PaleoBios 18(2&3):1-15, September 1998
A new species of Microcosmodon (Mammalia: Multituberculata)
from the Paleocene Tullock Formation of Montana, and an argument
for the Microcosmodontinae
ANNKWKII.
Department of Integrative Biology and University of California Museum of Paleontology,
University of California, Berkeley, CA 94720
ABSTRACT—A new species of the multituf erculatc genus Microcosmodon Jcpsen, 1930 from the Puercan Tullock
Formation of Garfield County, Montana is described. Study "fa relatively large sample of molars and premolars
from both upper and lower dentitions indicates that derived characters distinguishing Microcosmodon from other
multitubcrculate genera were developed b\ the early Paleocene. The presence of the new species .suggests that
microcosmodontine species richness in the Western Interior was as high at the beginning of the Paleocene as at its
end. Phylogenctic analysis of species assigned to the Microcosmodontinae confirms the close relationship originally
postulated between Microcosmodon and Pentacosmodon Jepsen, 1940, but indicates that Microcosmodon woodi
I Iolt/man and Wolbcrg, 1977 does not belong to the genus or subfamily. This is consistent with the hypothesis
that microcosmodontines diverged from other cucosmodontids before their dispersal from Asia to North America.
INTRODUCTION
In first describing the multitubcrculate mammal
Microcosmodon conns from the TifFanian Big Sand Coulee
locality of Wyoming, Jepsen (1930) noted that its "relatively
small and unique 1'4 serves sharply to distinguish this genus
from others" (p. 510). Later, however, in forming the
Eucosmodontidae, Jepsen (1940) grouped this distinctive
animal with Eucosmodon Matthew and Granger, 1921 and
Neoliotomus Jepsen, 1930 on the basis of their laterally
compressed lower incisors, on which the enamel is mostly
restricted to the anterolabial side. I Ie also included the newly
described Pentacosmodon pronusjepsen, 1940. Pcn.'acosmodon
pronnswAs described from a nearly complete right mandible,
found in the Tiffanian Polecat Bench assemblage of Wyo-
ming with M. conns, and it possessed a similar, rela lively small
P4. Icpsen (1940) recognized that Microcosmodon and
Pentacosmodon might be closely related, and Kielan-
Jaworowska (1974a) noted that they differed from other
members of the Kucosmodontidae. Holtzman and Wolbcrg
(1977) created a new subfamily for the two genera, the
Microcosmodontinae. The first comprehensive phyloge-
nctic analysis of the Multituberculata (Simmons 1993),
however, showed the two genera to be only distantly related,
and she recommended that the subfamily be dropped from
use.
Meanwhile, the geographic range and known diversity of
Microcosmodon was expanded. Microcosmodon conns was
recognized in the latest Tiffanian Badwater fauna of the
Fort Union Formation in central Wyoming (Kt ishtalka et
al., 1975; but see Krausc, 1977 regarding identification of
individual illustrated specimens), and Krause (1977) identi-
fied M. conns in the late Tiffanian portion of the Ravcnscrag
Formation in Saskatchewan. Microcosmodon woodi was de-
scribed from the Tiffanian of Wyoming, Montana, and
North Dakota (Holtzman and Wolbcrg, 1977; Holtzman,
1978), and M. rosci Krause, 1980 from the Clarkforkian
part of the Willwood Formation of Wyoming's Bighorn
Basin.
The stratigraphie range of the subfamily was extended to
the earliest Paleocene by Archibald's (1982) description of
Achcronodon garbani from a channel 2 to 4 m above the
base of the Z-coal (IrZ of Swisher et al., 1993) of
Montana's Tullock Formation. Subsequently, Johnston and
Fox (1984) described M. arcnatns from the Puercan por-
tion of the Ravcnscrag Formation in Saskatchewan. Al-
though both taxa are rare in their assemblages and are
known only from very limited material, they have tiny,
highly arched microcosmodontine P(s, and the ML mor-
phology of M. arcuatus justifies its generic assignment (but
sec results and discussion of phylogenctic analysis, below).
Fox (1989) also described and illustrated a
microcosmodontine P4 from the Long Fall horizon of the
Ravcnscrag Formation. Lerbekmo (1985) and Sloan (1987)
have argued that this horizon is of Paleocene age, while Fox
(1989, 1990) argues that the site is Lancian. If Long Fall is
indeed latest Cretaceous, this P4 represents the only
microcosmodontine known before the Tertiary.
Otherwise unpublished microcosmodontines were listed
by Fox (1990) as follows: a microcosmodontine from the
Puercan Croc Pot locality of the Ravcnscrag Formation,
Achcronodon n. sp. and Pentacosmodon n. sp. from the
Tiffanian Cochrane 2 locality of the Porcupine Hills For-
mation, M. woodi from the Tiffanian Hand Hills West and
DW2 localities of the Paskapoo Formation, and
Microcosmodon sp. from the Tiffanian Swan Hills .Site 1
locality of the Paskapoo Formation.
On the basis of a relatively large sample of mostly well
preserved premolars and molars, I herein describe a new
species of Microcosmodon from higher in the Tullock For-
mation, roughly contemporaneous with M. arcnatns
2                                                           PAI.liOMOS, VOL. I8y NUMBERS 26~3, 1998
(Johnston and Fox, 1984) and apparently distinct from
A. garbani. The presence of" another early Paleoccne spe-
cies in the northern Western Interior indicates that
microcosmodontines were at least as diverse at that time as
they were at the end of the Paleoccne. 1 have included this
new species in a phylogenetic analysis of selected
eucosmodontids, seeking to test the monophyly of the
genus Microcosmodon and of the subfamily
Microcosmodontinae, and to provide the first systematic
diagnoses of these taxa.
GEOLOGICAL CONTEXT
All known specimens of the new species were collected in the
Tullock Formation of Garfield County, Montana, in the area
mapped by Archibald (1982), and in the channel fades
between the Y and W-coals, referred to as the "Garbani
channel deposits" (Archibald, 1982). The fades consists of
buff to yellowish silts and sands deposited in stacked chan-
nels. Dramatic cross-cutting relationships among channels
of various sizes are visible in outcrop, and the degree of cross-
cutting suggests that bone deposited in this system may have
been reworked several times, resulting in time-averaged
deposits. The outcrop encompassing University of Califor-
nia Museum of Paleontology (UCMP) site V73080 and Ix>s
Angeles County Museum site 3099 (UCMP V72125), is
separated from UCMP sites V75193, V74122, V73120, and
V72130 by approximately 2 kilometers. The outcrops arc-
thus difficult to correlate, although Archibald (1982, Fig.
72) placed V73080 and V72125 lower in the section than the
other localities. The other sites, which occur in more or less
continuous outcrop, were deposited by different channels in
a single river system. Formation of modern badland topog-
raphy makes it impossible to determine cross-cutting rela-
tionships among the bone-bearing channels, but the lowest
points of all are less than 3 m above a thin, whitish marker
sandstone.
The sanidine rich 1 IFX-Coal, which underlies the lower-
most point of the Garbani channel faeies by about 16.5 m,
is dated at 64.77 ± 0.06 Ma, and a bentonite within the W-
Coal, which occurs about 5 m above the top of the channel
deposits, at 64.11 ± 0.02 Ma. The Garbani channel lies
entirely within magnetochron 29N (4"Ar/wAr dates and
paleomagnetic data from Swisher et al., 1993). Its fauna
belongs to the Pu2 or Pu3 interval-zone (Archibald et al.,
1987). It is younger than the Hell's Hollow channel from
which Acberonodon is known, which is dated between
65.16 ± 0.4 Ma and 64.95 ± 0.05 Ma (Swisher et al.,
1993).
'Fhe Garbani channel sites were intensively collected
from 1972 to 1980 and annually prospected in subsequent
years. Bryant (1989) described lower vertebrates from these
sites, and Simmons (1987) described the new
multituberculate ineniulabis lamberti from V73080. A
large, diverse, and mostly undescribed  multituberculate
fauna is known from the channel faeies. All localities are
recorded n the UCMP catalog.
METHODS
Specimens were measured as described and illustrated by
Eaton (1995), using a Leica Wild MZ8 microscope with
measuring reticle at 20x magnification.
All phylogenetic analyses were run on an Apple
Macintosh PowerBook 5300. The matrix used in phyloge-
netic analysis was entered in MacClade 3.04 ( Maddison and
Maddison, 1992). As the only extradcntal remains known
of the tax j under study are incomplete mandibles of three
species, t ic matrix consists solely of dental characters.
Some of the characters used are as described by Simmons
(1993), while others describing morphology peculiar to the
lower-level taxa under consideration have been added or
altered, aid multiple characters that are not independent of
large body size have been omitted (Weil, 1996). In particu-
lar, although cusp counts arc used in the description of
teeth, the/ are omitted from the analysis. A ratio of cusps/
mm of crown length, which better expresses the crowding
of cusps on the tooth, is used instead.
Only characters and states informative with respect to
the taxa considered are used in the matrix. States for
characters 5, 10, 12, 14-18, and 20-22 depend on numeri-
cal values that vary along a continuum. Wherever possible,
these states were defined by plotting the entire range of
values for a character for all taxa considered, and dividing
this range at natural break points.
PAUP 3.1.1 (Swofford, 1993) was used for phyloge
netic and bootstrap analyses. In the phylogenetic analysis
26 characters were used, no characters were ordered, and
outgroups were specified. Both ACCTRAN and
DELTRAS character optimizations were examined in us-
ing analys s results to construct taxonomic diagnoses. The
bootstrap analysis was made using the branch and bound
algorithm. 1000 replicates, and no topological constraints.
AutoDecay 3.0.1 (Eriksson and Wikstrom, 1995) was
used to determine decay indices. An heuristic run using
random addition sequences was made.
ABBREVIATIONS
I ACM—I os Angeles County Museum
UCMP—University of California Museum of Paleontology
Abbreviations for teeth follow the convention of the first
letter of the tooth's type, followed by a number indicating
its position, subscripted to indicate the lower dentition or
superscripted to indicate the upper dentition. Cusp counts
follow the convention of listing the number of cusps in the
most labial row first and progressing linguallv, the number
of cusps in each row being separated by a colon.
WEIL-NEWMICROCOSMODON SPECIES
3
PI IYI.OGF.NKTIC ANALYSIS
Description and Discussion of Characters
1.  Enamel ultrastriictnre
0    preprismatic
1    intermediate
2—gigantoprismatic
Simmons (1993) used iliis as a three-state character, for
preprismatic, small prismatic, and gigantoprismaic enamel.
She scored Microcosmodon with a "9," indicating missing
data (p. 159). In fact, data were available but ambiguous.
Carlson and Krause (1985) showed that M. corns and M.
woodi, the two species sampled, have enamel prisms inter-
mediate in size between those classified as small prismatic
and those that are gigantoprismatic, and likewise intermedi-
ate in prism structure, having a mixture of circular and
arcade-shaped prisms. This state is unique in Carlson and
Krausc's study, which treated 36 multitubercukte species,
and is scored here as character state 1. No taxa considered
in the current analysis have small prismatic enamel, so that
state is not scored here.
2.  Enamel covering of the lower incisor
0—is of uniform thickness
1—is thicker on the labial surface than on the lingual
surface
2—is restricted to the labial surface for all or most of the
incisor length
Restricted enamel and lateral compression of the lower
incisor are the principal characters used to unite the
Eucosmodontidae (Jcpscn, 1940). Holtzman and Wolberg
(1977) differentiate the subfamily Microcosmodontinae
from other eucosmodontids partly on the basis of
microcosmodontine enamel covering the entire tip of the
lower incisor as well as the labial or antero-external surface.
This may not be a distinction based on an actual difference
in enamel coverage, but instead a difference reflecting the
amount of wear. The type specimen of M. conus is a
mandible showing little or no wear, and so the enamel cap
on the lower incisor is preserved in its entirety. Wear
removes the enamel on the lingual side of the rip first, so
that even a moderately worn tooth, as are all those with
definite assignment to Eucosmodon and Styfjimys, appears to
have completely restricted enamel. Thus, character state 2
is written to reflect this uncertainty, rather than to force a
distinction that may be art if actual.
3.  Compression of the lower incisor
0—not laterally compressed
1 —laterally compressed
4.  P, present/absent/cusp count
0—present, one cusp
1    may be present or absent within a species
2—always absent
The P. in Microcosmodon is small or absent i.i M. conns
(Jepsen, 1940), and if dislodged from the mandible would
pass through a 30-mesh screen, usually the smallest mesh
size employed for the recover)' of mammal teeth. Perhaps as
a result, it is not known from other species. Its presence has
been inferred in M. woodi (Holtzman and Wolberg, 1977),
M. arcuatus (Johnston and Fox, 1984), and in M. barlcyi
(this paper) on the basis of the presence of an anterobasal
concavity of the P4. This may prove to be erroneous, as
Krause (1980) notes that M. rosci, known from a nearly
complete mandible, has a shallow anterobasal concavity but
no P.. On the other hand, Jepsen's (1940) description
leases open the possibility that the P. was shed during the
life of the animal, and if this was the case then M. rosci, too,
may have possessed it at an early stage. Scoring of this
character reflects published descriptions of the animals, but
the descriptions should be regarded as "best guesses" open
to future revision.
5.  P4 number of serrations/mm crown length
0—greater than or equal to 2.3 serrations/mm
1—fewer than 2.3 serrations/mm
6.  Shape of the P4 profile
0—crown is longer than it is high
1—crovvn height and length subequal; crown high and
rounded in appearance
2—crown high, anterior edge flat rather than rounded
7.  Lateral (internal and external) ridges of the P4
0—strong
1—weak or absent
This character describes only the relief of the ridges on
the P4, not their length. While the lateral ridges of M.
arcuatus are both short and weak, the lateral ridges of M.
woodi are described by Holtzman and Wolberg (1977) as
weak, although illustrations show them to be short with
high relief. M. woodi i.s therefore scored as having strong
lateral ridges.
8.  P4 roots
0—equal or subequal in size
1 —anterior root larger than the posterior root
9.  Prominence of the P4 posterolabial shelf
0—anterior crown width greater than posterior crown
width
1—posterior crown width greater than or equal to ante-
rior crown width
10.  Mt number of external cusps/mm crown length
0—2.0 or fewer cusps/mm
1—between 2.0 and 3.0 cusps/mm
2—3.0 or more cusps/mm
Cusp counts, useful in the description of
multituberculate teeth, are problematic as a character in
phylogenetic analysis, due to their strong relationship to
body size in the Taeniolabidoidea (Cole and Krause, 1988).
The same is true of tooth measurements; molar lengths arc-
not independent of body size or of each other. The propor-
tion of cusps/mm makes use of cusp counts and length
measurements, but describes the crowding and proportion-
ate size of the cusps, a feature that has phylogenetic
significance and i.s independent of body size. Cusps/mm is
scored separately for the M,, M1, M,, and M2 in this data
4
PALEOHIOS, VOL 18, NUMBERS 2&3, 1998
set. These characters have not yet been used sufficiently to
establish their independence. They appear dependent in
this ciata set, but the low number of taxa in this analysis and
low variability within the Microcosmodontinae make this
study a poor test of independence.
11.  M, cusp rows diverge posteriorly (or converge an-
teriorly)
0—rows are parallel; do not diverge
1—diverge
12.  M, external cusps/mm crown length
0-2.0 or fewer cusps/mm
1—more than 2.0 cusps/mm
13.  M, and M, with notch posterior to internal cusp
row
0—no notch
1—notch present
The outline of the occlusal surface of the tooth has a
notch or indentation at the base of the last internal cusp.
This character describes the "sigmoidal and strongly
angled" ridge of Krause (1977, p. 27), but is described here
as a notch rather than a ridge because the notch is visible
even in worn teeth, while the ridge delineating it may be
worn away. The best illustration of the feature is that of
Pentacosmodon (Jepsen, 1940, p. 340).
14.  P^/Mj crown length ratio
0—greater than or equal to 1
1—less than 1
15.  Mt/M, crown length ratio
0-less than 1.5
1 — 1.5 to 2.0, inclusive
16.  M1 median cusps/mm crown length
0—less than or equal to 2.0
1—more than 2.0
While external cusps are used for the cusp/mm ratio on
the lower molars, and internal cusps on the M', the median
row cusp count of the M1 is less variable among specimens
referred to a single species, and so is used for this ratio.
17.  Body size: M, length
0—less than 2.0 mm
1—2.0 to 3.0 mm, inclusive
2—more than 3.0 mm
Length of the M,  rather than of the M1 is used to
represent body size because Microcosmodon is best known
from complete lower mandibles. Upper and lower denti-
tions have never been found in association, and all assign
merits of upper molars to the genus arc tentative.
18.  Length of internal cusp row of the M1, as a per-
centage of total M1 crown length
0—internal row absent
1—less than 50% length of crown
2—50% to 80% length of crown, inclusive
19.  Median and external cusp rows of the M1 diverge-
anteriorly
0—rows are parallel; do not diverge
1—rows diverge anteriorly
20.  M2 internal cusps/mm crown length
0—1.9S or more cusps/mm
1—fewer than 1.99 cusps/mm
21.  P4/M' crown length ratio
0—0.8 or greater
1-less :han().8
22.  M'/M2 crown length ratio
0—1.5 or less
1— grearer than 1.5
23.  Molar cusp shape, (Ml/1)
0—conical
Tabic 1. Matrix of 10 taxa scored for 26 dental characters used in phylogenetit analysis. "*" indicates character states 0 and 1. Two
characters have two possible character states: enamel on the lower incisor of (.'. jiiditbac is not restricted, but its relative thickness is
unknown, and the internal cusp row of M's tentatively referred to M. harleyi may be greater or less than 50% of total tooth length.
Pentacosmodon pronus
Microcosmodon  conus
Microcosmodon  woodi
Microcosmodon  rosei
Microcosmodon arcuatus
Microcosmodon harleyi
Stygimys  kuszmauli
Nemegtbaatar gobiensis
Paracimexomys  robisoni
Cimexomys judi:hat ¦
			11111	11112	22222	2	
12345	678910		12345	67890	12345	(.	%complele
32120	1010	0	00110	71???	??200	0	77%
22] 10	1101	I	11111	1110?	77301	1	88%
72101	0000	¦p	?????	1710?	1721?	:	58%
22121	21?1	2	11711	11100	70371	0	85%
???00	1110	1	1?11?	?1?7?	77211	0	58%
???00	1011	1	01111	11*0?	17311	r:	81%
32120	0000	0	10001	02101	11111	l	100%
021 11	00?0	0	1000?	02101	11271	P	85%
? ? ? 1 ?	777?	1	10070	01010	00071	l	62%
3*000	0010	2	00000	00100	00011	0	100%
WEIL- NEWMICROCOSMODON SPECIES
1—quadrangular
2—subcrescentic
3—crcsccntic or recurved
Molar cusp shape may or may not be independent of the
cusp/mm ratio, depending on the group under study. In
general, crcsccntic cusps have anteropostcriorly elongated
bases, and therefore the cusp/mm ratio is low. The lower
molars of Microcosmodon, however, have proportionally
high cusps with narrow bases, better described as recurved
than crcsccntic. Therefore molar cusp shape md cusp/mm
ratios have been recognized as separate characters in this
study.
24.  Presence/absence of multiple accessor/ roots on
the Ml/1
0—present
1—not present
25.  M1 cusp arrangement
0 cusps in internal and external rows aligned across the
tooth
1—cusps staggered
26.  Crown ornamentation; molar cusps are grooved or
ridged
0—no ornamentation
I—grooves or ridges present
Omitted Characters
Several dental characters useful in determining
multituberculate relationships are omitted from this list, as
they do not vary among the taxa considered for this study,
in those instances in which the state is known. Among the
characters that vary considerably within the ( imolodonta is
the vertical position of the unworn serrate edge of the P , in
relation to the unworn molar row: in Pcntacosmodon promts,
Microcosmodon conns, Stygimys kiiszinaiili Sloan and Van
Valen, 1965, Nemegtbaatar gobiensis Kielar -Jaworowska,
1974a, and dmexomys judithae Sshni, 1972 the serrate edge
of the P4 is level with or below the level of the nolar occlusal
surface. Mention of the presence or absence of posterobasal
cusps on the 1'4 has also been omitted: posterobasal cusps are
present on the P(s of M. harleyi, S. kuszmaul':, N.gobiensis,
Paracimexomys robisoni Eaton and Nelson, 1991, and C.
judithae.
TAXON CHOICK
These characters were scored for Microcosmodon conns, M.
woodi, M. rosci, M. arcnatns, M. harlcyi, Pcntacosmodon
pronus, Stygimys knszmattli, Ncmcgtbaaiar gobiensis,
Cimexomysjudithae, and Paracimexomys robisoni(Table 1).
Acheronodongarbani'xs too incomplete to be useful in this
study, and was not included. .S'. kuszmauli, a non-
microcosmodontine eucosmodontid, is particularly useful
because of its North American distribution and 100% com-
pleteness with respect to this matrix. N. gobiensis was
included as an hypothesized ancestor (Kielan-Jaworowska,
75
A
•C
69
54
100
100
100
*C
100
100
Pcntacosmodon pron ns
Microcosmodon conns
M. rosci
M. harlcyi
M. arcnatns
M. woodi
Nemegtbaatar gobiensis
Stygimys kuszma nli
dmexomys judithae
Paracimexomys robisoni
Pcntacosmodon pron ns
M. con us
M. rosci
M. harlcyi
M. arcnatns
M. woodi
Nemegtbaatar gobiensis
Stygimys knszmanli
Cimcxomys judithae
Paracimexomys robisoni
Figure 1. (A) Strict consensus of 4 .shortest trees showing
microcosmodontinc relationships. Decay indices are shown above
the node, bootstrap values below. Node A = Microcosmodontinae;
Node C = Microcosmodon; "Microcosmodon " woodi falls outside the
Microcosmodontinae, and may be more closely related to other
eucosmodontids (Node B). (B) A 50% majority rule consensus
shows M. conns and M. rosci forming a more weakly supported
clade, in addition to those shown in the strict consensus.
1974b)  of Microcosmodon  and   other  North  American
eucosmodontids.
Designated outgroups are (,". judithae, which is 100%
complete with respect to this matrix, and /'. robisoni, which
is 58% complete. Simmons's (1993) study shows
Paracimexomys to be one of the most primitive members of
the Cimolodonta (Ptilodontoidca + Tacniolabidoidea),
clearly outside of the relatively derived clade that contained
the taxa here considered as the ingroup. Subsequent work
(Weil, in preparation) corroborates this finding, although
Paracimexomys is paraphylctic, and a single species, P.
robisoni, is used here. Cimcxomys is also a paraphylctic
taxon (Weil, in preparation). Although Simmons's (1993)
phylogeny positioned Cimcxomys within a clade that also
included the taxa considered here as ingroup taxa, she
combined characters of four species to score the genus.
When species of Cimcxomys are scored separately for a
dental data set, C. judithae clearly falls outside the clade
6
PALEOHIOS, VOL. 18, NUMBERS 26-3, 1998
containing this study's ingroup taxa, although it is more
derived within the Cimolodonta than P. robisoni (Weil, in
preparation).
RESULTS
An exhaustive search resulted in four shortest trees of 41
steps, with consistency indices of 0.683, retention indices of
0.683, and rescaled consistency indices of 0.466. The strict
and majority-rule consenses of these trees are shown in
Figure 1. The consensus trees indicate two significant
relationships. Firstly, as surmised by Johnston and Fox
(1984), specimens previously called Microcosmodon woodi
tall outside the group of species that constitute a possibly
monophyletic genus Microcosmodon, and should be given a
new generic name. Secondly, the subfamily
Microcosmodontinae is monophyletic if M. woodi is re-
moved.
Although neither consensus shows Microcosmodon ex-
clusive of M. woodi to be monophyletic, they do not
preclude it from being so. The polytomy at
Microcosmodontinae (node A, Figure 1A) is the result of
missing data for M. arcuatus, which is only 58% complete
with respect to this matrix. If it is assumed that M. arcuatus
has intermediate enamel ultrastructure (character 1, state 1)
and an M, with more than 2 cusps/mm crown length
(character 12, state 1) an exhaustive search produces two
shortest trees of 42 steps. Microcosmodon is monophyletic
in both. While the states introduced are not different from
what would be expected in an animal assigned to
Microcosmodon, the scoring of these two characters incor-
porates two assumptions about M. arcuatus that may be
incorrect. Therefore, M. arcuatus can only be questionably
referred to Microcosmodon (node C, Figure 1A) until more-
is known about it.
Hypotheses of relationships among multituberculates
are difficult to convincingly confirm or disprove by phylo-
genetic analysis, partly because most species are known only
from isolated teeth, which limits available data, and partly
because analyses using extra-dental characters, where they
are available, (Simmons 1993; Rougier et al., 1997) seem
to indicate high levels of homoplasy in dental data. Com-
pared with some published trees of living and fossil mam-
mals, the tree in Figure 1A is poorly supported. Three of
five nodes have decay indices of one, meaning that for each
node at least one tree that is only one step longer does not
contain the clade (Bremer, 1994). One of these three nodes
has no bootstrap support; the other two have bootstrap
values indicating that they are present in under 70% of trees
generated by random sampling of the character set
(Felsenstein, 1985). The tree is, however, better resolved
than and as well supported as other multituberculate phy-
togenies derived from strictly dental data sets. I therefore
feel justified in using the results of this study in revising the
diagnoses of the Microcosmodontinae and of
Microcosmodon.
Character state changes at nodes A and C (Figure 1A)
were determined using both the ACCTRAN and
DELTRAN character optimization options offered by
PAL'P. One character state transformation, the increase of
M1 cusp/length ratio from less than 2 cusps/mm to more
than 2 cusps/mm, may diagnose either node (taxon) de-
pending on which optimization is used, since the M1 of
Pcntacosmodin has not been described. The character is
thus not used in either diagnosis, because although the M'
cusp/length ratio of Microcosmodon is greater than 2.0,
that of Pcutacosmodon is unknown.
This analysis does not strongly support any hypothesis of
relationships vvithin the genus Microcosmodon, although all
possess at tapomorphies or combinations of
synapomorphies indicating that they are not metataxa. The
majority-rule consensus tree (Figure IB) shows that the
two youngest species, M. con us and M. rosci, may be sister
taxa, and that the new species described below, M. harlcyi,
may be more closely related to them than to M. arcuatus.
SYSTEMATIC PALEONTOLOGY
Suborder Taeniolabidoidea (Granger and Simpson,
1929)
Family Eucosmodontidae (Jepsen, 1940)
Subfamily Microc:osmodontinae Holtzman and Wolberg,
1977
Included taxa Acbcrouodon Archibald, 1982;
Microcosmodon Jepscn, 1930; Pentacosmodon Jepsen, 1940.
Revised Diagnosis—Eucosmodontids with the P4 re-
duced and shorter than the M,; P crown height subequal
to or greater than crown length; M, and M, with an
inflection of the enamel ridge at the posterior of the tooth,
forming a notch posterior to the ultimate cusp of the
internal row.
Discussion—Molt/.man and Wolberg (1977) defined the
Microcosmodontinae with a suite of characters, some of
which, like the "greatly enlarged I," (p. 5) apply to the
Eucosmodontidae in general, and others of which, such as
"poorly developed lateral ridges" (p. 5) apply to only one
of the genera initially included. Subsequent descriptions of
new species have disqualified the presence ol accessory
roots on the M( and M1 as a diagnostic character (Johnston
and Fox, 1984; this paper). The enlarged pterygoid fossa
may also diagnose the subfamily, but is only known in
specimens ot Pentacosmodon promts and Microcosmodon
conus.
All non-microcosmodontine eucosmodontids have an
elongated, proportionally large P, and lack the notch
described in the occlusal outline of the M. and M,, retain-
ing the primitive states of these characters. Holtzman and
Wolberg (1977) used "Eucosmodontinac" to group
"eucosmodontids exclusive of Microcosmodon and
Pentacosmodon^ (p. 4). As there is no evidence that these
taxa form a monophyletic group, I have avoided the use of
this term.
WA7/.~,V£1V MICROCOSMODON SPECIES
7
The diagnosis given above is based on a phylogenetic
analysis using 26 dental characters useful in multituberculate
systematics in general, and for this family in particular.
Although all described microcosmodontines have six or
fewer serrations on the P+, serration and cusp counts were
not used in this analysis and thus are not included in the
diagnosis. An M' median cusp/mm length ratio greater
than 2.0 may also diagnose this family, but the character
was not included in the formal diagnosis because it may
diagnose the genus Microcosmodon when different character
optimizations are considered.
Microcosmodon woodi falls definitively outside the
Microcosmodontinae, and should, as initially proposed by
Johnston and Fox (1984), be assigned to a different genus
and subfamily.
Microcosmodon Jepsen, 1930
Microcosmodon Jepsen, 1930; Krause, 1980
(non) Microcosmodon I loltzman and Wolberg, 1977 (in part)
Type species— Microcosmodon conus Jepsen, 1930
Included species -Microcosmodon rosei Krause, 1980;
Microcosmodon harleyi, new species.
Questionably referred species—Microcosmodon arcuatus
Johnston and Fox, 1984
Revised Diagnosis—Microcosmodontine with more
than two external cusps/mm crown length of the M,; P
posterior crown wider than the anterior crown; molar cusps
semicrescentic to strongly recurved in shape; enamel ultra-
structure intermediate between small prismatic enamel with
circular prisms and gigantoprismatic enamel with arcade-
shaped prisms.
Discussion—" Microcosniodoii" woodi Holt/.man and
Wolberg, 1977 is removed from the genus Microcosmodon
and from the subfamily Microcosmodontinae. Of the lower
dentition, only the P4 and 1( are known, and as noted by
Johnston and Fox (1984) the P4 is neither proportionally
small nor is the crown high and rounded, which sets it apart
from the Microcosmodontinae. Phylogenetic analysis shows
"A/." woodi grouping with Stycfimys and Newajtbaatar
(node B, Figure 1A), a node supported by transformations
in six characters.
Microcosmodon} arcuatus almost certainly belongs in
Microcosmodon, although its incompleteness with regard to
the matrix used in this study leads to a polytomy at
Microcosmodontinae (node A, Figure 1A). It i; therefore
questionably referred to Microcosmodon until further re-
search confirms more convincingly its membership in the
clade.
Microcosmodon harleyi, new species
Figure 2A, B, C; Figure 3; Figure 4; Table 2
Type Specimen-UCMP 144449, right P4
Type Locality-UCMP locality V75193, Yellow Sand
Hill 7, within the Garbani channel fades, Tullock Forma-
tion, Garfield County, Montana, USA.
Referred Specimens-P4: UCMP 124444 - 124446,
144434, 144435, 144448, 144449, LACM 32942. M,:
UCMP 124447, 144438, 144439 - 144442, 144450. M,:
UCMP 144451, 144452. P4: UCMP 144436, 144437,
144443 - 144445. In addition to the hypodigm, four upper
first molars are tentatively referred to M. harleyi: UCMP
143668, 144475 - 144477.
Occurrence—In addition to the type locality, UCMP
localities V73080, V73120, V72130, V74122, and V72125
(LACM locality 3099). All localities are within the Garbani
channel fades, Tullock Formation, Garfield County, Mon-
tana, USA, and are of Puercan age.
Etymology—Named in honor of Mr. Harley Garbani,
who has invested years collecting in the area from which
this species is known.
Diagnosis—A species of Microcosmodon distinguished
from other members of the genus by the presence of strong
lateral ridges on the P4 in combination with parallel rather
than posteriorly divergent cusp rows on the Mr
Description—
P4: (Figure 2A-C) The profile of the serrate crest is high
and rounded, departing from a true half-circle in having a
relatively long, flat anterior edge. All unbroken specimens
show six serrations, with the apex of the crest at approxi-
mately the third serration. The first and second serrations
are widely separated, the first occurring halfway between
the base of the crown and the third serration. The second
serration is closer to the third; the third and fourth serra-
tions are very close together. The fifth and sixth serrations
are prominent and cusp-like on all unworn specimens, and
the sixth leans slightly posteriorly.
Although some of the specimens are chipped or broken,
all probably had four internal ridges extending anteriorly
and inferiorly from the second through fifth serrations. The
first internal ridge is always short, extending from the
second serration and not reaching the first, and is faint on
some specimens but quite prominent on others. The other
ridges are prominent and relatively long, extending inferior
to the first serration; the third ridge is the longest. The first
ridge is directed more inferiorly than the others but never
intersects the second, the second and third ridges are
parallel, and the fourth is widely separated from the others
and subparallcl. On UCMP 124446 the fourth ridge bends
sharply away from the third, and is discontinuous.
As is the case with the internal ridges, four external
ridges were probably present on all specimens before wear,
extending anteriorly and inferiorly from the second through
fifth serrations. Again, the first ridge never extends below
the first serration, is not parallel with the other ridges, and
is of variable prominence. The other ridges are prominent,
and the third and fourth ridges are longest.
An anterobasal concavity for the P, is present on all
specimens, but is much more pronounced on some than on
others. There arc two roots, the anterior larger than the
posterior. The anterolabial lobe does not extend far inferi-
PALEOBIOS, VOL. 18, NUMBERS 2&3, 1998
D
Figure 2. Right P4s of microcosmodontines from the Tullock Formation: UCMP 144449, holotypc of Microcosmodon barleyisp nov.
in (A) labial; (B) lingual; and (C) occlusal views. UCMP 116953, type specimen of Acbcnmadonjinrbnni\n{V)) occlusal; (E) labial; and
(!•') lingual views. Scale bar = 0.5 mm.
WEIL-NEW MICROCOSMOOON SPECIES
Figure 3. Molars of Microcosmodon barleyisp. nov.: tentatively referred right M's in occlusal view arc (A) UCMP 144476 from V73080
and (15) UCMP 143668 from V74122, displaying the differences between teeth from these sites. (C) left M, UCMP 144440 in occlusal
view; (D) right M, UCMP 144452 in occlusal view; and (E) 3/4's lingual view, showing the ridge and notch posterior to the second
internal cusp. Scale bar = 0.5 mm.
10
PALEOHIOS, VOL. 18, NUMBERS 2&3, 1998
orly and is not inflated, sinking into the anterior root with
low relief. The posterolabial ledge is large in proportion to
the rest of the tooth, so that in occlusal view the posterior
of the tooth crown is as wide or wider than the anterior.
The least worn specimens show remnants of two cuspules
on the ledge, although this part of the tooth became worn
first in normal use.
M,: (figure 3C) Cusps 5:4. The average ratio of cusps/
mm is 2.42. Unworn cusps are proportionally high and
pointed, with the exception of the first external cusp. The
first external cusp is the smallest and first worn. It is conical
and directed anteriorly, and is not well separated from the
second external cusp, which is also conical and may be erect
or directed anteriorly. The third external cusp is conical and
directed slightly posteriorly, and there is a large gap be-
tween the third and fourth cusps. The fourth and fifth
cusps are tilted sharplv posteriorly and are joined for about
half their height.
The internal cusps are taller than the external cusps. The
first internal cusp is variable, being well separated from the
second in specimens UCMP 144450, 144442, 144438,
and 144441, but less distinct and more cuspule-like in
UCMP 124447 and 144440. The first internal cusp is
shorter than the second through fourth, conical, and erect
where separate from the second. The second through fourth
internal cusps are subcqual in height, evenly spaced, coni-
cal, and tilted posteriorly.
There are no ridges or grooves in the enamel. The tips of
the cusps wear before the central valley is deeply worn, and
the external cusp row wears flat well before the internal
row. In occlusal view the tooth is slightly waisted, indented
between the third and fourth external cusps and between
the second and third internal cusps. The central valley of
the tooth is slightly sinuous in unworn specimens, and is
closed posteriorly by a low ridge of enamel joining the
posterior cusps. This ridge is inflected immediately poste-
rior to the last internal cusp (the "sigmoidal and strongly
angled" ri.igc of Krause, 1977, p. 27), creating a notch.
There is no enamel ridge joining the anterior cusps. There
arc two roots, oval in cross-section and about equal in size.
Most of the M(s do not appear to have been tightly
appresscd with the P4, as they lack an anterior facet.
M,: (Figure 3D, E) Cusps 3:2. The average ratio of
cusps/mm is 2.2. Both specimens arc unworn, although
neither ha,' roots. 'The cusps arc high and strongly recurved.
The first external cusp is clearly separated from the second
and third, which are joined for more than half their height.
The internal cusps are better separated on UCMP 144452
than on 1^4451. Internal cusps are not ridged or grooved,
but there arc faint grooves at the bases of the external
cusps. The cusp rows are parallel, and there is a wide central
basin, clos.'d posteriorly by a low enamel ridge connecting
the posteriormost cusps.   This ridge is strongly inflected
Figure4. 1" of Microcosmodon haricyisp. now: UCMP 144444, a left P4, in (A) lingual; (B) labial; and (C) occlusal views. Scale bar-0.5 mm.
WEIL-NEW MICROCOSMODON SPECIES
11
Table 2. Measurements of isolated molars and premolars of Microcosmodon harkyisp. nov. Lengths and widths are given in millimeters.
" " indicates data missing due to breakage, " ¦" indicates the presence of a cuspule, "?" indicates thai cusps or cuspules are not countable
due to wear, and "6?" indicates that serration count is questionable due to breakage.
.Specimen
Site number       Side      I,cngth      Width     Cusps/serrations
1'4
I.ACM
UCMP
UCMP
UCMP
UCMP
uc:mp
UCMP
UCMP
M,
UCMP
UCMP
UCMP
UCMP
UCMP
UCMP
M2
t'CMl'
UCMP
P4
UCMP
UCMP
UCMP
UCMP
UCMP
M
UCMP
UCMP
UCMP
UCMP
32942
124444
124445
124446
144434
144435
144448
144449
124447
144438
144440
144441
144442
144450
144451
144452
144436
144437
144443
144444
144445
143668
144475
144476
144477
3099
V73080
V73080
V73080
V74122
V74122
V73120
V75193
V73080
V73080
V73080
V73080
V73080
V72130
V72130
V74122
V74122
V74122
V73080
V73080
V73080
V74122
V74122
V73080
V73080
1.
R
R
R
R
1.
R
R
I.
I.
I.
I.
R
1.
R
R
R
I.
R
I.
R
R
1.98
2.07
1.91
2.08
2.06
1.91
1.88
1.88
1.87
2.04
2.05
2.20
2.17
2.10
1.36
1.34
1.66
1.67
1.76
1.83
1.53
2.2
2.28
2.50
2.45
.64
.91
.52
.87
.77
.67
.69
.87
.92
.99
.99
1.04
1.01
1.14
1.16
.72
.69
.77
.74
.74
1.16
1.24
1.39
1.41
6 =
6
6;
5:4
5:4
5:4
5:4
5:4
5:4
3:2
3:2
2:4
2:4
2:4
2:4
l+:4
4+:5:r
4:5:?
5:5:3
5+:5:>
immediately posterior lo the second internal cusp, creating
a pronounced notch in the occlusal outline of the tooth.
P4: (Figure 4) Cusps 2:4, except for UCMP 144445, on
which the anterior anteroextcrnal cusp is reduced to a
cuspule. Cusps are all conical, with no grooves or ridges on
the cusps or sides of the crown. The third and fourth
medial cusps are slightly less well separated than the others.
The fourth medial cusp is the largest and highest on the
tooth, and is placed at about only two-thirds of the length
of the crown, so that in profile the tooth appears low and
posteriorly elongated. Morphology is variable posterior to
the last medial cusp. On the largest specimen, UCMP
144444, two ridges run posteriorly from the lasi medial
cusp, each culminating in a posterobasal cusp. The external
posterobasal cusp is placed higher on the crowr than is the
internal, bin they are linked posteriorly by a ridge, creating
a small but relatively deep basin. Other specimens follow
the same general plan, but the ridges and basin are less
accentuated.
In occlusal view the tooth shape is close to rectangular,
with slight waisting between the second and third cusps of
the medial row. There arc two roots, anteroposteriorly
compressed and oval in cross-section, and equal in size
except for some thickening ol the anterior root proximal to
the tooth crown. With wear, the internal posterobasal cusp
is worn down lo a flat facet, accompanied by wear on the
tip of the last medial cusp.
M1: (Figure 3A, B) Cusps 4-5+:5:3-r. The average ratio
of cusps/mm is 2.1. These teeth are only tentatively re-
ferred to M. barleyi, as they differ in the ratio of median
cusps/mm crown length and in the degree of cusp
cresccntism from other M's assigned to the genus
Microcosmodon. Within this admittedly small sample, the
M's from V73080 (Garbani Quarry) and V74122 (Biscuit
12
PALEOBIOS, VOL. 18, NUMBERS 2&3, 1998
Springs) sites differ in size, and also in the number of cusps
in the external row. On the teeth from locality V73080
(Figure 3A), which are larger, the first external cusp is low
and cuspule-likc, formed on a ridge running anteriorly from
the apex of the second external cusp. The second through
fifth cusps of the external row are conical in shape, evenly
spaced and well-separated, and increasing in height posteri-
orly. The fifth and last external cusp is joined to the last
median cusp by a ridge thai bears a single cuspule, which is
significantly larger on UCMP 144477. The smaller teeth
from V74122 (Figure 3B) have four external cusps, with a
tiny cuspule anterior to the first cusp and no cuspule
posterior to the last external cusp. The labial side of the
cusps is unornamented; there are some faint grooves in the
enamel at the cusp bases on the lingual side. On the least
worn specimen the valley between the external and median
rows is slightly sinuous, as the cusps of the rows arc
staggered in relative positions, and is closed at either end by
ridges joining the first and last cusps of the rows. The
median and external cusp rows are parallel, except for the
first external cusp of the teeth from Garbani Quarry, which
is medial to the rest of the external row. There are five
median cusps, with no ridges or grooves. The first may be
conical, but the second through fifth median cusps are
strongly cresccntic. The third median cusp has an
anteroposteriorly elongated base and is so strongly cresccn-
tic that it appears recumbent. The last median cusp is
connected to the internal cusp row by a low ridge of
enamel. The internal cusp row is about half the length of
the tooth, extending to the posterior or middle of the base
of the third median cusp. It is more a curved, crcnulatcd
ridge than a row of separate cusps, although the least worn
specimen (UCMP 144476) shows 3 cusps or cuspules.
The occlusal surface of the tooth is flat or slightly
concave. The anterior end of the tooth is square, showing
no appression facet where it would have been in contact
with the 1". There are two roots and no accessory roots.
The anterior root is round in cross-section and directed at
an angle labially, while the posterior root is very
anteroposteriorly compressed and is directed at an angle
lingually.
Discussion—The phylogenetic diagnosis does not allow
for strictly descriptive detail that may be used in "keying
out" specimens. As at least three microcosmodontine spe-
cies occur in a relatively narrow range of time and space, the
early Paleocene of eastern Montana and of southern
Saskatchewan, direct points of comparison are desirable.
The P( of Microcosmodon harleyimay be distinguished from
that of M. arcuatus by having a lower arc, less inflation of
the crown, much stronger and longer lateral ridges, and a
proportionally larger posterolabial shelf. The My of M.
harleyi may be distinguished by the parallel cusp rows,
lower cusp count than that of M. arcuatus, and the pres-
ence of recurved rather than semicrcsccntic cusps. The P4
of M. harleyi is distinguished from the single specimen of
Achcronodon aarbani (Figure 2D-F) by its larger average
length, significantly greater width, and proportionally
smaller posterolabial shelf. The P4 of M. harleyi cannot be
distinguished from that of the Long Fall
microcosirodontinc as described and illustrated by Fox
(1989).
All specimens of Microcosmodon harleyi are isolated teeth.
As is traditional when a more complete specimen is not
available, :he designated holotvpe is a particularly well
preserved isolated P4. This convention is awkward for two
reasons: (I) the diagnosis of M. harleyi depends on a
combination of characters of the P4 and M:; and (2) my
experience is that the molars of some multituberculate
species are as phylogenctically informative as, if not more
informative than, the 1'4. Where they have been described,
the M( ami M, of Microcosmodon species arc derived and
distinctive. If associations of isolated teeth have been made
correctly by myself and previous workers, the M, differs
sufficiently among the species from which it has been
described to be used in species identification.
DISCUSSION
Comparison to Previous Phylogenetic Studies
While the o Hiclusion that the subfamilv Microcosmodontinae
is monophy letic if"Microcosmodon" woodiis removed confirms
the results of some earlier studies (Jepsen, 1940; Kielan
f aworowska, 1974a; Holtzman and Wolberg, 1977); it is not
consistent with the phylogenetic analysis of Simmons (1993).
Her multituberculate phylogeny places Pentacosmodon in
"Group X" (p. 161) with Asian genera, while Microcosmodon
is more closely allied with North American Cimexomys and
Sty/jimys.    As    a    result,    she    recommended    that
Microcosmodontinae be removed from usage.
Two main factors contribute to our disparate results.
First, Simmons (1993) included many more taxa, as her
study was more comprehensive than this one and had
different objectives. One of her conclusions was that the
addition of more taxa, including incomplete taxa, affected
relationships, and from that standpoint her analysis is supe-
rior to this one. The taxa she used, however, were genera,
scored using combinations of characters obtained from
different species. This is not significant for Pentacosmodon,
which is monotypic, or for Stygimys, as she also used only S.
liuszmanli. For Cimexomys, however, she used a combina
don of characters from four species, and subsequent work
(Weil, in preparation) shows the genus to be paraphyletic.
To score Microcosmodon she used characters from three
species, including "A/." woodi, which this study indicates
does not belong to the genus. This inevitably resulted in
inaccuracies in the scoring of characters.
Second, she used a significantly different character set,
combining dental and cranial characters. Some of her char-
acters, such as cusp counts and multiple tooth length
measurements, arc very closely related to each other and to
body  size  (Weil,   1996).   More  significantly,  the  dental
WEII.-NEW MI(:ROCOSM()IX)N SPECIES
13
portion of her character set did not include characters
traditionally used to diagnose the Microcosmodontinae,
such as Pj/Mj ratio (proportional size of the P4), or P)
shape, nor did it include characters describing molar shape,
which proved to be significant in this study. Both
Microcosmodon and Pentacosmodon were extremely incom-
plete with regard to her matrix. Again, her study had less
specific objectives than this one, and the advisability of
making a specific recommendation to drop
Microcosmodontinae from use on the basis of the results of
such a study is debatable.
Rougier et al. (1997) adopted a modified version of
Simmons's (1993) character set, and accepted her
combinative and incorrect scoring of many characters. Their
analysis also indicates a distant relationship of
Microcosmodon and Pentacosmodon, although tie two gen-
era differ for only two characters; Mt cusp formula, which
was not used in this study, and presence of the P , which, as
discussed above, varies within Microcosmodon.
The Acheronodon Problem
Acheronodon has not been included in any phylogenetic
analysis due to its incompleteness. Archibald [1982) de-
scribed an isolated and damaged multituberculate P (shown
here in figure 2D-F) from the lowermost Tullock Forma-
tion. Noting his reservations on the inadvisable practice of
naming a new taxon on a single tooth, Archibald gave the
unique P the name Acheronodon garbani. He placed
Acheronodon in the Microcosmodontinae, and qualified his
designation of a new genus with a caveat that "fie recovery
of additional material of Acheronodon may show that it is
congeneric with Microcosmodon" (p. 70).
Microcosmodon harlcyi has since been identified from
multiple sites in the Tullock Formation. These sites, be-
longing to the "Garbani channel" mapped by Archibald
(1982), are more than 30 in higher than and several miles
to the southwest of the type locality of Acheronodon, and
occur in a channel complex that suggests considerable
time averaging. As noted in the description of M. harlcyi,
the lower fourth premolars of the two are different in
several particulars, but the hypodigm of Acheronodon is
inadequate to provide data on much morphology' or on any
variation. A larger sample of either species is needed, and a
sample containing teeth of Acheronodon other than the P4
would be particularly useful. Unfortunately, such a sample
may be difficult to obtain, as microcosmodontines arc never
the most common taxa in a fauna: of the thousands of
isolated mammal teeth recovered from the Garbani channel
facies, only 25 have been identified as microcostnodontine
and all are referred to M. harlcyi. It is possible that some or
all of the teeth here described as M. harlcyi belong to the
same species described as A. garbani. If this is the case, that
species belongs in the genus Microcosmodon, as Archibald
(1982) suspected.
Biogeography
Regardless of the generic designation ofAcheronodon, there
is no doubt that the P4 belongs to a microcosmodontine, as
does the P4 reported from the Long Fall horizon (Fox,
1989). Micro-cosmodontines thus appear in North America
at the beginning of the Paleocene or in the latest (Cretaceous;
that they are not found in the basal Paleocene, extremely
well-sampled Bug Creek assemblages (Lofgren, 1995) may
indicate that the single occurrence of A. garbani is close to
their first actual appearance in Montana. Microcosmodontine
species richness is as high at the beginning of the Paleocene
as at the end, suggesting that their diversification predates
their appearance in the fossil record. Their geographic range
is limited: aside from a questionably referred isolated incisor
from the Goler Formation of California (McKenna et al.,
1987) and a possibly misidentified isolated incisor from
England (Sloan, 1981), microcosmodontines are known
only from the northern portion of North America's Western
Interior.
Eucosmodontid multitubereiilates arc thought to have
dispersed from Asia to North America (Kielan-Jaworowska,
1974a; 1974b) after the divergence of microcosmodontines
from other eucosmodontids (Holtzman and Wolberg,
1977). The results of this study are consistent with this
hypothesis, although they should not be construed as strong
support. The timing of microcosmodontine divergence is
controversial. Kielan-Jaworowska (1974b) thought all
North American eucosmodontids derived from
Ncmegtbaatar, known from the late Campanian Barun
Goyot Formation of Mongolia. Holtzman and Wolberg
(1977), however, argued that the microcosmodontines
could not be derived from Ncmegtbaatar, and thus must
have diverged earlier. A third possibility was introduced by
Simmons's (1993) phylogenetic study, which indicated
that the Eucosmodontidae as originally conceived is not a
monophyletic group, and that Pentacosmodon and
Ncmegtbaatar &\x not closely related to Microcosmodon and
Eucosmodon. While this study indicates that Simmons's
(1993) results may not have been entirely accurate, it does
support the contention of Holtzman and Wolberg (1977),
as Ncmegtbaatar seems to be more closely related to "M."
woodi and Stygimys. When and whether
microcosmodontines could be derived from Ncmegtbaatar
or another eucosmodontid lineage is clearly a subject wor-
thy of future research.
ACKNOWLEDGMENTS
Dr. William (Clemens, Dr. Anthony Barnosky, and Dr.
Patricia Holroyd read and commented on early versions of
this manuscript. Dr. David Krause provided a helpful review.
This study was funded by NSF grant EAR 9505841 and
earlier grants to W. A. Clemens, and by grants from the
University of California Museum of Paleontology. Photomi-
crographs were made on the UCMP's Environmental Scan-
14
PALEOBIOS, VOL 18, NUMBERS 2&3, 1998
ning Electron Microscope with the assistance of Dr. Karen
Wetmore Grycewicz. This is UCMP Contribution No.
1672.
LITERATURE CITED
Archibald, J. D. 1982. A study of Mammalia and geology across
the Cretaceous-Tertiary boundary in Garfield County, Mon-
tana. University of California Publications in Geological Sci-
ences 122, 286pp.
Archibald, J. D., VV. A. Clemens, P. D. Gingerich, D. VV. Krause,
K. II. Lindsay, and K. D. Rose. 1987. First North American
Land Mammal Ages of the Cenozoic era. pp. 24-76 in M. O.
Woodburnc (cd.) Cenozoic Mammals of North America.
University of California Press, Berkeley.
Bremer, K. 1994. Branch support and tree stability. Cladistics
10:295-304.
Bryant, L.J. 1989. Non-dinosaurian lower vertebrates across the
Cretaceous-Tertian' boundary in Northeastern Montana. Uni-
versity of California Publications in Geological Sciences 134,
107pp.
Carlson, S. J. and D. W. Krause. 1985. Enamel ultrastructure of
multituberculate mammals: An investigation of variability.
Contributions from the Museum of Paleontology, The Uni-
versity of Michigan 27(1): 1-50.
Cole, T. M., Ill and D. W. Krause. 1988. Interspecific relation
ships   between   tooth   size  and  cusp   numbers  in   the
Multituberculata (Mammalia). Journal of Vertebrate Paleon-
tology 8(supp. to 3):12A.
Eaton, ). G. 1995. Cenomanian and Turonian (Early Late Greta
ccous) multituberculate mammals from southwestern Utah.
Journal of Vertebrate Paleontology 15(4):761-784.
Eaton, J. G. and M. K. Nelson. 1991. Multituberculate mammals
from the Lower Cretaceous Cedar Mountain Formation, San
Rafael Swell, Utah. Contributions to Geology, University of
Wyoming 29(1): 1-12.
Eriksson, T. and N. Wikstrom. 1995. AutoDecay, Version 3.0.1.
Felsenstein, J. 1985. Confidence limits on phylogenies: An ap-
proach using the bootstrap. Evolution 38(4):783-791.
Fox, R. C 1989. The Wounded Knee Local Fauna and mamma
lian   evolution   near  the   Cretaceous Tertiary   boundary,
Saskatchewan,  Canada.   Palaeontographica  Abteilung A
208:11-59.
Fox, R. C. 1990. The succession of Paleocene mammals in
western Canada, pp. 51-69 in T. M. Bown and K. D. Rose
(eds.). Dawn of the Age of Mammals in the northern part of
the Rocky Mountain Interior, North America. Geological
Society of America Special Paper 243.
Granger, W. and G. G. Simpson. 1929. A revision of the Tertiary
Multituberculata. Bulletin of the American Museum of Nam
ral History 56(9):601-676.
Holt/man, R. C. 1978. Late Paleocene mammals of the Tongue
River Formation, western North Dakota. Report of Investiga-
tion 65, North Dakota Geological Survey. 88 pp.
Holtzman, R. C. and D. L. Wolberg. 1977, The
Microcosmodontinae and Microcosmodon woodi, new
multituberculate taxa (Mammalia) from the I .ate Paleocene of
North America. Scientific Publication of the Science Museum
of Minn jsota new series 4( 1): 1 -13.
Jepsen, G. L.  1930. Stratigraphy and paleontology of the Pale
ocene of Northeastern Park County, Wyoming. Proceedings
of the American Philosophical Society 69:463-528.
Jepsen, G.  L.   1940.  Paleocene faunas of the Polecat  Bench
Formation, Park County, Wyoming. Proceedings of the Ameri
can Philosophical Society 83(2):217 339.
Johnston, I'.A. and R. C. Fox. 1984. Paleocene and Late Greta
ccous mammals from Saskatchewan, Canada.
Palaeontographica Abteilung A 186(1-6): 163-222.
Kiclan-Jaworowska, Z. 1974a. Multituberculate succession in the
Late Cretaceous of the Gobi Desert (Mongolia), pp. 23 44 in
Z. Kiclan-Jaworowska (cd.). Results of the Polish Mongolian
Palaeontological Expeditions—Part V. Palaeontologia Polonica
30.
Kielan Jaw< rowska, Z. 1974b. Migrations of the Multituberculata
and the -ate Cretaceous connections between Asia and North
America. Annals of the South African Museum 64:231-243.
Krause, D. W. 1977. Paleocene multituberculates (Mammalia) of
the Roche Percee Local Fauna, Ravcnscrag Formation,
Saskatchewan, Canada. Palaeontographica Abteilung A 159( 1-
3): 1-36.
Krause, D. W. 1980. Multituberculates from the Clarkforkian
Land-Mammal Age, Late Paleocene-Early Eocene, of Western
North America. Journal of Paleontology 54(6): 1163-1183.
Krishtalka, .., C. C. Black, and D. W. Riedel. 1975. Paleontol-
ogy and geology of the Badwater Creek area, central Wyoming
Part 10: A late Paleocene mammal fauna from the Shotgun
Member of the Fort Union Formation. Annals of Carnegie
Museum 45(9): 179-212.
Lerbekmo, I. F. 1985. Magnetostratigraphic and biostratigraphic
correlations of Maastrichtian to early Paleocene strata between
south-central Alberta and southwestern Saskatchewan. Bulle
tin of Canadian Petroleum Geology 33(2):213-226.
Lofgren, D   L. 1995.  The Bug Creek problem and the Creta-
ceous-Tertiary transition at McGuire Creek, Montana. Uni
versify o " California Publications in Geological Sciences 140,
185 pp.
Maddison, W. P. and D. R. Maddison. 1992. MacClade 3.04.
Sinauer Associates, Sunderland, Massachusetts.
Matthew, W. D. and W. Granger. 1921. New genera of Pale-
ocene Mammals. American Museum Novitates 13, 7pp.
McKenna, M. C, J. H. Hutchison, and J. II. Hartman. 1987.
Paleocene vertebrates and nonmarine Mollusca from the Goler
Formation, California, pp. 31-41 in B. F. Cox (cd.). Basin
analysis and paleontology of the Paleocene and Eocene Goler
Formation, El Paso Mountains, California. Society of Eco-
nomic Paleontologists and Mineralogists, Pacific Section, Los
Angeles.
WEJL-ME WMICROCOSMODON SPECIES
15
Rougier, G. W., M. J. Novacek, and D. Dashzeveg. 1997. A New
Multituberculate from the Late Cretaceous locality Ukhaa
Tolgod, Mongolia. Considerations on Multituberculate inter-
relationships. American Museum Novitates 3191, 26 pp.
Sahni, A. 1972. The vertebrate fauna of the Judith River Forma-
tion, Montana. Bulletin of the American Museum of Natural
History 147:321-412.
Simmons, N. B. 1987. A revision of Taeniolabis (Mammalia:
Multitubcrculata), with a new species from the Puercan of
Eastern Montana. Journal of Paleontology 61(4)794-808.
Simmons, N. B. 1993. Phylogcny of Multitubcrculata. pp. 146-
164 in F. S. Szalay, M. J. Novacek, and M.C. McKcnna
(eds.). Mammal Phylogeny: Mesozoic Differentiation,
Multituberculates, Monotremes, Early Therians, and Marsu-
pials. Springer-Verlag, New York.
Sloan, R. E. 1981. Systematics of Paleocene mult tuberculatcs
from the San Juan Basin, New Mexico, pp. 127-160 in S. G
Lucas, J. K. Rigby Jr., and B. S. Kues (eds.). Advances in San
Juan Basin Paleontology. University of New M:xico, Albu-
querque.
Sloan, R. E. 1987. Paleocene and latest Cretaceous mammal ages,
biozones, magnetozones, rates of sedimentation, and evoltion
pp. 165-200 in J. E. Fassett and J. K. Rigby, Jr. (eds.). The
Cretaceous-Tertiary Boundary in the San Juan and Raton
Basins, New Mexico and Colorado. Geological Society of
America Special Paper 209.
Sloan, R. E. and L. Van Valcn. 1965. Cretaceous mammals from
Montana. Science 148:220-227.
Swisher, C. C, III, L. Dingus, and R. F. Butler. 1993. 40Ar/wAr
dating and magnetostratigraphic correlation of the terrestrial
Cretaceous-Paleogene boundary and Puercan Mammal Age,
Hell Creek—Tullock formations, eastern Montana. Canadian
Journal of Earth Sciences 30(9): 1981-1996.
Swofford, D. L. 1993. Phylogenetic Analysis Using Parsimony,
Version 3.1.1. Computer program distributed by the Illinois
Natural History Survey, Champaign Illinois.
Weil, A. 1996. Lying through their teeth: Dental characters,
body size, and a phylogeny of North American
multituberculates. Journal of Vertebrate Paleontology
16(supp. to 3):71A-72A.