Preferred Citation: Lofgren, Donald L. The Bug Creek Problem and the Cretaceous-Tertiary Transition at McGuire Creek, Montana. Berkeley:  University of California Press,  1995. http://ark.cdlib.org/ark:/13030/ft3m3nb1tx/

The Bug Creek Problem and the Cretaceous-Tertiary Transition at McGuire Creek, Montana Donald L. Lofgren UNIVERSITY OF CALIFORNIA PRESS Berkeley · Los Angeles · Oxford © 1995 The Regents of the University of California

Preferred Citation: Lofgren, Donald L. The Bug Creek Problem and the Cretaceous-Tertiary Transition at McGuire Creek, Montana. Berkeley:  University of California Press,  1995. http://ark.cdlib.org/ark:/13030/ft3m3nb1tx/

Acknowledgments

This research effort was part of a continuing University of California Museum of Paleontology (UCMP) field program headed by Dr. Wiliam A. Clemens. I thank those associated with the UCMP who helped to see this project to completion, most notably Dr. Clemens who suggested and supervised the project, and the members of UCMP field parties during the years 1984 to 1989 especially: Mark Goodwin, Dr. J.H. Hutchison, Kyoko Kishi, Dr. Laurie Bryant, Allen Tedrow, Matthew Colbert, James Foster, Dr. Anthony Runkel, and Dr. Zhexi Luo. Special thanks to Mark Goodwin and Kyoko Kishi for fossil preparation and to those who sorted screenwash concentrate from McGuire Creek.

I am grateful to the many local ranchers who allowed access to their land, and for their friendship and hospitality, including Lester and Cora Engdahl, Robert and Jane Engdahl, Judd and Jay Twitchell, Clay Taylor, and their families. Don and Marge Beckman of Fort Peck contributed the use of a motorboat which made fieldwork and fossil collection in areas far from roads logistically feasible.

For their helpful comments and discussions I thank, Dr. Bryant, Dr. Runkel, Dr. J. David Archibald, Dr. Carol Hotton, Dr. Douglas Nichols, Dr. J. Keith Rigby Jr., Dr. Carl Swisher, Dr. Lowell Dingus, Dr. Jennifer Hogler, Dr. Zhexi Luo, Dr. Judd Case, Dr. Richard Fox, Gayle Nelms, and Robert Dundas. Special thanks to Dr. Hutchison for hours of stimulating discussion on various aspects of the Cretaceous-Tertiary transition in eastern Montana.

Dr. Archibald, Dr. J.A. Lillegraven, and Dr. P. Sadler carefully reviewed the original manuscript and provided many helpful suggestions. Dr. Hotton kindly allowed me to report the age of a number of rock samples from McGuire Creek based on her analyses and interpretations of palynofloras.

Specimens were loaned by Dr. L. Van Valen of the University of Chicago and Mary Ann Turner of the Yale Peabody Museum. Financial support was provided by the Department and Museum of Paleontology, University of California at Berkeley (Annie M. Alexander Endowment); the University of California at Berkeley (Regents Fellowship and the Alfred F. Moore and Chella D. Moore Scholarship); National Science Foundation Grant no. BSR-85-13253 to Dr. Clemens and the Board of Trustees of the Raymond M. Alf Museum.

I especially want to thank my parents Lionel and Delores Lofgren, and my wife Laura, for their support during my academic studies, which made completion of this project possible.

Abbreviations

Abstract

The "Bug Creek Problem" refers to questions concerning the nonmarine vertebrate biostratigraphy of the upper Hell Creek and lower Tullock formations, an interval where significant faunal turnover is evident. This stratigraphic interval, which spans the Cretaceous-Tertiary transition, contains three distinct types of vertebrate assemblages, Lancian-Bug Creek-Puercan, and perhaps the youngest known records of dinosaurs in North America. Bug Creek assemblages are faunally intermediate to Lancian (Cretaceous mammals/dinosaurs) and Puercan (Paleocene mammals/no dinosaurs) assemblages, which suggests that faunal turnover during the Cretaceous-Tertiary transition was not abrupt. However, several issues concerning Bug Creek assemblages are in dispute, including: (1) their geochronologic age(s); (2) whether they contain reworked fossils; and (3) whether they are temporally intermediate between Lancian (Cretaceous) and Puercan (Paleocene) faunas. Bug Creek assemblages are critical to Cretaceous-Tertiary extinction debates because they can be interpreted to support both gradual and catastrophic models of faunal turnover during the Cretaceous-Tertiary transition.

Lithofacies and biostratigraphic (including pollen) analyses of the upper Hell Creek and lower Tullock formations in the vicinity of McGuire Creek, McCone County, Montana, were employed to address contentious issues surrounding Bug Creek assemblages. Seven lithostratigraphic facies were recognized at McGuire Creek, and analysis of facies indicates that Bug Creek assemblages are restricted to lag deposits of large channel fills in the upper Hell Creek Formation which are deeply entrenched into floodplain deposits that yield in situ dinosaur remains. A locally traceable erosion surface was created by these channeling events. Because an enormous amount of older sediment was removed during channel entrenchment, the probability that Bug Creek assemblages contain reworked fossils is very high. Additional evidence indicating that the dinosaurian and Lancian mammal components of Bug Creek assemblages are reworked is: (1) Bug Creek assemblages are present only in channel fills that were entrenched into Lancian (Cretaceous) strata; and (2) when traced laterally, channel facies that contain Bug Creek assemblages (with dinosaurs) interfinger with floodplain deposits that lack dinosaurs.

Complete channel-fill sequences in the upper Hell Creek Formation are overlain

by a laterally extensive lignite, the McGuire Creek "Z" (MCZ) (base of the Tullock Formation), which is overlain by pond deposits.

Palynology was employed to distinguish Cretaceous and Paleocene strata at McGuire Creek. All channel fills containing Bug Creek assemblages that could be sampled for pollen yielded Paleocene palynofloras. Thus, the erosion surface created by the entrenchment of channels yielding Bug Creek assemblages into floodplain deposits with in situ dinosaurs is the delineation of the local Cretaceous-Tertiary boundary in a large part of the McGuire Creek area.

Palynology yields further evidence that reworking has occurred, because: (1) channel fills with Bug Creek assemblages are palynologically Paleocene, while floodplains with the highest in situ dinosaur remains yield Cretaceous palynofloras; (2) the lowest Paleocene floodplain deposits (identified palynologically) lack dinosaurs; and (3) dinosaur remains encased in mudstone that yielded a Cretaceous palynoflora are present in a Paleocene channel fill at Black Spring Coulee (see Lofgren et al., 1990). Thus, data from McGuire Creek indicate that the presence of dinosaurs in Paleocene channels is the result of reworking, not of their survival into the Paleocene.

Development of a biostratigraphic zonation for the upper Hell Creek Formation at McGuire Creek is limited because of complex channeling, and channel fills can be temporally ordered only in rare instances. Thus, biochronologic criteria form the main basis for temporally ordering the three types of assemblages as Lancian-Bug Creek-Puercan (oldest to youngest).

McGuire Creek local faunas are referred to the Lancian North American Land Mammal "Age" (NALMA) and the Pu0 and Pu1 interval zones of the Puercan NALMA (Archibald and Lofgren, 1990). Analysis of the informal subdivision of the Pu0 interval zone into biochrons (bk1-bk2-bk3) indicates that the type localities (Bug Creek Anthills-bk1; Bug Creek West-bk2;, Harbicht Hill-bk3) require additional sampling and new systematic treatments before these biochrons can be abandoned or formally adopted.

The survival rates of the dinosaurian and mammalian part of the latest Cretaceous vertebrate fauna may or may not be compatible with catastrophic terminal Cretaceous extinction, depending on the provenance of the Lancian component of Bug Creek assemblages. At McGuire Creek, the survival rates of mammal species (approximately 10%) and dinosaur genera (0%) are low across the Cretaceous-Tertiary boundary because the presence of dinosaurs and certain Lancian mammals in Bug Creek assemblages (palynologically Paleocene) are interpreted to be the result of reworking.

Introduction

The "Bug Creek Problem" refers to a number of unresolved issues concerning the Bug Creek vertebrate assemblages and nonmarine biostratigraphy of the uppermost Hell Creek and lower Tullock formations of eastern Montana. The mammalian component of Bug Creek assemblages is intermediate in composition between typical Late Cretaceous and Early Paleocene faunas, but detailed systematic and biostratigraphic study of these assemblages has never been completed. Also, it has not been determined whether the presence of dinosaurs and certain Cretaceous mammals in Bug Creek assemblages is a taphonomic artifact caused by reworking. This lack of data has fueled the debate concerning patterns of faunal turnover of the terrestrial biota during the Cretaceous-Tertiary transition. These uncertainties are illustrated by the fact that the Bug Creek faunal data have been used to support both catastrophic (Smit and Van der Kaars, 1984; Smit, 1985; Smit et al., 1987) and noncatastrophic models (Clemens and Archibald, 1980; Archibald, 1981, 1982, 1984, 1987c; Clemens, 1983; Archibald and Clemens, 1984) of vertebrate turnover during the K-T transition in eastern Montana.

Unresolved questions can be divided into four main issues:

(1) Do Bug Creek vertebrate assemblages represent the remains of animals that lived contemporaneously or were Cretaceous fossils eroded and redeposited with Paleocene vertebrate remains?

(2) Based on biostratigraphic and/or biochronologic criteria, are Bug Creek assemblages temporally intermediate between Lancian (Late Cretaceous) and Puercan (Early Paleocene) assemblages?

(3) Can Bug Creek assemblages themselves be placed in a temporal sequence that reflects the faunal differences between individual local faunas?

(4) What are the geochronologic age(s) of Bug Creek assemblages?

Geologic and Biostratigraphic Setting

The first paleontological-geological exploration of Garfield and McCone counties, Montana (Figures 1a-c), was undertaken by Barnum Brown of the American Museum of Natural History, who described the dinosaur-bearing "Hell Creek beds" (Brown, 1907). Later, Rogers and Lee (1923) and Thom and Dobbin (1924) recognized the Hell

Figure 1
a. Location of study area in northeastern Montana.
b. Location of sites of paleontological interest in Garfield County, Montana. Fossil localities/local
faunas are: (HE) Hell's Hollow, (FC) Flat Creek.
c. Location of sites of paleontological interest in McCone County, Montana. Fossil localities/local
faunas are: (HH) Harbicht Hill, (CBB) Chris' Bone Bed, (FR) Ferguson Ranch, (BA) Bug Creek
Anthills, (BW) Bug Creek West. Heavy dashed lines outline the McGuire Creek study area.
Plate 1 is a partial geologic map of the McGuire Creek area.

Creek "beds" and the Tullock "Formation" as members of the Lance Formation. Wilmarth (1938) elevated both to the rank of formation. Roland Brown (1952) proposed a formal lithostratigraphic boundary between the Hell Creek and Tullock formations at the base of the first laterally persistent lignite bed above the highest occurrence of dinosaurs. This lignite, termed the "Z" coal bed in McCone County by Collier and Knechtel (1939), was used by Roland Brown (1952) to approximate the local Cretaceous-Tertiary boundary, assuming that the passing of the last dinosaur signaled the close of the Cretaceous.

Paleontological research in eastern Montana was minimal for five decades following Barnum Brown's (1907) early work. In the meantime, North American vertebrate paleontologists developed a series of provincial time units based on nonmarine mammal-bearing rock units that encompassed the latter part of the Late Cretaceous and all Cenozoic epochs. Of interest here are those units which occur near the Cretaceous-Tertiary boundary: the Lancian, "Bugcreekian," and Puercan North American Land Mammal "Ages." North American Land Mammal "Ages" (abbreviated to NALMA) are not true ages, because in most cases they are based on aggregates of mammal genera, not on corresponding chronostratigraphic standards (see Lillegraven and McKenna, 1986, for detailed discussion of NALMAs). Savage (1962) advocated the placing of "Ages" in quotation marks to denote this fundamental difference between stage-ages and NALMAs.

The Puercan NALMA was proposed in 1941 by the Wood Committee (Wood et al., 1941), based on the mammalian fauna from the Puerco Formation (now lower part of the Nacimiento Formation), San Juan Basin, New Mexico. The Puercan NALMA was considered to be Paleocene in age. Dorf (1942) proposed the Lancian "Age" based on the Lance Formation at its type locality, Lance Creek, Wyoming, and suggested that the Lancian "Age" could be recognized by mammals and dinosaurs of the "Triceratops zone" and plants from the Lance Formation. The Lancian NALMA was considered to be Late Cretaceous in age (Wood et al. 1941). This was the status of North American nonmarine Cretaceous-Tertiary temporal units at the time of the Bug Creek discoveries.

In the early 1960's, several fossil-rich vertebrate sites from channel fills in the Hell Creek and Tullock formations in the vicinity of Bug Creek, McCone County, Montana (Figure 1c), were discovered and briefly described by R.E. Sloan and L. Van Valen (1965; Van Valen and Sloan, 1965). Ken's Saddle, a locality from the upper Hell Creek Formation was described as being of typical Late Cretaceous aspect (or Lancian in age) in that it contained a fauna similar to that from the Lance Formation (dinosaur remains and a Lancian mammalian assemblage). A Puercan mammal fauna, Purgatory Hill, from the lower Tullock Formation was also mentioned (Sloan and Van Valen, 1965) and briefly described (Van Valen and Sloan, 1965). Of special interest were three channel fillings in the uppermost Hell Creek Formation which yielded dinosaurs and mammals of Late Cretaceous aspect (Lancian), along with mammals whose descendants are characteristic of the Puercan (Sloan and Van Valen, 1965). The three sites, Bug Creek Anthills, Bug Creek West, and Harbicht Hill (Figure 1c), which were referred to as the Bug Creek Faunas, contained the first records of the multituberculates Catopsalis (first

North American record) and Stygimys , the "proteutherian" Procerberus , and the archaic eutherian ungulates ("condylarths") Protungulatum, Mimatuta, Oxyprimus , and Ragnarok (Table 1). Based on their mammalian assemblages and stratigraphic position, the Bug Creek Faunas (here designated Bug Creek assemblages) were inferred by Sloan and Van Valen (1965) to be younger (from a higher horizon) than the Lancian fauna, Ken's Saddle, but were considered to have been deposited before the early Puercan ("Ectoconus zone"). Therefore, the Bug Creek assemblages were intermediate in composition between previously known Lancian/Late Cretaceous and Puercan/Paleocene assemblages (such as those present at Ken's Saddle and Purgatory Hill) and it was implied that the Bug Creek assemblages represented a previously unsampled faunal interval situated between the Lancian and Puercan NALMAs (Sloan and Van Valen, 1965; Van Valen and Sloan, 1965). Sloan and Van Valen (1965) considered the Bug Creek assemblages to be Cretaceous in age, using two criteria: (1) the Bug Creek sites were stratigraphically below the coal bed (termed the "Z" coal) used to approximate the position of the local K-T boundary in Roland Brown's (1952) formula of first coal above the highest occurrence of dinosaur remains ("Z" coal was also used as the lithostratigraphic boundary between the Hell Creek and Tullock formations); (2) dinosaur remains were present in the Bug Creek assemblages.

Using the stratigraphic distance of each locality above or below the "Z" coal, Sloan and Van Valen (1965) placed the five sites in the following temporal sequence, from oldest to youngest: Ken's Saddle, Bug Creek Anthills, Bug Creek West, Harbicht Hill, and Purgatory Hill. With the five sites sequentially ordered in this fashion, Van Valen and Sloan (1977) argued that the typical Late Cretaceous vertebrate assemblage (or Triceratops community) was gradually replaced by an assemblage with Paleocene affinities (or Protungulatum community) and that the turnover began stratigraphically below the "Z" coal or during the latest Cretaceous.

Archibald (1982) proposed a model of vertebrate turnover in eastern Montana similar to Van Valen and Sloan's (1977), whereby two faunal-facies were present in the upper Hell Creek Formation prior to the close of the Cretaceous: a Hell Creek faunal-facies (the Triceratops community of Van Valen and Sloan, 1977) from siltstones, and a Bug Creek faunal-facies (the Protungulatum community of Van Valen and Sloan, 1977) from large sandstone channel fills (Archibald, 1981, 1982). From their stratigraphic position the two faunal-facies were interpreted to be in part coeval, but were separated by ecological differences. Limited magnetostratigraphic data indicated that the two faunal-facies occurred during the same period of reversed polarity (Archibald et al., 1982). Archibald (1981, 1982) concluded that prior to the close of the Cretaceous, faunal turnover occurred gradually within the Bug Creek faunal-facies concurrent with a rapid turnover within the Hell Creek faunal-facies.

Testing the proposal of an extraterrestrial cause for catastrophic, terminal Cretaceous extinctions (Alvarez et al., 1980; Alvarez, 1986) required greater precision in stratigraphic studies of the Bug Creek assemblages. The Bug Creek faunal data were in apparent conflict with a catastrophic extinction hypothesis because the changeover between Late Cretaceous (Triceratops community, or Hell Creek faunal-facies) and

Early Paleocene faunas (Protungulatum community, or Bug Creek faunal-facies) was evidently in progress before the close of the Cretaceous (Sloan, 1976; Van Valen and Sloan, 1977; Clemens et al., 1981; Archibald, 1982; Van Valen 1984).

An iridium enrichment, which is presumed to be the record of bolide impact (Alvarez et al., 1980), occurs within the "lower Z" coal (approximate K-T boundary) in western Garfield County (Smit and Van der Kaars, 1984). A number of typically Cretaceous palynoflora species are not found above the level of iridium enrichment when it is present in a local section (Hotton, 1984, 1988). Also, new appearances of Paleocene pollen species are rare in the first few meters above the iridium enrichment. Therefore, Paleocene palynofloras are distinguished from Cretaceous palynofloras by the extinction of Cretaceous species (Hotton, 1988). Because an iridium enrichment was not found in McCone County within the stratigraphic level comparable to that of the "lower Z" coal of Garfield County, an absence of Cretaceous pollen species was employed to determine the local K-T boundary in McCone County (in Montana, Bug Creek assemblages are known only from McCone County). Harbicht Hill, Bug Creek West, and a recently discovered Bug Creek site at Ferguson Ranch (Figure 1c) yielded palynofloras characterized by the absence of typical Cretaceous species and were therefore considered to be Paleocene in age (Smit and Van der Kaars, 1984; Smit, 1985; Sloan et al., 1986; Smit et al., 1987).

Subsequent debates on whether the Bug Creek faunal sequence began in the Cretaceous were centered on the Bug Creek Anthills locality. Bug Creek Anthills was argued to be Cretaceous (Archibald, 1984; Sloan et al., 1986; Rigby et al., 1987; Rigby, 1987; Newmann, 1988); Paleocene (Smit and Van der Kaars, 1984; Smit, 1985; Retallack et al., 1987); or of indeterminable age (Dingus, 1984; Fastovsky and Dott, 1986; Fastovsky, 1987; Smit et al., 1987). Detailed study of the geology in the vicinity of the Bug Creek Anthills locality indicated that correlation of this site is uncertain because of repeated episodes of incision and filling of fluvial channels (Fastovsky and Dott, 1986).

Because the Bug Creek assemblages were restricted to large sandy channel fillings, these assemblages were viewed by some workers as Paleocene, but containing reworked Cretaceous fossils (Smit and Van der Kaars, 1984; Bryant et al. 1986; Retallack et al., 1987). In contrast, others viewed dinosaur remains in these same channel fills as evidence of dinosaur survival into the Paleocene (Rigby, 1985, 1987; Rigby and Sloan, 1985; Sloan et al., 1986; Rigby, et al., 1987). However, Fastovsky (1987) concluded that provenance of the Bug Creek assemblages is uncertain because of taphonomic considerations imposed by the depositional setting of these vertebrate remains. Also, based on analyses of the sedimentary record spanning the K-T transition in eastern Montana, both Dingus (1983, 1984) and Fastovsky (1986, 1987) concluded that the chronostratigraphic resolution of these sediments is not sufficiently refined to resolve the K-T debates.

Therefore, while the age, provenance, and biostratigraphy of the original Bug Creek assemblages are central to K-T boundary debates, data from the original Bug Creek sites are not sufficient to resolve the issues being debated.

McGuire Creek Study Area

I conducted geologic and biostratigraphic analyses of the upper Hell Creek and lower Tullock formations in the vicinity of McGuire Creek because the area contains abundant Bug Creek-like vertebrate assemblages and extensive exposures of the stratigraphic interval in which the Cretaceous-Tertiary and Lancian-Puercan transitions occur. Thus, abundant faunal data with potential for good stratigraphic control made the McGuire Creek area ideally suited to address contentious issues concerning the Bug Creek assemblages. Results garnered from this study are divided into six parts:

1. Geological and Palynological Correlations : Using facies analysis and pollen analysis, I determined that channel facies with Bug Creek assemblages do not correlate to floodplain facies that yield in situ dinosaur remains. Also, I constructed a sequence of geologic events spanning the K-T transition at McGuire Creek which shows that developing a local biostratigraphic zonation for the upper Hell Creek Formation is extremely difficult because of multiple episodes of channel entrenchment.

2. Reworking of Fossils : I concluded that Bug Creek assemblages certainly contain reworked fossils, and that the survival of dinosaurs into the Paleocene is a taphonomic artifact caused by reworking.

3. McGuire Creek Biochronology and the "Bugcreekian Age" : From the results from 1 and 2 above, I concluded that current mammalian zonations spanning the K-T transition are problematic, but were developed on the best biochronologic data available.

4. Cretaceous-Tertiary Correlations : I determined that palynology is the best method to employ to recognize the terrestrial K-T boundary at McGuire Creek and that all Bug Creek assemblages for which data is available yield Paleocene pollen.

5. Faunal Turnover During the K-T Transition in Montana : I examined mammalian survivorship across the K-T boundary and concluded that survivorship rates vary greatly, depending on how the presence of Lancian mammals in Bug Creek assemblages and the age of Bug Creek assemblages are interpreted.

6. Appendices 1 and 2, on Mammalian Systematics and Faunal Data : I analyzed the systematic paleontology of McGuire Creek mammals (and a few other sites) and listed mammalian taxa present at each McGuire Creek locality to provide a biostratigraphic database for this study.

Materials and Methods

Geological and palynological investigations in the McGuire Creek area were limited to the stratigraphic interval of the upper 40 m of the Hell Creek Formation and the lower 20 m of the Tullock Formation. Field data were plotted directly onto standard USGS 7.5-minute quadrangles (Bug Creek and Nelson Creek Bay) that were enlarged 200% and later transferred to a 200% enlarged base map (Plate 1). Section thicknesses were measured in meters, using a hand level and jacob's staff.

Seven lithofacies (A-G) were recognized and mapped and the stratigraphic position of each is indicated in cross sections (Plates 2-4) and on a partial geologic map (Plate 1). Facies mapping was not completed in the northern and extreme southwestern part of the study area. For clarity in presentation, only the three channel-fill facies (B, E, F) and the coal-swamp facies (C) are indicated on the geologic map (Plate 1). Facies C lignites were used to correlate measured sections. Recognition and mapping of individual channel-filling events were completed where possible. Certain channel fills are traceable through laterally continuous exposures, while others are recognizable across disjunct exposures based on lithologic similarity and congruence of paleocurrent azimuths measured from crossbedded sandstones. The remainder of channel-fill exposures are mapped as undifferentiated, because they could not be correlated with the recognized channel fills. Although three types of vertebrate assemblages are found in channel fills at McGuire Creek (Lancian, Bug Creek, and Puercan; Table 1), faunal content was never employed to aid in recognition of individual channel fills.

Several techniques of fossil collection were employed at McGuire Creek, and the specific technique(s) used for each locality are listed in Appendix 2. All vertebrate sites were prospected for fossils by visual scanning of the available surface exposure. Hand quarrying was carried out at a few sites where recovery of articulated or associated skeletal material was desirable. Screenwashing techniques, described by McKenna (1965), were employed at many sites. Isolated teeth collected from mounds constructed by harvester ants were used to a limited degree. The ability of ants to concentrate fossils is well documented (Hatcher, 1896; Galbreath, 1959; Clemens, 1964; Clark et al., 1967; Macdonald, 1972), and anthill concentrates are useful for adding qualitative knowledge of the small vertebrate fauna (Clark et al., 1967). But anthill concentrations are problematic, because harvester ants can mix fossils from different stratigraphic

units. Transport experiments using glass or plastic beads indicate that ants can carry anthill material a minimum of 50 feet (Clemens, 1964). Also, anthill collections are biased toward small vertebrate elements because of the size selectivity of ants. Because of these biases, anthill collections were not included in quantitative analyses. Localities that include fossils from anthills are indicated in Appendix 2.

The University of California Museum of Paleontology (UCMP) records each vertebrate locality by using a V followed by numbers (e.g., V87030). Specimens are given a numerical code preceded by UCMP (e,g., UCMP 113465).

Unweathered rock samples analyzed for pollen and spore content were collected by digging beyond the weathered outcrop surface. Rock samples were prepared by standard palynological techniques and sent for analysis to Dr. Carol Hotton, who made a Cretaceous or Paleocene age assignment for each sample, based on comparison to stratigraphically controlled samples from Garfield County, Montana (Hotton, 1988).

Specific definitions of two terms, "Bug Creek vertebrate assemblage" and "local fauna" are required for clarity in later discussions. A Bug Creek vertebrate assemblage is defined as an aggregation of vertebrate fossils containing a mixture of Lancian or latest Cretaceous mammals and dinosaurs with mammals more typically indicative of a Puercan or Paleocene age (Table 1). Therefore, the term "Bug Creek vertebrate assemblage", or "Bug Creek assemblage" for short, refers to the aggregation of taxa listed in the middle column of Table 1, wherein mammalian taxa from both the left (Lancian) and right (Puercan) columns are present in some combination, along with dinosaur remains.

The term "local fauna" is employed at McGuire Creek in a manner similar to that advocated by Tedford (1970), but additional comments are necessary. As employed at McGuire Creek, vertebrate fossils within lag concentrations from individually mapped channel fills constitute a local fauna. Therefore, at McGuire Creek, local faunas are aggregates of fossil species collected from a restricted stratigraphic interval that is local in time and space (a distinct channel-filling event). However, fossil species collected from this depositional setting represent a thanatocoenose, rather than a biocoenose. Therefore, the term local fauna is used to describe the fossil species present in one channel fill, whether or not these animals actually coexisted.

A detailed taxonomic database is necessary for accurate assessment of biostratigraphic correlations and is especially critical when assessing the faunal turnover at the Cretaceous-Tertiary boundary. Therefore, a thorough but not exhaustive systematic review of the McGuire Creek mammalian faunas is given in Appendix 1. In many instances it was necessary to study specimens and epoxy casts in the UCMP collections from Bug Creek Anthills, Bug Creek West, and Harbicht Hill, in the upper Hell Creek Formation, Montana, and Mantua Lentil, in the Polecat Bench (Fort Union) Formation, Wyoming, to aid in identification of McGuire Creek species. Hence, some faunal data from these sites are included in this study.

The mammalian classification scheme used here follows that of Marshall et al. (1990) for marsupials, Novacek (1986) for eutherians, and Hahn and Hahn (1983) for multituberculates.

Works that involve Late Cretaceous-Early Tertiary mammals by Archibald (1982),

Clemens (1964, 1966, 1973), Sloan and Van Valen (1965), Lillegraven (1969), and Van Valen (1978) were followed where applicable. Modifications and additions to the morphological descriptions of taxa presented in these studies are given where appropriate, based on specimens recovered from McGuire Creek. No new species are named.

Only dental remains were studied. Isolated elements and fragments of postcranial material were recovered but are not included in this study.

Cusp terminology for therian teeth is that of Van Valen (1966: figure 1). Stylar cusps of marsupials are named in the manner outlined by Clemens (1966). Measurements of therian molars were taken following Lillegraven (1969: figure 5), and additions made for "condylarth" molars by Archibald (1982: figure 1). Multituberculate cusp terminology and measurements follow Clemens (1964) except for the /P4, where measurements were taken according to the system presented by Novacek and Clemens (1977).

Measurements of specimens are given in millimeters. Specimens were measured on an Ehrenreich Photo-Optical Industries "Shopscope" and rounded to the nearest one-hundredth of a millimeter.

Geological and Palynological Correlations

The Hell Creek Formation of eastern Montana was deposited as the continental facies of the last major Late Cretaceous regression of the Western Interior Epeiric Sea (Frye, 1969; Cherven and Jacob, 1985). The overlying Tullock Formation is also continental in origin and was deposited during a transgression of the same sea (Cherven and Jacob, 1985) (see Figure 2). Both formations were deposited by fluvial systems on the proximal part of deltas formed from debris eroded primarily from the Elkhorn Mountains volcanic arc in southwestern Montana (Gill and Cobban, 1973; Cherven and Jacob, 1985).

Fluvial paleoenvironments during the K-T transition in eastern Montana were reconstructed by Fastovsky (1987). The results of this analysis indicate that:

a broad fluvial plain, dissected by meandering, sediment-rich, suspended-load rivers, was aggraded by recycled, volcanic-rich clastic sediment. These were deposited by avulsive and sweeping channel migration, as well as by suspension settling during periodic floods. Vegetation was abundant, and collected in depressions, forming layers of matted plant debris. A high, fluctuating water table promoted the formation of weakly developed gleyed soils in moderately unstable flood-plain environments . . . A rise in water table, probably attributable to increased rainfall, induced a change in the paleoenvironments of this region at a time approximately coincident with that of the faunal and floral changes that define the K-P boundary. The elevated water table caused the gradual appearance of large ponded deposits and peat swamps, which resulted in discontinuous coal layers that characterize the K-P boundary. (Fastovsky, 1987:292)

The discontinuous coal layers produced by peat swamps are employed as a criterion for recognition of the lithostratigraphic boundary between the Hell Creek and Tullock formations (Calvert, 1912; Brown, 1952). In any single stratigraphic section, the formation boundary is defined as the first coal above the highest occurrence of dinosaurs (Brown, 1952). This "lowest" coal is termed the "Z," employing the letter terminology for designating coal beds in McCone County developed by Collier and Knechtel (1939). In this scheme, each stratigraphically higher coal bed receives a successive letter designation: Y, X, W, V, U. This system of lettering, especially the "Z" coal designation, has been employed in biostratigraphic studies in

Figure 2
Correlation chart (modified from Bryant, 1989: figure 2), showing lithologic
units in the study area (McCone County, Montana) and North and South
Dakota. Correlation of lithologic units, geochronologic units, and NALMA's
(North American Land Mammal "Ages" from Wood et al., 1941 and
Russell, 1975) based on Archibald and Lofgren, 1990, and this study.

eastern Montana (Sloan and Van Valen, 1965; and many others). However, use of this terminology in more than a limited area without qualification can cause confusion because it implies regional continuity of each coal bed. When originally proposed, it was stated that this lettering system was not suitable for application in regional correlation: "Bed Z, is probably not a continuous bed but rather a succession of lenses of coal in about the same stratigraphic position" (Collier and Knechtel, 1939:18). And with respect to coals of the Tullock Formation, "correlation of coal beds from place to place is difficult, and beds V, W, X, and Y, are all probably discontinuous, each being merely several lenses at about the same horizon" (ibid., p. 19). Therefore the employment of coal beds as regional correlation tools is problematic because coal beds in eastern Montana can be demonstrably discontinuous (Sholes and Cole, 1981; Archibald, 1982; Fastovsky, 1987).

At McGuire Creek the formational boundary is also recognized as the base of the first coal above the last dinosaurian fossils, but to eliminate potential confusion in correlation, this coal is mapped and designated the McGuire Creek Lignite (MCZ) (Plates 1-4), not the "Z" coal. Therefore, the MCZ is a local stratigraphic unit and has no implied regional significance.

In the McGuire Creek area, the upper Hell Creek Formation is composed primarily of gray and light to dark green sandstones, siltstones, and mudstones. Lesser amounts of purple or dark to light brown siltstones and mudstones are also present. Weakly lithified, very thick (15-20 m) gray sandstones, containing Bug Creek assemblages, are locally abundant in the uppermost 30 m. These sandstones are overlain by a thick and laterally persistent coal, the McGuire Creek Lignite (MCZ), which defines the base of the overlying Tullock Formation. Organic-rich sediments are rare in the upper Hell Creek Formation and usually consist of thin beds of black carbonaceous shale which have limited lateral continuity. However, a thick shaly coal, the Tonstein Lignite (TL), is present about 30 m below the MCZ and is laterally traceable for over 4 km (Plate 1). Studies of the Hell Creek Formation in eastern Montana and the Dakotas are presented by B. Brown, 1907; R. Brown, 1952; Jensen and Varnes, 1964; Frye, 1969; Moore, 1976; Butler 1980; Archibald, 1982; Fastovsky, 1987; and Rigby and Rigby, 1990.

The Tullock Formation at McGuire Creek is composed primarily of gray, yellow, and light brown sandstones, siltstones, and mudstones. In contrast to the Hell Creek Formation, well developed coal beds are common in the Tullock Formation. Tullock sediments are, in general, more resistant to erosion, more tabular and more laterally continuous than Hell Creek beds. The difference in erosive properties is clearly evident in topographic expression in the McGuire Creek area; major cliff development is commonly associated only with the Tullock Formation (Plate 1). Additional descriptions of the Tullock Formation in eastern Montana are provided by Rogers and Lee, 1923; Collier and Knechtel, 1939; Archibald, 1982; Fastovsky, 1987; and Rigby and Rigby, 1990.

Facies Analysis and Geological Correlations

Fastovsky (1987) has published a detailed facies analysis of vertebrate-bearing strata spanning the K-T transition in parts of eastern Montana. General classes of lithofacies

Figure 3
Comparison of terminology employed in this and other analyses of lithofacies in the upper
Hell Creek and lower Tullock formations of eastern Montana.

at Bug Creek (Fastovsky and Dott, 1986; Fastovsky, 1987), which is only 1 km north of the McGuire Creek study area, can be recognized and employed at McGuire Creek (see Figure 3). Additions or modification of these facies based on strata at McGuire Creek are given below when necessary. Two additional lithofacies (Facies B and G) were recognized at McGuire Creek and are described in detail.

Facies A

This facies is lithologically similar in part to the Siltstone Facies of Fastovsky, 1987, and Facies B of Fastovsky and Dott, 1986.

Description : Facies A is composed primarily of very thick to medium interbeds of green, gray, and purple, laminated to massive, siltstones and mudstones. Thick (up to 70 cm) lenticular sandstones that fine upward to siltstone or mudstone are also present but rare. Isolated and associated remains of larger-bodied vertebrates, primarily dinosaurs, are present. These occurrences usually are limited to isolated dinosaurian skeletal elements that are widely separated both geographically and stratigraphically. However, occurrences of more complete dinosaurian material were observed. The eroded and scattered remains of a partial ceratopsian skull are present in Section D (Plate 2). Mammalian remains were not recovered from this facies.

Interpretation : Facies A includes floodplain deposits that underwent soil development (Fastovsky, 1987; Fastovsky and McSweeney, 1987). Vertebrate fossils probably represent animal remains that were lying on the floodplain surface or in localized depressions that were subsequently buried by suspended sediment during flooding events. Fossils found in this depositional setting are unlikely to have been reworked from older strata, especially in the case of the ceratopsian skull cited above.

The absence of mammalian and other small-bodied vertebrate remains may be a reflection of the size of skeletal elements. Small bones have high surface-to-volume ratios and would be more readily destroyed during soil formation (Retallack, 1984).

Facies B

Description : Thin to medium interbeds of dark brown, brown, and gray sandstone, siltstone, and mudstone are present. Lenticular, fine-grained sandstones comprise 30-40% of this facies. Sandstone lenses are usually 15-20 cm thick and 15-20 m wide, and the largest observed is 1 m thick and 30 m wide. Sandstones also occur as thin laminated beds interbedded with siltstones and mudstones. Internally, sandstones exhibit ripple and small-scale tangential, planar, and trough (rare) cross-stratification sets. Siltstones and mudstones are laminated or massive, and all lithologies commonly have organic debris and plant macrofossils on laminae. Root casts are present but are rare. Siltstones and mudstones are more laterally continuous than sandstones, and usually can be traced laterally for over 25 m. Facies B fines upward and is overlain by Facies C.

The best exposures of Facies B near Section V (Plates 1, 2) indicate a channel-form morphology with steep bank margins which are deeply entrenched into Facies A beds. The base of Facies B is not exposed here, but a minimum of 6 m of relief is developed on this erosion surface, which dips 25-30 degrees to the east. The orientation of Facies B beds adjacent to this surface conforms to the eastward slope but levels out when traced toward the axis of the channel. This Facies B and Facies A contact trends 165 degrees and can be traced for 1.5 km to the southeast (Plate 1). Exposures of Facies B adjacent to Section V are over 150 m wide. East of these exposures are poorly exposed cross-stratified sandstones (Plate 1).

Articulated and associated remains of aquatic vertebrates including turtles, crocodiles, and champsosaurs are present in Facies B. Mammal and dinosaur remains were not recovered from this facies.

Interpretation : Facies B represents channels that were abandoned and slowly filled with suspended silt and mud. Lenses of cross-stratified sand indicate periodic influx of sand as channelized traction load during flooding events. The abundance of plant fossils but rarity of root traces suggests that the abandoned channel usually contained standing water. The presence of aquatic vertebrates supports this interpretation. Cross-stratified sandstones lateral (east) of Facies B outcrops (Section V) probably represent previously deposited thalweg channel-fill material.

Facies C

Facies C is lithologically similar to the Facies of Organic Accumulation of Fastovsky, 1987, and Facies D of Fastovsky and Dott, 1986.

Description : Facies C is composed of very thick to medium-bedded black to purple lignite and carbonaceous shale, containing 1-10 cm thick mud partings. Discontinuous or laterally persistent, 2-5 cm thick volcanic ash layers are present in both the McGuire Creek Lignite (MCZ) and Tonstein Lignite (TL). The MCZ and TL each contain a particularly persistent volcanic ash layer whose presence was useful for mapping across disjunct exposures. The MCZ is locally overlain by a 20-50 cm thick, massive, black to dark gray mudstone which is devoid of plant macrofossils and roots. This mudstone is useful for correlation of the MCZ between isolated exposures. Amber granules are abundant in the TL. Vertebrate remains were not recovered from Facies C.

Interpretation : Facies C represents the development of widespread coal-swamp deposits in the McGuire Creek area. Extensive coal-swamp development is evident because the MCZ and TL are laterally traceable over more than 20 and 4 sq. km, respectively (Plate 1). The extensive development of lignites in the McGuire Creek study area is noteworthy because coal horizons at or below the formation boundary can be extremely localized elsewhere (Archibald, 1982; Fastovsky and Dott, 1986; Fastovsky, 1987).

It appears that the TL, and possibly the MCZ, may have significant utility as correlation tools in biostratigraphic studies because they are evidently present in regions north and west of the McGuire Creek study area. A Hell Creek Formation lignite, the Null Coal, mapped within the Bug Creek drainage by Smit et al. (1987) and Rigby and Rigby (1990), almost certainly is equivalent to the TL based on elevation, stratigraphic position, and general lithologic similarity. However, the absence of ash layers in the Null Coal precludes definite correlation (NW quadrant of Section 16, Plate 1, and directly below the Bug Creek Anthills channel fill, NW quadrant of Section 9, T 22 N, R 43 E).

A coal in the upper Hell Creek Formation containing one or two ash layers and amber is exposed at Jawbone Coulee and Thomas Ranch, 5 and 15 km west of the McGuire Creek study area, respectively. The Jawbone Coulee lignite has two ash layers and amber, and closely parallels the TL in general stratigraphic position, lithologic character, and position of ashes within the coal (Section 13, T 22 N, R 42 E). In the opinion of Dr. Carol Hotton (pers. comm., 1989), the TL may also be equivalent to the Tonstein Coal at the Thomas Ranch locality (Hotton, 1988; Section 14, T 22 N, R 41 E) based on approximate stratigraphic position, presence of an ash layer, and abundant amber granules.

The MCZ is difficult to trace into the northern part of the study area, and correlations become questionable (Plate 1). Correlation between lignites is especially difficult to the northwest (Section HH) because of discontinuous exposures and thinning of coals. It appears that the MCZ and upper "Z" coal of the lower Bug Creek drainage (not including Russell Basin) are two different units (Section OO, Plate 4). However, mapping north of Section DD (Plate 4) indicates that the MCZ is traceable northward into the headwaters of the Bug Creek drainage and is equivalent to the uppermost lignite of the "Z" complex at Russell Basin (upper "Z" of Smit et al., 1987 and Rigby and Rigby, 1990). It appears that the upper "Z" and MCZ (Section OO) are two units of a series of thin lignites that comprise the upper "Z" complex of Russell Basin.

Facies D

Facies D is lithologically similar to the Variegated Facies of Fastovsky, 1987; Facies E of Fastovsky and Dott, 1986; and the "variegated" beds of Archibald, 1982.

Description : Facies D is composed of thin to medium interbeds of yellow-brown or gray laminated siltstone and mudstone. Thin individual beds can be traced for hundreds of meters laterally in continuous exposures. Facies D, as a whole, is laterally traceable for over 25 sq. km in the McGuire Creek area. Facies D contains isolated occurrences of articulated elements of turtles and champsosaurs.

Interpretation : Facies D represents extensive pond deposits (Fastovsky, 1987).

Facies E

Facies E is lithologically similar to the Facies of Epsilon Cross-Stratification of Fastovsky, 1987, and Facies C of Fastovsky and Dott, 1986.

Description : Facies E is composed of 3-20 m thick gray sandstones in elongate sheet morphologies whose base is scoured into Facies A or another Facies E unit (Plate 3). Basal scours are commonly flat but can also exhibit over 5 m of relief (Little Roundtop Channel, Sections Q and 861, Plate 3). Weakly to well-defined inclined heterolithic stratification (IHS) (following Thomas et al., 1987, this term replaces "epsilon cross-stratification") sets are present that exceed 15 m in thickness and exhibit dips of 10-20 degrees. IHS is defined by alternating beds of sandstone, siltstone, and mudstone. It extends down to the basal scour in some cases (Little Roundtop Channel, Plate 3), but in others (Second Level Channel, Plates 3 and 4) the basal 1-7 m is composed of thick, vertically stacked sets of trough cross-stratified sandstone (similar to Facies F) which fine upward in both grain size and scale of sedimentary structures before interfingering with IHS (Sections 861, Q, R, Plate 3; Section T, Plate 2; Sections R, 871, Plate 4; Sections O, P, Plate 4). Paleocurrents of trough cross-strata are approximately perpendicular to IHS surfaces (Figure 4). Complete Facies E sequences fine up to siltstone and mudstone and are overlain by Facies C, or Facies G and C.

Poorly sorted, clast-to matrix-supported lag concentrations of clayball conglomerate, carbonized plant debris, fossil wood, and vertebrate fossils are commonly present at the base of Facies E channels. Basal lags consist of discontinuous 30-100 cm thick lenses of poorly sorted, clast-supported, pebble-cobble conglomerates composed of rounded mudstone and siltstone clasts. Conglomerates also occur as lenses within or underlying trough cross-bedded, medium to fine grained sandstone. Fragments and impressions of carbonized plant debris are abundant, while fossilized wood is rare. Localized depressions in channel floors are commonly filled with conglomerates containing rich concentrations of microvertebrate fossils. These fossil remains usually consist of isolated teeth and disarticulated elements (complete elements are rare) of aquatic and terrestrial vertebrates which have undergone hydraulic transport and sorting. Fossils are commonly fragmentary and abraded, but many pristine bones and teeth were also recovered. Fossils are intermixed with pebble-sized intraclasts that apparently have similar settling velocities (this is not true for larger fossils). Isolated teeth of mammals and dinosaurs are common, and fragments of shed ceratopsian teeth com-

Figure 4
Trough cross-bedding orientations measured in the upper Hell Creek Formation at McGuire
Creek. Data presented are grouped by mapped channel fill except where noted N= number of
measurements; X= vector mean in degrees; S= circular dispersion in degrees; R= vector
magnitude. See Plate 1 for location of channel fills measured.

prise the great majority of dinosaurian dental remains. Mammal jaws are uncommon, and mammalian maxillary fragments are rare. In comparison to the number of dental remains at any one site, mammalian postcranial elements are under represented and usually consist of dense elements such as phalanges, carpals, tarsals, and distal or proximal fragments of other limb bones. Vertebrate fossils were not found on IHS surfaces or immediately above the base of Facies E sequences.

Facies E lag concentrations which produce Bug Creek vertebrate assemblages are the Brown-Grey, Second Level, and Little Roundtop channel fills. The Z-Line Channel does not cut deeply into Facies A beds and yields a Puercan assemblage. Well-preserved teeth and many heavily abraded fragments of dinosaur bone are present at locality V87072 (Brown-Grey Channel).

Interpretation : Facies E represents laterally accreted channel-fill deposits. IHS is interpreted as lateral accretion surfaces of point-bar deposits (Fastovsky, 1987), while thick trough cross-strata in the lower region of Facies E channel fills represent vertically accreted thalweg sandstones (Mossup and Flach, 1983). Facies E channels were

entrenched into pre-existing Facies A units. Superposed Facies E fills (Little Roundtop and Second Level channels, Plates 3 and 4) indicate that multiple channel cutting and filling events also occurred. Therefore, considerable reworking of pre-existing strata and their entombed vertebrate remains probably occurred.

Facies F

Facies F is lithologically similar to the Cross-Stratified Sandstone Facies of Fastovsky, 1987, and Facies A of Fastovsky and Dott, 1986.

Description : Facies F is composed of 10-20 m thick, multi-storied, gray sand bodies which are commonly overlain by Facies G and C, and usually sharply bounded below by Facies A. Facies F units truncate Facies E channels and also Facies D and C deposits above the MCZ (Plates 2-4). Relief on these erosion surfaces exceeds 15 m (Sections II and 811 or Sections O, P, 856, and 855, Plate 4; Sections H-B, Plate 2).

Individual stories are lenticular, have a sharp undulatory base, and are about 2-10 m thick and 10-50 m wide. Successive stories are commonly incised into preceding channel-fill cycles, and several erosion events can be seen in good exposures (for example, Section H, Plate 2). Limited three-dimensional exposure of some stories shows that they have channel-form geometries with subparallel axes. Paleocurrent data indicates a south-southeast flow direction with a markedly unidirectional trend (Figure 4). Mapped Facies F sand bodies exhibit widths of approximately 0.5 km perpendicular to paleocurrent means (Plate 1).

Relatively complete stories exhibit a decrease in scale of sedimentary structures from very thick or thick sets of trough and planar cross-stratification to medium or thin sets of trough, planar, and ripple cross-stratification, overlain by siltstone. Sediments within most stories fine upward from medium grained sandstone with minor amounts of mudstone pebble conglomerate at the base to organic rich fine-grained sandstone or siltstone at the top.

Basal scours of Facies F sand bodies commonly exhibit localized lag concentrations of poorly sorted, clast- and matrix-supported, clayball conglomerates which contain fossil wood, carbonized plant debris, and vertebrate fossils (see Facies E for a general description of the depositional setting of these vertebrate remains). In Facies F, vertebrate fossil remains were rarely found above the basal lag of the lowest story.

Facies F lag concentrations which produce Bug Creek assemblages are the Black Spring Coulee and possibly the Upper Tedrow and "Lower Tedrow" channel fills. Jacks Channel produces a Puercan vertebrate assemblage. Vertebrate fossil occurrences in lag concentrations usually consist of isolated elements, but abundant well-preserved dinosaur material is present at Black Spring Coulee (V87030; see the following chapter on Reworking of Fossils for discussion).

Interpretation : Facies F multi-storied sand bodies were probably developed through vertical aggradation in laterally stable channels of low sinuosity. The subparallel channel-form geometries of each story, low divergence of paleocurrents, and absence of lateral accretion features such as IHS support this interpretation. Upward decrease in grain size and scale of sedimentary structures within each story records waning paleoflow strength between flood events. Multiple scour and fill events occurred within channels, as evidenced by entrenchment of superposed stories. The deep entrenchment of Facies F channels into pre-existing sediments suggests that considerable reworking of fossils occurred.

Other channel fills, not assignable to a channel facies because of poor exposures, contain important occurrences of vertebrate fossils. The K-Mark Channel yields a Lancian vertebrate assemblage. Matt's Dino Quarry Channel produces disarticulated but closely associated hadrosaur skeletal elements and a complete turtle shell. The Shiprock and Up-Up-the-Creek sites yield Bug Creek assemblages.

Facies G

Description : Facies G is composed of tabular interbeds of thin to thick brown, gray, and greenish-gray sandstone, siltstone, and mudstone. Sandstones comprise 20-30% of this facies and are commonly laminated, but also exhibit ripple and small scale planar cross-stratification. Sandstones also occur as elongate lenses up to 10 cm thick within siltstones. Siltstones and mudstones are laminated and contain organic debris and well-preserved plant macrofossils on laminae. Complete leaves are extremely abundant locally in all laminated lithologies, and large palm fronds were observed. Fossilized in situ tree stumps are present at sections B and H (Plate 2). Root casts are present, but are not common. Individual sandstone, siltstone, and mudstone units are commonly traceable over tens of meters.

Thin black carbonaceous shales-mudstones composed of fissile mats of laminated plant debris are also present, and are traceable for hundreds of meters in continuous exposures. Articulated and associated remains of turtles, champsosaurs, and crocodiles are abundant in carbonaceous shales-mudstones. Neither mammals nor dinosaurs were recovered from this facies.

Facies G overlies floodplain facies (Facies A, east part of Plate 2) and overlies and interfingers with channel facies (Facies E and F, Plates 2-4). Facies G can be a continuum with Facies, A, E, and F and boundary recognition must be arbitrarily determined in many cases. However, with respect to Facies G and A, they differ in color, and Facies G beds are more tabular and laterally continuous. Also, Facies G sediments contain well-preserved plant macrofossils (including in situ tree stumps) and internal bedding features (laminations and ripples).

Interpretation : Facies G represents floodplain deposits developed in association with channel facies. These floodplain deposits comprise the culmination of fining upward sequences of channel facies (E and F), or the lateral floodbasin deposits of these channels. Extensive but short-lived swamps were developed on floodplains, as evidenced by the thin but laterally continuous carbonaceous shales containing abundant aquatic vertebrates. Cross-stratified sandstones probably represent minor crevasse-splay development on the floodplains. The presence of tree stumps, abundant plant macrofossils, and root traces indicates that these floodplains were probably heavily vegetated. However, paleosol development was generally not as advanced as in Facies A because delicate sedimentary structures such as laminations and whole-leaf plant fossils are extremely abundant and well preserved.

Geologic History of the McGuire Creek Area

A geologic history spanning the upper Hell Creek and lower Tullock formations at the McGuire Creek area can be constructed based on superposition, cross-cutting relationships, and interfingering between the facies described above. This sequence of events applies only to the study area at McGuire Creek.

Sediments within the upper but not uppermost Hell Creek Formation at McGuire Creek consist primarily of floodplain facies (Facies A; Plates 2-4). Floodplain facies also contain evidence of small-scale stream and extensive peat swamp development (e.g. TL: Tonstein Lignite). In situ dinosaur remains above and below the TL indicate that dinosaurs inhabited these floodplains.

The predominantly floodplain deposits of Facies A were then scoured by a series of major streams. Channels created by these streams were filled by sediments representing Facies B, E, and F. The thickness of IHS sets (up to 10 m; Sections 861, Q, R, Plate 3) and of the channel fills themselves (up to 20 m; Section II, Plate 4) indicate that these streams were very large. Over 15 m of relief is present on the contact between floodplains (Facies A) and channel-fill facies (Facies B, E, F; see Sections II through 811, Plate 4; and Sections H through B, Plate 2). The 15 m of relief indicates that stream channels were deeply entrenched into floodplains. As a result, the erosion surface developed on floodplain deposits (Facies A) is traceable for over 3 km (Sections K through R, Plate 3; Sections R through II and O through 855, Plate 4). An immense volume of previously deposited strata (mostly Facies A) was removed during these channeling events, and the likelihood that vertebrate remains were reworked is high. Channels were filled by vertically accreted (Facies F) or laterally accreted (Facies E) sandy sediment, while others were abandoned and filled primarily by finer-grained sediment that settled from suspension (Facies B).

A complex record of superpositional and cross-cutting relationships between laterally accreted (Facies E) and vertically accreted (Facies F) channels indicates the occurrence of multiple channel and fill events within this stratigraphic interval (below MCZ, above TL). Channels were deeply entrenched into pre-existing channel facies as well as floodplains (Facies A), and considerable reworking of vertebrate remains must have occurred. Eight mapped channel-fill units, six of which represent individual channel fills, are evidence of this complexity (Plates 1-4). Excellent exposures permit temporal sequencing of a limited number of these channeling events (a simplified version of channel-fill relationships is presented in Figure 5 and should be referred to in the following discussion). Superpositional and cross-cutting relationships indicate that the oldest demonstrable channel fill (of the six that were individually mapped) is the Little Roundtop Channel, which is scoured and overlain by the Second Level Channel (Plates 3, 4). The Second Level Channel is cut by both the Black Spring Coulee and Upper Tedrow Channel fills (Plates 3, 4). The Upper Tedrow channel cuts both the Brown-Grey and "Lower Tedrow" channel fills (Plate 3) ("Lower Tedrow" and Upper Tedrow channels may be equivalent, see section below on The Tedrow Area). These cross-cutting relationships reveal that both the Upper Tedrow and Black Spring Coulee channel fills are younger than the Little

Figure 5
Schematic stratigraphic relationships of lignites, channel fills yielding vertebrate faunas, floodplain deposits
with in situ dinosaur remains (horizontal long-bone figures), sediment samples analyzed for pollen, and age
interpretation of palynofloras. Horizontal distance depicted is approximately 5 km in NE-SW orientation.
Channel fills (or Local faunas) are: (MD) Matt's Dino Quarry; (KM) K-Mark; (BS) Black Spring Coulee;
(LR) Little Roundtop; (SL) Second Level; (SR) Shiprock; (UT) Upper Tedrow; (BG) Brown-Grey Channel;
(UU) Up-Up-the-Creek; (ZL) Z-Line; (JC) Jacks Channel. KM and MD are not superpositionally overlain by
the MCZ as shown, but are actually at the tops of isolated buttes.KM, MD, UU, and SR could be either
Facies E or F. Heavy line denotes the erosion surface created by successive channeling events.

Roundtop Channel. Temporal relationships that remain uncertain are those between the Black Spring Coulee and Upper Tedrow, Brown-Grey and Little Roundtop, and Brown-Grey and Second Level channel fills. Physical relationships of the Z-Line Channel to any of the other five channel fills cannot be determined.

Three types of vertebrate assemblages (Lancian, Bug Creek, and Puercan) were collected from lag deposits in Facies E, Facies F, or undifferentiated channel fills within this stratigraphic interval. Z-Line (Facies E) is the only channel fill that preserves a Puercan fauna below the MCZ (refer to Figure 5). Bug Creek assemblages are present in the Little Roundtop, Second Level, Black Spring Coulee, and Brown-Grey channel fills, and the Shiprock and Up-Up-the-Creek sites. A Lancian assemblage is present only in the K-Mark Channel, an erosional remnant of a channel fill of unknown size and facies (Plate 3). Matt's Dino Quarry yielded an associated hadrosaur skeleton in a channel fill of undetermined facies (Plate 3), but mammals were not recovered.

The presence of Lancian, Bug Creek, and Puercan vertebrate assemblages in nearly identical facies in the same stratigraphic interval undoubtedly indicates that some degree of faunal change was occurring at this time. However, because of discontinuous exposures and the lenticular nature of channel facies, temporal sequencing of the three assemblages by superpositional or cross-cutting relationships cannot be determined. Interpretations of faunal change at this stratigraphic interval are ambiguous because they cannot be clearly supported by a superposed stratigraphic sequence of faunas.

Channel fills (Facies B, E, F) commonly fine upward and/or grade laterally into floodplain deposits (Facies G). These floodplains were accreted in unison with channel migration (Facies E) or were developed after vertically accreted channels (Facies B, F) were filled. Facies G floodplains also overlie Facies A floodplains. Areas that preserve this relationship are Sections II and 811 (Plate 4; Figure 5, base of BS traced to left) and Sections B through H (Plate 3; Figure 5, just right of UU, then traced to right). These sections show that channels (Facies F) which produce Bug Creek assemblages (Black Spring Coulee Channel) interfinger with floodplain deposits (Facies G) that contain aquatic vertebrates but lack dinosaurs. In the absence of a facies bias, it appears that the lack of dinosaurs in Facies G floodplains indicates that they were extinct by this time, at least locally. Paradoxically, where it can be determined, floodplains (Facies G) that lack dinosaur remains appear to have accreted in unison with channels (Facies F) that yield dinosaurs in lag concentrations (Section II, Black Spring Coulee Channel, Plate 4; Sections H-B, Plate 2). However, the presence of dinosaur remains in these channel fills is probably the result of reworking (see Lofgren at al., 1990). Reworking would explain why disarticulated remains of dinosaurs are present in channel facies (Facies F) but are absent in contemporaneous floodplain facies (Facies G).

Based on topographic elevation, the stratigraphic level of the basal lags of channels (Facies E, F) which contain Bug Creek assemblages are equivalent to floodplains (Facies A) containing in situ dinosaurs. However, lateral correlation of Facies F channels containing Bug Creek assemblages indicates that these channels interfinger with Facies G floodplains, not Facies A floodplains (Sections II through 8111, Plate 4; Sections B through H, Plate 2). This is a reflection of the deep entrenchment of channels into older floodplain

deposits (Facies A), where as much as 20 m of relief is observed. Therefore, correlation of facies provides a more reliable method of determining relative age than elevation or relative (to laterally persistent coals) stratigraphic level. Analysis of facies relationships of vertebrate-bearing fluvial deposits is critical because faunas present at the same stratigraphic level or elevation could differ significantly in age.

In contrast to the examples where channels containing Bug Creek assemblages can be traced into age-equivalent floodplains, the relationship of channel fills bearing Puercan or Lancian vertebrate assemblages to age-equivalent floodplain deposits (Facies A or G) cannot be determined. The K-Mark Channel and Matt's Dino Quarry site are located at the top of isolated buttes, making lateral correlation to floodplain deposits impossible. The margins of the Z-Line Channel (Facies E) are not exposed in contact with floodplain deposits.

Complete channel-fill (Facies B, E, F) sequences fine upward into floodplains (Facies G) and are all overlain by a laterally extensive lignite (Facies C), the MCZ (base of the Tullock Formation). This indicates that, following the episode of major channeling and erosion, a large peat swamp was developed which blanketed most of the study area. This swamp was in turn replaced by development of extensive areas of standing water, represented by thick sequences of pond deposits (Facies D).

Pond deposits (Facies D) are overlain by fluvial sediments that aggraded in various depositional settings. Of interest are channels (Facies F) with lag concentrations containing vertebrate fossils that are deeply entrenched (up to 15 m) into pond deposits (Facies D) overlying the MCZ (Facies C). These channels, for example, Jacks Channel (Plates 2, 3), contain Puercan assemblages (which lack dinosaurs and typical Lancian mammals). Remains of dinosaurs are not present above the MCZ in the McGuire Creek area, and apparently dinosaurs were extinct by this time (but see Rigby, 1989).

Palynological Correlation of McGuire Creek Lithofacies

It has long been known that a palynological change that could be recognized in stratigraphic sequences in eastern Montana was roughly coincident with R. Brown's (1952) formula (first coal above highest dinosaur) for the terrestrial K-T boundary (Norton and Hall, 1969; Oltz, 1969; Tschudy, 1970). Based on more recent refinements, Cretaceous strata can be differentiated from Tertiary strata by the absence or great reduction in abundance of Cretaceous indicator species in the latter (Smit et al., 1987; Hotton, 1988). Therefore, palynology can be used for determining the relative age of vertebrate-bearing strata in the upper Hell Creek Formation where lithostratigraphic relationships are unclear, if one facies yields a Cretaceous palynoflora while another contains a Tertiary pollen assemblage. (For discussion of palynological recognition of the K-T boundary, see chapter below on Cretaceous-Tertiary Correlations). Thus, palynology is useful in correlating between disjunct exposures of channel (E, F, B) facies and/or floodplain facies (A, G). The results of palynological analyses of rock samples from McGuire Creek are given in Table 2.

Palynological correlation indicates that channel facies with Bug Creek or Puercan assemblages (Facies E, F) are younger than floodplain facies containing in situ dinosaur

remains (Facies A). Facies G floodplain deposits were not analyzed for pollen, but they are Paleocene, because they unequivocally overlie channel fills or Facies A floodplain deposits that produce Paleocene pollen. In contrast, Facies A floodplain deposits may be Paleocene or Cretaceous in age, depending upon the stratigraphic interval sampled. Where channel facies are deeply entrenched into Facies A floodplain deposits, these floodplain deposits are Cretaceous in age, based on pollen (Section FF, Plate 3, and Sections II and O, Plate 4). Where Facies G overlie Facies A floodplain deposits, the uppermost portion of Facies A deposits are palynologically Paleocene (Sections D and B, Plate 2). Smaller-scale channel entrenchment (less than 3 m) into Facies A floodplains containing Paleocene pollen occurs at Section N (Plate 2). Where sampled, Facies A beds that contain in situ dinosaurs produce Cretaceous pollen (Section O, Plate 4; the same bed in Section E from which rock sample 87DLL7-16-21 was collected produced a disarticulated ceratopsian skull 150 m to the south). In situ dinosaurs have not been found in Facies G floodplain deposits that yield Paleocene pollen (but these floodplain deposits yield turtles, champsosaurs, and crocodiles; Sections B-D, Plate 2). Without exception at McGuire Creek, in situ dinosaur remains have not been found in the stratigraphically lowest Paleocene floodplain deposits (as identified palynologically). This is also true in other K-T transition stratigraphic sequences of floodplain deposits in eastern Montana,

where the highest in situ dinosaur remains are never found within 2 m (below) of the palynological boundary (Archibald, 1987c).

Channel fills containing vertebrate fossils in the uppermost Hell Creek Formation were sampled for pollen when suitable lithologies were present. Otherwise, ages of channel fills were determined by superpositional or cross-cutting relationships with another rock unit whose age was determined palynologically. The results of these analyses are given in Table 3 and Figure 5.

All channels containing Bug Creek or Puercan vertebrate assemblages are demonstrated to be of Paleocene age where palynological determination is possible (Table 3; Plates 2-4, Figure 5). Therefore, channel fills containing Bug Creek or Puercan assemblages are palynologically equivalent to floodplain deposits that lack in situ dinosaurs (Facies A and G) and are palynologically dissimilar to floodplain deposits that contain in situ dinosaurs (Facies A).

Similar conclusions, based on lithostratigraphy and palynology, are reached concerning the timing of geologic events in the uppermost Hell Creek Formation at McGuire Creek. Lithofacies correlation indicates that channels (Facies B, E, and F) are entrenched into floodplain deposits (Facies A) containing in situ dinosaur remains (Figure 5; Plates 2-4). Limited lateral correlation of channels (Facies F) suggests interfingering with floodplain deposits that lack dinosaurs (Facies G) (Figure 5; Plates 2, 4). It appears, on the basis of palynology, that deep entrenchment of Paleocene channels into a pre-existing landscape composed of mostly Cretaceous and less commonly Paleocene floodplain deposits occurred; this erosion surface denotes the palynological K-T boundary (shown by heavy line in Figure 5). Correlation of channel facies (Facies E, F, B) laterally to floodplain facies (Facies G) is corroborated by palynological analysis. If present, the uppermost parts of Facies A floodplain deposits that lack in situ dinosaur remains are Paleocene. These strata may be remnants of floodplain deposits accreted lateral to Paleocene channels whose isolated vestiges (for example, the Little Roundtop or Brown-Grey channels) cannot be traced laterally into floodplain sequences because of repeated channeling.

In no case at McGuire Creek is a Bug Creek or Puercan vertebrate assemblage clearly associated with a Cretaceous palynoflora. The basal lags of these channels may contain reworked mud pebbles, cobbles, and boulders that produce Cretaceous pollen (Lofgren et al., 1990) but the channel fills yield Paleocene pollen where sampled (Table 3; Figure 5).

Relative age discrimination of individual channel fills containing Lancian, Bug Creek, or Puercan assemblages in the upper Hell Creek Formation at McGuire Creek is severely limited because all but one yield Paleocene palynofloras (Table 3; Figure 5). Matt's Dino Quarry Channel contains associated hadrosaur skeletal remains (but lacks mammals), yields a Cretaceous palynoflora, and is the only site that is demonstrably older than the others listed in Table 3. The only identifiable Lancian site, the K-Mark Channel, probably is palynologically Cretaceous, but this cannot be confirmed because the sediments are unsuitable for pollen analysis.

McGuire Creek Biostratigraphy

Development of biostratigraphic units at McGuire Creek that span the uppermost Hell Creek Formation is limited because of the lack of stratigraphic control on channel fills that yield fossil assemblages. References to Plates 2-4 are given to demonstrate biostratigraphic relationships in the following discussion (simplified version in Figure 5).

In spite of poor stratigraphic control, superpositional or cross-cutting relationships between channel facies or lignite beds (TL or MCZ) can be determined in some cases. The Jacks Channel local fauna (Puercan) is younger than the others listed in Table 3 because Jacks Channel lies entirely above the MCZ (Section EE, Plates 2 and 3). The other local faunas were collected from channel fills that are (or apparently were) capped by the MCZ prior to Holocene erosion.

Below the MCZ, limited temporal ordering of local faunas containing Bug Creek assemblages can be determined. The Little Roundtop Local Fauna is older than the Second Level Local Fauna because these respective channel fills are in superposition (Sections 861-R, Plate 3). The crosscutting relationship between the channel fills reveals that the Black Spring Coulee Local Fauna is younger than the Second Level Local Fauna (Section 871-II, Plate 4). The Upper Tedrow Channel cuts and/or overlies both the Second Level (Sections O-P-856-855, Plate 4) and Brown-Grey (Section GG, Plate 3) channels. The biostratigraphic utility of these relationships is limited, however, because only one mammal, Meniscoessus robustus , has been found in the Upper Tedrow Channel. The "Three Buttes Local Fauna" is an informal grouping of vertebrate assemblages found in the erosional remnants of channel fill(s) in an area adjacent to exposures mapped as Upper Tedrow or Second Level channels (Plate 1; NE quadrant, SE quadrant, Section 28, T 22N, R 43 E). However, lag deposits which produce fossils included within this "local fauna" are poorly constrained stratigraphically and cannot be correlated with either channel fill because of Holocene erosion of intervening exposures ("Three Buttes Local Fauna" is placed in quotations in Appendix 2 to denote this uncertainty).

The remaining channel fills containing the Lancian (K-Mark, Section BB, Plate 3), other Bug Creek (Shiprock, Section M1, Plate 3; Up-Up-the-Creek, Plate 1), and Puercan (Z-Line, Sections JJ-N-W, Plate 2) local faunas lack stratigraphic control, making their temporal relationships uncertain (other than relative to Jacks Channel). Usually lateral correlation of strata that cap each channel fill would aid in determining relative age, but all upper Hell Creek channel fills are (or apparently were) capped by the same laterally traceable stratigraphic unit, the MCZ.

Palynological data adds little information useful for more detailed biostratigraphic zonation of the uppermost Hell Creek Formation because all sites yielding vertebrates (with the exception of Matt's Dino Quarry Channel) yield similar palynofloras (Table 3, Figure 5).

In summary, it is virtually impossible to develop a local biostratigraphic zonation for the upper Hell Creek Formation at McGuire Creek that orders the three kinds of vertebrate assemblages. The uppermost 25 m of the Hell Creek Formation contains the records of a major faunal turnover, but the transition is recorded within channel-filling events that are not amenable to biostratigraphic subdivision.

Figure 6
Idealized cross-section of the Tedrow Area showing location of fossil and rock sample sites,
channel fills, and contacts between depositional units. The position of measured sections FF
and GG are shown on Plate 3. Rock samples analyzed for pollen are: (1) 86DLL7-29-1:
Cretaceous; (2) 86DLL7-29-2: Cretaceous; (3) 88DLL7-14-30: Paleocene; (4) 88DLL7-14-13:
Paleocene. Isolated mammal specimens are: (5)  Stygimys  incisors. (6) Ragnarok  lower molars.

The Tedrow Area

The complexity of channel cutting and filling in the uppermost Hell Creek Formation is shown in the Tedrow area (the term "Tedrow area" describes exposures adjacent to and including Sections FF-GG, Plate 3; Figure 6). The Brown-Grey, Upper Tedrow, and "Lower Tedrow" channel fills, and a mudstone lens containing Cretaceous pollen, are exposed in the Tedrow area, but determination of a sequence of depositional events for the channel fills and mudstone lens is difficult because Holocene erosion has removed critical information. These difficulties highlight the uncertainty inherent in biostratigraphic study of complex fluvial channel deposits, even in an area in which three-dimensional exposures are available.

The "Lower Tedrow" Channel is a trough cross-stratified sandstone containing a sparsely sampled vertebrate fauna consisting of associated ceratopsian skeletal remains (Figure 7) and a few isolated mammal teeth referable to Ragnarok and

Figure 7
Plan view of Tedrow Dinosaur Quarry (V87152). Letters on quarry map identify the following
 elements: (A) Sacrum: broken at base of neural arch over nearly entire length, entire dorsal
 part of sacrum absent. Transverse processes missing on many vertebrae. (B) Mandible (right):
 edentulous; dentary, surangular, and possibly angular present. Dental battery fractured-
weathered. (C) Dorsal rib #2 (left): distal end of shaft lost. (D) Squamosal (right): lateral edge
 and posterior portion of parietal suture present; broken medially at site of large ventro-medial
 depression. (E) Squamosal fragment (left): only posterior portion present. (F) and (G) Squamosal
 fragments (left): can be rearticulated; groove for quadrate (F) and heavy ridge of bone
 (across F-G) against which process of exoccipital abuts present; E-F-G from same squamosal.
 (H) Cervical vertebra #7: end of right transverse process lost. (I) Cervical vertebra #6. Note: J, K,
 L, M, and T not figured. (J) Cervical vertebra #8?: 50% of vertebra recovered as float; neural
 arch reconstructed. (K) Two unidentified fragments: wedged against sacrum; one fragment
 well rounded. (L) Frill? fragment (27×10 cm) nestled against sacrum. (M) Frill? fragment
 (14×10 cm). (T) Turtle shell fragments scattered throughout quarry; many from one individual
 of Adocus . Also present, but not shown, are many unidentifiable bone fragments (<12 cm length).

Stygimys . The dinosaur remains were quarried from the basal lag of the channel fill, which also contained clay balls, fragments of turtle shell, and carbonized plant debris (V87152, Figures 6, 7). Two molars of Ragnarok and two incisors of Stygimys were collected .7-.8 m and 1.5-1.7 m above the base of the channel fill, respectively (Figure 6). These specimens occur within troughs containing clay balls and other vertebrate fossils. It is uncertain if the dinosaur remains and the mammals occur within sediments deposited in the same depositional event (see discussion of Upper Tedrow Channel below).

Figure 8
Scenario 1 sequence. (1) Deposition of Lower Tedrow Channel fill with dinosaur remains in
basal lag, capped by mudstone lens with Cretaceous pollen. (2) Partial removal of 1 during
entrenchment of Upper Tedrow Channel. Dinosaur remains in Lower Tedrow Channel fill not
disturbed. (3) Deposition of Upper Tedrow Channel fill, including the mammals  Stygimys  and
Ragnarok . (4) Area as exposed today. In this scenario, the associated dinosaur remains
and the mammals were deposited in separate filling events or stories. Based on pollen,
the lower story with the dinosaur bones would be Cretaceous and the higher story with
the mammal teeth would be Paleocene.

Figure 9
Scenario 2 sequence. (1) Deposition of unfossiliferous sandstone channel fill, capped by
a mudstone lens containing Cretaceous pollen. (2) Partial removal of 1 during entrenchment
of Upper Tedrow Channel. (3) Deposition of Upper Tedrow Channel fill, with dinosaur remains
in basal lag and the mammals  Stygimys  and Ragnarok  higher in the fill. (4) Area as exposed
today. In this scenario, all the vertebrate remains in the "Lower Tedrow" Channel are Paleocene
unless reworked from Cretaceous sediments. Also, the "Lower Tedrow" Channel is a localized
depression in the Upper Tedrow Channel, and both were filled at about the same time.

The "Lower Tedrow" Channel could be Cretaceous or Paleocene, depending on whether it is part of the Upper Tedrow Channel. Situated 2 m above (topographically) the base of the "Lower Tedrow" Channel is a thick mudstone lens containing Cretaceous pollen (Lofgren and Hotton, 1988), which is overlain by the Upper Tedrow Channel (Figure 6). It is uncertain whether this mudstone ever extended across the part of the channel fill that produced the dinosaur remains (contra Lofgren and Hotton, 1988) and mammal teeth. Erosion has removed the critical area in which the Cretaceous mudstone lens would have been either in superposition above part of the "Lower Tedrow" Channel or clearly cut out by the Upper Tedrow Channel (Figures 8, 9). If the latter were true, then the "Lower Tedrow" Channel would represent a localized depression in the Upper Tedrow Channel (Figure 9, nos. 2-3).

The Brown-Grey Channel contains a well sampled Bug Creek assemblage in its basal lag (V87072, Figure 6) and its palynological age is Paleocene. This channel fill cuts the mudstone lens containing Cretaceous pollen. Lofgren and Hotton (1988) thought that the mudstone lens overlay the Brown-Grey Channel; this is now known to be incorrect.

Upper Tedrow is a multi-storied channel fill whose lowest story may include part or all of the "Lower Tedrow" Channel. The Upper Tedrow Channel is Paleocene because it cuts the Brown-Grey (Figure 6; Section GG, Plate 3) and Second Level channels (Sections P, 856, Plate 4), both of which produce Paleocene pollen (Table 3). Therefore, if part of the "Lower Tedrow" Channel is a localized depression of the Upper Tedrow Channel, it is probably Paleocene in age.

On the basis of preliminary palynological and geological investigations outlined in Lofgren and Hotton (1988), it was thought that the sequence of events in the Tedrow Area was: (1) deposition of the Brown-Grey Channel; (2) entrenchment of the "Lower Tedrow" Channel containing the dinosaur remains and subsequent filling; (3) deposition of the mudstone lens overlying "Lower Tedrow" during the Cretaceous; (4) entrenchment of Upper Tedrow into all three of the above. If this sequence of events were correct, then the "Lower Tedrow" and Brown-Grey channels would be Cretaceous and the Upper Tedrow Channel would be Paleocene. Trenching in the Tedrow area in 1989 revealed that the Brown-Grey Channel cuts the Cretaceous mudstone lens. This is corroborated by the presence of a Paleocene pollen assemblage in samples 88DLL7-14-13 and 88DLL7-14-30 from the Brown-Grey Channel, which also indicates that this channel is younger than the Cretaceous mudstone lens. However, the "Lower Tedrow" Channel appears to cut the Brown-Grey Channel when traced westward. This is impossible, because the Brown-Grey Channel cannot be cut by a channel ("Lower Tedrow") that is capped by a unit (mudstone lens) cut by the Brown-Grey Channel (Figure 6). Therefore, part of the "Lower Tedrow" Channel must be equivalent to the Upper Tedrow scouring event (that cut the Brown-Grey Channel and the Cretaceous mudstone lens). This fact negates a two-stage channeling hypothesis in which all vertebrate remains in the "Lower Tedrow" Channel are Cretaceous, and the "Lower" and Upper Tedrow channels are two distinct cut-and-fill events (contra Lofgren and Hotton, 1988).

Based on available geologic data from the Tedrow Area, two possible scenarios of

depositional events are presented in Figures 8 and 9, and neither can be falsified by direct field observation. If correct, the first hypothesis (Figure 8) would indicate a Cretaceous age for the dinosaur remains and a Paleocene age for the mammals. The second (Figure 9) would indicate a Paleocene age for both the dinosaur remains and the mammals, unless reworking of fossils occurred.

If the dinosaur remains were not reworked, then choosing between the two hypotheses has important implications for the timing of dinosaur extinction. The first hypothesis is more in line with conventional palynological-faunal correlations in which dinosaur remains indicate a Cretaceous age and the mammal Ragnarok indicates a Paleocene age. Ragnarok is only known to occur in clear association with Paleocene palynofloras at McGuire Creek and elsewhere in McCone County (Sloan et al., 1986; Rigby et al., 1987; Rigby, 1989). The associated ceratopsian remains would indicate a Cretaceous age for the base of the channel fill.

The second hypothesis is more controversial, because the presence of associated dinosaur remains in a Paleocene channel fill might suggest dinosaur survival into the Paleocene. The association of disarticulated skeletal material (Figure 7) might suggest that the bones were not reworked. However, reworking of dinosaur remains that persist in near association would be possible if the remains were eroded from nearby bank material and underwent little or no transport. Dinosaur remains have been found in channel deposits containing Paleocene pollen at McGuire Creek and elsewhere (Sloan et al., 1986; Rigby et al., 1987) but that these remains are not reworked from Cretaceous sediments has not been satisfactorily documented (Lofgren et al., 1990). Therefore, if the second hypothesis is correct, the dinosaur remains probably were reworked from nearby Cretaceous sediments into which the Upper Tedrow Channel was entrenched.

In summary, difficulties encountered in biostratigraphic study of complex fluvial deposits in the Tedrow Area indicate that determining a clear sequence of depositional events in even a limited area can be difficult. The complexity of channel cutting and filling in the upper Hell Creek Formation is imposing.

Reworking of Fossils

The most perplexing problem concerning vertebrate biostratigraphy of the K-T transition in eastern Montana is that of reworked fossils. As the data presented in the chapter above on Geological and Palynological Correlations show, the potential for reworking in the upper Hell Creek Formation is staggering. Channel fills containing Bug Creek assemblages were deeply entrenched into Lancian strata (up to 20 m), and a tremendous volume of sediment was eroded (see Figure 5). Exhumation and reburial of dinosaur and Lancian mammal fossils during the cutting and filling of these channels must have occurred, but to what extent? This question is especially critical in the uppermost Hell Creek Formation where there is a significant change in the vertebrate fauna preserved within a series of channel fills, as evidenced by the presence of Lancian, Bug Creek, and Puercan assemblages.

Originally, the Bug Creek assemblages (Bug Creek Anthills, Bug Creek West, and Harbicht Hill) were interpreted as being composed of the remains of animals that once lived contemporaneously (Sloan and Van Valen, 1965). Although these fossils were recovered from large channel fills, the possibility that some of them might be reworked was not considered, at least in print. Later, the depositional setting of these vertebrate remains and the potential for reworked fossils were more thoroughly appreciated (Smit and Van der Kaars, 1984; Bryant et al., 1986; Retallack et al., 1987; Fastovsky, 1987; Eaton et al., 1989; Bryant, 1989; Lofgren et al., 1990; Lofgren and Hotton, 1991a, 1991b).

The confusion over which fossils are reworked impinges on K-T correlations because many channels that produce dinosaurs and potentially reworked Lancian mammals yield Paleocene palynofloras. Some researchers argue that dinosaurs (but not Lancian mammals?) persisted into the Paleocene (Sloan et al., 1986; Rigby et al., 1987; Rigby, 1987, 1989), while others suggest that these remains are reworked (Smit and Van der Kaars, 1984; Bryant et al., 1986; Retallack et al., 1987; Bryant, 1989). At issue is whether Bug Creek assemblages represent (1) the remains of animals that lived contemporaneously, or (2) the remains of animals from temporally distinct assemblages that have been mixed through reworking.

Identification of Reworked Fossils

The problem then becomes one of identifying, if possible, which taxa have been reworked and to what extent reworking has extended their geologic range. Lancian mammal taxa whose presence in Bug Creek assemblages is probably the result of reworking, because they only occur in channel fills (many of which yielded Paleocene palynofloras) that are entrenched into Lancian sediments, are indicated by asterisks in Table 4. These taxa are identified as probably reworked. In all subsequent discussion of reworking, the term "Lancian mammals" will refer to those denoted by asterisks in Table 4.

To aid in recognition of reworked fossils, transport experiments on crocodile and dinosaur teeth were conducted, with mixed results (Argast et al., 1987; Ely and Rigby, 1989). These experiments compared laboratory controlled abrasion of fossil and extant vertebrate teeth with the abrasion evident on fossil teeth recovered from channel deposits (i.e., no abrasion=low potential of having been reworked; abraded=high potential of having been reworked).

Before reworking and/or transporting of vertebrate remains can be discussed further, the terminology employed needs to be clarified. Transported refers here to animal remains that are moved from one place to another, reworked here means to reprocess animal remains, in this case, those that were deposited, exhumed, and then reburied. Therefore, two very different components are involved: a spatial (transported: to move some distance) and a temporal component (reworked: to reprocess).

All fossils in channel deposits are transported to some degree, but not all fossils are reworked. Therefore, the amount of transport and the potential that the remains were reprocessed are interrelated only when reworked remains have undergone enough transport so that abrasion is noticeable. Elements that are transported a short distance usually will not be abraded, whether they are reworked or not (unless bone at the surface of a channel floor is sand-blasted). Conversely, elements that were transported and abraded before initial burial may not be reworked. Therefore, it is guesswork to link abrasion or absence of abrasion with reworking. Transport experiments on vertebrate remains are helpful as indicators of how to recognize a fossil that has been transported (Ely and Rigby, 1989; but see Argast et al., 1987), but they do not provide the insight necessary to reconstruct the taphonomic history of any particular fossil with confidence.

To illustrate this point, the basal lag of the Black Spring Coulee Channel (V87030, Section II, Plate 4) yields many well-preserved dinosaur elements which are dispersed within a conglomerate composed of cobble and boulder-sized angular blocks of mud-

stone. Some dinosaur bones are still encased in mudstone blocks, and one of these yielded a Cretaceous palynoflora (Lofgren et al., 1990). In contrast, a thin siltstone lens within this channel fill yielded a Paleocene palynoflora. Therefore, based on sedimentological and palynological analysis, these dinosaur remains were reworked from Cretaceous floodplain deposits into a Paleocene channel fill (Lofgren et al., 1990). This example demonstrates the potential difficulty in distinguishing reworked material from vertebrate remains that were initially buried during channel filling. Another example of the reworking of Cretaceous fossils into Tertiary deposits is provided by Eaton et al. (1989).

"Paleocene" Dinosaurs

Geological and faunal data from Bug Creek West, Harbicht Hill, Ferguson Ranch, Doverzee, and other sites in McCone County, eastern Montana, have been presented in support of the hypothesis that dinosaurs survived into the Paleocene (Sloan et al., 1986; Rigby et al., 1987; Rigby, 1987, 1989). Unabraded, isolated teeth of dinosaurs from Paleocene channel fills (containing Paleocene palynofloras or truncating beds of Paleocene age) have been interpreted as unreworked and cited as evidence in support of this assertion. However, these records are limited to channel deposits in the uppermost Hell Creek Formation that are deeply incised into Cretaceous dinosaur-bearing strata (see Rigby et al., 1987: figures 3, 5). In overbank stratigraphic sequences uncomplicated by major channeling events, the highest dinosaur remains occur approximately 2m below the palynological K-T boundary (Archibald, 1987c). If large dinosaur bones had been reworked into similar channel facies at McGuire Creek (Lofgren et al., 1990), would records of "Paleocene" dinosaurs based upon isolated teeth bear serious consideration? J.K. Rigby, Jr. (1987, 1989) provides the most complete argument favoring the existence of "Paleocene" dinosaurs. His points favoring this interpretation, here numbered 1-5, are evaluated individually.

1. Clay-pebble clasts and vertebrate remains in channel lag deposits probably would not survive lengthy transport; therefore, the presence of clasts associated with dinosaur remains indicates that neither was reworked (Rigby, 1987). Evaluation : If the clay clasts and vertebrate remains are reworked and not transported significant distances, this argument is nullified. Rigby argues that clay pebbles will not survive significant transport and therefore must be locally derived. If these clay clasts were eroded from bank material, the fact that they are preserved at all indicates little transport. This would also apply to the vertebrate remains eroded from the same bank material from which the clay pebbles were derived.

2. The potential source of reworked fossils is overbank deposits, which are not very fossiliferous and therefore are not likely to be the source of the extraordinary abundance of vertebrates seen in screen-washed concentrates (Rigby, 1987).

Evaluation : Eroded overbank facies are the source of the abundant reworked dinosaur remains in one channel fill analyzed by Lofgren et al. (1990). Rigby apparently made the assumption that eroded overbank deposits are the primary source for reworked dinosaur remains. In addition, erosion of pre-existing channel deposits, which are com-

mon in the upper Hell Creek Formation, would have been a major source for reworked fossils. Finally, the preservation potential of animal remains deposited in overbank and channel facies might be quite different. Diagenetic factors, especially soil formation, might preferentially destroy vertebrate remains in overbank facies either prior to, or after deep burial.

3. Many "Paleocene" dinosaur remains appear unabraded, and this would not be the case had these remains undergone transport and reworking (Rigby, 1987). Evaluation : As discussed above, fossils that are reworked and undergo little transport may not exhibit abrasion. Absence of abrasion does not exclude the possibility of reworking.

4. Many Lancian mammals do not appear associated with "Paleocene" dinosaur remains; if reworking occurred it must have been selective, which is improbable (Rigby, 1987). Evaluation : Bug Creek sites from McGuire Creek yielding Paleocene palynofloras contain dinosaurian remains associated with a diverse fauna of Lancian mammals (compare BG to HH and BA, Table 4). The Bug Creek West and Harbicht Hill sites, which yield Paleocene palynofloras and "Paleocene" dinosaurs (Sloan et al., 1986; Rigby et al., 1987), are not thoroughly sampled, nor are their vertebrate faunas completely described. If these sites were resampled and their faunas restudied, many Lancian taxa would probably be discovered. Many Paleocene McGuire Creek local faunas contain Lancian mammals as well as dinosaur remains. If dinosaur remains in Paleocene channel fills were not all reworked, it would be equally likely that Lancian mammals in Paleocene channel fills at McGuire Creek were not all reworked (contra Sloan et al., 1986; Rigby et al., 1987; Rigby, 1989).

Also, there are in McGuire Creek local faunas at V87072 and V87035-8 higher relative abundances of Lancian mammals than in Bug Creek Anthills (Table 5), which is presumably older (see the next chapter, on McGuire Creek Biochronology and the "Bugcreekian Age"). On the basis of arguments similar to those claiming the existence of "Paleocene" dinosaurs, it would seem that Lancian mammals also survived into the Paleocene because they become more abundant in younger channel fills, from Bug Creek Anthills (older) to V87072 and V87035-8 (younger). However, the reworking of vertebrate remains does not follow any predictable pattern, and the amount of reworked material will depend on the amount of previously deposited sediment eroded by the stream as well as the fossil content of the eroded sediment. These factors vary independently from the relative age of a channel fill.

5. The base of the Doverzee Channel fill lies just above the upper "Z" lignite (palynologically Paleocene) and yields dinosaur teeth but lacks Lancian mammals (Rigby, 1989). Therefore, the Doverzee Channel was never entrenched into Cretaceous sediment, and the dinosaur teeth in the channel fill are not reworked from Cretaceous beds. Evaluation : Independent mapping of the Doverzee channel fill (SW quadrant of Section 16, T 22 N, R 43 E) revealed the following points: (A) When the Doverzee Channel fill is traced southeast (downstream) across discontinuous exposures (based on paleocurrent azimuths and lithologic similarity) from the spot where the dinosaur teeth were reportedly collected (Rigby, 1989), it cuts out the upper "Z" lignite (Section OO, Plates 1 and 4). (B) Traced further southeast, the Doverzee Channel cuts into strata many meters below the "Z" lignite to a level presumably Lancian in age (contra Rigby,

1989). (C) Although southeast is the direction of paleoflow, if the Doverzee stream cut through the upper "Z" lignite downstream, it is probable (but not demonstrable) that this occurred upstream as well. Therefore, eroded Lancian strata could easily have been the source of the dinosaur teeth. (D) Given the history of claims that the absence of taxa is an artifact of small sample size (e.g., at Bug Creek West and Harbicht Hill), the claimed absence of Lancian mammals at the Doverzee locality cannot be evaluated until data on sample size and methods of collecting are given.

In summary, Rigby's (1987, 1989) arguments fall far short of substantiating the claim of dinosaur survival into the Paleocene.

Bug Creek Faunas Represent Contemporaneous Assemblages

Based on data from McGuire Creek, arguments similar to those supporting "Paleocene" dinosaurs could be proposed to support the hypothesis that Bug Creek faunas each represent contemporaneous assemblages. Pertinent evidence is as follows:

1. The excellent condition of Lancian mammal remains, especially those of marsupials, is evidence that these fossils were not all reworked. Three jaw and maxillary fragments containing 1-3 teeth of Pediomys and Glasbius were discovered at the Up-Up-the-Creek and Brown-Grey Channel localities. These are very delicate fossils and presumably would have been destroyed easily during erosion and redeposition by stream processes. Evaluation : These remains are found in a stream deposit and must have undergone some transport. Therefore, it is possible that they were reworked and underwent minimal transport before reburial. This argument would also apply to the presence of well-preserved dinosaur bones and teeth in Paleocene channels. In fact, it has been documented that well-preserved dinosaur remains were reworked into a Paleocene channel fill (Lofgren et al., 1990).

2. Lancian mammal and dinosaur remains are relatively abundant at certain McGuire Creek sites, such as V87072 (Brown-Grey Channel) and V87035-8 (Up-Up-the-Creek Local Fauna), more so than might be expected if reworking were the cause. In the case of the mammals, the marsupials Glasbius twitchelli and Pediomys elegans are especially abundant. Pediomys elegans is more abundant than any "condylarth" species at V87072, as is Glasbius twitchelli at V87035-8 (Table 6). Also, Lancian mammal taxa (denoted with * in Tables 5, 6, and 7) constitute over 5% of the total mammalian fauna in NISP (Number of Identifiable Specimens per Taxon) at both the generic and specific levels, and 16-19% in MNI (Minimum Number of Individuals) at the specific level from these two localities (Tables 6, 7). In a channel-fill depositional setting, relative abundance data are unlikely to reflect the relative abundances of animals that lived in the area when the deposit was formed. However, can the reworking hypothesis explain these apparently high relative abundances of Lancian taxa, especially when MNIs are considered? Evaluation : Vertebrate fossils in channel deposits are hydraulically sorted and transported, so elements identified as coming from one taxon have a low probability of being from one individual. Therefore, it is more appropriate in this deposi-

tional setting to use NISPs than MNIs (Badgley, 1986), because MNIs will overestimate the relative abundance of rare taxa. Flat Creek, the stratigraphically highest Lancian local fauna in the Hell Creek Formation (Figure 1b), produces about 66% Lancian taxa (NISP) at the generic level (Table 5). At McGuire Creek sites V87072 and V87035-8, the relative abundance of Lancian mammals is much lower (5%, NISP), which would be expected if reworking were responsible for their presence. Lancian mammal remains in McGuire Creek channel fills probably are fossils redeposited following erosion from facies containing abundant Lancian mammals.

The two arguments presented above are based on robust data from McGuire Creek, but fall short of demonstrating the contemporaneity of the components of Bug Creek assemblages (Lancian-Puercan mammals and dinosaurs).

Bug Creek Faunas Contain Reworked Dinosaurs/Lancian Mammals

Evidence suggesting that association of dinosaurs and Lancian mammals in Bug Creek assemblages at McGuire Creek is the result of reworking is outlined below:

1. Bug Creek assemblages from McGuire Creek are restricted to the basal lags of channels which were incised into Lancian strata (up to 20 m) (see chapter above on Geological and Palynological Correlations). This correlation of channel incision with the presence of Lancian mammals and dinosaurs in Bug Creek assemblages is not coincidental. Such a particular mixture of faunal components was caused by reworking. Similar arguments were proposed by Smit and Van der Kaars, 1984; Bryant et al., 1986; Retallack et al., 1987; and Bryant, 1989, based on the depositional setting of the original Bug Creek sites.

2. The reworking hypothesis can be tested with analyses of Facies G strata that overlie (or are laterally equivalent to) channel fills containing Bug Creek assemblages in the upper Hell Creek Formation at McGuire Creek. Facies G strata represent a low-energy, floodplain depositional setting where the possibility of reworking can be virtually eliminated. Unfortunately, this depositional setting usually is not conducive to producing microvertebrate concentrations. However, dinosaur remains commonly are present in floodplain deposits, and presumably would be present in Facies G floodplain deposits that are equivalent in age to channel facies that contain dinosaur remains. However, dinosaurs were not recovered from Facies G deposits. The absence of dinosaur remains in Facies G floodplains (which yield other vertebrates) is especially significant for laterally accreted channel facies (Facies E), where floodplain deposits and contemporaneous channel facies are juxtaposed (Figure 10). One would expect to find dinosaur remains in floodplain sediments (Facies G) if dinosaurs had been extant during the deposition of directly adjacent channel fills (Facies E). Because dinosaur remains occur in the basal lags of channels but are absent in age-equivalent floodplain deposits, their presence in these channels is probably the result of reworking.

3. Channel fills with nearly identical assemblages (those taxa not marked by * in

Figure 10
Taphonomic model adapted from Behrensmeyer, 1982, for occurrence of dinosaur remains in Paleocene
(or Pu0-Pu1) Facies E channel fills at McGuire Creek. The three possible sources of the dinosaur remains
(or Lancian mammals) are: (X) reworked from older strata through bank erosion, (Y) overland transport
from the floodplain surface into the channel, (Z) individuals that died in the channel. Dinosaur remains
are found in the basal lags of Facies E channels but not in age-equivalent floodplain deposits. The
absence of dinosaur remains in floodplain deposits which aggraded juxtaposed to laterally accreting
channels yielding dinosaurs suggests that the source of the remains is pathway X. If dinosaur bones
and teeth were entering the channel via pathways Y or Z, their remains would be expected to occur in
both the channel and floodplain deposits. Therefore, it is unlikely that dinosaurs lived during the
formation of these channel fills.

Table 4) sometimes produce dinosaur remains and Lancian mammals (BG, SR, UU) and sometimes do not (HE, ZL). Based on taxa which make their first appearance in the Puercan, it was concluded that these channel fills are approximately the same age but differ primarily in their Lancian components. Lancian taxa are found only in those channels that cut Lancian strata (Table 4, compare BG, SR, and UC with HE or ZL). These data indicate that the Lancian fossils were reworked.

The question then becomes one of deciding which argument is more compelling: reworked or not reworked? I suggest that the reworking argument is more compelling for the following reasons: (1) deep incision of channels (with Bug Creek assemblages) resulted in the subsequent erosion of a tremendous volume of Lancian strata; (2) exclusive association of Bug Creek assemblages with channels that demonstrably cut Lancian strata; (3) the presence of dinosaur remains in the basal lags of channels but their absence in age-equivalent floodplain deposits that yield other vertebrates; (4) the lack of dinosaurs in the stratigraphically lowest Paleocene floodplain deposits identified palynologically; and (5) the unequivocal occurrence of Cretaceous dinosaur remains in a Paleocene channel fill (Lofgren et al., 1990).

These data are hard to dismiss, especially considering the enormous potential for reworking in this depositional and stratigraphic setting. The burden then rests on those who argue for faunal contemporanity of dinosaurs and Lancian-Puercan mammals in Bug Creek assemblages, or "Paleocene" dinosaurs, to demonstrate that the fossils in question are not reworked.

To this end, a standard for recognition of "Paleocene" dinosaurs, consisting of four criteria, was proposed: (1) articulated dinosaur remains from a Paleocene channel fill; (2) in situ dinosaur from a Paleocene floodplain; (3) abundant dinosaur remains from a channel fill that did not scour Cretaceous sediments; and (4) dinosaur remains demonstrably reworked from Paleocene sediments (Lofgren et al., 1990). These criteria provide a standard for assessing the significance of Lancian mammal and dinosaur fossils in Paleocene rocks. Until at least one of these criteria is met, the reworking hypothesis should be dismissed as unsubstantiated.

In conclusion, the presence of dinosaurs and Lancian mammals in Bug Creek assemblages (and Paleocene channels) from the upper Hell Creek Formation at McGuire Creek is a result of reworking. Similarly, the claim of dinosaur survival into the Paleocene is rejected.

McGuire Creek Biochronology and the "Bugcreekian Age"

A "Bugcreekian" NALMA was proposed by Sloan (1987) and Archibald (1987a, 1987b) based on the time represented by the original Bug Creek faunas. A year later, I shared preliminary faunal data from McGuire Creek with J.D. Archibald which subsequently convinced him to abandon the "Bugcreekian" NALMA in favor of a less formalized mammalian zonation because of the difficulty of distinguishing earliest Puercan from "Bugcreekian" local faunas (see Archibald and Lofgren, 1990). Thus, the "Bugcreekian" NALMA was reduced in scale and became the Pu0 interval zone of the Puercan "Age," and McGuire Creek local faunas were referred to the Lancian and Puercan (Pu0 and Pu1 interval zones) NALMAs (Archibald and Lofgren, 1990) (Figure 11, Table 8). Additional faunal data from McGuire Creek (and sites nearby) further supports abandonment of a "Bugcreekian Age." These data and a discussion of the development of the "Bugcreekian" and the Pu0 interval zone (and subdivision into informal biochrons) are presented below.

Figure 11
Relationship of McGuire Creek sites to previously proposed K-T mammalian zonations.
Adapted from Archibald et al., 1987, and Archibald and Lofgren, 1990.

The "Bugcreekian Age"

The composition of their mammalian assemblages suggested that the Bug Creek faunas were younger than Lancian faunas (such as Ken's Saddle) within the Hell Creek Formation, but that they were deposited before the early Puercan (Sloan and Van Valen, 1965). Therefore, it was inferred that the Bug Creek faunas represented a previously unsampled faunal interval between the Lancian and Puercan "ages." Also, the Bug Creek faunas were placed in an ascending temporal sequence of Bug Creek Anthills-Bug Creek West-Harbicht Hill, partly on the basis of the gradual appearance of archaic ungulate mammal species (or "condylarths") (Bug Creek Anthills: Protungulatum donnae ; Bug Creek West: Protungulatum gorgun, Mimatuta morgoth ; Harbicht Hill: Oxyprimus erikseni, Ragnarok "harbichti" ) (Sloan and Van Valen, 1965; Sloan et al., 1986).

However, following their original descriptions, the Bug Creek faunas were generally considered to be Lancian in age (Sloan, 1970, 1976; Russell, 1975; Fox, 1978; Archibald, 1982). Johnston (1980) reported vertebrate assemblages from Saskatchewan (Frenchman 1, Frenchman Formation; Long Fall, Ravenscrag Formation) similar in composition to the Bug Creek faunas. Later, these assemblages were also interpreted to be Lancian in age (Johnston and Fox, 1984; Fox 1987, 1989, 1990).

The first proposal that the Bug Creek faunas represented a new temporal unit following the Lancian was provided by Van Valen and Sloan (1977:42): "We will refer to the time from the deposition of Bug Creek Anthills to that of the Z coal bed as Bug Creek time". Later, reference was made to a "Bugcreekian Age" (Sloan, 1983; Rigby et al., 1986; and Sloan et al., 1986), but neither definition nor characterization were given.

Formal definition of a "Bugcreekian Age" was presented by Sloan (1987) and Archibald (1987a, 1987b). Sloan (1987) defined the "Bugcreekian Age (Stage)" as equivalent to the Protungulatum zone or the Bug Creek faunal sequence (Bug Creek Anthills-Bug Creek West-Harbicht Hill) and its equivalents (which included Frenchman 1). Archibald (1987a, 1987b) defined the "Bugcreekian" NALMA as the time between the successive first appearances of the "condylarth" Protungulatum and the marsupial Peradectes . However, R.C. Fox (1989) rejected the concept of a "Bugcreekian" NALMA, preferring to recognize the Frenchman 1 (and Long Fall) fauna as representing latest Lancian time.

Archibald et al. (1987) subdivided the Puercan NALMA into a series of interval zones defined by successive first appearances (or FADs: First Appearance Datums) of unrelated taxa. The initial interval zone (Pu1) of the Puercan was defined as the time between the first occurrence of the marsupial Peradectes and that of the "condylarth" Ectoconus . As defined, the Pu1 or Peradectes / Ectoconus interval zone included the Hell's Hollow Local Fauna (lower Tullock Formation of Garfield County, Montana; see Figure 1b).

Pu0 and Pu1 Interval Zone of the Puercan Nalma

The "Bugcreekian Age" of Archibald (1987a, 1987b) and Sloan (1987) was accepted as being a previously unsampled faunal interval between the Lancian and Puercan "ages"

by Archibald and Lofgren (1990), but was rejected as the basis for proposing a new land-mammal age. Instead, the "Bugcreekian" was reduced in rank and defined as the initial interval zone of the Puercan or Pu0: Protungulatum / Peradectes interval zone. This change was necessitated by the discovery of Bug Creek assemblages at McGuire Creek which were transitional in composition between the "Bugcreekian Age" (now Pu0) and the earliest Puercan Pu1 interval zone. Prior to these discoveries, the youngest "Bugcreekian" fauna (Harbicht Hill) and the oldest Puercan local fauna (Hell's Hollow) were faunally distinct (approximately 50% overlap of species; see Archibald, 1982: table 59). The major faunal differences between the Hell's Hollow Local Fauna and the original Bug Creek sites were the disappearance of many Lancian taxa combined with the first appearance of Peradectes at Hell's Hollow (Archibald, 1982: table 59). Table 8 gives additional faunal data from the Bug Creek sites.

When considering the faunal data from McGuire Creek, it is difficult to draw a boundary between the "Bugcreekian" (or Pu0) and Pu1 interval zones, because they are similar in composition. Ignoring Lancian genera (which might be reworked), Pu0 faunas at LF, LR, SL, and BS are virtually identical to those at BG, SR, UU, and HE except that the latter localities yield Peradectes . Therefore, as defined, the Pu0 interval zone lacked "index" genera (Table 9) and contained only two possible "index" species, Catopsalis joyneri and Purgatorius ceratops . However, data from McGuire Creek indicate that Peradectes and C. joyneri occur in the Brown-Grey Local Fauna (Pu1) (Table 10), and the occurrence of P. ceratops at Harbicht Hill is suspect (Lillegraven and McKenna, 1986). Therefore, local faunas from the late "Bugcreekian" (Pu0) and early Pu1 interval zones can be nearly identical.

This uncertainty is compounded at McGuire Creek because Peradectes is rare. Therefore, the assignment of LR, SL, BS, BG, SR, and UU faunas to either the Pu0 or Pu1 interval zones might be influenced by sampling factors.

Finally, Pu0 and Pu1 sites are not well constrained biostratigraphically at McGuire Creek or elsewhere in the western interior. Documentation of the limited biostratigraphic control on Lancian, Bug Creek (Pu0: LR; Pu1: BG, SR, UU, SL?, BS?), and Puercan (Pu1: ZL, JC) local faunas at McGuire Creek was presented earlier. Therefore, while arguing that the Pu0 interval zone ("Bugcreekian") is a previously unsampled faunal interval, Archibald and Lofgren (1990) concluded that the potential difficulty in distinguishing Pu0 and earliest Pu1 faunas, stratigraphically (see Figure 12) and by faunal composition, is too weak a biostratigraphic basis from which to propose a new time unit on the scale of a NALMA.

The abandonment of the "Bugcreekian Age" does not mean that the Pu0 interval zone is not a distinct biochronologic unit; it is just a question of scale. The successive appearances of a number of new taxa (Catopsalis, Stygimys, Procerberus, Protungulatum, Mimatuta, Oxyprimus, Ragnarok ) in the Pu0 interval zone and the presence of these taxa in local faunas that contain Peradectes (Pu1 interval zone) support a biochronologic sequence of Lancian-Pu0-Pu1 (Archibald and Lofgren, 1990). The problem is one of discriminating between Pu0 and Pu1 faunas, because all of these newly appearing genera persist into the Pu1 interval zone (see Table 9) and Peradectes can be rare (see Table 10).

Figure 12
McGuire Creek biochronologic zonation based on biochronologic units proposed in Archibald et al., 1987, and
Archibald and Lofgren, 1990. Channel fills or local faunas as in Figure 5. Note: All three biochronologic units occur
within the same stratigraphic interval.

In addition to Peradectes , taxa whose first appearance are used to characterize the early Pu1 interval zone are Ragnarok engdahli, Catopsalis alexanderi, Acheronodon garbani , and Mimatuta minuial . However, A. garbani and R. engdahli are rare, which limits their use in regional correlations. The sample of A. garbani consists of the holotype, and R. engdahli is known only from the eight specimens described by Archibald (1982) from the Hell's Hollow Local Fauna (but see Rigby, 1989: figure 5).

Catopsalis alexanderi and Mimatuta minuial have a larger geographic distribution and sample size and therefore have greater utility for regional correlations. C. alexanderi is known from the Alexander locality, Denver Formation, Colorado, and Mantua Lentil, Polecat Bench Formation (Fort Union Formation), Wyoming, as well as the Hell Creek (this study) and Tullock formations, Montana (Middleton, 1982). M. minuial has a similar distribution, except that it is not known from the Denver Formation.

Recent work by Rigby (1989) indicates the occurrence of M. minuial, R. engdahli , and cf. Peradectes at the Doverzee locality (1 km north of the McGuire Creek study area). Doverzee and the local faunas from other localities under study by Rigby are probably referable to the Pu1 interval zone (FR, WT, DZ, VB, V1; Rigby, 1989: figure 5).

Informal Biochrons, bk1-bk2-bk3, of the Pu0 Interval Zone

Archibald and Lofgren (1990) subdivided the Pu0 Protungulatum/Peradectes interval zone into three informal biochrons defined by first appearances of the "condylarths" Protungulatum (bk1), Mimatuta (bk2), and Oxyprimus (bk3). The type localities for the three informal biochrons are the original Bug Creek sites: Bug Creek Anthills (bk1), Bug Creek West (bk2), and Harbicht Hill (bk3). The three biochrons and correlations based on them were tentative because of the lack of stratigraphic control on the three local faunas, and the differences in faunal composition between the localities could be a factor of either inadequate sample size or lack of recent taxonomic reviews. Therefore, Archibald and Lofgren (1990) chose to propose the biochrons informally until additional faunal data from these sites is available. Analyses of these faunas is presently underway by J. K. Rigby, Jr., and it is hoped that the results of this work will be available soon.

With the uncertainties surrounding the type localities of the bk1-bk2-bk3 biochrons, it might be better to abandon hope of recognizing different faunal levels within the Pu0 interval zone. However, the Bug Creek Anthills Local Fauna is especially significant in this regard because it is older than other Pu0 sites in the upper Hell Creek Formation. Thousands of mammal teeth and jaws have been collected from Bug Creek Anthills and all systematists who have studied these mammalian samples agree that the larger-sized "condylarths" Ragnarok spp. or Protungulatum gorgun present at other Pu0 sites in the upper Hell Creek Formation are not present at Bug Creek Anthills (Tables 4,7,9). Given the huge sample size and similarity in facies between Bug Creek Anthills and other Pu0 sites in the upper Hell Creek Formation, the lack of Ragnarok spp. and P. gorgun at Bug Creek Anthills indicates that this local fauna is older.

However, analysis of faunal differences between other Pu0 sites is complicated by the stratigraphic, taxonomic, and sampling uncertainties discussed below.

Lack of Stratigraphic Control

Stratigraphic relationships among the type localities for biochrons bk1 (Bug Creek Anthills), bk2 (Bug Creek West), and bk3 (Harbicht Hill) are uncertain. This is clearly the case with the Bug Creek Anthills locality, whose stratigraphic position has been debated at length (Smit and Van der Kaars, 1984; Fastovsky and Dott, 1986; Smit et al., 1987).

In spite of this, Sloan (1987:174) refers to the three original Bug Creek faunas as occurring within channel fills that are in superposition. In actuality, the Bug Creek West and Bug Creek Anthills local faunas occur within channel fills 2 km apart on opposite sides of the Bug Creek drainage and superposition of these channel fillings is not demonstrable (Fastovsky and Dott, 1986; Smit et al., 1987: figure 4; Rigby et al., 1987: figure 3). Also, the Harbicht Hill site is located over 20 km north of Bug Creek West and Bug Creek Anthills, and it is impossible to physically trace individual channel fills over 20 km in complex fluvial deposits such as those that comprise the upper Hell Creek Formation (Fastovsky and Dott, 1986; Fastovsky, 1987; this report).

As with the type localities, there is a general lack of superpositional or cross-cutting relationships among sites referable to the Pu0 interval zone. However, a series of localities in the Bug Creek drainage provides support for a bk1-bk2-bk3 succession. Tentative channel-fill relationships (Rigby et al., 1987) and preliminary faunal data (Rigby, 1989) indicate that the Big Bugger Channel yields Oxyprimus (bk3) and overlies the Bug Creek Anthills Channel, which yields only a single "condylarth," Protungulatum (bk1) (but see Luo, 1989, 1991). Also, the By George Channel yields Oxyprimus (bk3) and cuts the Scmenge Point Channel, which contains Protungulatum and Mimatuta but apparently not Oxyprimus (sensu bk2) (Rigby et al., 1987: figure 3; Rigby, 1989: figure 5). Therefore, channel fills with bk3 faunas overlie or cut channel fills with bk1 or bk2 faunas. The preliminary reports by Rigby et al. (1987) and Rigby (1989) are the only available biostratigraphic data which support a bk1-bk2-bk3 succession of informal biochrons.

Sampling Uncertainties

Biochronologic zonations based on small faunal differences are particularly susceptible to sampling uncertainties. Successive first appearances of the "condylarths" Protungulatum, Mimatuta , and Oxyprimus define the respective biochrons bk1-bk2-bk3 of the Pu0 interval zone. McGuire Creek faunal data (mainly Pu1 sites) indicate that Protungulatum, Mimatuta , and Oxyprimus are not rare and are represented in roughly approximate numbers of specimens (Table 10). It is assumed that collection of a large sample would yield these "condylarth" genera if they were extant in the immediate area at the time the deposit was formed. The sample size from the Bug Creek Anthills locality is again critical in this regard, because the absence of certain taxa which are

known to occur in other local faunas in the upper Hell Creek Formation strongly suggests that Bug Creek Anthills is older than other Pu0 sites.

However, the problem of insufficient sample size has been overlooked in previous biostratigraphic-biochronologic studies on the Bug Creek faunas. For example, a survey of the UCMP collection of mammals (n=300) from Harbicht Hill indicates that the marsupials Alphadon (UCMP 137313) and Glasbius (UCMP 137314) are present, as is the eutherian Gypsonictops (UCMP 137304) (Table 4). These taxa were thought to be extinct by the time the Harbicht Hill channel was filled (Sloan and Van Valen, 1965; Sloan, 1976). Similarly, the very small UCMP sample of mammals (n=10) from Bug Creek West yields an upper molar of Didelphodon vorax (Appendix 1). This is the first reported occurrence of Didelphodon from Bug Creek West. According to previous work, "Didelphodon vorax  . . . is quite unknown from any of the Paleocene localities in either Garfield or McCone counties" (sensu Bug Creek West and Harbicht Hill; Sloan and Rigby, 1986:1174). The former lack of records probably reflects small sample size because D. vorax and Glasbius are easily identified.

Last appearances of Lancian taxa such as Didelphodon, Pediomys, Glasbius, Alphadon , and others, and first appearances of certain "condylarths" from the Bug Creek West and Harbicht Hill sites have been accorded temporal significance (Sloan and Van Valen, 1965; Sloan, 1976; Van Valen and Sloan, 1977; Archibald, 1981, 1982, 1986, 1987c). The sizes of the samples from which these interpretations were made are not clear, although Sloan and Van Valen (1965) indicate that the minimum number of individuals (MNI) in their samples was 108 at Bug Creek West and 54 at Harbicht Hill. These samples are not large enough to include many of the rarer taxa. The sampling uncertainty is compounded by the fact that Bug Creek West and Harbicht Hill are channel-fill local faunas in which reworked fossils can occur. For example, samples from McGuire Creek indicate that virtually all Lancian taxa occur in Pu1 faunas (Table 4). However, many of these taxa are not found at Bug Creek West and Harbicht Hill which are Pu0 in age and presumably older than Pu1 sites. Is the presence of these taxa in McGuire Creek sites the result of reworking, or were Bug Creek West and Harbicht Hill not sufficiently sampled?

Based on the new faunal data presented above, the answer to the second question is yes. The reworking possibility was investigated earlier (see chapter above on Reworking of Fossils). In either case, it is apparent that the last appearance of Lancian taxa has limited value for biochronologic zonation. Also, the model of stepwise extinction of Lancian mammals (Van Valen and Sloan, 1977; Archibald, 1986, 1987c) within the Pu0 interval zone (Bug Creek Anthills-bk1, Bug Creek West-bk2, Harbicht Hill-bk3) is suspect because of their occurrence within the Pu1 interval zone (Tables 8, 9). Accordingly, Archibald and Lofgren (1990) abandoned last appearances of Lancian taxa as a basis for proposing or recognizing successive biochrons within the Pu0 interval zone.

Based on the examples discussed above, the available faunal lists of local faunas from the type localities of bk2 (Bug Creek West) and bk3 (Harbicht Hill) are far from complete, and what additional taxa these sites might eventually yield is unknown. The insufficient database from these localities raises doubt concerning the separability of the bk2 and bk3 biochrons.

Taxonomic Uncertainties

In 1965, announcement of the discovery and a brief description of mammalian components of the original Bug Creek faunas (Bug Creek Anthills, Bug Creek West, and Harbicht Hill) were published (Sloan and Van Valen, 1965). Other than "condylarths" from Bug Creek Anthills (Archibald, 1982; Luo, 1989, 1991), detailed systematic treatments of the mammalian faunas from the Bug Creek sites even now have yet to appear. Partly because of this lack of recent taxonomic review, proposal of a series of biochrons based on the mammal faunas from these localities was tentative (Archibald and Lofgren, 1990). Formal biochronologic zonations should be based on faunas that are well sampled, thoroughly described, and systematically analyzed. Therefore, mammals from the original Bug Creek sites require detailed description before informal biochrons (bk1-bk2-bk3) of the Pu0 interval zone can be either formally proposed or abandoned. Additional faunal data outlined below, based on UCMP collections support this contention.

Bug Creek Anthills: A survey of the multituberculates indicates that Cimexomys gratus (formerly C. hausoi ) is represented by a few isolated teeth (see Appendix 1). Previously, only the smaller species of Cimexomys, C. minor , was thought to occur at Bug Creek Anthills (Sloan and Van Valen, 1965; Archibald, 1982).

Luo (1989, 1991) argued that Mimatuta morgoth and Oxyprimus erikseni are both present at Bug Creek Anthills, though rare. If his views are accepted, the bk2 and bk3 biochrons will require redefinition. In that case, the successive biochrons should be redefined as the Protungulatum donnae/Protungulatum gorgun bk1, the Protungulatum gorgun/Ragnarok bk2, and the Ragnarok/Peradectes bk3 (pending complete descriptions of the Bug Creek West and Harbicht Hill local faunas).

Bug Creek West: The UCMP sample of mammals is small (n=10) and adds no new data beyond the occurrence of Didelphodon vorax discussed above.

Harbicht Hill: The UCMP mammal sample (n=300) yields a few specimens of Cimexomys gratus and a single specimen of Catopsalis which is referable to C. alexanderi (Appendix 1). In the original description of the Harbicht Hill fauna, Sloan and Van Valen (1965) list only the smaller species of Catopsalis, C. joyneri . Either both species are present at Harbicht Hill or specimens identified as C. joyneri in 1965 are now referable to C. alexanderi , a taxon erected later by Middleton (1982). Analysis of samples of Ragnarok nordicum from Mantua Lentil and R. harbichti from Harbicht Hill indicates synonymy of these species, with R. nordicum having priority (Appendix 1).

In summary, the mammalian faunas from the type localities of the bk1-bk2-bk3 biochrons of the Pu0 interval zone are insufficiently described and sampled. Results of ongoing studies by J.K. Rigby, Jr., are required before these biochrons can be formally adopted or abandoned.

Cretaceous-Tertiary Correlations

The Cretaceous-Tertiary Boundary

Correlation of the Pu0 and Pu1 interval zones with the K-T boundary must be preceded by a means to recognize Cretaceous and Tertiary sediments within both marine and nonmarine sections that span the transition. The K-T boundary is usually placed between the Maastrichtian and Danian stages (but see Voight, 1981). However, the upper limit of the type section of the Maastrichtian Stage (ENCI Quarry, St. Pietersburg, South Holland) is an erosional surface (Felder et al., 1980), while the lower limit of the Danian Stage (Stevns Klint and Faxse, Denmark) disconformably overlies chalks correlated with the Maastrichtian (Berggren, 1964). Subsequently, several sections in Europe were recognized as lacking a major hiatus between the Maastrichtian and Danian stages or as being "relatively complete" (see Dingus, 1983: part 1, for discussion). Recently, the K-T section at El Kef, Tunisia, was proposed as the K-T boundary stratotype by the International Geological Congress in 1989. The marine K-T boundary at El Kef is defined by the first appearances of planktonic forminifera and calcareous nannoplankton (Keller, 1989; with recent modification of FADs and LADs of planktic foraminifera by MacLeod and Keller, 1991a, 1991b).

Biostratigraphic correlation between nonmarine and marine zonations is difficult because in few instances are fossiliferous marine and nonmarine interdigitations preserved and well exposed. This is the case with latest Cretaceous-earliest Tertiary marine (Fox Hills and Cannonball formations) and nonmarine (Hell Creek and Tullock-Ludlow formations) stratigraphic units of the western interior in the northern United States (Figure 2). The Hell Creek Formation is underlain by the Fox Hills Formation, whose mollusk zonation (Cobban, 1958) has been correlated with the European mollusk zonation of the Maastrichtian Stage (Jeletsky, 1960, 1962). The Tullock Formation is laterally equivalent to part of the Ludlow Formation, which overlies the Hell Creek Formation in the Dakotas (Carlson and Anderson, 1966; Moore, 1976; Belt et al., 1984). The Ludlow Formation interfingers with the Cannonball Formation (Jeletsky, 1962), which contains Danian planktonic foraminifera (Fox and Olsson, 1969). Therefore, the marine K-T boundary would fall somewhere within the upper Hell Creek or lower Tullock formations in eastern Montana.

A more precise biostratigraphic determination of the marine K-T boundary in this nonmarine sequence is possible using paleomagnetic and radiometric correlation methods. The marine K-T boundary at El Kef occurs within paleomagnetic anomaly C29R (Keller, 1989). The uppermost Hell Creek and lower Tullock formations in Garfield and McCone counties were deposited in a period of reversed polarity that is correlated with anomaly C29R (Archibald et al., 1982). In eastern Montana, this stratigraphic interval includes the last records of dinosaurs and most genera of Lancian mammals and the first occurrence of Pu0 and Pu1 mammal assemblages (Archibald, 1982; Archibald and Lofgren, 1990). Chron 29R is usually considered to span approximately 500,000 years (Harland et al., 1982), which makes marine-nonmarine correlation still relatively imprecise. Radiometric dating techniques provide little additional precision because when applied to latest Cretaceous samples standard errors can exceed 500,000 years.

Because of difficulties in identifying the stratigraphic level in nonmarine sequences that temporally corresponds to the marine K-T boundary, a number of operational definitions have been employed to identify the K-T boundary in terrestrial sediments in eastern Montana. These are: (1) the base of the first coal zone (the "Z" coal) above the highest occurrence of dinosaur remains (Brown, 1952); (2) the stratigraphically highest occurrence of dinosaur remains (Archibald, 1982); (3) palynofloral extinction or disappearance (Tschudy, 1970); and (4) iridium enrichment (Smit and Van der Kaars, 1984; Smit et al., 1987). These criteria for boundary recognition are reviewed below.

"Z" Coal

This simple definition of using the first coal ("Z" coal) above the highest dinosaur was admittedly crude, but was proposed as a practical solution to end the prolonged debate during the first half of the twentieth century on where to place the K-T boundary in terrestrial sections in the western interior of North America (Brown, 1952). Subsequently, the "Z" coal was employed to approximate the K-T boundary in eastern Montana (Sloan and Van Valen, 1965; Van Valen and Sloan, 1977). However, the use of lithostratigraphic criteria to define a chronostratigraphic datum is both improper usage (Archibald, 1982; Fastovsky, 1987) and unrealistic. In a dynamic fluvial system, such as the one which deposited sediments spanning the K-T transition in eastern Montana, coal-swamp facies are not regionally persistent and coals are demonstrably discontinuous (Sholes and Cole, 1981; Archibald, 1982; Fastovsky, 1987).

This can be demonstrated in the McGuire Creek study area, where the TL and MCZ lignites are mappable for many kilometers but are not regional in extent. For example, the MCZ thins rapidly when traced to the northwest margin of the study area and grades laterally into black organic-rich mudstone (sections S, NN, OO, and HH, Plate 4), and correlations between sections OO and HH become highly questionable. Also, at Section OO (Plate 4), the "Z" coal of the Bug Creek drainage and the MCZ of the Black Spring Coulee drainage are apparently two different units (or part of a series of thin lignites that comprise the upper "Z" coal complex), and the "Z" coal is truncated by channeling. Therefore, both the concept of a regionally extensive "Z" coal bed and the use of the "Z" coal to define the K-T boundary should be discontinued.

Highest Occurrence of Dinosaur Remains

This means of boundary recognition assumes that the passing of the last dinosaur signaled the close of the Cretaceous (Brown, 1952). Whether the timing of dinosaur extinction and the appearance of microfossils used to define the marine K-T boundary are synchronous is unknown. Also, if this means of boundary recognition were to be employed, then all local faunas containing unreworked dinosaur remains would be Cretaceous in age. However, the possibility that dinosaur remains in Pu0 and Pu1 assemblages from channel facies are reworked (and the difficulty in identifying reworked fossils) suggest that the highest occurrence of dinosaur remains may actually significantly postdate dinosaur extinction. Conversely, in any local stratigraphic section composed of fine-grained deposits of floodplain origin, the highest preserved occurrence of dinosaur remains may significantly predate the actual time that dinosaurs succumbed to extinction. Finally, to avoid circular reasoning, it is desirable to determine the timing of dinosaur extinction using independent criteria. Therefore, the use of the highest occurrence of dinosaur remains is neither demonstrably precise nor logically appropriate to employ for K-T correlations.

Palynology

It has long been known that a palynological change could be recognized in local stratigraphic sequences in the western interior of North America that was roughly coincident with R. Brown's formula (1952) for the terrestrial K-T boundary (i.e., first coal above highest dinosaur) (Norton and Hall, 1969; Oltz, 1969; Leffingwell, 1970; Tschudy, 1970). More recently, discovery of an iridium enrichment at the marine K-T boundary in many sections worldwide (Alvarez et al., 1980; Orth et al., 1981, 1982; Alvarez et al., 1982, 1984; many others) has provided a means to globally correlate boundary events if the iridium enrichment represents the product of a single bolide impact (or another brief cataclysmic event) on earth. Recent K-T palynological research indicates that extinction of pollen species and an iridium enrichment are stratigraphically linked in many local sections throughout the western interior, from Saskatchewan and Alberta (Nichols et al., 1986; Jerzykiewicz and Sweet, 1986; Lerbekmo et al., 1987), Montana (Hotton, 1984, 1988; Bohor et al., 1984; Smit and van der Kaars, 1984; Smit et al., 1987), Wyoming (Bohor et al., 1987b), and Colorado-New Mexico (Orth et al., 1981; Tschudy et al., 1984). However, many other sections that are known to span the K-T transition in the western interior have been sampled for iridium but do not yield anomalous concentrations at the stratigraphic level that corresponds to the palynological K-T boundary.

The palynological K-T boundary in eastern Montana is recognized by the disappearance of Cretaceous indicator species, not by new appearances in the earliest Tertiary (Hotton, 1988). Absence of Cretaceous species can be employed to differentiate Cretaceous from Tertiary strata in local stratigraphic sections in eastern Montana (Sloan et al., 1986; Rigby et al., 1987; Smit et al., 1987; Hotton, 1988). Therefore, extinction of pollen species is an effective means to differentiate Cretaceous and Tertiary sediments, and was employed at McGuire Creek to determine the age of channel and floodplain facies yielding vertebrates.

Iridium

Iridium enrichment (or concentrations of iridium above background levels) may provide the ideal criterion for correlating marine and nonmarine K-T boundary sections worldwide (Berry, 1984). High concentrations of iridium were first reported by Alvarez et al. (1980) at the marine K-T boundary (as recognized then) in Italy and Denmark, and later elsewhere worldwide (Alvarez et al., 1982, 1984). Alvarez et al. (1980) proposed that the iridium was too concentrated to have been terrestrially derived and therefore must be the record of a bolide impact. The effects of this impact on latest Cretaceous organisms were interpreted to be the causal factor in terminal Cretaceous extinctions, in both marine and terrestrial realms.

According to an impact scenario, extraterrestrially derived iridium would have been dispersed into the air after impact, and eventually incorporated into sedimentary basins within 100 years or less. Iridium-enriched sediment at K-T boundary sections worldwide would be the record of this virtually instantaneous event in earth history, and would thereby provide a basis for worldwide correlation (Berry, 1984). However, others argued that K-T iridium enrichment was deposited during a period of intense volcanism lasting 10,000 to 100,000 years (Officer and Drake, 1985; Officer et al., 1987; Crocket et al., 1988). Reports of high concentrations of iridium during recent volcanic emissions from Kilauea Volcano, Hawaii (Zoller et al., 1983), support the possibility that K-T iridium enrichment may be terrestrially derived. Also, the concentrating effects of micro-organismal activity may have had a role in the formation of iridium anomalies (Dyer et al., 1989).

Shocked quartz (Bohor et al., 1984; 1987a) and microspherules (Smit and Klaver, 1981; Montanari et al., 1983; Montanari, 1986) are often associated with iridium anomalies and may represent impact-derived products. However, shocked quartz may originate from intense volcanism (Carter et al., 1986; but see Alexopoulos et al., 1988), and microspherules may have a volcanic origin as well (Naslund et al., 1986). Also, microspherules from three marine K-T sites have been interpreted to represent diagenetic infillings of organic spheres, not impact or volcanism products (Hansen et al., 1986).

The source(s) of the iridium, shocked quartz, and microspherules near the K-T boundary no doubt will be debated for years. In this study, iridium enrichment is used as a correlation tool, whatever its source. Iridium enrichment is present at the proposed K-T boundary stratotype at El Kef, Tunisia (Kuslys and Krahenbulh, 1983). It also has been reported from many of the terrestrial stratigraphic sections known to span the K-T transition in the western interior of North America (Orth et al., 1981, 1982; Smit and Van der Kaars, 1984; Nichols et al., 1986; Lerbekmo and St. Louis, 1986; Lerbekmo et al., 1987; Bohor et al., 1987b). Other sections spanning the K-T transition, most notably those in eastern Garfield and western McCone counties in eastern Montana (where Bug Creek local faunas are located), have been sampled for iridium, but have not yielded a high level of iridium enrichment. In any case, the critical premise is that the marine and nonmarine iridium enrichments are records of the same discrete event, which allows precise correlation.

Several challenges to worldwide iridium anomaly correlations have been proposed. For example, one hypothesis suggests that K-T extinctions were caused by multiple impacts, perhaps comet showers (Hut et al., 1987). Also, the marine (worldwide distribution) and terrestrial (North American distribution) K-T iridium anomalies may have been deposited by different mechanisms (Schmitz, 1988). If the marine and nonmarine anomalies were formed by different events, or if they were caused by multiple impacts, temporal correlation becomes suspect. However, iridium enrichment could still be an effective correlation tool for terrestrial K-T sections within North America.

Unfortunately, sections spanning the K-T transition in McCone County have been sampled for iridium enrichment without success (Smit et al., 1987). The nearest K-T iridium enrichment is 50 km to the west, at the Lerbekmo site in Garfield County. With the absence of records of iridium enrichment in McCone County, direct correlation of K-T sections to the marine K-T boundary cannot be accomplished. Pollen extinction then becomes the most precise correlation tool available for determining the local K-T boundary.

It has been implied that an iridium enrichment is present in the lower "Z" coal at Russell Basin, McCone County (Sloan et al., 1986; Rigby, 1989), a few kilometers north of the McGuire Creek study area. However, this iridium concentration is 40-80ppt, or only 2-3 times background (Fastovsky and Dott, 1986), and is similar to normal background levels such as those reported from Saskatchewan (5-60ppt, Nichols et al., 1986), North Dakota (25ppt, Johnson et al., 1989), and the Raton Basin in New Mexico (10-30ppt, Orth et al., 1981). A concentration of 40-80ppt is not anomalous, and reference to an iridium anomaly in McCone County should be discontinued.

Age of the Pu0 and Pu1 Interval Zones

Pollen extinction is employed to differentiate Cretaceous and Tertiary sediment at McGuire Creek in the absence of iridium enrichment. Criteria for palynological differentiation of Cretaceous and Tertiary strata were developed from independently dated sequences of overbank deposits from both Garfield and McCone counties (Hotton, 1988). The data presented in an earlier chapter on palynological correlations show that all Pu0 and Pu1 sites at McGuire Creek yield Tertiary palynofloras (Table 3). Therefore, at McGuire Creek, the Puercan-Lancian and the palynological K-T boundaries coincide at the crude level of precision that is available to us: first appearance of Protungulatum = Pu0 interval zone; first record of loss of Cretaceous indicator palynomorphs = Tertiary.

Because all Pu0 and Pu1 vertebrate assemblages were collected from channel facies, and the palynological K-T boundary criteria were developed in overbank sequences, it was important to test whether a facies bias might exist. The composition of the flora contributing to the pollen rain along channel margins might differ significantly from that in overbank facies. The absence of Cretaceous indicator species in a Cretaceous channel might be due to these species' preference for overbank environmental settings. Therefore, a sample from the channel fill containing Matt's Dino Quarry (Section AA, Plate 3), which yielded associated hadrosaur skeletal remains,

was analyzed for pollen. It yielded a Cretaceous palynoflora. This one sample suggests that palynological samples from channel facies are not biased by facies effects.

Erosion of Cretaceous strata and subsequent reworking of Cretaceous fossils into Paleocene channels occurred at McGuire Creek (Lofgren et al., 1990). A related issue is whether Cretaceous indicator palynomorphs were also reworked into these same channels. Paleocene channels yield a few occurrences of Cretaceous indicator species. These are usually less than 2% of total number counted, while the sample from Matt's Dino Quarry yielded 25% Cretaceous palynomorphs. For comparison, Cretaceous floodplain deposits yield 10-40% Cretaceous palynomorphs (pers. comm., C. Hotton, 1988). Therefore, it appears that if significant amounts of Cretaceous palynomorphs were eroded and redeposited, their numbers were diluted by large quantities of Tertiary pollen rain.

Reworking would explain the original conflict (now resolved) in age interpretation between the USGS palynologists and Carol Hotton concerning the Brown-Grey Channel. Thin siltstones in the Brown-Grey Channel were sampled and portions of a single sample were sent for analysis to two palynologists with experience in the K-T boundary problem. Rock sample 88DLL7-14-30 (original field no. 85H7-26-5, collected by J.H. Hutchison, 1985) was analyzed by palynologists at the USGS and subsequently assigned a Cretaceous age (pers. comm., Nichols and Wingate, March, 1988, USGS PL No. D6906). Sample 88DLL7-14-30 was also analyzed by Dr. Hotton, who concluded that it contains a Paleocene palynoflora (pers. comm., C. Hotton, December 1988).

This apparent conflict in age interpretation of 87DLL7-14-30 (=85H7-26-5) is a reflection of how the palynological boundary is defined. Earliest Paleocene palynofloras are depauperate and are recognized primarily on the absence or rarity of Cretaceous indicator species, not by new appearances in the earliest Paleocene (Hotton, 1988). Both analyses of the same rock sample indicate the presence of Cretaceous indicator species, but they are low in relative abundance (2.5-3.5% of the total sample), indicating a Tertiary age (pers. comm., C. Hotton, 1989). The USGS specialists made their Cretaceous interpretation based on the presence of Cretaceous indicator species. Therefore, the palynological boundary was recognized on different criteria by different palynologists: (1) the presence of Cretaceous indicator species no matter what their relative abundance; and (2) the relative abundance of Cretaceous indicator species.

In an attempt to resolve this disparity in approach, Dr. Nichols analyzed more samples of 88DLL7-14-30 and also 88DLL7-14-13, both from the Brown-Grey Channel. He concludes that because of the rare occurrence of characteristic Cretaceous palynomorphs in the samples, they still appear to be latest Cretaceous in age. However, the samples could be earliest Paleocene containing a few reworked Cretaceous palynomorphs. Identification of a species of pollen in the assemblage that may be restricted to the lowermost Paleocene in Wyoming and the Dakotas suggests this may be correct and earliest Paleocene is probably the most reasonable interpretation (pers. comm., D. Nichols, 1990). Therefore, Hotton and Nichols now agree that it is probable that the Brown-Grey Channel yields a Tertiary palynoflora with a low abundance of

reworked Cretaceous palynomorphs (pers. comm., C. Hotton, 1989, and D. Nichols, 1990). This is similar to what may be occurring with the vertebrate assemblage from the same channel (i.e., reworking of dinosaurs and Lancian mammals).

Elsewhere in McCone County, Montana, where vertebrate sites yielding Pu0 and Pu1 have been sampled for pollen, most Pu0 and Pu1 sites yield Paleocene palynofloras (Sloan et al., 1986; Rigby et al., 1987; Newmann, 1988). However, Cretaceous palynofloras are claimed to be associated with Pu0 assemblages from channel fills at Bug Creek Anthills (Newmann, 1988) and Grenda's Horn/Doc's Folly (Rigby, 1989). If this is true, then the Pu0 interval zone would span the K-T boundary. However, detailed stratigraphic or palynological data were not given in these reports. The position within the channel fill of the Cretaceous pollen or from what lithology the pollen was collected (mud clast, or a siltstone lens, etc.), are not stated, nor are abundances/ numbers of Cretaceous indicator species. Therefore, evaluation of claims of Cretaceous Pu0 assemblages (Newmann, 1988; Rigby, 1989) are not possible from the data given.

Similarly, two reportedly Pu0 assemblages (Frenchman 1 and Long Fall) from channel facies in Saskatchewan which yield dinosaurs and Lancian mammals are claimed to be Cretaceous in age, on the basis of relative stratigraphic position (Frenchman 1) or the presence or association with "unreworked" dinosaur and Lancian mammal remains (Frenchman 1 and Long Fall) (Johnston and Fox, 1984; Fox, 1987, 1989, 1990). However, Paleocene channels may contain reworked Cretaceous fossils (Lofgren et al., 1990). Also, palynological analysis of sediment within either channel fill has not been reported. Therefore, Pu0 assemblages from Frenchman 1 and Long Fall, as well as those from Bug Creek Anthills and Grenda's Horn/Doc's Folly, may be Cretaceous, but this remains to be demonstrated. If these Pu0 assemblages are Cretaceous, the Pu0 interval zone would span the K-T transition, and the Lancian-Puercan and Cretaceous-Tertiary boundaries would be diachronous.

Faunal Turnover during the K-T Transition in Montana

Bug Creek Assemblages (or Pu0 and Pu1 assemblages containing dinosaurs and Lancian mammals) are intermediate in composition when compared to unquestionably latest Cretaceous and earliest Paleocene vertebrate assemblages (Table 1). Estimates of vertebrate survival rates during the K-T transition in the western interior of North America can be significantly affected by two factors concerning these assemblages: (1) whether the Pu0 interval zone spans the K-T transition (or at least one Bug Creek assemblage is Cretaceous); and (2) whether the presence of dinosaurs and Lancian mammals in Pu0 and Pu1 assemblages is entirely the result of reworking.

The survival rate of non-dinosaurian lower vertebrates during the K-T transition is relatively high even if Bug Creek occurrences are omitted from the analysis. In eastern Montana, the survival rate of turtles spanning the Hell Creek-Tullock formations (and the K-T transition) is 84% at the generic level (Hutchison and Archibald, 1986). Similarly, the survival rate of species of non-dinosaurian lower vertebrates is at least 55% across the same stratigraphic interval (Bryant, 1989). If genera present elsewhere in the western interior during the Tertiary, but not yet found in the Tullock Formation, are included in the analysis, the survival rate of species could exceed 70% (Bryant, 1989).

In contrast, the survival rates of dinosaurian and mammalian vertebrates are significantly affected by the inclusion of Bug Creek occurrences in the analysis. Dinosaurs are represented by approximately 14-20 species in the Late Cretaceous Lancian "Age" (Sullivan, 1987; Archibald and Bryant, 1990). If dinosaurs did not persist into the Paleocene, then their survival rate is 0%. However, if bone fragments and isolated dinosaurian teeth present in channels that yield Paleocene palynofloras are accepted as evidence that dinosaurs survived into the Paleocene (an interpretation not favored here), the survival rate of dinosaur species would exceed 50% (13 of 20 species).

Mammalian survival rates are more complex because of the likelihood of reworking and the age uncertainty of the Bug Creek Anthills site. On the basis of data from eastern Montana (Table 11), if the Pu0 and Pu1 interval zones are entirely Paleocene and all the Cretaceous mammal species are reworked (i.e. those in question from Bug Creek assemblages), then the mammalian survival rate is 7% (2/28 species) or 13% (2/16 genera). In contrast, if all Cretaceous mammal species present in Pu0-Pu1 assemblages actually persisted into the Paleocene (were not reworked), then the survival rate

rises to 57% (16/28 species) or 75% (12/16 genera). Finally, if the Bug Creek Anthills locality were actually Cretaceous and this faunal data were included in the analysis, the survival rate of mammals would climb to 65% (24/37 species) or 82% (18/22 genera).

Bug Creek assemblages are frequently cited as evidence that rates of vertebrate extinction/origination were gradual or stepwise during the K-T transition and are not compatible with catastrophic extinction scenarios (Van Valen and Sloan, 1977; Clemens and Archibald, 1980; Clemens et al., 1981; Archibald, 1981, 1982, 1984, 1987c; Clemens, 1982; Archibald and Clemens, 1984; Van Valen, 1984). However, estimates of vertebrate survival rates during the K-T transition in eastern Montana can be significantly affected by how the dinosaurian and mammalian components of Pu0 and Pu1 assemblages (or Bug Creek assemblages) are interpreted. If dinosaurs and Cretaceous mammals persisted into the Paleocene (not reworked) and/or Bug Creek Anthills is Cretaceous, the species survival rate of dinosaurs would exceed 50% and that of mammals 65%. If this is true, catastrophic mass-extinction scenarios involving terrestrial vertebrate species during the K-T transition (e.g., Smit and Van der Kaars, 1984; Smit et al., 1987) would be effectively falsified.

Data from McGuire Creek indicate that Bug Creek assemblages are Paleocene and contain reworked Lancian mammals and dinosaurs. If this interpretation is correct, the survival rate of mammal species is approximately 10% and of dinosaurs 0%, making catastrophic mass-extinction scenarios more plausible. However, such scenarios are still not compatible with the high specific survival rates (55-71%) of non-dinosaurian vertebrates (Bryant, 1989) or the entire vertebrate fauna (exceeding 50%) (Archibald and Bryant, 1990).

Conclusions

The "Bug Creek Problem" refers to unresolved issues concerning the vertebrate biostratigraphy of the uppermost Hell Creek and lower Tullock formations of eastern Montana. This stratigraphic interval spans the K-T transition and yields three kinds of vertebrate assemblages: Lancian (Late Cretaceous), Bug Creek, and Puercan (Early Paleocene). Bug Creek assemblages are compositionally intermediate between Lancian and Puercan faunas, in that they contain Lancian mammals and dinosaurs along with mammals whose descendants are characteristic of the Puercan; thus they may be either Cretaceous, Paleocene, or both in age. Detailed systematic and biostratigraphic treatments of Bug Creek assemblages in concert with geologic analyses of the stratigraphic interval in which they occur were needed because these assemblages are central to debates concerning patterns of faunal turnover of the terrestrial biota during the Cretaceous-Tertiary transition.

Uncertainties concerning Bug Creek assemblages can be divided into four main issues, which are: (1) Are Bug Creek assemblages temporally intermediate between Lancian (Late Cretaceous) and Puercan (Early Paleocene) assemblages, on the basis of biostratigraphic and/or biochronologic criteria? (2) Can Bug Creek assemblages themselves be placed in a temporal sequence that reflects faunal differences between individual local faunas? (3) Do Bug Creek assemblages represent the remains of vertebrates that lived contemporaneously, or were Cretaceous (Lancian) fossils eroded and redeposited with Paleocene (Puercan) vertebrate remains? (4) What are the geochronologic age(s) of Bug Creek assemblages?

Lithofacies and biostratigraphic analyses of sections spanning the K-T transition at McGuire Creek, McCone County, Montana, were used to address the "Bug Creek Problem." Lithofacies analysis of the upper Hell Creek and lower Tullock formations indicates that Bug Creek assemblages are exclusively associated with lag deposits of large channel facies that are deeply entrenched into floodplains deposits which yield in situ dinosaur remains. A locally traceable erosion surface was created by these channeling events. Because of the enormous amount of older sediment eroded during channel entrenchment, the potential is very high that Bug Creek assemblages contain reworked fossils.

An indication that the dinosaurian component of Bug Creek assemblages is reworked comes from the lateral tracing of channel facies into floodplains containing in situ dinosaur remains. In two instances where channel fills containing dinosaur remains

can be physically traced into floodplain deposits, dinosaurs are absent in contemporaneous floodplain deposits.

At McGuire Creek, palynology is useful for the correlation of isolated channel fills containing Bug Creek assemblages with floodplain deposits yielding or lacking dinosaur remains. Where data are available, channel fills containing Bug Creek assemblages all yield Paleocene palynofloras, and floodplain deposits that contain the highest in situ dinosaur remains yield Cretaceous palynofloras. Also, the lowest Paleocene floodplain deposits identified palynologically lack dinosaurs. Therefore, both palynological and geological correlations indicate that channel fills containing Bug Creek assemblages are temporally equivalent to floodplain deposits which overlie those containing in situ dinosaurs. Also, the erosion surface developed between floodplain facies containing in situ dinosaurs and channel fills yielding Bug Creek assemblages is the local demonstration of the palynological Cretaceous-Tertiary boundary.

Because channel facies that yield Bug Creek assemblages are deeply incised into Lancian strata, considerable reworking of dinosaur and Lancian mammal remains must have occurred. However, fossils that were reworked can be identified only in exceptional instances, such as Black Spring Coulee, where many large dinosaur bones were reworked into a Paleocene channel fill (Lofgren et al., 1990). Because of the great potential for reworking, fossils representing all suspect Bug Creek taxa (dinosaurs and most species of Lancian mammals) should be assumed to be reworked unless at least one of these criteria are met: (1) articulated skeletal remains occur within a channel fill; (2) specimens are recovered from Paleocene floodplain deposits; (3) a channel that did not scour Lancian or Cretaceous sediments yields abundant specimens of suspect taxa; (4) specimens are demonstrably reworked from Puercan or Paleocene strata.

Development of a local biostratigraphic zonation for the uppermost Hell Creek Formation that subdivides strata in accordance with the presence of Lancian, Bug Creek, and Puercan assemblages is difficult because they occur in channel facies within a complexly channeled stratigraphic interval, and only in rare instances can channels be temporally ordered based on superposition or crosscutting relationships.

On the basis of previously proposed biochronologic zonations (Archibald et al., 1987; Archibald and Lofgren, 1990), McGuire Creek vertebrate assemblages are referred to the Lancian NALMA and Pu0 and Pu1 interval zones of the Puercan NALMA. Although most Pu0 and Pu1 sites are not well constrained stratigraphically, the successive appearances of new taxa (Catopsalis, Stygimys, Procerberus, Protungulatum, Mimatuta, Oxyprimus, Ragnarok ) in the Pu0 interval zone and the presence of these taxa into local faunas that contain Peradectes (Pu1 interval zone), support a biochronologic sequence of Lancian-Pu0-Pu1 (Archibald and Lofgren, 1990).

The informal division of the Pu0 interval zones into three informal biochrons (bk1-bk2-bk3; Archibald and Lofgren, 1990) was tentative because the mammalian faunas from the type localities of these biochrons (Bug Creek Anthills-bk1, Bug Creek West-bk2, Harbicht Hill-bk3) either require additional sampling or more recent systematic treatments. Analyses of these local faunas are presently underway, and results of these studies are required before these biochrons can be formally adopted or abandoned.

Extinction of Cretaceous palynomorph species has been employed to recognize earliest Tertiary sediments in eastern Montana (Hotton, 1988). Therefore, palynological criteria were employed to differentiate Cretaceous and Paleocene channel fills at McGuire Creek. Where palynological data are available, all channel fills that yield Puercan (Pu0 and Pu1) assemblages at McGuire Creek are Paleocene. The Brown-Grey Channel was originally considered to contain a palynoflora that could be interpreted, from independent palynological analyses of a single rock sample, as either Cretaceous or Paleocene. The conflict stemmed from different definitions of the palynological boundary: (1) a very low relative abundance of Cretaceous indicator species (Paleocene); or (2) the presence of such species no matter how rare (Cretaceous). While consensus on a Paleocene age assignment was reached, the original conflict suggests that the palynological K-T boundary can be recognized by multiple criteria.

The age of Bug Creek assemblages (or Pu0-Pu1 assemblages with dinosaurs and certain Lancian mammals), and whether or not these assemblages contain reworked dinosaur and Lancian mammal remains, are factors that can significantly affect estimates of the survival rates of these vertebrates during the K-T transition in eastern Montana. Survival rates of dinosaurs and mammals can be interpreted to be compatible with catastrophic extinction scenarios if Bug Creek assemblages are exclusively Paleocene and contain reworked Lancian mammal and dinosaur remains because the survival rate of mammals would be approximately 7% (2/28 species), and of dinosaurs 0%. These interpretations are supported by analyses at McGuire Creek, where all Bug Creek assemblages are palynologically Paleocene and appear to contain Cretaceous vertebrates (those with suspect Paleocene records) because of reworking.

Appendix 1—
Mammalian Systematics

Brief descriptions of mammalian components of the original Bug Creek assemblages were given by Sloan and Van Valen (1965), Van Valen and Sloan (1965), and Van Valen (1978). Except for the few studies noted here, detailed systematic treatments have yet to appear. Lillegraven (1969) provided additional description of Procerberus . Archibald (1982) described some of the mammals found in the Bug Creek assemblages, but the focus of his systematic work was on mammals from Garfield County. Novacek and Clemens (1977) studied the UCMP sample of Mesodma from the Bug Creek Anthills locality and argued that only a single species was discernable (contra Sloan and Van Valen, 1965). Luo (1989, 1991) studied the "condylarth" component of the UCMP sample from Bug Creek Anthills and argued that three species are present rather than the single species, Protungulatum donnae , described by Sloan and Van Valen (1965). Lupton et al. (1980) briefly described two partial dentaries of Protungulatum gorgun from Chris' Bonebed, a Milwaukee Public Museum locality in the upper Hell Creek Formation near Harbicht Hill (Figure 1c).

The original Bug Creek faunas (Bug Creek Anthills, Bug Creek West, Harbicht Hill) formed the basis for the "Bugcreekian" NALMA (Sloan, 1987; Archibald, 1987b), but are now recognized as comprising the initial interval zone (Pu0) of the Puercan Age (Archibald and Lofgren, 1990).

Lancian mammals whose presence in Puercan faunas (Pu0 or Pu1 interval zones) in eastern Montana may be entirely caused by reworking are: Meniscoessus robustus, Essonodon browni, Cimolodon nitidus, Pediomys hatcheri, P. krejcii, P. florencae, P. elegans, Pediomys sp. indet., Alphadon "wilsoni," A. rhaister, A. marshi, Alphadon sp. indet., Glasbius twitchelli, Didelphodon vorax, Gypsonictops illuminatus , and Batodon tenuis . These questionable occurrences in Puercan faunas are denoted by an asterisk (*) in the listed distribution of each species. Cimexomys minor and Neoplagiaulax burgessi are Lancian mammals whose present in Puercan faunas apparently is not the result of reworking from Cretaceous sediments.

Lithostratigraphic units that yield both Puercan and Lancian mammalian local faunas are the Hell Creek Formation in McCone County, Montana, and the Frenchman Formation in Saskatchewan, Canada.

Class MAMMALIA Linnaeus, 1758
Subclass ALLOTHERIA Marsh, 1880
Order MULTITUBERCULATA Cope, 1884
Suborder PTILODONTOIDEA (Gregory and Simpson, 1926) Sloan and Van Valen, 1965
Family NEOPLAGIAULACIDAE Ameghino, 1890
Mesodma Jepsen, 1940
Mesodma sp.

Comments : A large number (n>2200) of isolated and associated teeth from fragmentary dentaries and maxillas referable to Mesodma were recovered from localities in the upper Hell Creek Formation at McGuire Creek sampled by screenwashing techniques. Distinction of species of Mesodma is a difficult taxonomic problem (Sloan and Van Valen, 1965; Novacek and Clemens, 1977; Archibald, 1982). Because of the large number of fossils and the difficulties inherent in species identification, a thorough study of the sample of Mesodma from McGuire Creek is beyond the scope of this study. Although it is uncertain if one or more species are present, all specimens are referred to Mesodma sp. at this time.

Family CIMOLODONTIDAE Marsh, 1889a
Cimolodon Marsh, 1889a
Cimolodon nitidus Marsh, 1889a

Cimolodon nitidus Marsh, 1889a, p. 84 (see Clemens, 1964, p. 56, for synonymies).

Holotype . YPM 11776, left /M1 (Marsh, 1889a, pl. II, figs. 5-8).

Type locality . Mammal locality no. 1 of Lull (1915), UCMP loc. V5003, Lance Formation, Wyoming.

Referred specimen . 1 M1/, UCMP 133406, from loc. V87072.

Locality . UCMP loc. V87072.

Distribution . St. Mary River Formation, Alberta (Edmontonian); Lance Formation, Wyoming, and Scollard Formation, Alberta (both Lancian); Frenchman Formation, Saskatchewan, and Hell Creek Formation, Montana (both Lancian and *Puercan); Ravenscrag Formation, Saskatchewan (*Puercan).

Revised diagnosis . Clemens, 1964, p. 56.

Description . The single tooth referable to Cimolodon nitidus is an M1/. The cusp formula (5:7:5) and size (L: 4.51, W: 2.57) of this specimen agree with Clemens' (1964) description of the species.

Suborder TAENIOLABIDOIDEA (Granger and Simpson, 1929)
Sloan and Van Valen, 1965
Family EUCOSMODONTIDAE (Jepsen, 1940)
Sloan and Van Valen, 1965
Subfamily EUCOSMODONTINAE Jepsen, 1940
Stygimys Sloan and Van Valen, 1965
Stygimys kuszmauli Sloan and Van Valen, 1965
Tables 12-13; Figures 13-19

Eucosmodon gratus Jepsen, 1930, p. 499-500.

Stygimys gratus Sloan and Van Valen, 1965, p. 224.

Holotype . UMVP 1478, left lower jaw fragment with I and alveoli for /P4 and /M1 (Sloan and Van Valen, 1965, part of fig. 4).

Type locality . Bug Creek Anthills, Hell Creek Formation, Montana.

Referred specimens . Three /P4's, 4 /M1's, 2 M1/'s, 3 /M2's, 7 M2/'s from loc. V87035. Six P4/'s, 4 /P4's, 8 /M1's, 9 M1/'s, 2 /M2's, 9 M2/'s from loc. V87037. Five P4/'s, 12 /M1's, 13 M1/'s, 6 /M2's, 11 M2/'s, from loc. V87038. Two P4/'s, 1 /P4, 1 /M1, 9 M1/'s, 4 /M2's, 4 M2/'s from loc. V87051. Three P4/'s, 5 /P4's, 5 /M1's, 4 M1/'s, 10 /M2's, 6 M2/'s from loc. V87071. Eight P4/'s, 10 /P4's, 19 /M1's, 18 M1/'s, 14 /M2's, 22 M2/'s from loc. V87072. Six P4/'s, 3 /P4's, 7 /M1's, 4 M1/'s, 9 /M2's, 14 M2/'s from loc. V87074. One P4/, 4 /P4's, 1 /M1, 5 M1/'s, 2 M2/'s from loc. V87077. Three P4/'s, 2 /P4's, 12 /M1's, 6 M1/'s, 8 /M2's, 6 M2/'s from loc. V87098. Two P4/'s, 8 /P4's, 14 /M1's, 10 M1/'s, 9 /M2's, 13 M2/'s from loc. V87151. Twenty five other McGuire Creek localities yielded 1-6 isolated teeth per site.

Localities . V86031, V87028, V87030, V87034, V87035, V87036, V87037, V87038, V87040, V87049, V87051, V87052, V87070, V87071, V87072, V87073, V87074, V87077, V87078, V87084, V87086, V87098, V87101, V87108, V87109, V87114, V87115, V87124, V87151, V87152, V87153, V88038, V88039, and V88041.

Distribution . Upper Hell Creek and lower Tullock formations, Montana, and Polecat Bench Formation (Fort Union Formation), Wyoming (all Puercan).

Discussion and Description . When first proposed, Stygimys included three species formerly referred to Eucosmodon (E. gratus, E. jepseni, E. teilhardi ) and a new species, Stygimys kuszmauli , which was designated as the genotype (Sloan and Van Valen, 1965). Stygimys kuszmauli and S. gratus were reported from the Bug Creek sequence of channel deposits (Bug Creek Anthills, Bug Creek West, Harbicht Hill) in the upper Hell Creek Formation of Montana (Sloan and Van Valen, 1965). Sloan and Van Valen (1965) noted the presence of S. gratus at Harbicht Hill, which was previously known only from Mantua Lentil, Polecat Bench Formation, Wyoming (Jepsen, 1930). The new, smaller species of the genus, S. kuszmauli , was listed as occurring at Bug Creek Anthills (type locality) and Bug Creek West (Sloan and Van Valen, 1965).

The original diagnosis of S. kuszmauli is short and only refers to size differences between it and S. gratus : "This is the smallest species of the genus. Length /P4, 4.6 +/- .3mm (standard deviation of sample); 11 serrations. The specimens of Stygimys from Harbicht Hill are larger than those from Bug Creek Anthills and can be referred to Stygimys gratus " (Sloan and Van Valen, 1965, p. 224). The /P4's of S. gratus from Mantua Lentil have 11 serrations and lengths of 4.8mm and 4.9mm (Jepsen, 1940). These lengths fall within one standard deviation of the mean length of /P4's of S. kuszmauli from Bug Creek Anthills (Table 12). Therefore, the original diagnosis of S. kuszmauli does not distinguish it from S. gratus .

To resolve this problem, the samples of Stygimys in the UCMP collections from Bug Creek Anthills (V65127 and V70201) and Harbicht Hill (V71203) were compared to the small (n=6) sample of Stygimys gratus from Mantua Lentil described by Jepsen (1930; 1940). This analysis indicates that the holotype of S. gratus , PU 13373 a dentary

Figure 13
Bivariate plot (length vs. width) of /P4's of  Stygimys  from Bug Creek Anthills (V65127 and
V70201), Harbicht Hill (V71203), Tedrow Quarry D (V87072), and Up-Up-the-Creek 2 and 3
(V87037-38) (all from the Hell Creek Formation, McCone County, Montana), Worm Coulee
1 (V74111; from the Tullock Formation, Garfield County, Montana), and Mantua Lentil (from
the Polecat Bench Formation, Wyoming). One of the two /P4's from Mantua Lentil
(PU 14496) is encased in sediment, and only length could be measured; its width was
estimated at 1.75 mm, based on the similar length of PU 14419, which has a width of 1.72 mm.

fragment with /M2, represents a different taxon than the remainder of its hypodigm. The holotype (PU 13373) is much smaller than the two other dentary fragments from Mantua Lentil referred to S. gratus . Also, although the /M2 from the holotype is heavily worn, its morphology differs significantly from /M2's of S. kuszmauli from Bug Creek Anthills. Comparison of the holotype of S. gratus with that of Cimexomys hausoi from the Tullock formation of Garfield County, Montana, indicates that these two specimens represent the same taxon. Therefore, PU 13373 becomes the namebearer for that species, now recognized as Cimexomys gratus (further discussion in section on C. gratus ).

Because the holotype of "Stygimys gratus " becomes the namebearer of Cimexomys gratus , the remainder of the Mantua Lentil sample of Stygimys is without species referral. Since the original diagnosis of Stygimys kuszmauli did not distinguish it from the Mantua Lentil sample of Stygimys , comparison of samples from the two type localities, Mantua

Figure 14
Bivariate plot (length vs. width) of /M1's of  Stygimys  from various localities (see Figure 13).

Lentil and Bug Creek Anthills, was required. The UCMP sample of Stygimys from Harbicht Hill was included in the analysis because when Stygimys was described initially, the Harbicht Hill sample was referred to "Stygimys gratus " (Sloan and Van Valen, 1965).

/P4: Both /P4's (PU 14496, PU 14419) from Mantua Lentil have 11 serrations. All /P4's from Harbicht Hill have 11 serrations. Three of 36 /P4's from Bug Creek Anthills have 10 serrations, the rest have 11. Based on size, /P4's from Mantua Lentil and Bug Creek Anthills are, on average, nearly identical (Table 12). A bivariate graph of length vs. width indicates that the two /P4's from Mantua Lentil fall within a cluster of specimens of S. kuszmauli from Bug Creek Anthills (Figure 13). The /P4 samples from Harbicht Hill and Bug Creek Anthills have similar ranges in size, but the mean /P4 length of specimens from Harbicht Hill is much larger than that from Bug Creek Anthills (Table 12). However, the difference in mean lengths of /P4's from Harbicht Hill (V71203) and Bug Creek Anthills (V65127 only) was tested statistically, and was not significant (T test: T=-1.23; df 16+ 2.12).

/M1: The single /M1 (PU 14417) from Mantua Lentil has a cusp formula of 6:5. Fifteen percent (7 of 45) of /M1's from Bug Creek Anthills have 6 cusps on the external row, the remainder have seven. All /M1's from Bug Creek Anthills have 5 cusps on the

Figure 15
Bivariate plot (length vs. width) of /M2's of  Stygimys  from various localities (see Figure 13).

internal row. All /M1's from Harbicht Hill (n=10) have 5 cusps on the internal row, and all but one have 7 cusps on the external row; the one exception has 6. Based on size, all three samples are indistinguishable (Table 12). Plotted graphically by length vs. width, /M1's from Harbicht Hill and Mantua Lentil compare closely with those from Bug Creek Anthills (Figure 14).

/M2: The Mantua Lentil sample yields an /M2 (PU 13373) but it is referable to Cimexomys (this report), not Stygimys (contra Jepsen, 1930; sensu Sloan and Van Valen, 1965). Harbicht Hill /M2's (n=10) have a cusp formula of 4:2, as do Bug Creek Anthills /M2's (n=46) except for three, which have 5 cusps on their external rows. Based on size, /M2's from Harbicht Hill are, on average, slightly larger than those from Bug Creek Anthills (Table 12). However, when the two samples are plotted graphically, the small differences in mean length and width are apparently insignificant (Figure 15).

P4/: The Mantua Lentil sample does not contain P4/'s. Harbicht Hill P4/'s (n=3) have a cusp formula of 2-3:7-8:1, and those from Bug Creek Anthills (n=38) 2-3:7-10:1. Based on size, the samples are nearly identical (Table 12). A bivariate length vs. width plot indicates that the Harbicht Hill sample falls completely within the sample from Bug Creek Anthills (Figure 16).

Figure 16
Bivariate plot (length vs. width) of P4/'s of  Stygimys  from various localities (see Figure 13).

M1/: The two M1/'s (PU 14420A, PU 14420B) from Mantua Lentil are significantly larger than the mean of those from Bug Creek Anthills and Harbicht Hill (Table 12) (PU 14420 refers to both M1/'s, here separated for discussion by A and B designations). However, the total sample from both Harbicht Hill and Mantua Lentil consists of only four specimens (Table 12). From their size alone, these small samples could be interpreted to represent separate species (Figure 17). When larger samples, such as that from Bug Creek Anthills, are included in the analysis, specimens from Mantua Lentil and Harbicht Hill fall within the range of size variation of the Bug Creek Anthills sample. The largest specimens from Bug Creek Anthills are of similar size or larger than those from Mantua Lentil (Figures 17, 18). Therefore, Stygimys kuszmauli and "S. gratus " are not distinguishable by size. In fact, because of the small samples from Harbicht Hill and Mantua Lentil, the strongest argument for recognizing two species of Stygimys on the basis of size would be those from Harbicht Hill and Mantua Lentil, not Bug Creek Anthills and Mantua Lentil.

The cusp formula of M1/'s of Stygimys is highly variable. The two M1/'s from Mantua Lentil have cusp formulas of 7:7:5 (PU 14420A, Figure 18) and 7:7:3 (PU 14420B, not figured), each with 1 or 2 cuspules on the anterior part of the internal cusp row. The two M1/'s from Harbicht Hill have cusp formulas of 8-7:8-7:5-6. The large

Figure 17
Bivariate plot (length vs. width) of M1/'s of  Stygimys  from various localities (see Figure 13).

sample of M1/'s from Bug Creek Anthills exhibits cusp formulas of 8-7:8-7:3-8, with the number on the internal row being highly variable. Those from Bug Creek Anthills usually have 3-5 cusps on the internal row, but many cusp counts depend in part on whether tiny cuspules are counted as cusps. The cusp morphology and formula of M1/'s of S. kuszmauli compare closely to those from Mantua Lentil. For example, UCMP 73131 has a cusp formula of 7:7:2 with 4 cuspules (Figure 18), and UCMP 103864 8:8:7 with 1 cuspule (not figured). UCMP 37131 and 103864 are two of the largest M1/'s of Stygimys from BCA and are similar in size to those from Mantua Lentil (PU 14420A and 14420B L: 5.23-5.25, W: 2.55-2.57; UCMP 103864 L: 5.35, W: 2.74; UCMP 73131 L: 5.16, W: 2.44). The samples from Bug Creek Anthills and Harbicht Hill are also similar in morphology. Therefore, on the basis of both size and morphology, samples of M1/'s of Stygimys from Bug Creek Anthills, Harbicht Hill, and Mantua Lentil are not distinguishable.

M2/: The Mantua Lentil sample does not contain M2/'s. Harbicht Hill M2/'s have a cusp formula of 1-2:3:3-4, and those from Bug Creek Anthills 1-3:3:3-4. A third external cusp is present only on a single M2/ from Bug Creek Anthills. Size measurements and a bivariate length vs. width plot show that the samples are very similar (Table 12, Figure 19).

From this analysis, the samples of Stygimys from Bug Creek Anthills, Mantua Lentil, and Harbicht Hill cannot be separated with confidence on the basis of size or

Figure 18
Stygimys kuszmauli  Sloan and Van Valen. (A) Isolated right
M1/, UCMP 73131, Bug Creek Anthills (locality V65127),
occlusal view. (B) Isolated right M1/, PU 14420A, Mantua
Lentil, occlusal view. Scale bar = 4 mm.

morphology. Smaller samples of Stygimys (i.e., Mantua Lentil and Harbicht Hill) may exhibit morphological differences, but when larger samples are analyzed (i.e. Bug Creek Anthills), differences appear to have little significance. The morphology of Stygimys is highly variable, which can only be fully appreciated with large samples. The Bug Creek Anthills, Mantua Lentil, and Harbicht Hill samples apparently represent a single, highly variable species of Stygimys that rightfully retains the name S. kuszmauli , the genotypic species.

The sample of Stygimys from the upper Hell Creek Formation at McGuire Creek was compared to those in the UCMP collections from Harbicht Hill (V71203) and Bug Creek Anthills (V65127 and V70201), upper Hell Creek Formation, and Worm Coulee 1 (V74111), lower Tullock Formation. Archibald (1982) studied the Worm Coulee 1 sample and referred it to Stygimys aff. S. kuszmauli . Two of the largest samples from McGuire Creek (V87072, V87037-38) are shown in comparison with Bug Creek Anthills, Harbicht Hill, and Worm Coulee 1 (remeasured for this study) (Figures 13-17, 19). The Stygimys samples from all McGuire Creek sites, Harbicht Hill, Bug Creek Anthills, and Worm Coulee 1 exhibit a wide range of morphological variation, but these differences fade when large samples are analyzed. Fossils from these samples are nearly identical in size and morphology (Tables 12, 13, Figures 13-17, 19), and probably belong to a single species. Using these comparisons and the absence of distinguishing characters, I refer the samples from McGuire Creek sites and Worm Coulee 1 to Stygimys kuszmauli .

Fox (1989) described a new species of Stygimys, S. cupressus from the Long Fall

Figure 19
Bivariate plot (length vs. width) of M2/'s of  Stygimys  from various localities (see Figure 13).

locality of the Ravenscrag Formation in Saskatchewan. The species is diagnosed by the /P4, whose cutting edge has a symmetrical arc with a long, low, inclined anterior edge leading to the first serration (Fox, 1989). Also, specimens in the University of Alberta collections reveal that the /P4 of S. cupressus (n=4) is longer than that of S. kuszmauli (n=9) (Fox, 1989). However, lengths of /P4's from Long Fall (4.9mm-5.4mm) fall within the upper part of the range of variation of the large UCMP sample from Bug Creek Anthills (n=34; see Figure 13 and Table 12). A direct comparison of a large sample of S. kuszmauli from Bug Creek Anthills and the Long Fall sample of S. cupressus would provide an interesting test as to the validity of the latter species because of the highly variable morphology of S. kuszmauli .

The morphology of Stygimys kuszmauli has been described in detail (Archibald, 1982, formerly Stygimys aff. S. kuszmauli ), but some new information is provided by the fossils from McGuire Creek and Harbicht Hill.

Almost all /P4's from McGuire Creek (n=40) and Harbicht Hill (n=8) have 11 serrations. Three of these from the large McGuire Creek sample have only 10 serrations (UCMP 134739, loc. V87072; UCMP 132666, loc. V87037; UCMP 132803, loc. V87078). The P4/'s from Harbicht Hill and McGuire Creek have a cusp formula of 2-3:6-10:1.

McGuire Creek and Harbicht Hill /M2's have a cusp formula of 3-5:2. Ninety percent of the /M2's (n=60) have 4 external cusps. One specimen (UCMP 134725, loc. V87072) has a large fifth external cusp, which is weakly separated from the fourth external cusp by a shallow lingual groove. The M2/'s have a cusp formula of 1-3:3:3-5. Specimens with 4 or 5 internal cusps are rare (5 of 60). When an additional cusp is developed, it is either anterior to and not well separated from the first internal cusp or it is developed posterior to the third internal cusp and deflected medially to occupy a position at the posterior end of the medial valley. One specimen (UCMP 132138, loc. V87151) has cusps developed in both of these positions, giving the tooth 5 cusps in the internal row.

The /M1's have a cusp formula of 6-8:5. When an eighth external cusp is present, it is developed as a robust cusp on a labial bulge within the notch separating cusps 4 and 5, or as a small cusp on a low ridge that extends posteriorly from cusp 4.

The M1/'s of Stygimys kuszmauli from McGuire Creek and Harbicht Hill have a cusp formula of 7-8:7-8:3-7. When cusp 8 is present on the external or medial row, it is small and positioned on the anterior margin of the tooth. Three to 7 cusps comprise the internal row. When more than 5 cusps are present, the internal row extends anteriorly to, or slightly past, the midline of the tooth. Cusps on the internal row become progressively smaller anteriorly, with cusps 6 and 7 more appropriately described as tiny cuspules. This is similar to the condition found in M1/'s of S. kuszmauli from Mantua Lentil (formerly S. gratus ).

Family TAENIOLABIDIDAE Granger and Simpson, 1929
Catopsalis Cope, 1882a

Comments: Catopsalis is a paraphyletic taxon (Simmons and Desui, 1986) known from Upper Cretaceous strata in Asia (Kielan-Jaworowska and Sloan, 1979) and Puercan to Tiffanian strata in North America (Archibald et al., 1987; Archibald and Lofgren, 1990). Two species of the genus, C. joyneri and C. alexanderi , have been reported from eastern Montana (Sloan and Van Valen, 1965; Middleton, 1982). C. joyneri is best known from the Bug Creek sequence of channel fills (Bug Creek Anthills, Bug Creek West, Harbicht Hill) in the upper Hell Creek Formation of McCone County, Montana. The type of C. alexanderi comes from the Alexander locality in the Denver Formation, Arapahoe County, Colorado. The hypodigm of C. alexanderi included specimens originally referred to C. foliatus (see Middleton, 1982, p. 1198). Two specimens also referred to C. alexanderi in the hypodigm were collected from the Tullock Formation of Garfield County, Montana. These fossils (UCMP 116954, M1/ fragment, UCMP loc. V74111; UCMP 124404, M2/, UCMP loc. V74110) were originally referred to Catopsalis cf. C. foliatus (see Archibald, 1982).

Species of Catopsalis display certain evolutionary trends when set in decreasing chronologic order (Kielan-Jaworowska and Sloan, 1979; Middleton, 1982). One of these is a gradual increase in size: Catopsalis joyneri is smaller than C. alexanderi , which in turn is smaller than C. foliatus .

Catopsalis joyneri , and the distinctly larger C. alexanderi , are both present at McGuire Creek, but never at the same locality. This might be a factor of the small sample size (n=10), or may reflect temporal differences between sites.

Catopsalis joyneri Sloan and Van Valen, 1965
Table 14

Catopsalis joyneri Sloan and Van Valen, 1965, p. 225.

Type . UMVP 1494, right maxilla with complete palate, M1/, roots of P4/, alveoli of P3/ and M2/.

Type locality . Bug Creek Anthills, Hell Creek Formation.

Referred specimens . One P4/ from loc. V87071. One M2/, 1 /M1 from loc. V87151. One M2/ from loc. V87153.

Localities . UCMP locs. V87071, V87151, and V87153.

Distribution . Upper Hell Creek Formation, Montana (Puercan); possibly upper Frenchman and Ravenscrag formations, Saskatchewan (both Puercan).

Description . In 1965, Sloan and Van Valen published a brief description of C. joyneri based on fossils collected at Bug Creek Anthills, McCone County, Montana. A more complete description of the species has yet to appear. Therefore, a brief description of the small McGuire Creek sample (n=4) is given below.