and moving water were conducted in order to interpret orientation patterns
of the fossil bones found at Thomas Farm. I observed bone hydraulic behavior in a series of
experiments using modern skeletal elements. "Fresh" deer and horse bones that retained grease,
as well as elements that had been naturally weathered, were placed in water, and water uptake
rates were monitored by recording weight gains to bones after varying periods of immersion. The
initial tendency for an entire bone or one end of a bone to float was also noted, as was the bone's
settling orientation through standing water.
Hydraulic behavior of bones in moving water was noted in two different experimental
procedures. In one, conditions similar to those that occur in periodic filling and draining of a
pond or sinkhole were simulated using a large (67 cm wide by 3 m long by maximum 0.5 m deep)
sink. The sink had a sloping bottom (degree of slope 10"), and a drain 15 cm in diameter at the
lower end. The bottom of the sink was covered with a 5 cm thick layer of fine sand. After the
sink was slowly filled with water, I dropped selected skeletal elements into it. The original
positions of the bones at rest were noted. Release of the plug created a strong current. Any
movements of the bones in response to this current were measured and changes in bone positions
noted. Each experiment was repeated three times.
I also observed fluvial effects on bone orientation and dispersal by dropping elements into
a small creek on the University of Florida campus. Flow velocity ranged from 35 to 50
cm/second along the length of the stream used in these experiments. Skeletal elements of
Recent deer and horses were placed in the stream in selected groups corresponding to their
Voorhies dispersal groups (Voorhies 1969), and their transport was observed. Each element's
position during transport and its final resting orientation were noted. Each set of bones was
observed in three trials. Results of these studies were consulted in analysis of fossil skeletal
abundances and bone orientations.
Numbers of identifiable specimens per species (NISP; Badgley 1986a, 1986b) of
mcgafaunal taxa both within each major sedimentary unit and for the recently excavated portion
of the site as a whole were determined. Broken fossils were examined to see if contacts with
other broken elements of the same type could be made. Minimum numbers of individuals (MNI;
Shotwell 1955) per species were calculated, based on the most abundant identifiable element.
The similarities in faunal composition and relative abundances of megafaunal taxa between layers
were assessed by the row by columns test of independence using the G-test (Sokal and Rohlf
1981:745). Faunal samples were considered significantly different from one another at the p <
Skeletal Element Abundance
The abundances of various types of elements found in an assemblage are of use in
determining the agents involved in forming a particular fossil assemblage (Voorhies 1969; Wolff
1973; Korth 1979; Maas 1985). Skeletal-part frequencies (Behrensmeyer and Boaz 1980; Badgley
1986a) of Parahippus leonensis elements were calculated as shown below:
where n = the number of each element type, and total # represents the total number of
identifiable elements represented.
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 45
The relative representation of an element was calculated by determining the number of
each clement present relative to the number expected based on the minimum number of
individuals present. The formula is:
100 X n obs./n exp. (based on MNI)
(Voorhics 1969; Wolff 1973; Korth 1979). Both abundance calculations were determined for
elements of Parahippus leonensis within each major bone-bearing layer, and in some cases results
for several similar layers were grouped to obtain larger sample sizes. Relative abundance ranks
of skeletal elements obtained from fossiliferous sedimentary layers were compared statistically to
ranks of abundances of elements from various modern bone assemblages using the Spearman
rank-order correlation test (Sokal and Rolf 1981). Representation patterns of fossil and Recent
accumulations were considered to be correlated at the p = 0.05 level. Abundances of element
types were also compared with those characteristic of fluvial deposits (Voorhies 1969;
3chrensmeyer 1975; Korth 1979; Hanson 1980).
The comparison of a bone to a quartz grain for which fluvial behavior has been determined
empirically (Allen 1965) allows one to predict the range of current velocities necessary to cause
transport of that bone (Behrensmeyer 1975; Korth 1979; Pratt 1979). The diameter of a grain (or
sphere) of quartz that should be transported by a current also of sufficient velocity to cause
transport of the bone is determined using the following formula:
dqn = (Pb-1) ; (1.91 Vb/ 1.65)
where dq,, is the diameter, in centimeters, of the equivalent quartz grain, Pb is the bone density,
and Vb is the bone volume (Behrensmeyer 1975). I determined the volume of Parahippus
leonensis elements using the water displacement method. The density of a fossil bone is
frequently much greater than its density prior to fossilization, so the density values used in the
calculation were determined for bones of modern mammals. I calculated densities of modern
Equus elements by weighing both dry bones and bones that had been allowed to soak in water for
a day. Hypothetically, weight gains caused by water uptake may change the transport potential of
a bone. The "wet" or "dry" weight of the element divided by its volume (determined by water
displacement of waterlogged elements) provided both "dry" and "wet" density values. If I was not
able to determine density of a particular element, I used the density calculations given by
Behrensmeyer (1975:570) for zebra elements.
Orientation patterns of fossil skeletal elements were determined using a number of
techniques originally used by geologists but now also employed by taphonomists (Voorhies 1969;
Hill and Walker 1972; Saunders 1977; Andrews et al. 1981; Shipman 1981; Shipman et al. 1981).
The presence of preferred bone orientation is inferred by construction of a mirror-image rose
diagram. The rose diagram records bone compass orientations in 10-degree increments from 0 to
359 degrees east of North (Shipman 1981). Rose diagrams were constructed for all bones with
orientation data, long bones with orientation data, bones from specific sedimentary layers, and
for various types of elements. Orientation patterns of bones were compared to the pattern of
elements expected in a uniform, or non-preferred orientation pattern (random orientation
pattern of Shipman 1981) using the Chi-square goodness of fit test (Sokal and Rohlf 1981:710).
Orientation patterns were considered significantly different from uniform at thep < =0.01 level.
Frequency of orientation to a preferred end was also assessed by determining percentages of
orientations of ends of bones to each of the four compass quadrants (NE, SE, SW, NW). Bone
bearings and plunges were also plotted on stereonets using the technique for plotting bearings
and plunges of lines (Voorhies 1969; Ragan 1973:95).
46 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
RESULTS AND DISCUSSION
Previous Excavations and Interpretations
Although the Thomas Farm locality was discovered over fifty years ago
and has been worked intermittently since that time, procedural information
concerning the excavations is limited. This historical account of the site has
been obtained from various publications, notes, photographs, and anecdotal
Field work at Thomas Farm was initiated in 1931 by geologists Clarence
Simpson and G. M. Ponton. An account of the results of their work, later
published in George Simpson's discussion (1932) of the Thomas Farm
mammalian fauna, included a description of a 6-foot stratigraphic section
showing 9 sedimentary layers. The section was obtained from a trench that
was evidently dug on the western edge of the site, as a major feature of the
eastern portion of the site, the boulder layer, is not mentioned in the
G. Simpson (1932) offered two hypotheses concerning the mode of
deposition of the locality. In the first, he suggested that the fossils had
accumulated as a result of the breakdown of the marine Hawthorn Formation
and subsequent reconcentration of the bony elements in the underlying Ocala
Limestone. This idea was later discarded, as no Miocene age marine fossils
are known from the locality. The alternate interpretation, one that would be
repeated and elaborated upon in subsequent publications, was dubbed "the
sinkhole hypothesis." Simpson noted that the fossiliferous sediments were
located in a large, roughly circular area about 250 feet (80 m) in diameter and
roughly 15 feet (5 m) lower in elevation than the surrounding ground surface
(G. Simpson 1932). He suggested that the depression marked the site of an
ancient sinkhole. According to G. Simpson, this interpretation was supported
by the fact that the vertebrate fossils were concentrated in specific layers, and
characteristics indicative of fluvial transport and deposition were lacking.
Following a 6-year hiatus, excavation at the site was renewed in the late
1930s by field parties from the Museum of Comparative Zoology at Harvard
University. White (1942:Plate 14) published a map and a stratigraphic section
of the site based on analysis of auger samples taken by C. Simpson. The exact
placement of the early excavation cannot be identified with certainty, as neither
White's map, nor the original map from which the published version was
constructed, contain any landmarks. Based on the stratigraphic section and
photographs of the locality taken in the 1940s (Barbour 1944:11), it seems
likely that much of the excavated region was in the vicinity of the boulder layer.
PRATY: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 47
The stratigraphic section published by White (1942) does not contain fine-
scale sedimentological details compared with what is known about the
sedimentary layers at present, but it does provide important documentation of
the early excavations undertaken by MCZ workers. Several of the sedimentary
layers pictured in White's section were not encountered in subsequent
excavations. The sedimentary unit referred to by White as the clayball layer,
located on the easterly surface of the boulder bar, was evidently not extensive
and was removed in early excavations. White also discussed the presence of a
cap of phosphatic sandstone at the southeastern edge of the site. He identified
this layer as the marine Hawthorn Formation (Hawthorn Group of Scott
1988), an assignment rejected by later workers (Bader 1956). Although no
sediments assignable to the Hawthorn Group are found within the locality
today, a recent core taken by the FGS in February 1988 approximately 100 m
to the south of the site (Thomas Farm #3) encountered a massive unit of
phosphatic marine sediments. Preliminary investigation indicates that these
sediments are not part of the Hawthorn Group, but their extent and their
relationship to Thomas Farm sediments is not clear (T. Scott pers. comm.).
White (1942) did not subscribe to C. Simpson and Ponton's earlier
hypothesis (G. Simpson 1932) that the site was formed as a result of sediment
deposition in a sinkhole, suggesting instead that the sediments represented a
sequence of several cycles of cutting and filling by a fluvial system. He cited
the presence of lenticular clays, the lack of sorting of the sediments, and the
presence of the boulder layer as support for this interpretation. White
proposed that a stream initially carved a channel in the Ocala limestone, and
eventually silted up, depositing sands and bones as the current slowed.
According to his interpretation, the stream was later rejuvenated, and current
velocities increased sufficiently to cause transport and eventual deposition of a
bed, approximately 2 m in thickness, composed primarily of cobble-sized
limestone boulders (White 1942:Plate 14). White also stated that subsequent
loss of stream competence resulted in deposition of the lower energy clayball
and sand layers that were later truncated by intrusion of marine Hawthorn
A. S. Romer (1948) of the MCZ summarized all previously published
interpretations concerning the mode of deposition of the Thomas Farm locality
and expressed support for the fluvial hypothesis. He also reported an areal
extent for the site that approached an acre, and a depth of unexcavated
fossiliferous sediments exceeding 40 feet. This information was evidently
obtained by analysis of cores, presumably those taken by C. Simpson in 1941,
although Romer did not identify the source of his information. It appears that
he overestimated the area, but not the depth of the site. A recent FGS core
sample (Thomas Farm #2) taken in February 1988 from the southwestern
portion of the site contained over 10 m (33 ft) of fossiliferous sediments.
48 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
In 1956, Robert Bader of the UF Department of Zoology, who
participated in the Thomas Farm excavation for several years and published an
analysis of the equids, reviewed the hypotheses concerning formation of the
locality (Bader 1956). Bader favored the sinkhole interpretation, but also
suggested that the sinkhole had been stream-fed. He rejected White's
identification of the uppermost layer as the Hawthorn Formation, contending
that the area had not been covered by the mid-Miocene sea.
The years 1955 through 1957 were very active field seasons for the MCZ
and the FGS. During that period, a permanent topographic map of the site
was constructed based on a USGS benchmark located on State Highway 129,
some two miles away. A number of auger samples were taken in 1956 and
analyzed by FGS geologists (Puri and Vernon 1964; Puri et al. 1967).
Photographs taken by Bryan Patterson of the MCZ show the locations of
major excavation trenches (Fig. 4). Several of these trenches have been
relocated in recent work at the site.
Stanley Olsen, a fossil preparator at the MCZ, joined the FGS in 1956.
He expanded Bader's (1956) view of site formation into a more elaborate
interpretation of the mode of deposition of the Thomas Farm, describing the
site as a combination sinkhole-stream-cave complex (Olsen 1959, 1962). This
interpretation placed more emphasis on the vertebrate fauna known to occur
at the locality than on sedimentological evidence. Walter Auffenberg of the
FLMNH (1963a, 1963b), who agreed with Olsen's views, published an
educational booklet containing an illustrated reconstruction of the site in which
he diagrammed the events involved in sinkhole formation. Richard Estes
(1963) described the salamander and lizard component of the herpetofauna
and suggested, without first-hand knowledge of the site, that deposition had
taken place in a spring-fed, seasonally water-filled sinkhole.
Two FGS publications released in the 1960s (Puri and Vernon 1964; Puri
ct al. 1967) contained maps and stratigraphic descriptions of the Thomas Farm
locality, prepared principally by S. J. Olsen. Puri and Vernon (1964)
interpreted several sedimentological features of the site as indicative of fluvial
deposition. They described three major, supposedly fluvial, sedimentary facies
at the locality: a water-lain blue-green clay, a point bar composed primarily of
limestone boulders (the "boulder bar"), and calcareous sand layers that they
described as "crossbedded." In addition, based on the presence of bat fossils at
the locality, they suggested that the stream was directly associated with a
From the late 1950s through the 1960s, fieldwork at Thomas Farm was
conducted very sporadically. The site was cleared in preparation for the 1964
Society of Vertebrate Paleontology field trip. The Field Trip Guidebook
published for the meeting (Auffenberg et al. 1964) discussed the major
sedimentary and faunal characteristics of Thomas Farm and reviewed the
hypotheses of site formation. Thomas Patton of the FLMNH, who provided
PRATT: TAPIONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA
Figure 4. Photographs of the Thomas Farm excavation taken during the 1956 field season. (A)
View to southwest showing drainage ditch. (B) Trench in northeastern region of the site. Note
dipping beds by rock hammer. Photographs by Bryan Patterson, courtesy of Museum of
Comparative Zoology, Harvard University.
50 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
much of the material for the Guidebook, made a small collection from the site
in the middle 1960s.
In the summers of 1971 through 1973, Auffenberg and FLMNH crews
spent several weeks in the field at the Thomas Farm locality. Notes taken by
Auffenberg and David Frailey during the 1973 field season show the location of
their dig for that year. Auffenberg and Frailey used the limestone outcrop in
the northern edge of the site as a reference point for the northwestern stake of
a grid system, and a point 30 feet due south of the original northern gatepost as
the northeastern stake. The grid was composed of 10-foot by 10-foot squares
designated by the letters A through F from west to east, and by the numbers 1-
5 from north to south. It is unclear whether the grid possessed 6 or 7 west-east
squares. The map of the grid in the field notes lists only 6 (A through F);
however, the distance between the northwestern and northeastern corners of
the grid exceeds 70 feet, indicating either an error of well over 10 feet was
made in measuring, or that the grid system actually contained seven east-west
squares. Upon completion of their dig, they covered the excavated area with a
sheet of bright blue plastic.
Although the grid measurements obtained from the field notes are not
exact, the approximate location of the 1973 grid and excavation is shown on a
Figure 5. Location of Auffenberg field party 1973 grid (solid lines) superimposed on grid outline
uscd in this study (dashed lines). Cross-hatched area indicates region excavated in 1973 summer
PRATY: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA
- -- .o
i -'^ '.' *
Figure 6. Photographs taken during the 1989 summer field season. (A) View from northeast of
Art Poycr and Richard Hulbert excavating unit 6. Bar in foreground marks the northern
boundary of 15N x 10E grid square. (B) Richard Hulbert demonstrating use of the measuring
system. (C) Close-up of fossils in unit 6. Dental pick (approximately 15 cm long) for scale.
52 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
map of the site (Fig. 5). Several of Auffenberg's trenches, indicated by
remnants of blue plastic, have been relocated in the most recent excavation,
although none of the marking stakes were recovered in place. Sediment levels
in Auffenberg's dig were designated by i, ii, and iii from highest to lowest
levels. The criteria used to identify these levels are not stated in the field
notes, and I was not able to correlate the three levels with my described units.
From the middle 1970s to early 1980, no excavations were undertaken at
Thomas Farm. In 1980, Thomas Farm was extensively cleared in preparation
for a field trip conducted during the 40th annual meeting of the Society of
Vertebrate Paleontology (Webb and MacFadden 1980). Excavation for the
current study began in 1981 and continued on a regular basis through 1985.
Photographs in Figure 6 show the site as it looked in 1989.
In the course of the most recent excavation, I employed the digging and
collecting techniques described above in the Materials and Methods section.
Information obtained on the sediments excavated was used to construct a
detailed stratigraphic section for the site, shown in Figure 7. Recent field work
has brought to light a number of features which must be taken into
consideration in the interpretation of the environment of deposition. Several
of these features are related to the geology of the locality, while others pertain
to the taphonomic agents involved in forming and modifying the Thomas Farm
Environment of Deposition
The Thomas Farm sediments are very localized and most cannot be
traced laterally over distances greater than 20 m. The geology of the Thomas
Farm locality is complex and study of the sediments is hindered by the fact that
large portions of sedimentary layers were removed in previous excavations.
The composite stratigraphic section of the site (Fig. 7) shows a number of
features that warrant detailed consideration in a geological analysis:
1. Extensive boulder layers of the upper unit of the Eocene
Crystal River Formation and the Oligocene Suwannee
Limestone are found in the central portion of the site (units 8
and 11). In-place limestone outcrops are located to the north
and west of the deposit.
2. Sedimentary layers dip to the south-southwest.
3. The sediments in the lower portion of the section consist of
alternating sequences of clay and sand layers (units 1 through
8a). These layers are rich in large vertebrates, but
microvertebrates are comparatively uncommon.
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 53
4. The upper boulder layer is overlain by massive beds of
calcareous sand containing abundant microvertebrates (units
12 through 15). Large vertebrate remains are rare in these
1. In-Place Limestone and Limestone Boulder Layers
The limestones at the Thomas Farm locality, both the in-place Paleogene
limestone and that redeposited in the Miocene boulder layers, provide
information concerning the mode of deposition. Although limestones of three
different ages are present within the site, only the oldest is still found in place
anywhere in the region (Puri and Vernon 1964; Puri et al. 1967). The exposed
outcrop of limestone located on the northwestern side of the northern drainage
ditch extends approximately 3 m above the present ground surface, and
contains the invertebrate fossils Oligopygus wetherbyi (an echinoid) and
Amusium ocalanum (a pectinid bivalve), index fossils of the middle unit of the
late Eocene Crystal River Formation (Puri and Vernon 1964; Williams et al.
1977). A core sample taken to the west of the excavated area (Thomas Farm
#1) in February 1988 hit limestone at 2.5 m below the surface. The limestone
in the core cannot be identified, but because it exceeds 5 m in thickness, it is
most likely in-place middle Crystal River limestone. White (1942) also
reported encountering limestone at about 7 ft below the ground surface in the
western portion of the excavation.
Unit 8, the lower massive boulder layer within the site itself (Figs. 7, 8) is
composed primarily of the upper unit of the Crystal River Formation of latest
Eocene age, as evidenced by presence of the annelid worm Rotularia vemoni
(Nicol and Jones 1982). This layer is thickest in the easternmost region of the
excavation (over 1 m in thickness), but pinches out toward the west. The
boulders vary from completely unweathered angular fragments to weathered
subangular boulders. The surrounding matrix is a fine limestone sand,
although thin clay drapes are found in some regions of the boulder layer.
Excavation of this level has not been extensive, and it is not clear at this point if
mean boulder size changes within the layer. It appears that many of the
boulders have also undergone post-depositional breakdown as the result of
ground water solution. The vertebrate fossils found in the boulder layer have
been crushed and flattened by post-depositional compaction.
In addition to the rubble layer, two large fragments of upper Crystal River
limestone have been found near the southern boundary of the excavation. One
was encountered in the course of excavation of a north-south trench through
the eastern section of the 10N,10E and 5N,10E squares. This fragment
measures nearly 2 m across, but its thickness has not yet been determined.
54 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
5 Limestone Pebbles
15 ~~" D Clay
Figure 7. Composite stratigraphic section of sediments excavated in this study. Scale = 1 m.
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 55
Another huge fragment of upper Crystal River limestone was removed by a
backhoe in the excavation of the southeastern portion of the eastern drainage
trench. This "boulder" exceeds 3 meters in diameter and may have been
broken off the piece of limestone exposed in 5N,10W. Neither of the 2 huge
fragments were in situ, as they both were recovered from regions of the site
that are considerably lower than the top of the older in-place middle Crystal
The upper boulder layer, unit 11, is composed almost entirely of the
friable Oligocene Suwannee Limestone, characterized by the irregular echinoid
Figure 8. Vertical section showing lower boulder layer (unit 8), sand and clay layers (units 9 and
10), and upper boulder layer (unit 11). North to right side of page, scale (white line) = 10 cm.
56 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
Rhyncholampus gouldii (Puri and Vernon 1964). This boulder layer is similar
to the underlying unit 8 rubble layer, except that many more of the boulders
are rounded and show evidence of extensive solution. A number of the
boulders are quite large in size (over 25 cm in diameter). The thickness of this
boulder layer thins to the west where it is replaced by a layer of fine white
limestone sand. Large numbers of vertebrate fossils are found in unit 11,
although they are generally badly crushed.
Neither the upper unit of the Crystal River Formation, the Rotularia
vemoni Zone of Williams et al. (1977), nor the Suwannee Limestone occur in-
place today at Thomas Farm or in the surrounding region. Based on the
presence of Suwannee Limestone in the boulder layer and its absence from
nearby in-place outcrops, White (1942), and later Puri et al. (1967), stated that
the boulders had been carried to the locality from some distance away by a
fluvial system of extremely high competence. These workers apparently were
unaware that actually two boulder layers are present, and that the lower is
composed of upper Crystal River-age limestone. The thickness of the boulder
layers, coupled with the large size of the boulders and limestone fragments,
indicate that fluvial transport is unlikely to have been responsible for
deposition of these sedimentary layers. Rather, during the early Miocene a
complete stratigraphic section of middle Crystal River Formation through
Suwannee Limestone existed at the site. The only remnants of the upper
Crystal River Formation and the Suwannee Limestone are those that fell into
the deposit and were preserved. The remaining in-place sections of these two
limestone units were removed by erosion. The nearest measured section
containing a reasonably complete sequence of these units occurs in a quarry
near Mayo, in Lafayette County, about 75 km northwest of Thomas Farm.
Using the thickness of the Mayo section and the type section of the Crystal
River Formation (Puri and Vernon 1964) in Citrus County, Florida, as general
guides, the in-place limestone sequence at Thomas Farm in the early Miocene
may have been as much as 30 m thick.
Numerous previous workers (Simpson 1932; White 1942; Bader 1956;
Olsen 1959; Puri and Vernon 1964) have noted the presence of limestone at
Thomas Farm and have attempted to relate it to the mode of deposition. In
fact, the entire north-central region of Florida is underlain by limestone that
has been subject to solution since the late Oligocene (Cooke 1945; Puri et al.
1967; Lane 1986), and sinkholes and caves are common features of the north
Florida karst terrain (Davis 1930; Puri and Vernon 1964; Puri et al. 1967,
Williams et al. 1977; Sinclair et al. 1985; Beck and Sinclair 1986; Lane 1986). It
is likely that Thomas Farm is indeed the site of an ancient sinkhole, perhaps of
the collapse doline variety. A sinkhole that forms through limestone collapse
generally possesses high, steep sides and may act as a trap, while a sinkhole
that forms through solution of limestone and the weight of overlying non-
limestone sediments tends to be funnel-shaped, and may form a shallow pond
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 57
near the ground surface (Stringfield et al. 1974; Bogli 1980; Sinclair et al. 1985;
Trudgill 1985; Beck and Sinclair 1986; Lane 1986). The boulder layers
obviously represent collapse events, but probably do not mark the initial
collapse phase that caused formation of the sinkhole, as these layers lie above
the fossiliferous sand and clay sediments. Recovery of a deeper, extensive
boulder layer at the base of the fossiliferous sediments would substantiate the
interpretation of Thomas Farm as a collapse sink. If the unit 8 and 11 boulder
layers are remnants of a collapsed cave roof, then it must be assumed that the
sediments below unit 8 were deposited in a closed cave system. However, the
absence of speleothems, as well as several other features discussed later,
suggest that such was not the case.
2. Dip of Beds
One of the most obvious features of the Thomas Farm sediments is that
all beds excavated exhibit a pronounced dip to the south-southwest (Fig. 9A).
Steepness of dips of the various beds range from 12 to 22 degrees, averaging
about 15 degrees. The beds strike from 255 degrees east of North to 310
degrees east of North, with the majority striking 270 degrees, almost due west
(Table 2). The dip of the beds may be clearly seen in White's stratigraphic
reconstruction of the deposit (White 1942:Plate 14), and in photographs taken
during the 1956 field season (Fig. 4); however, among previous papers on the
Thomas Farm, only the 1964 SVP Guidebook (Auffenberg et al. 1964) makes
reference to this characteristic. Other workers (Puri and Vernon 1964; Puri et
al. 1967) incorrectly interpreted the dipping beds as crossbedding.
The Thomas Farm sediments exhibit two major directions of dip. The
lower layers dip southwest at about 220 degrees east of North. A change in dip
direction occurs in the upper part of the massive sand layer designated as unit
5. Upper unit 5 and higher layers dip on average 180 degrees east of North.
Since there is no evidence of substantial regional dips in north-central Florida,
the presence of steeply dipping beds requires explanation.
One hypothesis, involving tilting by post-depositional tectonic activity, is
quite unlikely for a number of reasons. Such phenomena are rare in Florida's
geological history. Evidence for subsidence would be provided by large-scale
faulting over an extensive area. There are small faults in several regions of the
site, but they appear to be associated with local post-depositional compaction
and are hardly part of a regional pattern. The fact that there are two different
sets of bed orientations within the site also indicates that these are local
intraformational structural features.
An alternative explanation is that the dip of the beds reflects their original
position at the time of deposition. Acceptance of this hypothesis limits the
58 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
Figure 9. (A) Vertical section showing dipping beds at Thomas Farm. North to left of page,
scale (black line) = 10 cm. (B) Debris/sediment cone model formed under laboratory
conditions. Note dipping sedimentary layers. Scale = 5 cm.
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 59
number of geologic conditions under which the fossiliferous sediments may
have formed. Fluvial deposits do not commonly possess sets of beds that
uniformly dip steeply over distances of several meters, and beds laid down in
standing water are ordinarily level except for slumping. The two most common
non-marine geologic conditions under which beds show consistent dip are
deltas and alluvial fans (Twenhofel 1932). There is no evidence to suggest that
either of these two features are represented at Thomas Farm. Both require
large streams or rivers, and the latter is characteristic of arid mountainous
regions. A third means by which large sets of beds may be deposited on an
angle is seen in regions of karst and involves formation of a debris cone at the
bottom of a sinkhole (Sweeting 1973; Brain 1981). If the opening to a sinkhole
is relatively constricted, sediment influx is confined to a limited area and
sediment accumulates primarily in a cone-shaped pile on the floor of the sink.
A small-scale debris cone model formed in an aquarium (see Materials and
Methods) shows that sedimentation under the above described conditions does
result in a cone-shaped structure composed of uniformly dipping sediment
layers (Fig. 9B).
Debris cone deposition is generally initiated in a collapse sinkhole when
the limestone roof of a closed chamber (cave) within the limestone collapses.
This cave-in results in a pile of boulders that forms the base of the cone, and
also provides an opening through which sediment from the surface may be
Table 2. Strikes and dips taken on major sedimentary units, s.d. = standard deviation of mean.
Sediment No. of Mean Strike
Unit Observations degrees E of N s.d. Mean Dip s.d.
1 1 314 24 SW
2 2 314 9 SW 2.0
3 3 311 19.6 13 SW 2.6
4 2 309 13 SW 0.5
5 (lower) 5 312 4.8 16 SW 1.2
5 (upper) 8 280 6.9 15 SW 3.1
6 1 272 22 SW
7 12 265 9.7 21 SE 5.2
8 10 279 13.6 18 SW 4.9
9 3 261 7.3 20 SE 1.7
10 7 266 7.2 23 SE 4.8
11 4 289 32.1 14 SW 2.8
12 7 282 7.9 19 SW 3.4
14 8 281 5.2 23 SW 4.0
15 3 273 12.6 22 SW 3.7
60 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
Under ideal conditions, a debris cone that forms in the center of a sink is
conical and exhibits a number of recognizable features. In a natural
environment, perfect symmetry is neither expected nor seen, but a number of
characteristics pertaining to the dip of the beds and general shape of the cone
can be identified.
As one moves across the surface of a cone at any given depth, the
compass direction of dip should change. This relationship is illustrated in
Figure 10A. Dip direction also changes more rapidly near the apex of the cone
(distance 1 in Fig. 10A) than near the base of the cone (distance 2 in Fig. 10A).
Applying this model to the Thomas Farm locality, it is expected that directions
of maximum dips recorded in the eastern regions of the site would be more
easterly than those taken in more western sections of the site. Due to the
relatively small area excavated, the maximum distance across which a given bed
can be measured seldom exceeds 5 m. Those layers in which several strikes
and dip measurements were taken at east-to-west distances of 5 m or greater
are units 7 and 8. These beds show an easterly to westerly trend in dip
directions, although there is also some minor local variation. The total
difference in dip direction in unit 7 across an east-to west line 8.3 m long is 31
degrees, and in unit 8, the dip direction changed by 14 degrees over a 5 m east-
west distance. Based on the amount of change in east-to-west dip directions of
the various layers, the circumference of the cone taken at the level units 7 and
8 in 15N,10E Mil-15 may be estimated using the following formula:
Circumference = a (i)
where a is the maximum east-west distance between two dip readings, and b is
the change in dip direction across a. The circumference estimates of the cone
in units 5, 7, and 8 range from 97.5 to 128 m. Therefore, if a complete cone
formed in the middle of the sink, the bottom of the sinkhole may have been as
wide as 30 to 40 m, well within the size range of collapse sinkholes found today
in north Florida (Lane 1986).
Figure 10B illustrates another characteristic of dipping beds on a debris
cone. In heading north along a level line that passes through the central
longitudinal axis of a cone from a given point on the surface of the due south
face, dip direction should suddenly change to the north if the cone is
symmetrical. Assuming the above estimated calculations of debris cone size
are reasonably accurate, the length of line xy measured from point x on the
south face of the cone in unit 7 to point a point (y) on the same level of the
north face in unit 7 is 30 m. It is unlikely that north-dipping sediments, if any
exist, will ever be exposed. A pole barn and a driveway are located 17 m to the
north of the excavation, and the present elevation of this region is over 5 m
higher than that of the dig.
It is possible that the debris cone did not form in the center of the sink,
but that factors governing sediment deposition caused buildup of a partial cone
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA
Figure 10. Changes in dip directions of beds deposited on a sediment cone. (A) Diagrammatic
representation of degree of change in dip direction across various levels on the surface of a
sediment cone. (B) Diagram illustrating the 1800 change in dip direction on either side of the
cone's central axis.
62 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
or talus slope against the northern wall of the sinkhole, and thus the northern-
dipping beds were truncated. The map of the locality (Fig. 1, point 1) shows
the location of the pinnacle of middle Crystal River limestone on the northern
edge of the pit. This outcrop is probably a remnant of the northern wall of the
sink. An FGS auger sample taken approximately 15 m to the north of the site
in 1956 (Puri and Vernon 1964; AS-297) reportedly hit limestone at an
elevation of 11.9 m (39 feet) above sea level (7.6 m below ground surface).
However, as the type of limestone was not identified, it cannot definitely be
stated whether it was in place or part of a rubble layer.
As noted previously, there is a change in dip attitude of all beds above
unit 5a (Table 2). Dip directions of the lower layers average 220 degrees to the
southwest. Starting with the upper layers of massive unit 5, the dips change to
a more southerly (175-190 degrees east of North) orientation, and average this
reading throughout the remainder of the stratigraphic sequence. The change
in dip direction may be the result of a change in the location of initial sediment
input, as shown in Figure 11, which schematically illustrates results of
laboratory experiments with the sinkhole model. Either an increase in the size
of the opening through which sediment falls, or displacement of the position of
the opening toward the west results in the formation of a second cone that
partially overlies the first. Dip directions of beds forming the upper cone are
shifted eastward relative to dips of beds in the lower cone. It is not unlikely
that the source of sediment input at Thomas Farm may have shifted slightly to
the west, either by the formation of a larger opening at the top of the sink, or
by blockage of the previous opening, causing a change in the direction of dip of
the beds. Presumably, enlargement of the opening would involve introduction
of limestone rubble into the site; however, there is no evidence that a rubble
fall occurred at the time the bed dip shifted. Sediments of units 5a and 5b are
very similar in lithology and neither contain large boulders of limestone. The
original opening may have been constricted by slippage of a large fragment of
limestone that initially blocked the eastern portion of the sink entry. This huge
boulder may have subsequently fallen and formed the lower boulder layer in
Evidence that the Thomas Farm sediments form a debris cone is also
provided by features of the two boulder layers. In the eastern region of the site
where boulders are thickest (Fig. 12A) the sand layer is extremely thin, but it
thickens gradually toward the west, attaining a maximum thickness of about 25
cm. Both boulder layers thin toward the west. This type of relationship was
reproduced in the laboratory with the sinkhole model (Fig. 12B). Boulders pile
up just below the point of input but pinch out along the lower surface of the
cone. Sandy sediments that fall on top of the boulder layer are thickest along
the the lower surfaces and near the base of the cone. The introduction of a
second boulder layer on top of the sand causes compaction of the sand at the
point of initial input, and some of the sand makes its way into the interstitial
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA
Figure 11. Possible mechanism by which dip of beds may change at one point or location on the
sediment cone. Sediment input at point A results in formation of sediment layers whose dip at
point X is 2200 E of N (dashed line). A westerly shift in sediment input (point B) results in
deposition of sediments with more easterly dip directions at point X (white line).
spaces between the lower boulders. However, the sand layer remains thick on
either side of the apex. This difference in the relative thickness of sand and
boulder layers is also due in part to the differing angles of repose achieved by
the two sediment types. The relationship of the sand and boulder layers
corroborates other evidence of a complex debris cone within the site.
3. Clay Layers
The sediments found below the boulder layers are predominantly clay and
sandy-clay layers (Fig. 13). Many of these beds are composed of numerous
64 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
7Un it 11
Figure 12. The relationship of sand and boulder layers at Thomas Farm. (A) Photograph of
boulder layers showing position of sand layer between them. Scale = 10 cm. (B) Sediment cone
model showing thinning of boulder layers and thickening of sand layers on slope of debris cone.
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA
*-^ ^^ ^l11 2 1 i'-....
figure 13. Vertical section showing laminated sand and clay layers at Thomas Farm. Shown is
unit 5, scale = 30 cm.
fine laminae, characteristic of water-lain sediments (Dunbar and Rodgers
1957). Vertebrate fossils of medium- to large-size animals are found
concentrated in specific layers. Units 1 through 4 have been exposed only
recently, and were not extensively excavated in this study. Units 5a and 5b are
both laminated orange sandy-clays. The two portions of unit 5 were given
separate letter designations because of the change in dip direction that occurs
between them. Total thickness of unit 5 averages about 43 cm. It is more
sandy than some of the upper layers, and closer inspection reveals that this unit
is composed primarily of thin (1-2 mm) laminae, each composed of a fining
upward sequence of quartz sand to yellow or gray clay. The repetitive nature
of the sediments indicates a repetitive pattern of deposition, but whether it
signifies a daily, seasonal, or yearly cycle cannot be determined. The method
by which these laminae formed is also unclear. Finely laminated sediments are
characteristic of sinkhole and cave deposits (Laury 1980; Agenbroad 1984;
White et al. 1985). Milske et al. (1983) attributed deposition of laminated
66 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
sediments in Mystery Cave, Minnesota, to variations in current velocity of the
cave drainage system. At least two clay laminae in unit 5 contain well-
preserved plant remains. These fossils are currently under study; however,
preliminary analysis indicates that the remains are predominantly aquatic
plants (Newsom pers. comm.).
Unit 5 is overlain by unit 6, a gray clay layer with numerous cobble- to
pebble-size fragments of middle Crystal River limestone. Vertebrate fossils
are numerous in the lower portion of this unit. Unit 6 grades upward into a
somewhat less fossiliferous sandy-clay. Total thickness of unit 6 varies
somewhat over the area excavated, ranging from 10 to 18 cm. Unit 7 is very
similar to unit 6, and is composed of a clay-pebble layer that grades into a sand
layer. Unit 7 is about 15 cm thick. The lowest portion of unit 8 is composed of
clay-sand layers, rich in vertebrate fossils. In the upper level of unit 8, large
limestone boulders are surrounded by fine sand.
Although there is no evidence of bioturbation of any of the layers and
freshwater molluscs and ostracodes are absent, the laminated nature of the
sediments, and the presence of aquatic plant remains indicate that the sand
and clay sediments were water-lain. In addition, excavation revealed that
bones and pebbles had caused deformation of the sediment layers, a
phenomenon that occurs when sediments are wet or moist (Schrock 1948).
Calcite structures previously interpreted as "dripstone" are also found in the
most superficial layers of these units; however, these formations are probably
of recent origin and are not speleothems.
4. Calcareous Sand Layers
The sediments above the boulder bar (Fig. 14) consist of several thick
beds of limestone sand composed of breakdown products of the middle
member of the Crystal River Formation (M. McKinney pers. comm.). The
grains are not weathered, and the sand contains the characteristic forams
Lepidocylina and Operculinoides. The layers above the upper boulder layer
have been designated as units 12 through 17. Each of these beds is a massive
lime-sand layer separated from the overlying unit by a thin clay lamina or a
layer of calcite. Microfaunal remains are amazingly abundant and well-
preserved in these layers, particularly in unit 15. A complete section of the
layers above the boulder bar was exposed in a trench dug through the southern
edge of the 5N,10E square. The total thickness of these layers exceeds 1.5 m,
and the method by which they were formed is not known. However, based on
the unique nature of these sediments when compared to the other sediment
units of the site, it is clear that the environment of deposition changed after
collapse of the final boulder layer.
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA
Figure 14. Vertical section showing massive calcareous sand layer unit 15. Length of pick handle
= 42 cm.
Sediment Deposition and Site Formation: Summary
The features of the Thomas Farm sediments and associated limestone
suggest that in the early Miocene, the site was a large sinkhole or cave.
Although connection to surface drainage features such as above-ground
streams is not indicated, it is likely that solution joints in the limestone
provided internal or deep drainage, as is seen in present-day Florida sinkholes
(Williams et al. 1977; Lane 1986).
The "sediment cone" formed as sediment, perhaps introduced from the
ground surface above as well as entering the sinkhole through the underground
drainage network, piled up on the floor of the sink. Sediments also formed
from breakdown of the limestone itself. The presence of reworked fragments
and fossils (i.e. Chlamys spillnani) of middle Crystal River age throughout the
various layers of the site indicate that ground water was conducted through
68 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
Figure 15. Proposed sequence of deposition at the Thomas Farm locality. (A) Formation of a
chamber in middle Crystal River limestone. (B) Collapse of chamber roof forms base of debris
cone. (C) Deposition of sands and clays, followed by wall collapse of upper Crystal River
limestone (UCR). (D) Collapse of Suwannee Limestone (SL). (E) Formation of cave and
deposition of calcareous sand outwash from cave onto debris cone.
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 69
solution joints in this limestone. Most artesian flow in north-central Florida
today is carried through Crystal River limestones (Fergusen et al. 1947;
Rosenau et al. 1977). The absence of speleothems from the fossiliferous clays
and boulder layers suggests but does not necessarily prove that these sediments
were not deposited in a cave. Caves in Florida that develop under certain
environmental conditions (for example, water-filled caves) are not
characterized by secondary calcite formations (Davis 1930; Lipchinsky 1963;
Williams et al. 1977). If the sediments were deposited in a cave setting, it
would appear that some agent must have transported bones to the locality, as
will be discussed in the following section.
Although the exact mode of formation or deposition may never be
known, Figure 15 presents a possible scenario of events important in the
development and filling of Thomas Farm. Initial solution and collapse of
limestone caused the formation of a chamber (or cave) within the limestone
(A). With the subsequent collapse of the chamber roof, a sinkhole was
formed, and deposition of bones and sediments began (B, C). The sinkhole is
pictured in Figure' 15 as jug-shaped, as this is a characteristic shape in early
stages of formation of a collapse sinkhole. Early in the history of its
development, the Thomas Farm sinkhole was water-filled, perhaps seasonally.
The lower clay layers were water-lain. The sinkhole was fed and drained by
underground drainage systems in the middle Crystal River Formation, possibly
in response to fluctuating water table levels. Some event, perhaps partial
collapse of a wall or overhang, caused a change in sediment input location and
a shift in dip of beds. The first major collapse of upper Crystal River limestone
(C) evidently occurred at a lower zone of weakness between it and the middle
member of the Crystal River Formation and along the upper unconformity
between the Crystal River and Suwannee Limestones (Cooke 1945; Puri and
Vernon 1964). The huge boulders of upper Crystal River-age may also have
fallen at this time. Shortly after the first collapse, the unsupported overhanging
portion of Suwannee Limestone also broke apart and fell onto the debris cone
(D). The reasons for the limestone collapse are not clearly known. One causal
factor may have been lowered water table levels. Lowering of water levels may
leave overhanging sections of limestone unsupported (Bogli 1980, Beck and
Sinclair 1986; Lane 1986), which then break at zones of solution or weakness.
Massive calcareous sand layers mark the final stages of deposition of the
Thomas Farm locality. The sedimentological and faunal differences between
the upper sand units and the lower clay layers indicates that following the two
major limestone collapse events, a major change in mode of deposition
occurred. The fact that large numbers of microvertebrates, in particular bats,
are present in these layers, and that the sand is composed of breakdown
products of the middle Crystal River Formation, provide evidence that the
upper sediments may have been outwash from a cave in the middle unit of the
Crystal River Formation (E). It is clear that caves were present in the area
70 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
prior to the major collapse event, as bat remains are found even in the lowest
layers of the site; however, collapse of the sides of the sinkhole perhaps
exposed the opening to a cave in the immediate vicinity of deposition. The
cave was evidently frequented by mammalian and avian carnivores as well as
bats (Pratt 1986, 1989). However, conditions no longer favored preservation of
remains of large vertebrates.
The results of geological analysis indicate that several of the previously
presented ideas concerning the mode of depositon of the Thomas Farm
deposit are at least partially correct. This interpretation, based on additional
data, presents a more complete view of site formation and shows that several
of the depositional events were sequential rather than concurrent, as had been
suggested by some earlier workers. Sedimentary evidence argues against the
presence of a fluvial system of high competence. Taphonomic data compiled
for the vertebrate fossils is presented in the next section and provides
additional detailed information regarding the mechanisms by which the entire
locality and bone concentrations formed.
Taphonomic investigations are undertaken to determine the potential
sources of a fossil concentration and the taphonomic factors that have acted to
modify the assemblage. Features of the environment of deposition can also
often be inferred by taphonomic analysis of the megavertebrate fossils found at
a locality. Recent taphonomic studies (Korth 1979; Bown and Kraus 1981;
Behrensmeyer 1982, 1988; Maas 1985; Badgley 1986a) have demonstrated that
formation of bone concentrations, even in seemingly homogeneous
environments of deposition, is an extremely complex process. Based on
geologic evidence, Thomas Farm clearly represents some type of karst-related
deposit, most likely a cave or sinkhole. Figure 16 lists possible mechanisms
that may have been important both in forming the bone accumulation in a cave
or sinkhole and in its further modification prior to excavation. The bone
source within the sinkhole or cave was provided by animals either falling in or
being transported in by some mechanism. A variety of agents such as
weathering and scavenging potentially modified the assemblage prior to its
recovery as a fossil deposit. In order to determine which events may have been
crucial in the formation of the deposit, there are numerous aspects of a
megavertebrate bone assemblage that may be examined from a taphonomic
viewpoint. The four following areas of investigation have been shown in
previous studies to provide the highest quality taphonomic information.
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA
LOSS Live Animals LOSS
Weathering Dead Animals Scavenging
Figure 16. Summary of potential taphonomic pathways leading to formation of the Thomas Farm
vertebrate fossil assemblage.
72 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
1. Surface features of the bones provide evidence of physical
taphonomic factors such as weathering caused by subaerial
exposure (Behrensmeyer 1975, 1978; Korth 1979; Andrews
and Cook 1985), and abrasion characteristic of fluvially
transported bone (Korth 1979; Shipman 1981; Shipman et al.
1981: Behrensmeyer 1982). Diagnostic breakage and surface
modification may be caused by carnivores and scavengers
(Bonnischen 1973; Haynes 1980, 1983; Hill 1980; Binford
1981) while other surface features are indicative of trampling
(Andrews and Cook 1985; Behrensmeyer et al. 1986).
2. Faunal diversity and relative abundances of taxa provide
information concerning the environment of deposition and
possible sources of the bone concentration, in addition to
indicating paleohabitats near the site of fossilization. In this
study, comparison of faunal components between the various
bone-bearing layers are made to determine if taphonomic,
environmental, or depositional changes occurred during the
course of site formation.
3. Activities that disperse or cause sorting of skeletal remains
can be discerned by assessment of relative abundances of
skeletal element types. Bones accumulated by carnivores and
other biotic means often show recognizable patterns of
representation (Behrensmeyer and Boaz 1980; Brain 1981;
Binford 1981, Blumenschine 1986). High representations of
hydrodynamically similar skeletal elements may indicate the
presence of moving water and also provide estimates of
minimum and maximum current speeds (Voorhies 1969;
Behrensmeyer 1975; Korth 1979; Hanson 1980; Badgley
4. Patterns of bone orientation can be extremely useful in
taphonomic analysis (Voorhies 1969; Saunders 1977; Hunt
1978; Andrews et al. 1981; Shipman 1981; Shipman et al.
1981; Maas 1985). Observations on the presence of skeletal
articulation or association as well as compass orientations of
long bones and their positions within the enclosing sediment
provide information concerning rates of burial, evidence of
scavenging or trampling, and presence or absence of water-
Megafaunal vertebrate remains from the six most extensively excavated
levels (units 5, 6, 7, 8, 11, and 15) are analyzed. Units 9 10, 12, 13, and 14 were
the least fossiliferous layers and therefore were not extensively worked.
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 73
Types of Bone Modification
The physical condition of a fossil bone can provide information
concerning taphonomic processes and depositional environment of a fossil
locality. A bone may be altered both prior to its burial and after it has been
buried and fossilized. The types of skeletal modification that occur after the
animal's death and before the bone is either completely destroyed or buried
provide documentation of events that occurred shortly after the death of the
animal. Post-depositional changes are caused by processes that affect buried
bone, either prior to or following its fossilization. Hill (1980) and Binford
(1981) categorized types of post-mortem damage to bone. Types of
destruction can also be grouped by causative taphonomic factors
(Behrensmeyer 1978; Korth 1979; Pratt 1979; Binford 1981; Haynes 1980, 1983;
Bchrensmeyer et al. 1986).
A bone may be destroyed by weathering if it is exposed to climatic factors
prior to its burial. Behrensmeyer (1978) documented six stages of weathering
on Recent bone, ranging from stage 0 (no modification) to stage 5 (nearly
complete breakdown). Bone destruction attributable to weathering provides
evidence that the element was exposed on the ground surface before being
covered by sediment. Within a fossil locality, recovery of elements exhibiting
all stages of weathering indicates that the assemblage probably formed
attritionally; however, even if the fossils do not appear to be weathered, an
attritional assemblage cannot be ruled out. The degree of weathering is
dependent on the type of climate, or even the microclimate at the ground
surface, as well as the length of time of exposure (Behrensmeyer 1978). Bones
that are exposed in the open prior to burial may also exhibit fine grooves and
striations caused by roots and fungi (Haynes 1980; Andrews and Cook 1985).
Carnivores can be important agents of bone modification. Hill (1980)
characterized the various types of damage attributable to predators;
Bonnischen (1973), Binford (1981), Shipman (1981), and Haynes (1980, 1983),
described bone destruction caused by various mammalian carnivores. Many of
the diagnostic features are caused by a predator's teeth.
Large mammals can modify, damage, and even disperse skeletal elements
by trampling them (Behrensmeyer and Boaz 1980; Andrews and Cook 1985;
Behrensmeyer et al. 1986). Bones subject to trampling are characterized by
diagnostic striations and, depending on the element, may be broken or crushed
(Behrensmeyer and Boaz 1980).
Post-depositional damage, in particular breakage, is sometimes difficult to
distinguish from pre-burial damage. The most diagnostic post-depositional
break, a smooth stress fracture that occurs perpendicular to the long axis of the
bone (Shipman 1981) is caused by sediment compaction. Bones may also be
crushed by the weight of overlying sediments. Re-exposure of a bone, either
74 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
prior to or following fossilization, also has the potential to cause modification
and destruction. Water-worn or abraded bones are characteristic of fluvial
environments. It is not the action of the water alone that causes bone
destruction, but the waterborne sand particles that wear and polish the ends
and processes of skeletal elements. The degree of water-wear on a bone is a
function of the amount of time the element is exposed to the action of sand
particles and the strength of the current involved. The bone may be modified
as it is transported in an aquatic system or be abraded in situ as entrained
particles move past (Behrensmeyer 1982). Therefore, this type of bone
modification, while easily recognizable, is difficult to quantify. With the
exception of a study by Korth (1979) on the effects of water-wear on
microvertebrate remains, and observations by Shipman et al., (1981) and
Behrensmeyer (1982), little quantitative information exists concerning the
amount of abrasion seen on a bone and the length of time required to produce
it. For the purposes of this study the degree of water-wear is assigned to four
classes (Fig. 17). Unworn bone bears no obvious signs of water-wear.
Minimal wear is indicated by slight rounding of ends of bones and processes
and some pitting by sand grains on the surface of the bone. Moderately water-
worn bone has rounded articular ends and processes, with cancellous bone
visible at articular ends. The morphological features of the bone are still
evident, so the bone is identifiable. Severely water-worn bone is characterized
by completely abraded ends and processes. Distinguishing features have been
Surface Modification of Thomas Farm Fossils
Compaction of sediments and repeated wetting and drying of the
fossiliferous clays has caused extensive post-fossilization breakage of the
Thomas Farm megavertebrate remains, as shown in Figure 18. Fossils near
the present-day surface are subject to solution by ground water or are etched
by root acids. These post-depositional factors have acted to obscure pre-
fossilization features on a great number of the bones collected. Observations
therefore pertain to bones that have been relatively undamaged by these
Unit 5. Vertebrate fossils recovered from unit 5 may be divided into two
groups: those that show characteristic water-wear, and those that are
unmodified. Complete elements, although not commonly found in this unit,
show no signs of pre-burial destruction and, except for post-fossilization
compaction breaks, are perfect. No bones found in unit 5 show signs of
weathering, modification by carnivores, or trampling. Bones that have been
abraded in running water are common and include both long bone fragments
and epiphyses, podials, and phalanges. Water-wear ranges from minimal to
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA
A 1cm B
Figure 17. Bone water-wear classes, demonstrated on equid astragali. (A) Minimal wear,
indicated by slight pitting of bone surface. (B) Moderate wear, characterized by early stages of
surface abrasion and exposed cancellous bone. (C) Severe wear, indicated by high degree of
rounding and by loss of distinguishing surface characteristics.
Figure 18. Partially prepared left dentary of Parahippus leonensis. Cracks are due to post-
depositional compaction and sediment drying.
76 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
moderate, with podials showing the highest degree of abrasion. The presence
of both complete, unaltered bones and of water-worn bones indicates that
skeletal elements were derived from two sources. Complete bones may have
been introduced into the sink from the immediate vicinity and buried rapidly,
while the water-worn elements may have been transported to the sink from a
more distant location. Alternately, the elements showing signs of water-wear
may have simply been exposed, in situ, to sand abrasion. The lack of large
quantities of severely abraded bones argues against the presence of a stream of
high competence, and indicates that the water-worn bones were neither
transported more than a few kilometers, nor exposed to abrasive sediments for
any great length of time (Shipman et al. 1981; Behrensmeyer 1982).
Units 6 Through 8. Units 6 and 7 and the lower clay layer of unit 8 are
very similar to one another in lithology and faunal content. Bones recovered
from these beds show predominantly two types of destruction, weathering and
water-wear, although the majority of the bones appear unmodified.
In unit 6, weathered elements are uncommon, and none of the bones
collected shows evidence of climatic destruction beyond stage 1 of
Behrensmeyer (1978), characterized by fine cracks running parallel to the long
axis of the bone. The most notable alteration of the unit 6 fossils is abrasion
caused by sand grains entrained in running water. Water-worn elements of the
most numerous megafaunal species, Parahippus leonensis and Archaeohippus
blackbergi, are common in unit 6. Roughly 66% of the podials, in particular
scaphoids, lunars, and astragali, shows evidence of minimal to moderate water-
wear. The larger bones show signs of wear at the articular ends, where the
surface of the bone has been worn away to reveal the cancellous bone beneath.
This feature may also be seen on the ends of weathered bones; however, the
fossils have none of the other obvious characteristics (dessication cracks,
flaking, etc.) associated with weathering. No evidence of bone destruction by
trampling or by activities of carnivores was noted. Although carnivores
regularly chew the ends of long bones, exposing cancellous portions
(Bonnischen 1973; Haynes 1980, 1983; Binford 1981), this agent of destruction
may be recognized by the jagged or uneven edges that result. The presence of
water-worn bones in unit 6, as in unit 5, indicates that moving water was one of
the taphonomic factors associated with the site during its formation.
Weathered bone is a more common feature of units 7 and 8 than of units
5 or 6, although bones from these upper layers do not show evidence of
extreme weathering destruction. A small number of complete elements, in
particular those found in unit 8, are more weathered on the side found facing
upward in the deposit. Behrensmeyer (1978) noted that the surface of a bone
that is facing upward is subject to greater climatic destruction than the side
resting on the ground. The number of whole bones found with the more
weathered surface upward may indicate that this type of modification occurred
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 77
after the bones had attained their final resting positions and prior to immersion
in water or burial, as exposure to air is a requirement for weathering.
Therefore, the bottom of the sink may have been dry for an unknown period of
time. The fact that the first sequence of limestone collapse took place shortly
after deposition of this layer provides some support for this observation. Block
collapse of limestone is reported to occur during dry or drought periods when
water table levels are significantly lowered (Lane 1986). Small bones and bone
fragments with weathered surfaces did not show any relationship between side
found facing upward in the field and regions of heaviest weathering damage. It
is possible that these bones were exposed subaerially elsewhere before falling
into the sink, or were displaced by some agent, such as a scavenger or moving
water, from their original positions within the sink.
Bone damage caused by carnivores is not apparent on fossils from units 7
and 8, with the exception of two elements of Parahippus leonensis. A
depressed puncture hole, similar to that made by a mammalian carnivore
canine (Haynes 1980, 1983; Hill 1980, Binford 1981; Shipman 1981) was found
on a very damaged complete calcaneum from unit 7. The broken edge of a
mandibular fragment from unit 8 has numerous depression fractures similar to
those seen on chewed bone (Hill 1980; Binford 1981). The entire surfaces of
both bones are pitted. It is possible that these two elements were initially
derived from scat of a large mammalian carnivore. However, the rarity of
bones with this type of modification implies that carnivores or scavengers did
not have access to the majority of bones found in these layers.
Abrasion caused by water-wear is evident on a few elements from units 7
and 8. The astragalus, an element noticeably subject to fluvial abrasion,
exhibits minimal to moderate water-wear in 3 of 5 specimens in unit 7, and on
1 of 14 specimens from unit 8. The bones obtained from the upper sand and
boulder layer of unit 8 are often complete, although most have been crushed by
post-depositional compaction. Little evidence of pre-burial climatic or biotic
destruction can be discerned on these elements.
Unit 11. In unit 11, the upper boulder layer, large numbers of complete
skeletal elements, including skulls and mandibles, are found. Unfortunately,
the majority of the elements have been so crushed or deformed by post-
depositional compaction that postmortem, pre-burial damage cannot be
assessed. A few of the less damaged bones show weathering features
equivalent to those of early stage 1 of Behrensmeyer (1978). Very little water-
worn bone is found in this unit. Those bones that do show evidence of water-
wear are very light in color, unlike the majority of darkly-colored bones found
at the locality. The reason for this feature is not known, but does suggest a
different taphonomic history for these elements.
No bones from unit 11 show signs of having been modified by trampling,
and only a few show evidence of carnivore/scavenger activities. A complete
78 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
horse innominate possesses several puncture holes probably caused by a small
canid, but the rest of the bone is not damaged. A broken piece of a horse
mandibular ramus has what appears to be a chewed edge as well as scoring
marks on the surface (Binford 1981; Haynes 1983). However, the large
number of complete bones present in unit 11 and the rarity of elements
showing results of carnivore predation indicate that predators were involved
only to a very minor extent in the accumulation of the mammalian remains
found in the upper boulder layer.
Unit 15. Unit 15 is depauperate in megafauna and no complete limb
bones of megafaunal taxa were recovered. Larger vertebrates are represented
by long bone fragments, teeth, and footbones. As in unit 11, the few water-
worn bones found in unit 15 are very light in color. Evidence of bone
destruction by carnivores is most pronounced in this layer. A number of small
bone fragments have chipped edges, a type of damage frequently caused by
gnawing (Binford 1981), and several phalanges and distal calcanea possess
puncture holes almost certainly caused by mammalian carnivores. As the
microfaunal assemblage from unit 15 appears to be of scatological origin (Pratt
1986, 1989), it is likely that these small elements of megafaunal taxa were
ingested by a predator.
Results of the analysis of physical modification of Thomas Farm bones
are summarized in Table 3. Although elements from the various layers show
minor differences with regard to the types and degrees of modification, the
overall dominant pattern is one of similarity.
In all units analyzed, weathering features are minimal. Lack of
weathering may be explained in two ways. It is possible that most of the bones
were never exposed to the forces of subaerial modification. If remains are
buried shortly after the animal's death, as might occur in a catastrophic event
(Behrensmeyer 1978), or if the carcass is immediately deposited into an
aquatic system, the bones will not weather. Hill (1980) described the latter
situation for Recent hippopotamus carcasses in a lake in the Amboseli
National Park, Kenya. The floating carcasses were macerated by the lake
waters, and the bones were presumably deposited on the lake floor. The
alternate explanation is that Thomas Farm bones were exposed for some time
in a terrestrial environment, but climatic conditions were not of sufficient
severity to cause obvious weathering features. Behrensmeyer (1978)
demonstrated that elements in wooded or moist environments are less likely to
reach advanced weathering stages than bones exposed in open, arid areas.
Andrews and Cook (1985) reported that bones of a Recent cow exposed in a
temperate climate reached only stage 1 of weathering after 8 years, and Hill
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 79
Table 3. Types of surface modification of Thomas Farm fossils. Relative abundances of elements
possessing features are assessed using the following scale: VR, very rare (less than 1%); R, rare
(less than 10%); U, uncommon (less than 20%); C, common (30-50%); A, abundant (greater than
50%). Weathering stages range from 0 unweatheredd) to 5 (severely weathered), as described in
Behrensmeyer (1978). Degrees of water-wear are: Un, unmodified; Mn, minimal; Mo, moderate;
and Se, severe. For a complete description of stages of water-wear see text and Figure 17.
TYPE 5 6 7 8 11 15
Maximum Weathering 0 1(R) 1(U) 1(U) 1(U)
Trampling Evidence no no no no no no
Carnivore Chew/Bite no no VR VR VR C
Water-wear Stage Mn(C) Mn(C) Mn(U) Mn(U) Mn(VR) Mn(VR)
Mo(U) Mo(R) Mo(U)
(1980) noted that microclimatic factors and presence of protective vegetation
may act to retard the weathering process. For these reasons, it cannot be
definitely stated that Thomas Farm fossils showing no evidence of weathering
were not subject to exposure, nor is it possible to estimate the length of time
that bones showing evidence of weathering were exposed. The problem also
remains as to whether these bones were modified in situ, or were introduced
into the site from some other source. As discussed previously, it is likely that
some of the bones from unit 8 were weathered in place, and it appears that the
smaller weathered bones made their way into the deposit after laying out on
the surface elsewhere.
The lack of both trampling marks and surface features indicative of
actions of carnivores also suggest that the bones forming the assemblage were,
in general, inaccessible to these agents of modification. Trampling of bones
occurs in regions of "high traffic," such as trackways and watering holes
(Behrensmeyer and Boaz 1980; Conybeare and Haynes 1984; Andrews and
Cook 1985). Lack of trampling indicates that the sinkhole may have acted as a
trap rather than as a watering hole. Bone assemblages modified by carnivores
and scavengers fall into two general categories; kill sites and dens (Binford
1981), each of which may be recognized by particular taphonomic signatures.
Bones taken to dens tend to be more heavily gnawed than are those abandoned
at a kill site (Binford 1981). Haynes (1983) pointed out that the activities of
carnivores and scavengers may not be indicated by gnawing, but as a rule other
features, such as toothmarks or scrapes, will generally be present. The dearth
of bones showing any evidence of carnivore activity and the absence of gnawed
80 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
bones at Thomas Farm does not totally rule out the possibility that carcasses
were disturbed or modified by carnivores and scavengers, but does suggest that
the megavertebrate remains probably do not represent a den or lair
The presence of water-worn bone, particularly in the lower sedimentary
levels, indicates that these layers were associated with an aquatic environment
and that current velocities were sufficient to cause transport and abrasion of
the smaller elements. The fact that within the lower levels of the site the
assemblage is a mixture of larger, unworn elements and smaller bones
exhibiting features of water-wear suggests that the smaller elements may have
been transported to the site from elsewhere and were deposited in the sinkhole
as the current slowed. The lack of water-worn bone in the upper boulder layer,
unit 11, suggests that the influence of moving water was less important in the
later stages of site formation.
Species diversity and abundances of the vertebrate fauna represented at a
fossil locality can provide evidence of factors responsible for formation of the
bone assemblage. Only in rare cases, such as those involving catastrophic
death and immediate burial do assemblages reflect living abundances of the
living fauna. In most other instances the difference between fossil abundance
and diversity and the diversity of the living fauna is due to any number of
Thomas Farm Faunal Composition
Relative abundances of taxa were estimated by two methods: by
determining NISP (Badgley 1986a, 1986b), and by calculating the MNI per
taxon (Shotwell 1955; Voorhies 1969; Wolff 1973; Grayson 1978; Damuth
1982). These abundance values do not reflect the relative abundances of the
living taxa (Western 1980; Damuth 1982), and the values must be corrected if
relative abundances of the once-living populations are to be estimated for
paleoecological reconstruction. However, in taphonomic analysis, the
abundance values provide a useful means of comparing the faunal
compositions of the various sedimentary units. Relative abundances of
megafaunal taxa in units 5, 6, 7, 8, 11, and 15 are shown in Table 4 and Figures
19 and 20. Values have been converted to percentages to facilitate
comparisons between the sedimentary units.
The data demonstrate two major features of the Thomas Farm
megafauna. Table 4 shows that sedimentary units 5 through 8 have a higher
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 81
proportion of specimens of the aquatic taxa Alligator olseni and Pseudemys sp.
than do the upper layers 11 and 15. Numbers of individuals of alligators are
low because the specimens found (teeth and osteoderms) are not highly useful
in MNI calculations. The presence of aquatic forms in the lower layers
supports the sedimentological evidence that suggests these units were subject
to water-mediated deposition. However, the relative rarity of aquatic turtles,
both in abundance and number of species represented, contrasts dramatically
with their great abundances in other aquatic sites in Florida, such as the
McGeehee Farm (Rose and Weaver 1966; Jackson 1976), the Love Site (Webb
et al. 1981), and the Leisey Shell Pit (Hulbert and Morgan 1989). Other
components of the aquatic fauna well-represented at the above localities but
virtually absent from Thomas Farm are bony fish. Fish are represented in the
microfauna, but are extremely rare (Pratt 1986, 1989). The low representation
of fish suggests that the water source at Thomas Farm was in some way
isolated from aquatic environments such as surface streams and ponds, where
fish are generally common in Florida. Laury (1980) attributed the relative
rarity of fish at the Mammoth Springs site, South Dakota, to warmth of the
water. Although thermal springs do occur in Florida today (Rosenau et al.
1977), water temperatures are not high enough to exclude fish. It is possible
that high mineral content of the water might have caused conditions unsuitable
for some aquatic organisms, but under such conditions it would be expected
that all aquatic forms be absent. The presence not only of turtles and alligators
but also of several species of presumably aquatic or semi-aquatic amphibians
(Holman 1965, 1967; Estes 1963; Pratt 1986, 1989) suggests that this was not
the case. It is intriguing that all aquatic or semi-aquatic members of the
herpetofauna are capable of overland dispersal, and frequently make their way
from one water source to another by this method (R. Franz pers. comm.).
A second obvious feature of the Thomas Farm megafauna is the
numerical dominance of Parahippus leonensis in all but the uppermost
sedimentary unit (Table 4). The high abundance of this species may reflect the
animal's abundance in the area. Equids characteristically form herds, and
Hulbert (1984) has suggested that the social structure of P. leonensis may have
involved small herds. The site also may have attracted horses. Behrensmeyer
and Boaz (1980) report that during periods of drought in the Amboseli Basin,
Kenya, over 50% of the living mammalian megafauna in the vicinity of a water
source is composed of migrating zebra and wildebeest. Although the Thomas
Farm sinkhole may have served as a water source, to date there is no evidence
that watering sites were rare or that the region was subject to prolonged
The high abundances of P. leonensis also may be a function of the actions
of taphonomic agents such as weathering, prey selection by carnivores, stream
transport, or the trapping abilities of the site itself, which may cause a size bias
82 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
Table 4. Relative abundances of megafaunal species from the Thomas Farm locality. (A) Unit 5;
(B) Unit 6; (C) Unit 7; (D) Unit 8; (E) Unit it; and (F) Unit 15. See text for abbreviations.
% of % of
Taxon NISP total MNI total
A. UNIT 5
Chelonia (total) 27 6.3 5 19.3
Geochelone tedwhitei 13 3.0 2 7.7
Pseudevys sp. 14 3.3 3 11.6
Alligator olseni 180 41.7 3 11.6
Carnivora (total) 28 6.5 6 23.1
Amphicyon longiramus 2 0.5 1 3.8
Cynodestus ianmonensis or 15 3.5 2 7.7
small canid 1 0.2 1 3.8
small mustelid 9 2.1 1 3.8
Hemiicyon johnh/enryi 1 0.2 1 3.8
Artiodactyla (total) 30 6.9 6 23.1
Nothokemas floridanus or 5 1.2 2 7.7
Prosynthetoceras texanus 2 0.5 1 3.8
Blastomneryx floridanus 15 3.5 2 7.7
Machaerometry gilchristensis 8 1.8 1 3.8
Perissodactyla (total) 166 38.5 7 26.9
Parahippus leonensis 112 25.9 3 11.4
Archacohippus blackbergi 51 11.8 2 7.7
Anchitheritum clarenci 2 0.5 1 3.8
Floridaceras white 1 0.2 1 3.8
Total Unit 5 431 100.0 26 100.0
B. UNIT 6
Chelonia (total) 49 6.3 4 9.3
Geochelone tedwhitei 28 3.6 2 4.7
Pseudemys sp. 12 1.5 2 4.7
Unidentified 9 1.1 -
Alligator olseni 207 26.4 3 6.9
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 83
Table 4 Continued
% of % of
Taxon NISP total MNI total
Carnivora (total) 58 7.4 10 23.3
Amphicyon longiramus 7 0.9 2 4.7
Cynclos caroniavorus 2 0.3 1 2.3
Cynodesmus iamonensis or 36 4.5 2 4.7
small canid 2 0.3 1 2.3
Leptarctus ancipidens 4 0.5 2 4.7
small mustelid 5 0.6 1 2.3
Hemicyon johnhenryi 2 0.3 1 2.3
Artiodactyla (total) 83 10.6 9 20.9
Nothokemas floridanus or 9 1.1 2 4.7
Prosynthetoceras texanus 14 1.8 1 2.3
Blastometyx floridanus 38 4.9 3 6.9
Machaerometyx gilchristensis 21 2.7 2 4.7
Merycoidodon sp. 1 0.1 1 2.3
Pcrissodactyla (total) 387 49.4 17 39.5
Parahippus leonensis 253 32.3 9 20.9
Archaeohippus blackbergi 130 16.6 7 16.3
Floridaceras white 4 0.5 1 2.3
Total Unit 6 784 100.0 43 100.0
C. UNIT 7
Chclonia (total) 46 6.4 4 12.1
Geochelone tedwhitei 20 2.8 2 6.1
Pseudemys sp. 12 1.7 2 6.1
Unidentified 14 1.9
Alligator olseni 136 18.9 3 9.1
Carnivora (total) 42 5.8 6 18.1
Amphicyon longiramus 5 0.7 1 3.0
Cynodesmus iamonensis or 22 3.1 2 6.1
Leptarctus ancipidens 2 0.3 1 3.0
small mustelid 5 0.7 1 3.0
Hemicvon johnhemyi 1 0.1 1 3.0
Unidentified 7 1.0
84 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
Table 4 Continued
% of % of
Taxon NISP total MNI total
Nothokemas floridanus or
Total Unit 7
D. UNIT 8
Cynodesmus iamonensis or
Nothokemas floridanus or
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 85
Table 4 Continued
% of % of
Taxon NISP total MNI total
Pcrissodactyla (total) 356 62.9 12 35.3
Parahippus leonensis 276 48.7 7 20.6
Archaeohippus blackbergi 77 13.6 3 8.8
Anchitherium clarenci 2 0.4 1 2.9
Floridaceras white 1 0.2 1 2.9
Total Unit 8 566 100.0 34 100.0
E. UNIT 11
Chelonia (total) 18 2.7 2 5.4
Geochelone tedwhitei 18 2.7 2 5.4
Alligator olseni 13 1.9 2 5.4
Carnivora (total) 67 10.2 9 24.3
Amphicyon longiramus 7 1.1 1 2.7
Cynelos caroniavonis8U 2 0.3 1 2.7
Cynodesmus iamonensis or 25 3.8 3 8.1
small canid 1 0.2 1 2.7
Leptarctus ancipidens 2 0.3 1 2.7
small mustelid 7 1.1 1 2.7
Hemicyon johnhenryi 4 0.6 1 2.7
Unidentified 19 2.8
Artiodactyla (total) 79 12.0 9 24.3
Nothokemasfloridanus or 42 6.4 3 8.1
large camelid 1 0.2 1 2.7
Prosynthetoceras texanus 11 1.7 2 5.4
Blastomeryx floridanus 8 1.2 1 2.7
Machaeromeryx gilchristensis 12 1.8 1 2.7
Desmathyus sp. 2 0.3 1 2.7
Unidentified 3 0.2
Perissodactyla (total) 481 73.1 15 40.5
Parahippus leonensis 350 53.2 9 24.3
Archaeohippus blackbergi 108 16.4 4 10.8
Anchitherium clarenci 2 0.3 1 2.7
Unidentified equid 19 2.8
Floridaceras white 2 0.3 1 2.7
Total Unit 11
86 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
Table 4 Continued
% of % of
Taxon NISP total MNI total
F. UNIT 15
Cynodesmuis iamonensis or
Total Unit 15
in the fauna represented. Badgley (1986a) showed that size distributions
within fossil faunas from the Siwaliks can be attributed in part to taphonomic
factors. Table 5 shows body size distribution for the Thomas Farm
A common feature of bone assemblages that form attritionally is the
underrepresentation of small (less than 15 kg) taxa (Behrensmeyer et al. 1979;
Behrensmeyer and Boaz 1980). Due primarily to higher birth and death rates
(turnover), over a given period of time more small vertebrates die than larger
vertebrates (Western 1980; Damuth 1982). However, 1dss of remains of small
animals from a bone assemblage can be attributed to the greater effect of
taphonomic agents on small bones with the highest surface area-to-volume
ratios (Behrensmeyer et al. 1979; Behrensmeyer and Boaz 1980). Although
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 87
Table 5. Abundances of Thomas Farm mammals by size category. See Pratt (1986) for body
BODY MASS (KG)
1-15 16-100 101-200 > 200
LEVEL MNI % MNI % MNI % MNI %
5 7 37 9 47 2 11 1 5
6 11 30 23 64 1 3 1 3
7 7 26 15 63 1 4 2 7
8 9 30 18 60 2 7 1 3
11 11 29 24 63 2 5 1 3
15 11 74 4 26 0 0 0 0
remains of small mammals at Thomas Farm are more poorly represented than
those of mammals in the next largest size category, their abundance is
considerably higher than that observed by Badgley (1986a) in a fluvial deposit
from the Siwaliks in Pakistan. She proposed that low abundances of fossils in
the 1-15 kg size range were due to winnowing of lighter elements by fluvial
transport. The moderate representation of smaller vertebrates from Thomas
Farm suggests that transport by moving water did not serve to dramatically
decrease abundances of bones of smaller taxa.
Mammals in the 16-100 kg size category are the most highly represented
at Thomas Farm, in part due to the fact that P. leonensis falls within this size
range (Hulbert 1984). The high abundance of this form and others of similar
size, and the relative rarity of larger mammals could be indicative of a
predator-selected assemblage. Rosenzweig (1966) demonstrated the
correlation between predator size and maximum prey size. Based on this
relationship, the most likely predators of P. leonensis were the large carnivores
Hemnicyon johnhenryi and Amphicyon longiramus (Pratt 1986). Although this
hypothesis will be examined further in a later section, based on the rarity of
bone showing evidence of carnivore modification, it is unlikely that carnivores
were the sole or major contributors to the bone accumulation.
It is also possible that the site itself acted as a size-selective trap. The low
representation of larger members of the fauna (over 100 kg) may indicate that
they were too large to have gained access to the sinkhole, although this
hypothesis is not supported by evidence on skeletal dispersal (see following
section). The significance of the composition of the Thomas Farm megafauna
awaits future discoveries of other early Miocene localities from Florida with
88 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
which this fauna may be compared. The high representation of P. leonensis
and lower representations of larger mammals cannot be fully explained at this
Although it is not possible to compare the Thomas Farm fauna with other
Florida faunas of similar age, comparison of faunal composition from the
various layers of the site itself can provide evidence both of changes in
environment and of taphonomic events that occurred during the course of
deposition. Comparisons of faunal abundances in the lower sand and clay units
show that relative abundances of taxa in unit 5 are different from those of units
6, 7, and 8 (Figs. 19 and 20). Alligator specimens are higher in relative
abundance and equid remains are lower in relative abundance in unit 5
compared to units 6 though 8. Units 6 through 11 appear generally similar in
terms of relative NISP per taxon, while values for unit 15 are clearly different
from those of other layers (Fig. 19). If relative NISP per taxon from each layer
are compared statistically with those of other layers using the R x C test of
independence using the G-test (Sokal and Rohlf 1981), relative abundances in
each unit, except 6 and 7, are significantly different from those in the over-lying
and under-lying units at thep = 0.01 level (Table 6).
If relative abundances in terms of MNI per taxon per sedimentary level
are compared (Fig. 20), the differences in relative abundances of taxa between
the units are minimized. The main reason for the apparent increase in
similarity of different levels if MNI rather than NISP are used is caused by
species that possess large numbers of elements that are not diagnostic in MNI
calculations (for example, alligator). Application of a R x C analysis of
independence using the G-test (Sokal and Rohlf 1981) shows that in terms of
relative MNI per order of megafaunal taxa, no sedimentary unit, with the
exception of level 15, is significantly different from the one above or below it at
thep = .01 level (Table 6).
Badgley (1986a, 1986b) discussed the relative merits of NISP and MNI
calculations, and the circumstances under which each should be employed. It
is obvious that for the Thomas Farm fauna, NISP overestimates the differences
between levels, and MNI underestimates these differences. Nevertheless, both
methods indicate that in terms of fauna preserved, the factors governing
formation of the bone assemblage were fairly constant until deposition of the
final layer, unit 15. The fact that the same species are represented in all units
of the site indicates that neither a dramatic change in climate, environment,
nor the vertebrate fauna took place during the period of time the site was
sampling the fauna.
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA
Figure 19. Relative abundances of identifiable specimens per order of megafaunal taxa. (A)
Unit 5. (B) Unit 6. (C) Unit 7. (D) Unit 8. (E) Unit 11. (F) Unit 15. Abbreviations: AR,
Artiodactyla; CA, Carnivora; CH, Chelonia; CR, Crocodilia; PE, Perissodactyla.
90 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
Figure 20. Relative abundances of minimum numbers of individuals (MNI) per order of
megafaunal taxa. (A) Unit 5. (B) Unit 6. (C) Unit 7. (D) Unit 8. (E) Unit 11. (F) Unit 15.
See Figure 19 for list of abbreviations.
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 91
Table 6. Statistical comparison of relative abundances of megafaunal taxa from different
sedimentary units at Thomas Farm, based on (a) NISP and (b) MNI. Reported are calculated G
values from the R x C test for association (Sokal and Rohlf 1981:599). Abundance values are
considered significantly different at the 0.01 level if G > 13.28 at 4 degrees of freedom (*). For
explanation of test and abbreviations see text.
Unit 6 7 8 11 15
5 (a) *27.5 -
6 (a) 10.3
(b) 0.2 -
7 (a) *21.5
8 (a) '20.1
11 (a) *49.3
Skeletal Articulation or Association
The extent of articulation of vertebrate skeletal remains provides insight
into the method by which bones have accumulated and the possible span of
time involved in the formation of the assemblage. Completely articulated
skeletons are rare in the fossil record, and generally indicate that only a short
time had elapsed between death and burial, such that the remains were not
greatly modified or dispersed by taphonomic events. If articulated remains are
not present, then the abundances of bones and their positions within the
sediment have the potential to provide clues concerning the taphonomic factors
important in forming and modifying the assemblage.
Thomas Farm Skeletal Asssociations
Figures 21A and 21B, depicting locations of fossil bones within two
representative square meters of Thomas Farm, clearly show that megafaunal
remains are not articulated. Bones of Parahippus leonensis are extremely
abundant, and it is assumed that at least some of these elements were from
associated skeletons. However, with the exception of a femur and tibia and an
astragalus and calcaneum, no definite articular matches could be made.
Efforts to match postcranial elements of a single individual were hampered not
only by the large number of P. leonensis elements found in every square
excavated, but also by the fact that many bones were deformed or crushed by
post-depositional compaction. Bones of taxa that are relatively rare, such as
92 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
Figure 21. Bone plots of representative square meters at Thomas Farm. (A) 10N x 10E, M1l,
unit 11. (B) 15N x 10E, M20, unit 6.
PRATT: TAPIIONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 93
rhinoceros and amphicyonid, do seem to be concentrated in areal extent within
certain layers, suggesting the possibility that these elements may be parts of the
same skeleton. However, to date no strong evidence (such as an articular
match between elements) has been found to suggest that these elements were
all derived from one individual. Behrensmeyer and Hill (1984) have suggested
that even in the absence of articulation, associations of skeletal parts may be
indicated by the similar, high relative abundances of elements that disarticulate
late in the disarticulation sequence. However, as sequences of disarticulation
vary considerably depending on the type of animal and the environment (arid,
moist, aquatic) in which decomposition occurs, this method cannot be used to
indicate association of P. leonensis.
Among the lower vertebrates, evidence of skeletal association is rare.
Turtle and tortoise shells are almost always disarticulated. With the exception
of two portions of a broken pleural bone found several meters apart within the
same sedimentary layer, no matches were found between disarticulated parts
of chelonian shells. Kenneth Dodd (pers. comm.) has noted that shells of
aquatic turtles disarticulate completely within 8 months when exposed in
modern Florida terrestrial environments, and presumably less time is required
for disarticulation under aquatic conditions, assuming the shells are not buried.
The conclusion drawn from the absence of articulated skeletons is that
burial of the megafaunal remains at Thomas Farm was not taking place under
conditions that favor the preservation of complete skeletons. Therefore, events
related to disarticulation of skeletons and modification of the bone
concentration were undoubtedly responsible for biasing bone representations
within the assemblage.
Representation assessments of skeletal elements provide an estimate of
the relative abundance of each element type recovered from a fossil deposit.
Given that taphonomic factors may bias a bone assemblage by leading to
increases or deceases in bone abundances of the original death assemblage, it
is hypothetically possible to recognize causative taphonomic agents by patterns
of skeletal element abundance. Any combination of the events listed in Figure
16 may have been responsible for the final assemblage, and therefore
representation patterns of bones may be correspondingly complex
(Behrensmeyer 1982; 1988). However, if one taphonomic agent was of
overriding importance, then the composition of the fossil bone assemblage may
record this fact. For this reason, comparisons of bone relative abundances in a
fossil assemblage with those in Recent bone accumulations of known origin can
serve to indicate possible factors involved in modifying the fossil assemblage.
94 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
In this study, the Thomas Farm assemblage was compared with modern bone
accumulations, listed below, that were formed or modified by taphonomic
agents similar to those proposed to have been involved in the formation of the
Thomas Farm deposit (Fig. 16):
1. Attritional death assemblage: Behrensmeyer and Boaz
(1980) recorded bone abundances for a large collection of
carcasses in the Amboseli National Park, Kenya. The
assemblage, formed primarily of mammals dying of natural
causes, was modified to some extent by carnivores and
scavengers, trampling, and weathering. This impressive body
of data provides a representative example of the ways in
which a death assemblage, exposed on the ground surface,
can be modified in terms of bone representation.
2. Carnivore kill site: Areas where animals congregate, such as
watering holes, are often sites where predators stalk and kill
prey. After feeding, carnivores may abandon bones at these
locations (Binford 1981). Although bone representations vary
due to a large number of factors such as type of predator, size
and age of prey, and time of year (Binford 1981;
Blumenschine 1986), a kill site assemblage often possesses a
diagnostic pattern of element representation (Behrensmeyer
and Boaz 1980; Binford 1981; Blumenschine 1986). Records
of Recent bone assemblages formed in this manner include a
study by Binford (1981) on remains of caribou killed by
wolves, and Hill's data (from Binford 1981) on skeletal
abundances of topi killed by a variety of predators. Although
other studies on bone assemblages formed by carnivores are
known (e.g. Blumenschine 1986), the two examples cited here
are considered most suitable for several reasons. Thomas
Farm carnivores capable of killing large prey, such as
Parahippus leonensis, include two species of canid, two
species of amphicyonid, and an ursid. Several of these taxa
are very wolf-like in their morphology (Pratt 1986). Although
caribou are considerably larger than P. leonensis, topi, with an
average body mass of about 80 kg (Binford 1981), are within
the proposed size range of the fossil horse species (Hulbert
1984). Finally, these Recent samples provide quantitative
bone counts against which fossil bone abundances may be
3. Carnivore den or lair: Predators and scavengers may remove
portions of a carcass from the scene of a kill and transport
them to their living quarters. Accumulations of bones formed
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 95
in this way by wolves (Binford 1981) and hyaenas
(Behrensmeyer and Boaz 1980) are compared with the fossil
4. Moving water: Bones may be added to or removed from an
accumulation by the action of moving water. An assemblage
consisting of elements that may have either resisted transport
or been carried from elsewhere, may be recognized by the
numbers and types of bones present. Voorhies (1969)
established transport groups for bones of the mammalian
skeleton by demonstrating that certain bone types behave in a
similar fashion in running water. Behrensmeyer (1975)
showed that the ability of a bone to be moved or resist
transport can be assessed quantitatively by treating the
element as a sedimentary particle. Hanson (1980) suggested
that the role of moving water in forming and modifying an
assemblage may be evaluated by comparing relative
abundances of bones from each of the various transport
groups. A generalized transport sequence for bones of
Parahippus leonensis based on previous studies (Voorhies
1969; Behrensmeyer 1975; Korth 1979; Pratt 1979), stream
experiments conducted in association with this study, and
estimations of quartz grain equivalents, is shown in Table 7.
Element Representation at Thomas Farm
Relative abundances of skeletal elements may be assessed by two
different methods. The first, skeletal-part frequency (Badgley 1986a), refers to
the percentage of the total bone assemblage represented by each element type
(Table 8). Comparison with the percentage frequency of the same element in
a complete skeleton indicates which fossil bones are present in higher or lower
proportions than expected in an unmodified death assemblage. Obviously, the
greatest differences in terms of percentage representation between a fossil
assemblage and a complete skeleton are likely to be seen in elements such as
vertebrae, teeth, and phalanges that possess the highest skeletal-part frequency
values in a living mammal. Conversely, lowest differences between fossil and
recent assemblages in terms of skeletal-part percentages tend to be elements
such as skulls that are represented by only one element in the complete
An alternative method of quantifying skeletal element abundance involves
dividing the number of each fossil element type present by the number
expected based on the MNI (Wolff 1963; Voorhies 1969; Korth 1979), to arrive
at a percentage value termed the relative representation (Table 8). In
actuality, this number reflects the abundance of each bone type relative to the
96 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
most abundant element. If factors have acted to cause the enrichment of one
element in the assemblage, then representation values of other elements may
appear deceptively low. Although each method of calculating bone
abundances has certain shortcomings, use of the two methods together
provides the most informative picture of bone representation patterns. In the
comparison of abundances of fossil bones with those in a modern
accumulation, differences in relative abundances of individual bone types are
to be expected, and should not be weighted too heavily. Rather, similarities or
differences in overall abundance patterns provide more information in
comparing Recent and fossil assemblages. The analysis of element
preservation at the Thomas Farm locality was restricted to elements of
Parahippus leonensis for several reasons. P. leonensis is the most abundant
member of the larger (over 5 kg) mammalian fauna at the locality, and its
easily identifiable remains are found in all of the major units. Because it is
difficult to assign many of the postcranial elements of the Thomas Farm
carnivores and artiodactyls to species, MNI calculations for each species of
these groups must be based on cranial material, and relative representations of
elements cannot be determined. In addition, the size range of P. leonensis is
small (Hulbert 1984), so that in consideration of fluvial transport groups,
effects of size variations within element types need not be considered.
Unit 5. Table 8 lists skeletal part frequencies and relative representations
of Parahippus leonensis elements recovered from unit 5. Differences in
frequency values between unit 5 elements and those of a complete skeleton are
Table 7. Predicted transport groups for elements of Parahippus leonensis. Arrows indicate
elements that due to shape, or differences in wet and dry densities, may belong to more than one
dispersal group. Range of calculated quartz grain equivalent diameters listed in parentheses. See
text for discussion of quartz grain equivalents.
(<3 mm) (>3 mm) (>10 mm) (>20 mm)
I I/II II II/III & III
small podials cheekteeth humerus dentary
phalanges incisors radius <---skull
vertebrae astragalus 4---femur
ribs calcaneum tibia
1. metapodial prox. ulna 4--pelvis---
*---large podials m. metapodial
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 97
Table 8. Relative abundances of elements of Parahippus leonensis from major sedimentary units
of the Thomas Farm locality. Abbreviations: No., number of each element; RI. rep, percent
relative representation; S-PF, skeletal-part frequency; Ectocun., ectocuneiform; Entocun.,
entocuneiform; Mesoentocun., mesoentocuneiform; D., distal; M., medial; Pr., proximal. See text
for discussion of calculations.
UNITS MNI=3 UNIT 6 MNI=9
ELEMENT No. S-PF Rl. rep No. S-PF RI. Rep
Pr. Phalanx III
M. Phalanx III
D. Phalanx III
98 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
Table 8 Continued
UNIT 7 MNI= 5 UNIT 8 MNI= 7
ELEMENT No. S-PF RI. rep No. S-PF RI. Rep
Pr. Phalanx III
M. Phalanx III
I). Phalanx III
65.0 22 0.080
75.0 8 0.029
30.0 4 0.014
47.5 15 0.054
30.0 4 0.014
2.5 3 0.011
10.5 35 0.127
PRATIT: TAPIONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 99
Table 8 Continued
UNIT 6-8 MNI = 17 UNIT 11 MNI= 9
ELEMENT No. S-PF RI. rep No. S-PF Rl. Rep
Pr. Phalanx III
M. Phalanx III
D. Phalanx III
100 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
Table 8 Continued
UNIT 15 MNI = 1 SKELETON
ELEMENT No. S-PF RI. rep No. S-PF
Pr. Phalanx III
M. Phalanx III
D. Phalanx III
TOTAL 25 14
PI'I'VA : TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 101
Table 9. Values of rs for Spearman rank-coefficient test (Sokal and Rolf 1981). Number of
ranks in sample indicated by n.
A. Comparison of skeletal-part frequency ranks of Thomas Farm bone assemblages from
different sedimentary levels to one another and to a complete skeleton.
5 6 7 8 6-8 11 15
n 16 16 16 16 16 16 16
Skeleton **.878 **.829 **.915 *.756 **.900 **.787 *.776
Unit 5 **.807 **.946 **.779 **.864 **.680
Unit 6 *.879 **.898 **.849 *.607
Unit 7 *.851 **.926 *.629
Unit 8 .830 *.537
Unit 11 *.488
B. Skeletal-part frequency ranks of Thomas Farm bone assemblages and a complete skeleton
compared to Recent bone assemblages. (a) wolf kill site (Binford 1981 Table 5.01, col. 25); (b)
wolf den (Binford 1981, Table 5.01, col. 27); (c) surface assemblage (Behrensmeyer and Boaz
1980, Table 5.6); (d) hyaena den (Behrensmeyer and Boaz 1980, TabTe 5.6).
(a) (b) (c) (d)
n 10 10 8 8
Skeleton .491 .406 *.780 .434
Unit 5 .558 *.760 .428 .381
Units 6-8 .433 .545 .404 .333
Unit 11 *.645 *.685 .571 .667
Unit 15 .118 .254 .470 .339
C. Relative representation values of Thomas Farm bone assemblages compared to (a) wolf kill
site (Binford 1981, Table 5.01 col. 26); (b) wolf den (Binford 1981, Table 5.01, col. 28), and (c)
topi bone assemblage (Hill's data, from Binford 1981, Table 5.02, col. 2).
Unit 5 .257 .330 .217
Units 6-8 .090 .078 -.213
Unit 11 .527 .527 .062
Unit 15 .106 -.021 -.250
**significant at 0.01 level
*significant at 0.05 level
102 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
demonstrated in Figure 22A. Placement of each element type from left to
right along the X axis of this graph (and subsequent graphs, Figs. 22B-D, 23,
and 24) corresponds to its equivalent quartz grain diameter, from largest to
smallest, and therefore to the element's predicted tranportability in moving
The unit 5 bone assemblage is characterized by higher proportions of
lateral metapodials and proximal and medial phalanges, and lower
representations of vertebrae and distal phalanges than those of a complete
skeleton (Fig. 22A). Relative representation values are based on dentaries
(Table 8 and Fig. 23A) and with the exception of the astragalus and calcaneum,
it is the lightest, smallest elements such as vertebrae, patellae, and distal
phalanges that show the lowest representation. Statistical comparison of unit 5
skeletal-part frequencies with those of a complete skeleton using the Spearman
rank-coefficient test, which provides a pairwise comparison of abundance
ranks (Sokal and Rolf 1980), shows that the pattern of abundance is highly
correlated with that found in a living animal (Table 9). Unit 5 skeletal-part
frequencies show no significant correlations with kill sites (Binford 1981:Table
5.01, col. 25; Table 5.02, col. 2) or with an attritional assemblage
(Behrensmeyer and Boaz 1980; Table 9). The Recent bone assemblages are
similar to one another, and somewhat different from the unit 5 assemblage
(Fig. 24), in the high relative representations of skulls, pelvic girdle elements,
and vertebrae and low relative abundances of podials and phalanges (Binford
1981; Behrensmeyer and Boaz 1980). Blumenschine (1986) reported that some
large carnivores may selectively remove or chew on distal appendages, thereby
destroying podials and phalanges. Behrensmeyer and Boaz (1980) also
attributed low representations of these elements in the surface assemblage to
the actions of carnivores and scavengers, but noted that trampling can cause
rapid burial of podials and other small bones. The fairly high representations
of these bones in unit 5 suggests that the assemblage does not represent a
surface accumulation formed or modified by carnivores and scavengers.
Comparison of skeletal-part frequencies in unit 5 with those of carnivore
den assemblages (Binford 1981; Behrensmeyer and Boaz 1980) shows a
correlation in frequency rank between the fossil assemblage and the den
assemblage formed by wolves (Binford 1981: Table 5.01, col. 27). The
assemblages are similar in the higher-than-expected proportions of limbs and
podials and the low abundances of vertebrae (Fig. 22A, 24C). In terms of
relative representations, the patterns of abundance appear similar (Fig. 23A,
Fig. 24D), although the correlation between the two assemblages is not
significant (Table 9C).
Moderate to high relative representation of all but the lightest elements in
the unit 5 fossil assemblage suggests that the bones were not transported as a
group by moving water. The presence of elements having a wide range of
quartz grain equivalent diameters from extremely large to small would require
PRA1T: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA
0 0 n n0
I 0I ooo-lu n
L 00 10
I 20 0 0 202
Figure 22. Graphic depiction of differences between skeletal-part frequencies of Parahippius
Iconensis remains from Thomas Farm and skeletal-part frequencies of a complete P. leonensis
skeleton [(Fossil SPF) (Complete SPF)]. (A) Unit 5. (B) Units 6-8 combined. (C) Unit 11.
(D) Unit 15. Abbreviations: A/C, astragalus and calcaneum; Chth, cheekteeth; Den, dentary; Hf,
distal phalanx (hoof); Inc, incisor; Lb, limbs (humerus, radius, femur, tibia); LM, lateral
metapodials; Max, maxilla; MM, medial metapodials; Pat, patella; Phi, proximal and medial
phalanges; Pod, podials except astragalus and calcaneum; S/P, scapula and pelvis; Uln, proximal
ulna; Vert, vertebrae.
ssi sy ^? ?-
Figure 23. Relative representations of skeletal elements of Parahippus leonensis. (A) Unit 5.
(13) Units 6-8 combined. (C) Unit 11. (D) Unit 15. See Figure 22 for list of abbreviations.
104 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
El F -1 2 a>57
_ o -
" 02 4 02
*^ 2 \ 2
Figure 24. (A and C) Differences between skeletal-part frequencies of bones abandoned by
wolves at a kill site (A), or brought to a wolf den site (C) and skeletal part frequencies of a
complete artiodactyl skeleton. (B and D) Relative representations of bones abandoned by wolves
at a kill site (B), or brought to a wolf den (D). All data from Binford (1981). See Figure 22 for
list of abbreviations.
that the agent of transport be a stream of high competence, capable of carrying
the entire assemblage from elsewhere. The lack of sedimentological evidence
indicating fluvial transport and the absence of water-wear on elements
belonging to transport groups III and II suggest that the majority of the
elements were not brought into the sinkhole in this fashion. However, the low
representations of the lightest elements do indicate that the bone assemblage
in the sink was modified by the action of moving water, and the lightest
elements may have been removed by winnowing. Figure 23A shows that
transport groups III and II elements possess the highest mean relative
representation values, followed by group I/II elements, and finally group I
elements. This pattern is similar to the 1B model of Hanson (1980, Fig. 9.3).
Units 6-8. Units 6, 7, and 8 are similar not only in lithology but in faunal
content and skeletal representation as well (Tables 4, 8). It has been suggested
in previous sections that the bone assemblages in these units were formed
under very similar circumstances. For these reasons, the bone counts from
units 6 through 8 were combined to form a large sample (Table 8). The
calculated MNI decreases from 21, if the MNI for each layer are summed, to
x avI NO
PRATF': TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 105
17 if the 3 layers are treated as one. However, lumping the sample from these
three sedimentary units does not affect the order of representation of elements
from most to least abundant.
Skeletal-part frequency ranks of combined units 6-8 elements are highly
correlated with those of a complete skeleton (Table 9A, Fig. 22B). The bone
assemblage is similar to that of unit 5 and unlike Recent kill site accumulations
modified by carnivores (Table 9, Fig. 24A) in higher-than-expected frequencies
of podials and phalanges, and lower-than-expected frequencies of vertebrae.
Bone frequencies also are not highly correlated with those of carnivore den
sites (Table 9B, Fig. 24). Units 6-8 differ from the lower sedimentary level in
the higher representations of elements, such as astragali, that are classified as
transport group I/II elements (Fig. 23B). The apparent enrichment of these
elements in the accumulation may be attributed to transport by moving water.
It is possible that water moving through underground drainage networks
transported these lighter elements from another more distant bone source,
perhaps another sinkhole, and added to them to the assemblage within the
Thomas Farm sinkhole. This interpretation is supported by the observation
that the majority of abraded, water-worn bones within units 6, 7, and 8 are
those possessing quartz grain equivalent diameters of 6 mm or less. Slowing of
current velocity within the sinkhole would cause deposition of all but the
lightest group I elements, which were then carried out of the sink by the
somewhat reduced current.
Unit 11. The assemblage of Parahippus leonensis elements in unit 11 is
the most problematical, because not only do skeletal-part frequency ranks
resemble those in a complete skeleton, they are also similar to bone
abundances of a carnivore kill site and a den site (Table 9A & B, Figs. 22C,
24). In terms of relative representations of elements, no significant
correlations with either type of carnivore-modified accumulation are indicated
(Table 9C). The pattern of relative representation values more closely
resembles that of a carnivore den site, in which the largest densest elements,
such as dentaries and limbs, are represented in higher abundances than smaller
and less dense elements (Figs. 23C, 24D). This pattern also resembles the type
2A model of Hanson (1980, Fig. 9.3) of a lag deposit; however, there is less
evidence of aquatic influence in unit 11 than in the lower clay and sand layers.
The fact that the layer is composed of large limestone boulders might lead to
the assumption that the remains were initially deposited in a cave by
scavengers, and fell into the site as the cave walls and floor collapsed.
However, the lack of bone showing any evidence of destruction by gnawing
suggests that this interpretation is not the most acceptable. The small sample
size of cranial material from unit 11 does not allow testing of the hypothesis
that the horses were killed in a catastrophic event as the wall fell (Hulbert pers.
comm.), but the presence of bones throughout all levels of the boulder layer
106 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
suggests that neither the boulders nor the bones were deposited
instantaneously. The composition of the sediment itself may have been the
dominant factor in determining the composition of the bone assemblage.
Losses of the lighter, less dense elements may have been due to crushing and
compaction of these bones by large boulders as the layer was deposited, as well
as by subsequent post-depositional compaction.
Unit 15. The number of equid specimens in unit 15 (n =63) is too low to
allow for statistical analysis of skeletal percentage preservation. Megafaunal
remains within unit 15 are dominated by incisors, phalanges, podials, and
vertebrae (Fig. 22D, 23D). Limb bones, with the exception of small fragments
or epiphyses, are rare. Based on high abundances of microfaunal elements, it
is probable that remains in this layer represent a coprocoenosis formed by
small canids (Pratt 1986, 1989).
Skeletal-Part Preservation: Summary
Although skeletal-part compositions of the bone accumulations in several
sedimentary layers do show correlations with either carnivore kill sites or dens,
bone abundances in the various layers are most similar to one another and to
bone abundances in a complete skeleton (Table 9). Therefore, it is logical to
conclude that the original death assemblages were relatively unmodified by
weathering, scavenging, or transport. If the site acted as a trap, the above
conclusion would imply that the carcasses were rapidly made innaccesible to
modification by these agents. However, the lack of skeletal articulation
indicates that sufficient time must have elapsed prior to burial to allow for
complete disarticulation. Agenbroad (1984) suggested that disarticulated
mammoth remains in Hot Springs, South Dakota, were deposited from rocky
talus slopes near the base of the sink, or from carcasses floating in the water of
the spring. It has been shown that carcasses submerged in water may float,
and as ligaments and tendons decompose, bones fall to the bottom of the water
body (Dodson 1973; Hill 1980). This scenario is a plausible one for explaining
the source of many of the Thomas Farm elements. The similarity of several of
the Thomas Farm bone assemblages (unit 7 and 11) to carnivore den deposits
is more likely due to the fact that several very different types of taphonomic
agents all may cause loss or destruction of the lightest, smallest elements
(Hanson 1980; Behrensmeyer 1982; Andrews and Nesbit Evans 1983). The
lack of gnawed or chewed bone suggests that moving water, rather than
carnivores, was responsible for the removal of group I elements. This
conclusion is consistent with previous observations that have suggested the
presence of an underground water source associated with the sink. The
hypothesis that moving water played a role in formation of the Thomas Farm
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 107
assemblage can be further tested by consideration of bone orientations,
discussed in the following section.
Recent experimental work has shown that moving water is not only
capable of transporting skeletal elements, but also of influencing their
orientations within the sediment (Voorhies 1969; Saunders 1977; Hunt 1978;
Shipman 1981, Maas 1985). Direction of flow and, to some degree, current
velocity may be inferred by recording directional and positional orientations of
bones in a fossil assemblage. Orientations of the long axes of limb bones are
the most reliable indicators of current direction (Voorhies 1969). The degree
of dip of a bone within a sedimentary layer measured along the element's long
axis (plunge) may also be related to paleocurrent strength (Voorhies 1969).
The shape of a bone influences its preferred orientation or stable resting
position in the presence of moving water. A fluvial system with a current
velocity sufficient to cause movement or transport of long bones (required
velocities vary depending on the size of the bones and physical characteristics
of the stream) is recognized by a rose diagram similar to that illustrated in
Figure 25A (Voorhies 1969; Shipman 1981). Long bones are positioned with
their long axes parallel to the prevailing current direction and their less dense
articular ends facing downstream (Voorhies 1969). Shorter bones, or
cylindrical elements with ends of more or less uniform density, tend to align
perpendicular to the current. Figure 25B shows a proposed orientation pattern
caused by sheetwash down an incline. Bones are oriented with the long axes
parallel to maximum dip of the slope, and on either side of maximum dip. In
Figure 25C, a uniform or non-preferred orientation pattern is illustrated, which
in most circumstances indicates absence of a strong, unidirectional current. A
uniform orientation pattern also may be produced by multidirectional flow
such as that produced by an artesian spring (Saunders 1977), or by flow that
periodically changes direction. It should be remembered that the rose
diagrams pictured in Figure 25 represent orientations of hypothetical bone
assemblages, and variation from these patterns does not necessarily signify the
absence of water-related orientation. Rose diagrams based on actual data are
usually more complex than predictive models, so all aspects of bone position,
not just bearing, should be considered in analysis of a fossil bone assemblage.
Nevertheless, if bone orientation has been influenced by a non-random
directional factor, then the distribution of bone orientation direction tends to
be significantly different from a distribution in which no preferred orientation
pattern is evident (Shipman 1981). The presence of a non-uniform orientation
pattern can be determined by use of the Chi-square goodness of fit test (Sokal
108 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
and Rohlf 1981). Bone orientations are considered significantly different from
non-preferred or uniform at the p=0.01 level.
Thomas Farm Bone Orientations
Entire Site. The rose diagram showing orientations of bones collected
from all fossiliferous levels of the site (Fig. 26A) does not resemble that of a
fluvial environment (Fig. 25A). The pattern of bone orientations (arranged in
100 increments) is not significantly different from that expected if the bones
were distributed uniformly (Table 10); however, there are significantly more
bones aligned toward the northeast/southwest quadrants (62%) as opposed to
the southeast/southwest quadrants (Table 10).
By considering bone plunge as well as bearing, a stereonet may be
constructed in which each bone is indicated by a point whose coordinates are
bearing direction and degree of plunge (Voorhies 1969; Shipman 1981).
Figure 26B shows the stereonet of bearings and plunges for all elements on
which these readings were taken (n = 516). The predominance of points in the
southwest quadrant does not reflect the actual bone plunge direction, but the
dip of the sediments, which tend to obscure the plunge of the bone relative to
the bed. In order to determine the true plunge of the bone relative to the
sediment layer, its degree of plunge must be corrected for the dip of the bed in
which it is found (Shipman 1981:78). Once this step is accomplished, a
stereonet constructed using the corrected values shows that a number of bones
actually plunge northeast within the sediments (Fig. 26C). This pattern is
unlike that seen in a fluvial deposit (Voorhies 1969). Voorhies suggested that
in a fluvial environment the majority of bones behave like boulders by
imbricating and dipping upstream. In this case, the majority of bones should be
dipping in one direction. No such trend is evident in the corrected stereonet.
Significantly greater numbers of bones plunge either northeast or southwest as
opposed to northwest or southeast, but there is no significant difference in the
numbers of bones plunging northeast as opposed to southwest (Table 10).
Degrees of plunge are shallow, and in most cases do no exceed 15.
Elimination from the rose diagram and stereonet of elements shown
experimentally (Voorhies 1969; Korth 1979; Pratt 1979) to have preferred
orientations in directions other than parallel to the prevailing current
scapulaee, carpals, phalanges, pelves, patellae, tarsals, vertebrae) results in the
rose diagram and corrected stereonet shown respectively in Figures 27A and B
(n = 337). The significant northeast/southwest trend of the rose diagram is
evident, with over 70% of the elements' long axes aligned with this compass
direction, as opposed to 30% aligned to the southeast/northwest (Table 10).
However, the orientation pattern (in 100 increments) is not significantly
different from a uniform, non-preferred orientation distribution (Table 10).
The stereonet of long bone bearings and plunges corrected for the dip of the
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 109
Table 10. Chi-square values for compass orientations of skeletal elements. Orientation patterns
considered significantly different from a uniform pattern if P<0.01. Abbreviations: D.F.,
degrees of freedom; Met., metapodials; N, number in sample; P, probability value; Prox..
CATEGORY N D.F. X2 P
All Elements/Entire Site 913 17 9.24 0.900
All Elements (by quadrant) 913 3 53.06 <0.005
Long Bones/Entire Site 337 17 8.70 0.950
Long Bones (by quadrant) 337 3 54.04 <0.005
Direction of Plunge (NE vs. SW) 187 1 1.18 0.600
All Elements/Unit 5 99 17 33.76 0.010
Long Bones/Unit 6 71 17 59.57 <0.005
Long Bones/Unit 7 60 17 76.63 <0.005
Long Bones/Unit 8 72 17 45.70 <0.005
Long Bones/Unit 11 96 17 32.40 0.125
umceri/Entire Site 46 17 40.83 <0.005
Tibiae/Entire Site 62 17 87.38 <0.005
Medial Met./Entire Site 86 17 34.31 0.010
Lateral Met./Entire Site 84 17 33.07 0.125
Prox. Phalanges/Entire Site 69 17 34.11 0.010
Figure 25. Hypothetical bone orientation patterns. (A) Fluvial system with prevalent current
direction NE/SW. (B) Sheetwash down an incline showing bone orientations following maximum
SW slope of beds. (C) Uniform or non-preferred orientation pattern (from Shipman 1981).
110 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
10% of bones
Figure 26. Orientation patterns of megafaunal elements at Thomas Farm. (A) Rose diagram
showing orientations of all bones measured at Thomas Farm. (B) Stereonet, not corrected for
dip of beds, showing bearings and plunges of all bones measured. (C) Stereonet, corrected for
dip of beds. showing corrected bearings and plunges of all bones measured.
10% of bones
Figure 27. Orientation patterns of long bones. (A) Rose diagram showing orientations of all
long bones measured. (B) Stereonet, corrected for dip of beds, showing bearing and plunges of
all long bones measured.
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 111
beds shows that the majority of the bones plunge northeast or southwest, but
there is no significant difference in the numbers of bones plunging northeast as
opposed to southwest (Table 10). The high uniformity in direction and degree
of bone plunge and the relatively low angles of plunge indicate that it is
unlikely that bones were trampled into the soft sediment. Bones trampled in
soft sediment reportedly exhibit high degrees of plunge (Andrews et al. 1981).
Treating the Thomas Farm site as a single homogeneous deposit results
in the misleading conclusion that the bones do not show a significant
orientation pattern, although the majority of limb elements do show bearing
directions to the northeast/southwest. If each major sedimentary unit is
considered separately, more definite orientation patterns emerge.
Unit 5. Fossils on which it was possible to obtain orientation
measurements were not very numerous in unit 5 (n = 99). The rose diagram
of all bone bearings in unit 5 shows a northeast/southwest trend in
orientations. The greatest numbers of bones are aligned from from 40-70/220-
250 degrees east of North, with over 30% of the bones located within this 300
span. The distribution of bone directions in Unit 5 is significantly different
from uniform (Table 10). The sample size of limb bones (long bones) from
unit 5 is too small to test statistically for evidence of preferred bearing
direction (n = 22). Over 25% of the elements are oriented with their long axes
pointing 70/260 degrees east of North (Fig. 28A), suggesting that some factor
was causing alignment of the long bones toward this direction.
Unit 6. Long bones collected from unit 6 display a significant pronounced
orientation trend (Fig. 28B, Table 9). Most of these elements are oriented
from 10-40/190-220 degrees east of North, with a smaller peak at
approximately 900 to the major orientation peak. The bimodal shape of the
rose diagram bears some similarity to that of a fluvial or current-influenced
Unit 7. Unit 7 bone orientations are similar to those of unit 6, although
the spread of preferred direction is wider. The orientation pattern of long
bones only (Fig. 28C) is significantly different from that of a uniform array
(Table 10). Based on the similarity of the rose diagrams for units 6 and 7, it is
evident that factors responsible for bone alignment did not change dramatically
during the time period in which these two layers were deposited.
Unit 8. In unit 8, skeletal element orientations are not as clearly defined
as in the lower two layers, 6 and 7. The more diffuse pattern seen in Figure
28D is probably related to the fact that unit 8 is not uniform
sedimentologically. The distributions of long bones (n = 72: Fig. 28D) do
show a statistically significant orientation pattern (Table 10). Although the
112 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
10% of bones
Figure 28. Rose diagrams of bearings of long bones collected from major sedimentary units. (A)
Unit 5. (B) Unit 6. (C) Unit 7. (D) Unit 8. (E) Unit 11.
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 113
preferred bone direction is not as obvious as in units 6 or 7, the majority of
long bones (73%) are oriented to the northeast/southwest.
Units 11 and 15. If long bones only (n = 96) from unit 11 are plotted,
(Fig. 28E) the dominant preferred orientation is between 40-80/220-260
degrees east of North. However, the orientation pattern is not significantly
different from uniform at the 0.01 level (Table 10). No rose diagrams have
been plotted for unit 15, as the number of bones collected with complete
orientation data was very low (n = 8).
The presence of preferred bone orientations to the northeast/southwest
clearly shows that some agent, perhaps moving water, was responsible for bone
alignment. If a current is to be implicated as a major cause of bone alignment,
it is necessary to determine if evidence of unidirectional flow, either from
northeast to southwest, or from southwest to northeast, can be determined.
The compass bearing of a skeletal element provides only a portion of the
information that can be obtained concerning possible current direction and
environment of deposition. To accurately determine the speed and direction of
paleocurrent flow, a bone's specific behavior as a paleocurrent indicator must
be examined. This procedure requires experimental analysis using modern
bones as means of documenting the hydraulic behavior of skeletal elements in
a variety of aquatic situations.
Orientation and Positions of Selected Element Types
Skeletal elements may be separated into groups according to their
usefulness as paleocurrent indicators. Based on data from previous flume
studies (Voorhies 1969; Pratt 1979), and recently conducted fluvial experiments
using skeletal elements of Recent deer, which are similar in general size and
shape to Parahippus leonensis elements, I have classified skeletal elements
according to their usefulness as paleocurrent indicators:
Category A: Bones that show a preferred axis of
orientation, and usually possess one end that repeatedly
orients downcurrent. Elongate limb elements are included in
this group and are of most use in determining direction of
Category B: Elements that do not necessarily indicate
prevailing current direction, but may exhibit predictable
stable orientations or resting positions in running water.
Included in this group are the pelvis, astragalus, calcaneum,
and proximal phalanges, and alligator osteoderms. Positions
of these bones may provide evidence that deposition occurred
under conditions of moving water, as opposed to standing
114 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
water, and indicate, in a general way, strength of current, as a
bone may assume a stable resting position at current velocity
just below critical transport velocities for the bone.
Category C: Bones that do not show any preferred
orientations. Most of the elements in this group are small in
size and are not elongated in any one dimension. Carpals,
tarsals, and distal phalanges have limited usefulness as
paleocurrent indicators, and are included in this category.
For obvious reasons, category C will not be discussed in the
In the following discussion, each sedimentary unit will not be described
individually, as numbers of each bone type from any one level are too small to
allow for statistical analysis. The overall similarity of bone orientation
direction throughout all units studied justifies lumping the samples from these
layers to obtain a large, statistically significant sample.
Orientation Behavior of Category A Elements. Most long bones,
particularly limb elements, tend to orient with their long axes parallel to the
direction of running water (Voorhies 1969; Hanson 1980). Some bones may
also orient at right angles to the current. Field positions and orientations of
fossil bones are compared with orientations observed for Recent bones in
controlled and natural fluvial environments (Voorhies 1969; Boaz and
Behrensmeyer 1976; Pratt 1979, 1986). Voorhies (1969) observed as a general
rule that the larger end of a long bone tends to orient in the downstream
direction. Pratt (1979), showed that it is not necessarily the larger end that
points downcurrent, but the less dense (frequently air-filled) end that is moved
into the downstream position. This finding is particularly true of long bones
that have been exposed to the drying effects of weathering.
Humerus. Long axes of fossil humeri from Thomas Farm (n = 46)
clearly align with the predominant preferred direction (Figure 29A). The
majority of these elements are found lying between 40-80/220-260 degrees east
of North. The directional distribution of humeri is significantly non-uniform
(Table 10). The directions that the ends of the humeri point may be assessed
by examining complete fossil humeri. Of thirteen whole humeri, 2 were found
with the proximal end pointing northeast, 5 with the proximal southeast, 5 with
proximal southwest, and 1 with proximal northwest (Table 11). Voorhies'
flume experiments (1969) showed that the "larger" (interpreted here as
proximal) ends of humeri of sheep, coyotes, and badgers oriented downstream
in about 85% of the runs. In flume experiments on rabbit to raccoon-sized
skeletal elements (Pratt 1979), proximal ends of humeri oriented downstream
about 65% of the time (37/56 trials). Experiments conducted in conjuction
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 115
with this study showed that fresh and weathered deer humeri placed in a small
stream were transported, and came to rest, with the proximal ends facing
downstream in 5 of 5 trials. Humeri of Recent deer dropped from a horizontal
position into standing water fell with the denser distal end first (provided both
epiphyses were present). If the proximal end was air-filled, as was observed in
several trials, the humerus floated in a vertical position with the proximal end
floating at the surface. The lighter proximal end often leads the way as the
element is transported downstream.
Distal portions of humeri are the most abundant parts of this element
found at the Thomas Farm. Little experimental work has been conducted on
orientation patterns of broken skeletal remains, although Boaz and
Behrensmeyer (1976) recorded the final orientation of a hominid distal
humerus in a flume as distal downstream. Of the 25 distal humeri in the fossil
deposit, 14 (56%) are oriented with the distal end pointing to the southwest,
indicating that distal humeri exhibit a preferred orientation, probably in the
downcurrent direction. More experimental work is needed to confirm the
correctness of this statement.
The positions in which the humeri are found in the field also indicate that
certain orientations are more stable than others. Complete humeri are most
often found medial or posterior side up, while distal portions only are most
frequently found posterior side up.
Radius. The orientations of 35 fossil radii are shown in the rose diagram
in Figure 29B. Although this element is too poorly preserved at Thomas Farm
to be of use in statistical determination of a preferred current direction, almost
80% (28) of the radii point either northeast or southwest, documenting the
bone's tendency to be aligned parallel to the prevailing current direction.
The differences in size and density of the two ends of the radius are not as
pronounced as are those of the humerus, however, 5 of 8 (63%) complete radii
were found with the distal end pointing southwest (Table 11). Voorhies (1969)
found that the larger ends (he did not specify the end, it is assumed that he was
referring to the distal end) of sheep and coyote radii oriented downstream 72%
of the time. Flume experiments on radii of small mammals (Pratt 1979),
showed no significant preferred end orientation for this element. Recent deer
radii dropped from a horizontal position into standing water fell with the
proximal end first, but the distal end did not float.
Broken proximal and distal ends of fossil radii do not show a clear pattern
of end orientation, and no conclusions can be drawn from these data. The
majority of radii were found either with the anterior or posterior surface up.
This position is to be expected, particularly for Parahippus radii, which are
flattened in the anterior-posterior plane.
116 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
10% of bones
Figure 29. Rose diagrams of bearings of selected "category A" long bones. (A) Humerus. (B)
Radius. (C) Femur. (D) Tibia. (E) Medial metapodial. (F) Lateral metapodial.
I'PRA'T: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 117
Table 11. Positional orientations of Thomas Farm elements. Data are for complete elements
only. Shown are proportions of elements with predicted downstream end pointing to each
compass quadrant. For discussion of downstream indicators and abbreviations, see text.
% with predicted
Element n end NE SE SW NW
Iumerus 13 proximal 15.4 38.5 38.5 7.6
Radius 8 distal 12.5 12.5 62.5 12.5
Femur 12 distal 16.7 8.3 50.0 25.0
Tibia 26 proximal 7.7 7.7 76.9 7.7
Medial 35 distal 28.5 14.3 42.9 14.3
Femur. Few femora were collected on which orientation data could be
taken (n = 27), and it is not possible to conduct a Chi-square test to determine
if their pattern of orientation, shown in Figure 29C, is significantly different
from that expected in a uniform distribution. Roughly 33% of the femora are
aligned within the 40-60/220-240 degree range of bearings.
Of 11 whole femora, 6 were found in the field with the distal end facing
southwest, 2 with the distal northeast, 1 with the distal end southeast, and 3
with the distal northwest (Table 11). Voorhies (1969) showed that the larger
(it is assumed he meant distal) end of the femur oriented downstream in 78%
of the flume runs. Femora of smaller mammals were also found aligned with
the distal end facing downcurrent in 72% of the runs (Pratt 1979). Recent deer
femora usually possess distal ends that are lighter than the proximal ends, and
although both ends may initially float, the proximal end invariably becomes
waterlogged first. In one experiment, a deer femur was placed in water. The
proximal end sank immediately, but the distal end remained floating for five
hours. In a natural fluvial environment, a weathered femur was transported,
bobbing along the water surface for a distance of over 150 m, until the distal
end became waterlogged. In all trials conducted, the distal end faced
downstream when the bone ceased movement.
Fossil proximal and distal femur portions are not abundant enough to
indicate clearcut trends in orientation, although the majority of incomplete
elements were recovered with the broken end pointing toward the southwest.
Experiments in the laboratory on distal femora showed that the broken end
oriented downcurrent in 66% of the trials. The preferred resting orientation of
a complete femur is posterior surface up. A total of 6 complete femora were
found in this position, and 3 were found anterior surface up.
118 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
Tibia. The orientation of fossil tibiae from the Thomas Farm (n = 62) is
significantly different from uniform at the P= 0.001 level (Table 10). The rose
diagram of tibia orientations (Figure 29D) shows the preferred axis of
orientation at 40-70/220-250 degrees east of North, similar to that of the other
limb bones discussed above. Voorhies (1969) noted that the orientation of the
mammalian tibia was a reliable indicator of paleocurrent direction, as this
element most often comes to rest with the long axis parallel to the current and
the large proximal end facing downstream. In Voorhies' flume experiments,
the tibia was found in this position in over 90% of the trials. Flume studies on
small mammal bones (Pratt 1979) showed that the tibia oriented with the
proximal end downstream 87% of the time. Recent deer tibiae placed in a
small stream resisted movement in most cases, and in other trials rolled
perpendicularly to the current. An orientation of a long bone at right angles to
the current most often results when the water depth is not sufficient to totally
immerse the bone (Voorhies 1969). The proximal end of the tibia, in addition
to being a good deal larger than the distal end in size, may also frequently be
air-filled in a naturally dried skeleton, as evidenced by studies on deer tibiae.
Tibiae of adult deer, if dropped from a horizontal position, always fell through
standing water distal end first. Weathered bones sometimes initially floated
vertically with the distal end down.
The tendency for the fossil tibiae from Thomas Farm to orient with the
proximal end pointing predominantly in one direction is clearly seen. Of 26
complete tibiae, 20 (over 77%) were recovered with the proximal end pointing
southwest, providing strong evidence that water moved through the deposit
from the northeast towards the southwest (Table 11). No definite orientation
trend is evident for the broken proximal or distal ends of the tibia.
The preferred resting orientation of the whole tibia is medial side up.
Over 38% of the tibiae (10 of 26) were found in this position at Thomas Farm.
The bone may also come to rest with the posterior, but rarely the antero-
lateral surface, facing upward.
Metapodial. Relatively little experimental work has been done
concerning the behavior of metapodials in hydrodynamic conditions. Fossil
metapodials will be discussed as two groups, the "medial" metapodials, such as
metacarpal III and metatarsal III of horses and metacarpal and metatarsal
III/IV of artiodactyls, and the "lateral" metapodials of three-toed horses.
Metapodials of carnivores are not included, as orientation data were taken on
relatively few of these elements. Medial metapodial orientations (n = 84) are
shown in Figure 29E. Over 30% of these bones are aligned in a 20-50/200-230
degree wedge on the rose diagram. The orientation pattern is significantly
different from that expected in a uniform distribution (Table 10). Recent deer
metapodials dropped from a horizontal position into standing water showed no
PRATT: TAPIONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 119
preferred settling orientation, presumably because both ends are similar in size
and density. It is therefore not expected that this element would exhibit a
preferred end orientation. Deer metapodials in a stream may roll
perpendicularly to the current or be moved parallel to it. When placed in
running water, these elements came to rest with the distal end facing
downstream in 66% (3 of 5) of the trials. Of 35 complete medial metapodials
from the Thomas Farm, 15 (43%) are oriented with the proximal end pointing
southwest and 10 (28%) with the proximal end pointing the opposite direction.
The remaining 10 are equally divided; 5 with proximal ends facing northwest
and 5 with proximal southeast (Table 11). The lack of difference in density
between the two ends of a metapodial limits the usefulness of this bone as an
indicator of direction of prevailing current.
If distal portions of medial metapodials only are considered, 45% (14 of
31) point distal southwest, as opposed to 25% (8) pointing northeast, 19% (6)
northwest, and 11% (3) southeast. Proximal fragments are most often found
with the proximal end facing downslope; of 15 proximal ends, 53% (8) are
directed proximal southwest, and 33% (5) proximal southeast. Incomplete
metapodials appear to be more useful in determining current direction than
are complete metapodials.
The majority of medial metapodials are found lying either anterior or
posterior up, although occasionally one may be found on its side. The nearly
circular cross-section of a metapodial increases the likelihood that this bone
will not show any preferred side-up resting position.
Lateral metapodials of three-toed horses are fairly well-represented at the
Thomas Farm locality (n = 84). Most lateral metapodials have bearings
between 30-40/210-220 degrees east of North (Figure 29F), although a large
number of lateral metapodials are also found with their long axes lying
between 60-90/260-270 degrees. The orientation pattern is not significantly
different from uniform at the P < 0.01 level (Table 10). It is not possible to
conduct comparative studies on these bones using Recent specimens, as these
elements are absent in modern horses. Complete fossil lateral metapodials (n
= 23) were recovered with proximal ends pointing equally to the northeast,
southeast, and southwest. Only 2 were aligned with the proximal end pointing
No trend is seen for proximal ends of lateral metapodials, but 41% of
distal fragments (9 of 22) were found with the distal end facing southwest.
Lateral metapodials are almost always found lateral or medial surface up.
Orientation Behavior of Category B Elements. Although skeletal
elements belonging to category B are not necessarily found with one axis
consistently aligned in the direction of the prevailing current, these bones do
exhibit some type of predictable orientation in the presence of running water.
120 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
Pelvis. Data were obtained on only 16 complete fossil pelves or pelvis
fragments. Numbers of bearing readings taken on innominates are too few to
show a directional trend, however, 13 of the 16 (81%) innominates mapped
were found with the lateral acetabularr) surface up. In flume studies on large
mammals (Voorhies 1969; Korth 1979) transport of complete pelves joined at
the pubic symphysis has been observed. Complete pelves are usually rare at
fossil localities, so these studies are of limited use in analysis of hydraulic
behavior of fossil innominates. My studies on flume transport of innominates
of small mammals (Pratt 1979) showed that innominates introduced into the
flume with the medial surface up would eventually flip over and come to rest
with the lateral surface up. Even under conditions of increased current
velocity, these elements resisted further change in medial-lateral orientation.
The majority of pelvic elements found at the Thomas Farm locality in the
fluvially-stable position may indicate that current velocities were sufficient to
cause these elements to be moved from a less stable position and come to rest
with the medial surface down.
Astragalus and calcaneum. Ungulate astragali are cube-like in shape and
do not show a dominant preferred directional orientation; however, 31% of the
astragali found at the Thomas Farm were oriented with the proximal end
facing southwest (14 of 44). The preferred resting orientation of the astragalus
of Parahippus is with the trochlear or plantar surface up, as these sides are
wider and flatter than the medial or lateral sides.
Of 44 calcanea, 14 (31%) were found with the proximal-distal axis
oriented from 100-120/280-300 degrees east of North. The calcaneum
evidently orients with its long axis perpendicular to the current direction, as
well as parallel to it (Figure 30A). Studies on the behavior of the calcaneum
in running water show that modern horse calcanea rotate about the distal ends,
usually ending with the proximal articular ends pointing downstream, while
deer calcanea most often point distal downstream. The fossil calcanea from
the Thomas Farm were primarily equid calcanea and showed the former trend;
29 of the 44 specimens were found with the proximal end facing in the
downslope direction. The majority of calcanea were found resting either on
their medial or lateral sides, rather than on the narrow anterior or posterior
surfaces. The tendency of the calcaneum to orient at right angles to, as well as
parallel to the current direction limits its usefulness as a paleocurrent direction
indicator, although the bimodal distribution of its orientations can serve to
denote the presence of moving water.
Proximal phalanges. Proximal phalanges are common at the Thomas
Farm locality. The digit elements found at the locality are primarily proximal
phalanges of digit III of Parahippus leonensis and Archaeohippus blackbergi.
Their orientation pattern, as shown on the rose diagram (Figure 30B),
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 121
possesses two peaks, one parallel to the proposed current direction at about
30-50/210-230 degrees east of North, and a second, less well-defined peak at
about 90-120/270-300 degrees, oblique to the first peak. Proximal phalanges
can be classified as "rollers" (Hanson 1980), small, cynlindrical objects that
orient either parallel or perpendicular to the current direction. The
orientation pattern for proximal phalanges from Thomas Farm is significantly
different from random at thep = 0.01 level (Table 10). The proximal end of a
proximal phalanx of a modern deer is lighter than the distal end, and is swung
into the downcurrent or downstream direction by moving water. In a stream,
proximal phalanges of deer and horses are readily transported and always
come to rest with the proximal end facing downstream. At the Thomas Farm,
proximal phalanges are oriented primarily with the proximal ends oriented
downslope (28 pointing proximal southwest, 24 pointing proximal southeast),
suggesting their alignment reflects a paleocurrent gradient. Although
orientations of phalanges are often ignored in taphonomic studies, these
elements are sometimes abundant in fossil localities and can of some use as
indicators of current direction.
Alligator osteoderms. Alligator osteoderms are rectangular or oval, so
bearings are measured along the central keel. No preferred pattern of
directional orientation is evident (Figure 30C). The notable feature of the
fossil osteoderms is that over 75% (47 of 63) of the specimens were found with
the external (keeled) surface up. Osteoderms of modern alligators dropped
vertically through standing water most often land with the internal (non-
keeled) surface up. In the presence of moving water, osteoderms resting on
the keel are pulled over onto the flat internal surface. Once these bones have
come to rest with the external surface up, they seem to resist further
movement. The fact that most of, but not all, the fossil osteoderms are found
in fluvially-stable positions indicate that an intermittent current such as that
caused by the rapid draining of water through an opening in the bottom of a
sinkhole, may have been responsible for rotating most of the osteoderms to a
more stable resting position prior to their burial.
Bone Orientations: Summary
Long axis bearings of skeletal remains from all sedimentary units of
Thomas Farm (except unit 15) clearly show an orientation trend to the
northeast/southwest quadrants. The hypothesis that this pattern was produced
primarily by moving water is supported by the positions of category A elements
(Table 10), which without exception show a preferred "downstream"
orientation toward the southwest. The predominance of category B elements
showing fluvially stable resting positions or characteristic orientations are also
indicative of influence by moving water. The proposed direction of
122 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
10% of bones
Figure 30. Rose diagrams of bearings of selected "category B" bones. (A) Calcaneum. (B)
Proximal phalanx. (C) Alligator osteoderm.
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 123
paleocurrent is consistent with the orientations of solution joints within the
Ocala limestones, which form along northeast/southwest axes (Williams et al.
1977). If the Thomas Farm sinkhole/cave had been connected to an
underground drainage system, water would have flowed through joints in a
The fairly wide span of preferred orientations within each level suggests
that the current was either not strong, or if strong may have been of short
duration. Flow velocity in karst systems varies greatly depending on climatic
conditions and height of the water table (Lane 1986). The absence of highly
plunging bones also suggests that the current was not of sufficient strength to
cause high degree of imbrication (Voorhies 1969), and the uniformity of low-
degree dips to either the northeast or southwest indicates that once the bones
were deposited, they probably were not disturbed or moved by trampling or by
other biotic agents.
Sampling Duration of the Deposit
The time span of a fossil locality is a measure of the length of time that
deposit has sampled the living fauna. Schindel (1980) discussed the use of
sedimentation rates as a means of determining the amount of time represented
within a particular locality. Behrensmeyer (1982) pointed out that a bone
assemblage may accumulate over a longer or shorter span of time than the
sediments in which it is deposited. For that reason, knowledge of taphonomic
factors important in the formation of the bone accumulation are also important
in determining the time resolution of a faunal assemblage.
Relatively little is known concerning rates of sedimentation in Florida
sinkholes or caves. Evidence from other parts of the country (Laury 1980)
indicates that sinkholes may fill rapidly (within 300 to 500 years). At Thomas
Farm, rates of sedimentation of unit 5, the laminated sand layers (Figure 13),
may be estimated. If each couplet of sand and clay laminae represents a yearly
depositional cycle, this unit could have formed within a 200 year period.
However, it is more likely that laminae were deposited more frequently, so 200
years represents a maximum value. Based on sedimentation rates in recent
pond and lake deposits (Schindel 1980) it is not unreasonable to estimate that
the Thomas Farm sediments were deposited within a 1,000 year period.
Behrensmeyer (1982) showed that bone assemblages may represent long
spans of time, particularly if elements comprising the assemblage have been
held in primary storage prior to final deposition. Assemblages of this type are
characterized by low representations of small, light elements, and enrichment
of the heaviest, densest elements. While it is possible that Thomas Farm bones
may have been held in storage, lack of evidence of weathering or abrasion
124 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
suggests that the majority were deposited soon after the death of the animal.
Evidence that the bone source did not form over a considerably shorter time
than the sediments is provided by the fact that the entire bone assemblage, at
least in terms of Parahippus leonensis, is attritional, rather than catastrophic
The best evidence that the duration of the site was relatively short is
provided by the stage of evolution of several of the mammalian taxa. The fact
that rapidly evolving groups such as the Equidae (Hulbert 1984) and the
Heteromyidae are similar in the lowermost and uppermost sedimentary levels
of the site indicate that the time sampled was relative short, on the order of
1,000 years or less.
SUMMARY AND CONCLUSIONS
The presence of limestone surrounding and within the fossiliferous
sediments of Thomas Farm provides convincing evidence that deposition
occurred in a structure or complex of structures characteristic of karst terrains.
The lack of speleothems, and the presence of both aquatic plants and aquatic
vertebrates suggests that throughout most of its depositional history, the site
was a sinkhole rather than a cave. Initially, the walls of the sinkhole were
probably high and quite steep. The presence of a large sedimentary or debris
cone within the sinkhole is indicated by the extensive and uniform dip of the
sedimentary layers, and by the presence of two rubble layers composed of
Source of the Bone Assemblage
Based on numerous lines of evidence, the source of the megavertebrate
assemblage at Thomas Farm was almost exclusively autochthonous. Bones
may have been derived primarily from two sources: from live animals that
climbed or fell into the sinkhole and were either killed by the fall, or unable to
make their way out, or from skeletons that accumulated and disarticulated
either on the surface surrounding the sinkhole or on talus slopes along the
walls of the sinkhole. In either case, modification of bone by agents such as
weathering or carnivores was not a significant factor. The high degree of
similarity of the compositions of the bone assemblages from the various
sedimentary units to those of a complete skeleton suggest that following
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 125
disarticulation, burial must have been relatively rapid, and removal or
destruction of skeletal elements by various taphonomic agents did not greatly
alter the assemblages. Further support for this hypothesis is provided by the
fact that mammals with body masses of 15 kg or less are well-represented at
the site, compared to other sites where smaller members of the fauna are
under-represented due to taphonomic biases acting against their preservation
(Behrensmeyer and Boaz 1980; Badgley 1986a). The numerical dominance of
the three-toed horse Parahippus leonensis cannot be fully explained at this
time. The great abundance of this taxon may have been due to its social or
migratory behavior. There is little evidence to suggest that carnivores played a
major role in the formation of a bone assemblage composed predominantly of
one species. The site may also have functioned to some degree as a size-
selective trap. Following the two limestone rubble falls, it appears that the
trapping ability of the site was diminished, and the nature of the deposit was
radically changed (Pratt 1989).
The presence of water-lain clays, aquatic plants, and aquatic vertebrate
taxa indicates that the lower sedimentary units of the site were deposited under
aquatic conditions. Although several lines of evidence, such as the lack of large
aquatic frogs (Estes 1963; Pratt 1989), and the presence of in situ weathering in
unit 8, suggest that the water was neither deep nor permanent. The lack of
evidence that bones were disturbed by trampling indicates that for the most
part large groups of animals did not have access to the water source. The
rarity of fish and the absence of fresh-water molluscs, ostracods, and other
aquatic invertebrates suggests that the site was not connected to other aquatic
environments by surface streams. It is possible that water conditions may not
have been suitable for some forms of aquatic life, similar to the situation noted
by Laury (1980) for Hot Springs.
The presence of significant bone orientation patterns in the lower units of
the site provide evidence that water within the sinkhole was not stagnant, but
was moving through the sinkhole toward the southwest. The current may have
been caused by artesian groundwater flow through solution joints in the Crystal
River limestone. Flow velocity therefore varied with variations in the water
table; however, the high representation of all but the most transportable
skeletal elements indicates that flow velocity probably did not exceed 35
The Terrestrial Environment
Little is known concerning climates and habitats of the early Miocene of
Florida. In the case of Thomas Farm, faunal evidence suggest that the region
surrounding the sinkhole was forested rather than open terrain. All ungulate
126 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
herbivores from the site have brachyodont dentitions, with hypsodonty indices
(Janis 1984) of less than 1 (Pratt 1986). These animals, including Parahippus
leonensis, most likely were browsers on leafy or herbaceous vegetation
(Hulbert 1984). The most common carnivores found at the site are two canids
with skull proportions similar to the coyote, but with extremely short limbs.
Limb ratios for these carnivores are most similar to those of the South
American bush-dog Speothos venaticus (Pratt 1986), suggesting that they were
ambush hunters in wooded habitats.
The most abundant rodents of the early Miocene of Florida were
quadrupedal, brachydont heteromyids most similar to the Recent forest-
dwelling Heteromys and Liomys. Two species of sciurid, one an arboreal form,
and the other related to Tamias, were also present at the site (Pratt and
Morgan 1989). Spermophiline sciurids are not known from Thomas Farm.
The presence of several reptile and bat taxa that today are restricted to tropical
regions (Morgan pers. comm.) indicate that the climate of the region in the
early Miocene was more tropical than the current climate of northern Florida.
Additional evidence concerning modes of bone deposition in sinkholes and
early Miocene habitats in Florida awaits further discoveries of informative
early Miocene fossil localities.
Agcnbroad, L.D. 1984. Hot Springs, South Dakota: Entrapment and taphonomy of Columbian
mammoth. Pp. 113-127 in P.S. Martin and R.G. Klein (eds.). Quaternary Extinctions, a
Prehistoric Revolution. Univ. Arizona Press, Tuscon.
Allen, J.R.L. 1965. A review of the origin and characteristics of recent alluvial sediments.
Andrews, P., and J. Cook. 1985. Natural modifications to bones in a temperate setting. Man
Andrews, P., and E.M. Nesbit Evans. 1983. Small mammal bone accumulations produced by
mammalian carnivores. Paleobiology 9:289-307.
Andrews, P., G.E. Meyer, D.R. Pilbeam, J.A. Van Couvering, and J.A.H. Van Couvering. 1981.
The Miocene fossil beds of Mabuko Island, Kenya, geology, age, taphonomy, and
paleontology. J. Human Evol. 10:35-48.
Auffenberg, W. 1963a. The fossil snakes of Florida. Tulane Stud. Zool. 10:131-216.
1963b. Present problems about the past. Biological Sciences Curriculum Study, Univ. of
Florida. Pamphlet No. 6:1-35.
,W.T. Neill, T.H. Patton, and S.D. Webb. 1964. 1964 field trip in central Florida.
Guidebook, Annual Meeting, Soc. Vert. Paleo., Gainesville, Florida. 50 p.
Bader, R.S. 1956. A quantitative study of the Equidae of the Thomas Farm Miocene. Harvard
Univ., Mus. Comp. Zool. Bull. 115:49-78.
Badglcy, C. 1986a. Taphonomy of mammalian fossil remains from Siwalik rocks of Pakistan.
.1986b. Counting individuals in mammalian fossil assemblages from fluvial environments.
Harbour, T. 1944. That Vanishing Eden: A Naturalist's Florida. Little, Brown and Co., Boston.
Beck, B., and W.F. Sinclair. 1986. Sinkholes in Florida: An Introduction. Report 85-86-4, The
Florida Sinkhole Research Inst., Univ. Central Florida, Orlando. 18 p.
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 127
Behrensmeyer, A.K. 1975. The taphonomy and paleoecology of Plio-Pleistocene assemblages
east of Lake Rudolph, Kenya. Harvard Univ., Mus.Comp. Zool. Bull. 146:473-578.
S1978. Taphonomic and ecological information from bone weathering. Paleobiology
1982. Time resolution in fluvial vertebrate assemblages. Paleobiology 8:211-227.
S1988. Vertebrate preservation in fluvial channels. Palaeogeogr., Palaeoclimat.,
and D.E. Dechant Boaz. 1980. The recent bones of Amboseli National Park, Kenya. Pp.
72-92 in A.K. Behrensmeyer and A.P. Hill (eds.). Fossils in the Making, Vertebrate
Taphonomy and Paleoecology. Univ. Chicago Press, Chicago.
,and A.P. Hill. 1984. Disarticulation patterns of some modern East African mammals.
K.D. Gordon, and G.T. Yanagi. 1986. Trampling as a cause of bone surface damage and
pseudo-cutmarks. Nature 319(6056):768-771.
D. Western, and D.E. Dechant Boaz. 1979. New perspectives in vertebrate paleoecology
from a recent bone assemblage. Paleobiology 5:12-21.
Binford, L.R. 1981. Bones, Ancient Men and Moder Myths. Academic Press, New York, 320 p.
131umenschine, R.J. 1986. Carcass consumption sequences and the archaeological distinction of
scavenging and hunting. J. Human Evol. 15:639-659.
Boaz, N.T., and A.K. Behrensmeyer. 1976. Hominid taphonomy; transport of human skeletal
parts in an artificial fluviatile environment. Amer. J. Phys. Anthropol. 45:53-60.
Bogli, A. 1980. Karst Hydrology and Physical Speleology. Springer Verlag, Berlin, 284 p.
Bonnichsen, R. 1973. Some operational aspects of human and animal bone alteration. Pp. 9-25
in B.M. Gilbert (ed.). Mammalian Osteo-Archaeology: North America. Missouri
Archaeol. Soc., Columbia.
Bown, T.M., and M.J. Kraus. 1981. Vertebrate fossil-bearing paleosol units (Willwood
Formation, Lower Eocene, northwest Wyoming, U.S.A.): Implications for taphonomy,
biostratigraphy, and assemblage analysis. Palaeogeogr., Palaeoclimatol., Palaeoecol. 34:31-
Brain, C.K. 1981. The Hunters or the Hunted? An Introduction to African Cave Taphonomy.
Univ. Chicago Press, Chicago, 365 p.
Conybeare, A., and G. Haynes. 1984. Observations on elephant mortality and bones in water
holes. Quat. Res. 22:189-200.
Cooke, C.W. 1945. Geology of Florida. Florida Geol. Surv. Bull. 29:1-339.
Damuth, J. 1982. Analysis of the preservation of community structure in assemblages of fossil
mammals. Paleobiology 8:434-446.
Davis, W.M. 1930. Origin of limestone caverns. Geol. Soc. Amer. Bull. 41:475-628.
Dodson, P. 1973. The significance of small bones in paleoecological interpretation. Contrib.
Geol., Univ. Wyoming 12:15-19.
Dunbar, C.O., and J. Rodgers. 1957. Principles of Stratigraphy. John Wiley and Sons, New
York, 356 p.
Estes, R. 1963. Early Miocene salamanders and lizards from Florida. Quart. J. Florida Acad.
Fergusen, G.E., C.W. Lingham, S.K. Love, and R.O. Vernon. 1947. Springs of Florida. Florida
Geol. Surv. Bull. 31:1-196.
Frailey, D. 1978. An early Miocene (Arikareean) fauna from northcentral Florida (the SB-1A
local fauna). Occ. Pap. Mus. Nat. Hist., Univ. Kansas 75:1-20.
1979. The large mammals of the Buda local fauna (Arikareean: Alachua County,
Florida). Bull. Florida State Mus., Biol. Sci. 24:124-173.
Grayson, D.K. 1978. Minimum numbers and sample size in vertebrate faunal analysis. Amer.
I anson, C.B. 1980. Fluvial taphonomic processes: Models and experiments. Pp. 156-181 in
A.K. Behrensmeyer and A.P. Hill (eds.). Fossils in the Making, Vertebrate Taphonomy
and Paleoecology. Univ. Chicago Press, Chicago.
I laynes, G. 1980. Evidence of carnivore gnawing on Pleistocene and Recent mammalian bones.
128 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
S1983. A guide for differentiating mammalian carnivore taxa responsible for gnaw damage
to herbivore limb bones. Paleobiology 9:164-172.
Ilill, A.P. 1980. Early postmortem damage to the remains of some contemporary East African
mammals. Pp. 131-155 in A.K. Behrensmeyer and A.P. Hill (eds.). Fossils in the Making,
Vertebrate Taphonomy and Paleoecology. Univ. Chicago Press, Chicago.
,and A. Walker. 1972. Procedures in vertebrate taphonomy. J. Geol. Soc. London
Slolman, J.A. 1965. Early Miocene anurans from Florida. Quart. J. Florida Acad. Sci. 28:69-82.
1967. Additional Miocene anurans from Florida. Quart. J. Florida Acad. Sci. 30:121-140.
Ilulbert, R.C., Jr. 1984. Paleoecology and population dynamics of the early Miocene
(Hemingfordian) horse Parahippus leonensis from the Thomas Farm site. Florida J. Vert.
and G.S. Morgan. 1989. Stratigraphy, paleoecology, and vertebrate fauna of the Leisey
Shell Pit local fauna, early Pleistocene (Irvingtonian) of southwestern Florida. Pap.
Florida Paleontol. No. 2:1-19.
Hunt, R.M. 1978. Depositional setting of a Miocene mammal assemblage, Sioux County,
Nebraska (U.S.A.). Palaeogeog., Palaeoclimat., Palaeoecol. 24:1-52.
Jackson, D.R. 1976. The status of the Pliocene turtles Pseudemys caelata Hay and Chrysemys
carri Rose and Weaver. Copeia 1976(4):655-659.
Janis, C.M. 1984. The use of fossil ungulate communities as indicators of climate and
environment. Pp. 85-104 in P. Brenchley, (ed.). Fossils and Climate. John Wiley and
Sons, New York.
Korth, W.W. 1979. Taphonomy of microvertebrate fossil assemblages. Carnegie Mus. Nat.
Hist., Annals 48(15):235-285.
Lane, E. 1986. Karst in Florida. Florida Geol. Surv. Spec. Pub. 29:1-100.
Laury, R.L. 1980. Paleoenvironment of a late Quaternary mammoth-bearing sinkhole deposit,
Hot Springs, South Dakota. Geol. Soc. Amer. Bull. 91:465-475.
Lipchinsky, Z. 1963. A study of the origin of limestone caverns in Florida. Masters thesis, Univ.
Florida. 62 p.
Maas, M.C. 1985. Taphonomy of a late Eocene microvertebrate locality, Wind River Basin,
Wyoming (U.S.A.). Palaeogeogr., Palaeoclimat., Palaeoecol. 52:123-142.
Milkse, J.A., E.C. Alexander, Jr., and R.S. Lively. 1983. Clastic sediments in Mystery Cave,
southeastern Minnesota. Nat. Speleol. Soc. Bull. 45:55-75.
Nicol, D., and D.S. Jones. 1982. Rotularia ventoni, an annelid worm tube from the Eocene of
peninsular Florida. Florida Sci. 45:139-142.
Olscn, S.J. 1959. Fossil Mammals of Florida. Florida Geol. Surv. Spec. Pub. 6:1-75.
1962. The Thomas Farm fossil quarry. Quart. J. Florida Acad. Sci. 25:142-146.
Pratt, A.E. 1979. Taphonomic studies of small mammals. Masters thesis, Univ. Massachusetts,
Amherst. 139 p.
.1986. The taphonomy and paleoecology of the Thomas Farm local fauna (Miocene,
Hemingfordian), Gilchrist County, Florida. Ph.D dissertation, Univ. of Florida. 487 p.
.1989. Taphonomy of the microvertebrate fauna from the early Miocene Thomas Farm
locality, Florida (U.S.A.). Palaeogeogr., Palaeoclimat., Palaeoecol. 76:125-151.
and G. Morgan. 1989. New Sciuridae (Mammalia: Rodentia) from the early Miocene
Thomas Farm local fauna, Florida. J. Vert. Paleont. 9:89-100.
Puri, H.S., and R.O. Vernon. 1964. Summary of the geology of Florida and a guidebook to the
classic exposures. Florida Geol. Surv. Spec. Pub. 5:1-312.
Puri, H.S., J.W. Yon, Jr., and W.P. Oglesby. 1967. Geology of Dixie and Gilchrist counties,
Florida. Florida Geol. Surv. Bull. 49:1-155.
Ragan, D.M. 1973. Structural Geology, second edition. John Wiley and Sons, New York, 208 p.
Romer, A.S. 1948. The fossil mammals of Thomas Farm, Gilchrist County, Florida. Quart. J.
Florida Acad. Sci. 10:1-11.
Rose, F.L., and W.G. Weaver, Jr. 1966. Two new species of Chrysemys (=Pseudemys) from the
Florida Pliocene. Tulane Stud. Geol. 5:41-48.
Rosenau, J.C., G.L. Faulkner, C.W. Hendry, Jr., and R.W. Hull. 1977. Springs of Florida.
Florida Geol. Surv. Bull 31(revised):1-461.
Roscnzwcig, M.L. 1966. Community structure in sympatric Carnivora. J. Mammal. 47:601-612.
PRATT: TAPHONOMY OF THOMAS FARM LARGE VERTEBRATE FAUNA 129
SAS Institute, Inc. 1982. SAS User's Guide: Basics. 1982 edition. Cary, North Carolina, 923 p.
Saunders, J.J. 1977. Late Pleistocene vertebrates of the western Ozark Highlands, Missouri.
Illinois State Mus., Rep. Invest. 33:1-118.
Schindel, D.E. 1980. Microstratigraphic sampling and the limits of paleontologic resolution.
Schrock, R.R. 1948. Sequence in Layered Rocks. McGraw-Hill Book Co., New York. 507 p.
Scott, T.M. 1988. The lithostratigraphy of the Hawthorn Group (Miocene) of Florida. Florida
Geol. Surv. Bull. 59:1-148.
Shipman, P. 1981. Life History of A Fossil--An Introduction to Taphonomy and Paleoecology.
I larvard Univ. Press, Cambridge. 22 p.
Shipman, P., A. Walker, J.A. Van Couvering, P.J. Hooker, and J.A. Miller. 1981. The Fort
Ternan hominoid site, Kenya, geology, age, taphonomy, and paleoecology. J. Human Evol.
Shotwell, J.A. 1955. An approach to the paleoecology of mammals. Ecology 36:327-337.
Simpson, G.G. 1932. Miocene land mammals from Florida. Florida Geol. Surv. Bull. 10:741.
Sinclair, W.C., J.W. Stewart, R.L. Knutilla, A.E. Gilberg, and R.L. Miller. 1985. Types, features,
and occurrences of sinkholes in the karst of west-central Florida. U.S. Geol. Surv. Water
Resources Rep. Invest. 85-4126:1-81.
Sokal, R.R., and F.J. Rohlf. 1981. Biometry, 2nd Edition. W.H. Freeman and Sons, San
Francisco. 859 p.
Stringfield, V.T., P.E. LaMoreaux, and H.E. LeGrand. 1974. Karst and paleohydrology of
carbonate rock terranes in semiarid and arid regions with a comparison to humid karst of
Alabama. Geol Soc. Alabama Bull. 105:1-106
Seeing, M.M. 1973. Karst Landforms. Columbia Univ. Press, New York. 362 p.
Trudgill, S. 1985. Limestone Geomorphology. Longman, Gp., London. 196 p.
Twenhofel, W.H. 1932. Treatise on Sedimentation. The Williams and Wilkins Co., Baltimore,
Voorhies, M.R. 1969. Taphonomy and population dynamics of an early Pliocene vertebrate
fauna, Knox County, Nebraska. Univ. Wyoming Contrib. Geol. Spec. Pap. 1:1-69.
Webb, S.D. 1981. The Thomas Farm fossil site. Florida Paleo. Soc., Plaster Jacket 37:6-25.
and B.J. MacFadden. 1980. Field trip to North-Central Florida. Soc. Vert. Paleont.
Guidebook, Gainesville, Florida. 29 p.
B.J. MacFadden, and J.A. Baskin. 1981. Geology and paleontology of the Love Bone Bed
from the late Miocene of Florida. Amer. J. Sci. 218:513-544.
Western, D. 1980. Linking the ecology of past and present mammal communities. Pp. 41-54 in
A.K. Behrensmeyer and A.P. Hill, (eds.). Fossils in the Making, Vertebrate Taphonomy
and Paleoecology. Univ. Chicago Press, Chicago.
While. T.E. 1942. The lower Miocene mammal fauna of Florida. Harvard Univ., Mus. Comp.
Zool. Bull. 92(1):1-49.
White, W.B., B.E. Sheetz, S.D. Atkinson, D. Ibberson, and C.A. Chess. 1985. Mineralogy of
Rohrer's Cave, Lancaster County, Pennsylvania. Nat. Speol. Soc. Bull. 47:17-27.
Williams, K.E., D. Nicol, and A.F. Randazzo. 1977. The geology of the western part of Alachua
County, Florida. Florida Geol. Survey Rep. Invest. 85:1-98.
Wolff, R.G. 1973. Hydrodynamic sorting and ecology of a Pleistocene mammalian assemblage
from California. Palaeogeogr., Palaeoclimat., Palaeoecol. 13:91-101.
130 BULLETIN FLORIDA MUSEUM NATURAL HISTORY, BIOL. SCI. VOL. 35(2)
Thomas Farm Measuring System
The fossil measuring system devised for use at Thomas Farm (Figs. 3 and 6) was designed
to fulfill several requirements. Because the system was intended to be relatively permanent, it
was constructed of weather-resistant materials. For the sake of convenience, it is easily
assembled and disassembled. The system was also relatively inexpensive to construct.
The east-west axes of the system consist of 5 m lengths of rust-proofed conduit pipe or
PVC pipe, 1.9 cm in diameter, to which 5 m lengths of centimeter-ruled fiberglass measuring tape
are attached. Before these pipes are anchored in place, they are fitted with several PVC "T"
attachments that are loose enough to slide along the length of the pipe.
The horizontal axis is attached to vertical pipes driven into the ground at selected eastern
and western 5 m quadrat borders. The vertical pipes are fitted with snug-fitting "T" joints into
which the horizontal pipe is set. When this east-west axis pipe is leveled, the supporting vertical
"T" joints are anchored in place using hose clamps (see Fig. 3).
The north-south component of the measuring system consists of 1 m lengths of conduit or
PVC pipe to which 1 m lengths of fiberglass measuring tape have been attached. A pocket line
level is also fixed to these pipes. The north-south "rulers" fit into the sliding "T" fittings on the
cast-west pipes, and therefore can slide the length of the 5 meter axis at right angles to it. Using
this system and a plumb bob dropped to the desired measuring point, the east-west and north-
south coordinates of that point (in cm) can be determined.
Contributions to the BULLETIN OF THE FLORIDA MUSEUM OF NATURAL
HISTORY, BIOLOGICAL SCIENCES, may be in any field of biology. Manuscripts dealing with
natural history or systematic problems involving the southeastern United States or the New
World tropics are solicited especially. Manuscripts should be of medium length-circa 35 to 200
pages (10,500-60,000 words). Examination for suitability is made by an Editorial Board.
The BULLETIN is distributed worldwide through institutional subscriptions and
exchanges. It is considered the responsibility of the author to distribute his paper to all inter-
ested individuals. To aid in this the authors) receives) 50 copies free, and he(they) may
purchase additional separates at cost if ordered when page proof is returned. The author is also
responsible for any charges incurred for alterations made by him on galley or page proofs. The
Museum will send an invoice to the author for this amount upon completion of publication.
PREPARATION OF MANUSCRIPT
Contributors should consult recent numbers of the BULLETIN for preferred style
and format. Highly recommended as a guide is the CBE Style Manual, 3rd Edition, 1972
(American Institute of Biological Sciences, Washington, D.C.).
Manuscripts must be submitted in duplicate and satisfy the following minimal
requirements. They must be double-spaced throughout, including tables, figure captions, and
literature citations. Figure legends and tables should be typed on separate sheets. Also, please
submit a copy of the complete text and tables on one or more SV4" flexible diskettes.
All illustrations are referred to as figures. They must comply with the following
standards: Photographs should be sharp, with good contrast, and printed on glossy paper. If the
background of photographs (especially those of specimens) is not desired, amberlith should be
cut out and used to cover the background. Drawings should be made with dense black
waterproof ink on quality paper or illustration board. All figures should have a cover sheet. All
lettering will be medium weight, sans-serif type (e.g. Futura Medium, News Gothic) in cutout, dry
transfer, or lettering guide letters. Make allowance so that after reduction no lower case letter
will be less than 1 mm high (2 mm is preferred) nor any capital letter greater than 5 mm high.
The maximum size for figures is 9" x 14" (twice BULLETIN page size); figures should not be less
than typepage width (4"W). With soft lead pencil on the back of each illustration, designate the
top and identify each by author's name, manuscript title, and figure number.
All manuscripts not submitted in BULLETIN format will be returned to the author
Manuscripts and all editorial matters should be addressed to:
Managing Editor of the BULLETIN
Florida Museum of Natural History
University of Florida
Gainesville FL 32611-2035